Introduction to Geotechnical

Earthquake Engineering1.1 INTl irODUCTIONEarthquake

engineering deal s with the effects of earthquakes on peopl e and their environment
and with methods of reducing those effects. It is a very young discipl ine, many of
itsmost important devel opments having occurred in the past 30 to 40 years.
Earthquake engineeringis a very broad fiel d, drawing on aspects of geol ogy,
seismol ogy, geotechnical engineering, structural engineering, risk anal ysis, and
other technical fiel ds. Its practice al sorequires consideration of social ,
economic, and pol itical factors. Most earthquake engineershave entered the fiel d
from structural engineering or geotechnical engineering backgrounds,a fact that is
refl ected in the practice of earthquake engineering. This book coversgeotechnical
aspects of earthquake engineering. Al though its primary audience is geotechnical
engineering students and practitioners, it contains a great deal of information
thatshoul d be of interest to the structural engineer and the engineering
seismol ogist.1.2 BACKGROUNDThe study of earthquakes dates back many centuries.
Written records of earthquakes inChina date as far back as 3000 years. Japanese
records and records from the eastern Mediterraneanregion go back nearl y 1600 years.
In the United States the historical record of12 Introduction to Geotechnical
Earthquake Engineering Chap. 1earthquakes is much shorter, about 350 years. On the
seismical l y active west coast of theUnited States, earthquake records go back onl y
about 200 years. Compared with the mil l ionsof years over which earthquakes have
been occurring, humankind's experience withearthquakes is very brief.Today,
hundreds of mil l ions of peopl e throughout the worl d l ive with a significantrisk to
their l ives and property from earthquakes. Bil l ions of dol l ars of publ ic
infrastructureare continuousl y at risk of earthquake damage. The heal th of many
l ocal , regional , and evennational economies are al so at risk from earthquakes.
These risks are not unique to theUnited States, Japan, or any other country.
Earthquakes are a gl obal phenomenon and a gl obal probl em.Earthquakes have occurred
for mil l ions of years and wil l continue in the future as theyhave in the past. Some
wil l occur in remote, undevel oped areas where damage wil l be negl igibl e.Others wil l
occur near densel y popul ated urban areas and subject their inhabitantsand the
infrastructure they depend on to strong shaking. It is impossibl e to prevent
earthquakesfrom occurring, but it is possibl e to mitigate the effects of strong
earthquake shaking: to reduce l oss of l ife, injuries, and damage.1.3 SEISMIC HAZARDS
A number of natural l y occurring events, such as earthquakes, hurricanes, tornados,
andfl oods, are capabl e of causing deaths, injuries, and property damage. These
natural hazardscause tremendous damage around the worl d each year. Hazards
associated with earthquakesare commonl y referred to as seismic hazards. The
practice of earthquake engineeringinvol ves the identification and mitigation of
seismic hazards. The most important seismichazards are described in the fol l owing
sections.1.3.1 Ground Shal (ingWhen an earthquake occurs, seismic waves radiate away
from the source and travel rapidl ythrough the earth's crust. When these waves reach
the ground surface, they produce shakingthat may l ast from seconds to minutes. The
strength and duration of shaking at a particul arsite depends on the size and
l ocation of the earthquake and on the characteristics of the site.At sites near the
source of a l arge earthquake, ground shaking can cause tremendous damage.In fact,
ground shaking can be considered to be the most important of al l seismic hazards
because al l the other hazards are caused by ground shaking. Where ground shaking
l evel s are l ow, these other seismic hazards may be l ow or nonexistent. Strong
ground shaking,however, can produce extensive damage from a variety of seismic
hazards.Al though seismic waves travel through rock over the overwhel ming majority
of theirtrip from the source of an earthquake to the ground surface, the final
portion of that trip isoften through soil , and the characteristics of the soil can
greatl y infl uence the nature of shakingat the ground surface. Soil deposits tend to
act as "fil ters" to seismic waves by attenuatingmotion at certain frequencies and
ampl ifying it at others. Since soil conditions oftenvary dramatical l y over short
distances, l evel s of ground shaking can vary significantl ywithin a smal l area. One
of the most important aspects of geotechnical earthquake engineeringpractice
invol ves eval uation of the effects of l ocal soil conditions on strong groundmotion.
In this book, Chapter 3 presents methods for quantifying the most importantSec. 1.3
Seismic Hazards 3characteristics of strong ground motions, and Chapters 4 through 7
provide the backgroundand techniques for site-specific ground motion prediction.
1.3.2 Structural HazardsWithout doubt the most dramatic and memorabl e images of
earthquake damage are those ofstructural col l apse. From the predictabl e col l apse of
the unreinforced masonry and adobestructures in which many residents of
underdevel oped areas of the worl d l ive (Figure 1.1) tothe surprising destruction of
more modern construction (Figures 1.2 to 1.4), structural damageis the l eading
cause of death and economic l oss in many earthquakes. However, structuresneed not
col l apse to cause death and damage. Fal l ing objects such as brick facings and
parapets on the outside of a structure or heavy pictures and shel ves within a
structure havecaused casual ties in many earthquakes. Interior facil ities such as
piping, l ighting, and storagesystems can al so be damaged during earthquakes.Over
the years, considerabl e advances have been made in earthquake-resistant designof
structures, and seismic design requirements in buil ding codes have steadil y
improved. Asearthquake-resistant design has moved from an emphasis on structural
strength to emphaseson both strength and ductil ity, the need for accurate
predictions of ground motions hasincreased. In current design practice, the
geotechnical earthquake engineer is often responsibl efor providing the structural
engineer with appropriate design ground motions. In thisbook, Chapter 8 describes
the effects of l ocal soil conditions on ground motions and providesguidance for the
devel opment of site-specific design ground motions.Figure 1.1 Damage to buil dings
in Huaras, Peru fol l owing tbe 1970 Peru eartbquake.The adobe structures in the
foreground were destroyed, but tbe reinforced concretestructure in tbe background
suffered l ittl e damage (photo by G. Pl afker, courtesy of USGS).4Figure 1.2
Col l apsed portion of the reinforced concrete Hospital Juarez in Mexico City
fol l owing the 1985 Mexico earthquake (photo by E.V. Leyendecker, courtesy of EERI).
Figure 1.3 Effects of col umn fail ures at Ol ive View Hospital in the 1971 San
Fernandoearthquake. Col l apse of the canopy in the foreground pinned the ambul ances
beneaththem, rendering them usel ess (courtesy of EERI).Sec. 1.3 Seismic Hazards 5
1,,3.3 LiquefactionFigure 1.4 Reinforced concrete col umn atOl ive View Hospital
fol l owing the 1971San Fernando earthquake. Insufficienttransverse reinforcement was
unabl e toprovide adequate confinement (courtesy ofUSGS).Some of the most
spectacul ar exampl es of earthquake damage have occurred when soil deposits have l ost
their strength and appeared to fl ow as fl uids. In this phenomenon, termed
l iquefaction, the strength of the soil is reduced, often drastical l y, to the point
where it isunabl e to support structures or remain stabl e. Because it onl y occurs in
saturated soil s, l iquefactionis most commonl y observed near rivers, bays, and other
bodies of water.The term l iquefaction actual l y encompasses several rel ated
phenomena. Fl ow fail ures,for exampl e, can occur when the strength of the soil drops
bel ow the l evel needed tomaintain stabil ity under static conditions. Fl ow fail ures
are therefore driven by static gravitational forces and can produce very l arge
movements. Fl ow fail ures have caused the col l apseof earth dams (Figure 1.5) and
other sl opes, and the fail ure of foundations (Figure 1.6).The 1971 San Fernando
earthquake caused a fl ow fail ure in the upstream sl ope of the LowerSan Fernando Dam
(Figure 1.7) that nearl y breached the dam. Thousands coul d have been'kiIl ed in the
residential area immediatel y bel ow the dam. Lateral spreading is a rel ated
phenomenoncharacterized by incremental displ acements during earthquake shaking.
Dependingon the number and strength of the stress pul ses that exceed the strength
of the soil , l ateral spreading can produce displ acements that range from negl igibl e
to quite l arge. Lateral spreading is quite common near bridges, and the
displ acements it produces can damage theabutments, foundations, and superstructures
of bridges (Figures 1.8 and 1.9). Final l y, the6Figure 1.5 Liquefaction fail ure of
Sheffiel d Dam fol l owing the 1925 Santa Barbaraearthquake (K. Steinbrugge
col l ection; courtesy of EERC, Univ. of Cal ifornia).Figure 1.6 Liquefaction-induced
bearing capacity fail ures of the Kawagishi-choapartment buil dings fol l owing the
1964 Niigata earthquake (courtesy of USGS).Sec. 1.3 Seismic Hazards 7Figure 1.7
Lower San Fernando Dam fol l owing l iquefaction fail ure of its upstreamsl ope in the
1971 San Fernando earthquake (K. Steinbrugge col l ection; courtesy ofEERC, Univ. of
Cal ifornia).Figure 1.8 Effect of l ateral spreading on a smal l bridge in Japan
fol l owing the 1952Tokachi-Oki earthquake. Lateral spreading of the soil at the
abutment buckl ed the bridgedeck (K. Steinbrugge col l ection; courtesy of EERC, Univ.
of Cal ifornia).8 Introduction to Geotechnical Earthquake Engineering Chap. 1Figure
1.9 The Showa Bridge fol l owing the 1964 Niigata earthquake. Lateral spreading
caused bridge pier foundations to move and rotate sufficientl y for simpl ysupported
bridge spans to fal l (courtesy of USGS).phenomenon of l evel -ground l iquefaction
does not invol ve l arge l ateral displ acements butis easil y identified by the
presence of sand boil s (Figure 1.10) produced by groundwaterrushing to the surface.
Al though not particul arl y damaging by themsel ves, sand boil s indicatethe presence
of high groundwater pressures whose eventual dissipation can producesubsidence and
damaging differential settl ements.Liquefaction is a compl icated phenomenon, but
research has progressed to the pointwhere an integrated framework of understanding
can be devel oped. Chapter 9 of this bookpresents the basic concepts with which the
susceptibil ity, triggering conditions, and effectsof al l l iquefaction phenomena can
be understood, together with practical procedures foreval uation of l iquefaction
hazards.1.3.4 Landsl idesStrong earthquakes often cause l andsl ides. Al though the
majority of such l andsl ides aresmal l , earthquakes have al so caused very l arge
sl ides. In a number of unfortunate cases,earthquake-induced l andsl ides have buried
entire towns and vil l ages (Figure 1.11). Morecommonl y, earthquake-induced
l andsl ides cause damage by destroying buil dings, or disruptingbridges and other
constructed facil ities (Figures 1.12 and 1.13). Many earthquakeinducedl andsl ides
resul t from l iquefaction phenomena, but many others simpl y representthe fail ures of
sl opes that were marginal l y stabl e under static conditions. Various types ofseismic
sl ope fail ures, their frequency of occurrence, and procedures for their anal ysis
aredescribed in Chapter 10.Sec. 1.3 Seismic Hazards 9Figure 1.10 Sand boil in rice
fiel d fol l owing the 1964 Niigata earthquake (K.Steinbrugge col l ection; courtesy of
EERC, Univ. of Cal ifornia).1,,3.5 Retaining Structure Fail uresAnchored bul kheads,
quay wal l s, and other retaining structures are frequentl y damaged inearthquakes.
Damage is usual l y concentrated in waterfront areas such as ports and harbors(Figure
1.14). Because such facil ities are often essential for the movement of goods upon
which l ocal economies often rel y, the business l osses associated with their fail ure
can go farbeyond the costs of repair or reconstruction. The seismic design of
retaining structures iscovered in Chapter 11.1,,3.6 Lifel ine HazardsA network of
facil ities that provide the services required for commerce and publ ic heal th canbe
found in virtual l y any devel oped area. These networks, which incl ude el ectrical
powerand tel ecommunications, transportation, water and sewage, oil and gas
distribution, andwaste storage systems, have col l ectivel y come to be known as
l ifel ines. Lifel ine systemsmay incl ude power pl ants, transmission towers, and
buried el ectrical cabl es; roads, bridges,harbors, and airports; water treatment
facil ities, reservoirs and el evated water tanks, and buriedwater distribution
systems; l iquid storage tanks and buried oil and gas pipel ines; andmunicipal sol id
waste and hazardous waste l andfil l s. Lifel ine systems and the facil ities that
comprise them provide services that many take for granted but which are essential
in modernindustrial areas. Lifel ine fail ures not onl y have severe economic
consequences but can al soadversel y affect the environment and qual ity of l ife
fol l owing an earthquake.10(a)(b)Figure 1.11 Vil l age of Yungay, Peru, (a) before and
(b) after being buried by a giantl andsl ide in the 1970 Peruvian earthquake. The
same pal m trees are visibl e at the l eftside of both photographs. The l andsl ide
invol ved 50 mil l ion cubic meters of material thateventual l y covered an area of some
8000 square kil ometers. About 25,000 peopl e werekil l ed by this l andsl ide, over
18,000 in the vil l ages of Yungay and Ranrahirca (K.Steinbrugge col l ection; courtesy
of EERC, Univ. of Cal ifornia).Figure 1.12 A wing of Government Hil l School in
Anchorage, Al aska, straddl ed thehead scarp of the Government Hil l l andsl ide in the
1964 Good Friday earthquake (K.Steinbrugge col l ection; courtesy of EERC, Univ. of
Cal ifornia).Figure 1.13 Earthquake-induced l andsl ide al ong rail road tracks near
Ol ympia,Washington (photo by G.W. Thorsen). II12 Introduction to Geotechnical
Earthquake Engineering Chap. 1Figure 1.14 Fail ure of a quay wal l on Rokko Isl and in
Kobe, Japan in the 1995 HyogoKenNanbu earthquake (photo by S. L. Kramer).Lifel ine
fail ure can cause disruption and economic l osses that greatl y exceed the costof
repairing facil ities directl y damaged by earthquake shaking. The 1989 Loma Prieta
and1994 Northridge earthquakes caused economic l osses estimated at $8 bil l ion and
$30 bil l ionin the state of Cal ifornia al one. These l osses had severe l ocal and
regional repercussions buthad onl y minor effects on most U.S. citizens. The 1972
Managua earthquake, on the otherhand, caused l osses of $2 bil l ion, 40% of
Nicaragua's gross national product that year. Thehigh costs of reconstruction
produced a national debt that triggered infl ation, increasedunempl oyment, and
eventual l y contributed to the destabil ization of the Nicaraguan government.More
recentl y, the Hyogo- Ken Nanbu earthquake devastated the city of Kobe, Japan;total
damages have been estimated in excess of $100 bil l ion.Lifel ine fail ures can al so
hamper emergency response and rescue efforts immediatel yfol l owing damaging
earthquakes. Most of the damage in the 1906 San Francisco earthquake,for exampl e,
was caused by a fire that coul d not be fought properl y because of brokenwater
mains. Eighty-three years l ater, tel evision al l owed the worl d to watch another fire
in San Francisco fol l owing the Loma Prieta earthquake. These fires were caused by
brokennatural gas pipes, and again, firefighting was hampered by broken water
mains. The LomaPrieta earthquake al so caused the col l apse and near col l apse of
several el evated highwaysand the col l apse of a portion of the San Francisco-Oakl and
Bay Bridge. Loss of these transportationl ifel ines caused gridl ock throughout the
area. Some of the el evated highways werestil l out of service five years after the
earthquake.Sec. 1.3 Seismic Hazards 131.3.7 Tsunami and Seiche HazardsRapid
vertical seafl oor movements caused by faul t rupture during earthquakes can produce
l ong-period sea waves cal l ed tsunamis. In the open sea, tsunamis travel great
distances at highspeeds but are difficul t to detect-they usual l y have heights of
l ess than I m and wavel engths(the distance between crests) of several hundred
kil ometers. As a tsunami approaches shore,however, the decreasing water depth
causes its speed to decrease and the height of the waveto increase. In some coastal
areas, the shape of the seafl oor may ampl ify the wave, producinga nearl y vertical
wal l of water that rushes far inl and and causes devastating damage (Figure1.15).
The Great Hoei Tokaido--Nonhaido tsunami kil l ed 30,000 peopl e in Japan in 1707. The
1960 Chil ean earthquake produced a tsunami that not onl y kil l ed 300 peopl e in
Chil e, but al sokil l ed 61 peopl e in Hawaii and, 22 hours l ater, 199 peopl e in
distant Japan (Iida et aI., 1967).Earthquake-induced waves in encl osed bodies of
water are cal l ed seiches. Typical l ycaused by l ong-period seismic waves that match
the natural period of oscil l ation' of thewater in a l ake or reservoir, seiches may
be observed at great distances from the source ofan earthquake. The 1964 Good
Friday earthquake in Al aska, for exampl e, produced damagingwaves up to 5 ft high in
l akes in Louisiana and Arkansas (Spaeth and Berkman, 1967).Another type of seiche
can be formed when faul ting causes permanent vertical displ acementswithin a l ake or
reservoir. In 1959, vertical faul t movement within Hebgen Lake produceda seiching
motion that al ternatel y overtopped Hebgen Dam and exposed the l akebottom adjacent
to the dam in 1959 (Steinbrugge and Cl oud, 1962).Figure 1.15 Tsunami damage in
Kodiak, Al aska, fol l owing the 1964 Good Fridayearthquake (courtesy of USGS).14
Introduction to Geotechnical Earthquake Engineering Chap. 11.4 MITIGATION OF
SEISMIC HAZARDSUl timatel y, the goal of the earthquake engineer is to mitigate
seismic hazards. For new construction,hazard mitigation is embedded in the process
of earthquake-resistant design.Detail s of earthquake-resistant design of structures
are beyond the scope of this book, butsome aspects of earthquake l oading of
structures are described in Chapter 8. Earthquakeresistantdesign of sl opes, dams,
embankments, and retaining structures is based on topicspresented in Chapters 9 to
11. Mitigation of existing seismic hazards is al so very important.The important
topic of remediation of soil deposits for seismic hazard mitigation is coveredin
Chapter 12.1.5 SIGNIFICANT HISTORICAL EARTHQUAKESEarthquakes occur al most
continuousl y around the worl d. Fortunatel y, most are so smal l that they cannot be
fel t. Onl y a very smal l percentage of earthquakes are l arge enough tocause
noticeabl e damage, and a smal l percentage of those are l arge enough to be
consideredmajor earthquakes. Throughout recorded history, some of these major
earthquakes can beregarded as being particul arl y significant, either because of
their size and the damage theyproduced or because ofwhat scientists and engineers
were abl e to l earn from them. A partial l ist of significant earthquakes, admittedl y
biased toward U.S. earthquakes and earthquakeswith significant geotechnical
earthquake engineering impl ications, is given in Tabl e 1-1.TABLE 1-1 Significant
Historical EarthquakesDate780 B.C.A.D. 79893155617551783LocationChinaItal yIndia
ChinaPortugal Ital yMagnitude8.0 (est.)8.6Deaths180,000530,00060,00050,000CommentsOne
of the first rel iabl e writtenaccounts of a strong earthquake;produced widespread
damage westof Xian in Shaanxi ProvinceSixteen years of frequent earthquakes
cul minating with the
eruption ofMt. Vesuvius, which buried the cityof PompeiiWidespread damage; many
kil l ed incol l apse of earthen homesOccurred in densel y popul ated regionnear Xian;
produced thousands ofl andsl ides, which kil l ed inhabitantsof soft rock caves in
hil l sides;death estimate of questionabl eaccuracyLisbon earthquake; first scientific
description of earthquake effectsCal abria earthquake; first scientificcommission
for earthquakeinvestigation formedSec. 1.5 Significant Historical Earthquakes 15
Tt~BLE 1-1 Significant Historical Earthquakes (continued)1811-1812 Missouri
7.5,7.3,7.8 Several Three l arge earthquakes in l ess thantwo months in New Madrid
area;fel t al l across central and easternUnited States1819 India 1,500 Cutch
earthquake; first wel l documentedobservations offaul ting1857 Cal ifornia 8.3 Fort
Tejon earthquake; one of thel argest earthquakes known to havebeen produced by the
San AndreasFaul t; faul t ruptured for 250 mil es(400 km) with up to 30 ft (9 m)offset
1872 Cal ifornia 8.5 27 Owens Val l ey earthquake; one of thestrongest ever to have
occurred inthe United States1886 South Carol ina 7.0 110 Strongest earthquake to
strike eastcoast of United States; producedsignificant l iquefaction1906 Cal ifornia
7.9 700 First great earthquake to strikedensel y popul ated area in UnitedStates;
produced up to 21 ft (7 m)offset in 270-mil e (430-km) ruptureof San Andreas Faul t;
most damagecaused by fire; extent of groundshaking damage correl ated togeol ogic
conditions inpostearthquake investigation1908 Ital y 7.5 83,000 Messina and
surrounding areadevastated; Ital ian governmentappointed engineering commissionthat
recommended structures bedesigned for equival ent staticl ateral l oads1923 Japan 7.9
99,000 Kanto earthquake; caused majordamage in Tokyo-Yokohama area,much due to fire
in Tokyo andtsunami in coastal regions; strongl yinfl uenced subsequent design in
Japan1925 Cal ifornia 6.3 13 Santa Barbara earthquake; causedl iquefaction fail ure of
Sheffiel dDam; l ed to first expl icit provisionsfor earthquake resistance in U.S.
buil ding codes16 Introduction to Geotechnical Earthquake Engineering Chap. 1TABLE
1-1 Significant Historical Earthquakes (continued)1933 Cal ifornia 6.3 120
Considerabl e buil ding damage;school s particul arl y hard-hit, withmany chil dren
kil l ed and injured;l ed to greater seismic designrequirements in buil ding codes,
particul arl y for publ ic school buil dings1940 Cal ifornia 7.1 9 Large ground
displ acements al ongImperial Faul t near EI Centro; firstimportant accel erogram for
engineering purposes was recorded1959 Montana 7.1 28 Hebgen Lake earthquake;
faul tingwithin reservoir produced l argeseiche that overtopped earth dam1960 Chil e
9.5 2,230 Probabl y the l argest earthquake everrecorded1964 Al aska 9.2 131 The Good
Friday earthquake; causedsevere damage due to l iquefactionand many earthquake-
inducedl andsl ides1964 Japan 7.5 26 Widespread l iquefaction causedextensive damage
to buil dings,bridges, and port facil ities inNiigata; al ong with Good Friday
earthquake in Al aska, spurredintense interest in the phenomenonof l iquefaction1967
Venezuel a 6.5 266 Caused col l apse of rel ativel y newstructures in Caracas;
il l ustratedeffects of l ocal soil conditions onground motion and damage1971
Cal ifornia 6.6 65 San Fernando earthquake; producedseveral exampl es of
l iquefaction,incl uding near col l apse of LowerSan Fernando Dam; causedcol l apse of
several buil dings andhighway bridges; many structural l essons l earned, particul arl y
regarding need for spiral reinforcement of concrete col umns;many strong motion
recordsobtained1975 China 7.3 1,300 Evacuation fol l owing successful prediction saved
thousands ofl ivesin Haicheng, Liaoning Province1976 China 7.8 700,000 Thought to be
the most deadl yearthquake in history; destroyedcity of Tangshan, Hebei Province;not
predicted; death estimate ofquestionabl e accuracySec. 1.5 Significant Historical
Earthquakes 17TABLE 1-1 Significant Historical Earthquakes (continued)1985 Mexico
8.1 9,500 Epicenter off Pacific Coast, but mostdamage occurred over 220 mil es(360
km) away in Mexico City;il l ustrated effect of l ocal soil conditions on ground motion
ampl ification and damage;subsequent studies l ed to betterunderstanding of dynamic
properties of fine-grained soil s1989 Cal ifornia 7.1 63 Lorna Prieta earthquake;
extensiveground motion ampl ification andl iquefaction damage in SanFrancisco Bay
area1994 Cal ifornia 6.8 61 Northridge earthquake; occurred onpreviousl y unknown
faul t beneathheavil y popul ated area; buil dings,bridges, l ifel ines extensivel y
damaged; produced extraordinaril ystrong shaking at several l ocations1995 Japan 6.9
5,300 Hyogo-Ken Nanbu earthquake;caused tremendous damage toKobe, Japan; widespread
l iquefaction in recl aimed l andsconstructed for port of Kobe;l andsl ides and damage
to retainingwal l s and underground subwaystations al so observed2Seismol ogy and
Earthquakes2.1 INTRODUCTIONThe study of geotechnical earthquake engineering
requires an understanding of the variousprocesses by which earthquakes occur and
their effects on ground motion. The fiel d of seismol ogy(from the Greek seismos for
earthquake and l ogos for science) devel oped from aneed to understand the internal
structure and behavior of the earth, particul arl y as they rel ateto earthquake
phenomena. Al though earthquakes are compl ex phenomena, advances inseismol ogy have
produced a good understanding of the mechanisms and rates of occurrenceof
earthquakes in most seismical l y active areas of the worl d. This chapter provides a
briefintroduction to the structure of the earth, the reasons why earthquakes occur,
and the terminol ogyused to describe them. More compl ete descriptions of these
topics may be found ina number of seismol ogy texts, such as Gutenberg and Richter
(1954), Richter (1958),Bul l en (1975), Bath (1979), Bul l en and Bol t (1985), Gubbins
(1990), and Lay and Wal l ace(1995). A very readabl e description of seismol ogy and
earthquakes is given by Bol t (1993).2.2 INTERNAL STRUCTURE OF THE EARTHThe earth is
roughl y spherical , with an equatorial diameter of 12,740 km (7918 mil es) anda pol ar
diameter of 12,700 km (7893 mil es), the higher equatorial diameter being caused by
18Sec. 2.2 Internal Structure of the Earth 19higher equatorial vel ocities due to
the earth's rotation. The earth weighs some 5.4 x 102 1tons (4.9 x 1024 kg), which
indicates an average specific gravity of about 5.5. Since the specificgravity of
surficial rocks is known to be on the order of 2.7 to 3, higher specific gravities
are impl ied at greater depths.One of the first important achievements in seismol ogy
was the determination of theinternal structure of the earth. Large earthquakes
produce enough energy to cause measurabl eshaking at points al l around the worl d. As
the different types of seismic waves travel through the earth, they are refracted
and refl ected at boundaries between different l ayers,reaching different points on
the earth's surface by different paths. Studies of these refractionsand refl ections
earl y in this century reveal ed the l ayered structure of the earth and provided
insight into the characteristics of each l ayer.2.2.1 Seismic WavesWhen an
earthquake occurs, different types of seismic waves are produced: body waves and
surface waves. Al though seismic waves are discussed in detail in Chapter 5, the
briefdescription that fol l ows is necessary to expl ain some of the concepts of
Chapters 2 to 4.Body waves, which can travel through the interior of the earth, are
of two types: p-waves and s-waves (Figure 2.1). P-waves, al so known as primary,
compressional , or l ongitudinal waves, invol ve successive compression and rarefaction
of the material s throughwhich they pass. They are anal ogous to sound waves; the
motion of an individual particl ethat a p-wave travel s through is paral l el to the
direction of travel . Like sound waves,p-waves can travel through sol ids and fl uids.
S-waves, al so known as secondary, shear, ortransverse waves, cause shearing
deformations as they travel through a material . Themotion of an individual particl e
is perpendicul ar to the direction of s-wave travel . The directionof particl e
movement can be used to divide s-waves into two components, SV (vertical
Compressions/ "(a)" /RarefactionsI.. ~IWavel engthUndisturbed medium/ "(b)Figure 2.1
Deformations produced by body waves: (a) p-wave; (b) SY-wave. FromEarthquakes by
Bol t. Copyright © 1993 by W.H. Freeman and Company. Used withpermission.20
Seismol ogy and Earthquakes Chap. 2pl ane movement) and SH (horizontal pl ane
movement). The speed at which body wavestravel varies with the stiffness of the
material s they travel through. Since geol ogic material sare stiffest in compression,
p-waves travel faster than other seismic waves and are thereforethe first to arrive
at a particul ar site. Fl uids, which have no shearing stiffness, cannot sustains-
waves,Surface waves resul t from the interaction between body waves and the surface
andsurficial l ayers of the earth. They travel al ong the earth's surface with
ampl itudes thatdecrease roughl y exponential l y with depth (Figure 2.2). Because of
the nature of the interactionsrequired to produce them, surface waves are more
prominent at distances fartherfrom the source of the earthquake. At distances
greater than about twice the thickness of theearth's crust, surface waves, rather
than body waves, wil l produce peak ground motions.The most important surface waves,
for engineering purposes, are Rayl eigh waves and Lovewaves. Rayl eigh waves,
produced by interaction of p- and SV-waves with the earth's surface,invol ve both
vertical and horizontal particl e motion. They are simil ar, in somerespects, to the
waves produced by a rock thrown into a pond. Love waves resul t from theinteraction
of SH-waves with a soft surficial
l ayer and have no vertical component of particl emotion.Undisturbed medium/ "
Wavel engthI.. ..I(b) III1IIIIIIIUndisturbed medium/ "Figure 2.2 Deformations
produced by surface waves: (a) Rayl eigh wave; and (b) Lovewave. From Earthquakes by
Bol t. Copyright ©1993 by W.H. Freeman and Company.Used with permission.2.2.2
Internal StructureThe crust, on which human beings l ive, is the outermost l ayer of
the earth. The thickness ofthe crust ranges from about 25 to 40 km (15 to 25 mil es)
beneath the continents (al thoughit may be as thick as 60 to 70 km (37 to 44 mil es)
under some young mountain ranges) to asthin as 5 km (3 mil es) or so beneath the
oceans-onl y a very smal l fraction of the earth'sdiameter (Figure 2.3). The internal
structure of the crust is compl ex but can be representedby a basal tic l ayer that is
overl ain by a granitic l ayer at continental l ocations. Since it isSec. 2.2 Internal
Structure of the Earth 21Upper mantl e ---------;;~Lower mantl eInner coreRadius
-~---/Outer core(l iquid)Figure 2.3 Internal structure of the earth.exposed to the
oceans or the atmosphere, the crust is cool er than the material s bel ow it(Figure
2.4). In addition to being thinner, the oceanic crust is general l y more uniform and
more dense than the continental crust.A distinct change in wave propagation
vel ocity marks the boundary between the crustand the underl ying mantl e. This
boundary is known as the Mohorovicic discontinuity, orthe Moho, named after the
seismol ogist who discovered it in 1909. Al though the specificnature of the Moho
itsel f is not wel l understood, its rol e as a refl ector and refractor of seismic
waves is wel l establ ished. The mantl e is about 2850 km (1770 mil es) thick and can
beoo~oooC')oooC\JMil es bel ow surfaceooo,....,, , I
• : : : : : : : : : : : : : : : : : : : : : : : : : r: : : : : : : : : : : : : : : : : t: : : : : ~: : : : : : : : : : : : : : : : : : : : : : : : : : r: : : : r
: : : : : : : : : : : : : : : : : : : : ________________. ~ -__l -----~- . ~- ~ - --~--iii i •
--------------------------j--------------------------~---
---------------i----: -------------------------------------------- i
-----------------T-----r-------------: -------------r----;---------------------
'..: : : : : : : : : : : : : : : : : : : : : : r: : : : : : : : : : : : : : : : : r: : : : r: : : : : : : : : : : : : : : : : : : : : : : : r: : : : r: : : :
: : : : : : : : : : : : : : : i Mantl e iOuter core i 'Inner core--------------------------
r-------------------~-----r--------------------------r----~--------------------
: : : OL.-------'J ------'---: '------'---J--.• '-------'o8000E6000 e: : : : s~ 4000c.ECD
F 2000Figure 2.4 Estimated variation of temperature bel ow the surface of the earth.
(AfterVerhoogan, 1960.)22 Seismol ogy and Earthquakes Chap. 2divided into the upper
mantl e (shal l ower than about 650 km (404 mil es)) and the l owermantl e. No
earthquakes have been recorded in the l ower mantl e, which exhibits a uniform
vel ocity structure and appears to be chemical l y homogeneous, except near its l ower
boundary.The mantl e is cool er near the crust than at greater depths but stil l has
an average temperatureof about 4000°F. As a resul t, the mantl e material s are in a
viscous, semimol tenstate. They behave as a sol id when subjected to rapidl y appl ied
stresses, such as those associatedwith seismic waves, but can sl owl y fl ow l ike a
fl uid in response to l ong-term stresses.The mantl e material has a specific gravity
of about 4 to 5.The outer core, or l iquid core, is some 2260 km (1400 mil es) thick.
As a l iquid, it cannottransmit s-waves. As shown in Figure 2.5, the s-wave vel ocity
drops to zero at the coremantl eboundary, or Gutenberg discontinuity; note al so the
precipitous drop in p-wavevel ocity. The outer core consists primaril y of mol ten
iron (which hel ps expl ain its high specificgravity of 9 to 12). The inner core, or
sol id core, is a very dense (specific gravity up toabout 15), sol id nickel -iron
material compressed under tremendous pressures. The temperatureof the inner core is
estimated to be rel ativel y uniform at over 5000°F.Figure 2.5 Variation of p- and s-
wavevel ocity and density within the earth. (AfterEiby, 1980.),,, -----,------,-, ,,
,,,,----r----- ---- ,,,,,____ L __1000 2000 3000 4000 5000 KM,,,,I I I I
--,------r-----'------r-----,----I I I I II I I I II I I I II I I I II I I I I14U
CIl ..!!1E : ..~·u0~014~CO>E..!: !.9.>..- ·iiieCIl C00CrustFigure 2.6 shows the
infl uence of the earth's structure on the distribution of seismicwaves during
earthquakes. Since wave propagation vel ocities general l y increase withdepth, wave
paths are usual l y refracted back toward the earth's surface. An exception is atthe
core-mantl e boundary, where the outer core vel ocity is l ower than the mantl e
vel ocity.Sec. 2.3 Continental Drift and Pl ate Tectonics103°~---_/Focus23Innercore
OutercoreFigure 2.6 Seismic wave paths il l ustrating refl ection and refraction of
seismic wavesfrom the source (focus) of the earthquake by the different l ayers of
the earth. Note thatp- and s-waves can reach the earth's surface between 0 and
103°, but the l iquid nature ofthe outer core al l ows onl y p-waves to reach the
surface between 143 and 180°. In theshadow zone between angl es of 103 and 143°,
onl y paths refl ected from the inner corecan reach the earth's surface. (From
Sumner, 1969.)2.3 COrl iTINENTAL DRIFT AND PLATE TECTONICSAl though observations of
simil arity between the coastl ines and geol ogy of eastern SouthAmerica and the
western Africa and the southern part ofIndia and northern part of Austral iahad
intrigued scientists since the seventeenth century (Gl en 1975; Kearey and Vine,
1990),the theory that has come to be known as continental drift was not proposed
until the earl ytwentieth century (Tayl or, 1910; Wegener, 1915). Wegener, for
exampl e, bel ieved that theearth had onl y one l arge continent cal l ed Pangaea 200
mil l ion years ago. He bel ieved thatPangaea broke into pieces that sl owl y drifted
(Figure 2.7) into the present configuration ofthe continents. A more detail ed view
of the current simil arity of the African and SouthAmerican coasts is shown in
Figure 2.8.The theory of continental drift did not receive much attention until
about 1960, whenthe current worl dwide network of seismographs was abl e to define
earthquake l ocationsaccuratel y, and to confirm that l ong-term deformations were
concentrated in narrow zonesbetween rel ativel y intact bl ocks of crust. Al so,
expl oration of the ocean fl oor did not beginin earnest until after Worl d War II,
when new techniques such as deep-water echo sounding,seismic refraction, and piston
coring became avail abl e. Thb geol ogy of the ocean fl oor isyoung, representing onl y
about 5% of the earth's history (Gubbins, 1990), and rel ativel ysimpl e. Its detail ed
study provided strong supporting evidence of the historical movementof the
continents as assumed in the theory of continental drift. Within 10 years, the
theoryof continental drift had become widel y accepted and acknowl edged as the
greatest advancein the earth sciences in a century.24 Seismol ogy and Earthquakes
Chap. 2(a)(b)(c)Figure 2.7 Wegener's theory of contineotal drift: (a) 270 mil l ion
years ago; (b) 150mil l ion years ago; (c) 1 mil l ion years ago. (After Verney, 1979.)
2.3.1 Pl ate TectonicsThe original theory of continental drift suggested images of
massive continents pushingthrough the seas and across the ocean fl oor. It was wel l
known, however, that the oceanfl oor was too strong to permit such motion, and the
theory was original l y discredited bymost earth scientists. From this background,
however, the modern theory ofpl ate tectonicsbegan to evol ve. The basic hypothesis
of pl ate tectonics is that the earth's surface consistsof a number of l arge, intact
bl ocks cal l ed pl ates, and that these pl ates move with respect toSec. 2.3
Continental Drift and Pl ate Tectonics 25Key• Overl apGapFigure 2.8 Statistical
spherical fit of several continents using the continental sl opesrather than the
coastl ines. (After Bul l ard et al ., 1965, with permission of the Royal Society.)each
other. The earth's crust is divided into six continental -sized pl ates (African,
American,Antarctic, Austral ia-Indian, Eurasian, and Pacific) and about 14 of
subcontinental size(e.g., Caribbean, Cocos, Nazca, Phil ippine, etc.). The major
pl ates are shown in Figure2.9. Smal l er pl atel ets, or micropl ates, have broken off
from the l arger pl ates in the vicinityof many of the major pl ate boundaries but are
not shown here. The rel ative deformationbetween pl ates occurs onl y in narrow zones
near their boundaries. This deformation ofthe pl ates can occur sl owl y and
continuousl y (aseismic deformation) or can occur spasmodical l yin the form of
earthquakes (seismic deformation). Since the deformation occurspredominantl y at the
boundaries between the pl ates, it woul d be expected that the l ocationsof
earthquakes woul d be concentrated near pl ate boundaries. The map of earthquake
epicenters shown in Figure 2.10 confirms this expectation, thereby providing strong
supportfor the theory of pl ate tectonics.The theory of pl ate tectonics is a
kinematic theory (i.e., it expl ains the geometry ofpl ate movement without
addressing the cause of that movement). Something must drivethe movement, however,
and the tremendous mass of the moving pl ates requires that theNe-Subduction zone
Strike-sl ip (transform) faul ts",00;°&-'l >-~.c: ''0At\a"'~Antarctic Pl ateUncertain
pl ate boundaryRidge axisMexicotrenchCocos,Pl ateKermadec-Tonga : ttrench 'CKey
Pacific'", \.Pl ate ""Antarctic Pl ate, . JjMacquarie l \ic ridge': \.'7>- ~ridge ~~.~d
paC'Marianastrench'" '"Figure 2.9 The major tectonic pl ates, mid-oceanic ridges,
trenches, and transform faul ts of the earth. Arrows indicate directions of pl ate
movement. (After Fowl er,1990.)hi""Figure 2.10 Worl dwide seismic activity. The dots
represent the epicenters of significant earthquakes. It is apparent that the
l ocations of the
great majority ofearthquakes correspond to the boundaries between pl ates. (After
Bol t, 1988.)28 Seismol ogy and Earthquakes Chap. 2driving forces be very l arge. The
most widel y accepted expl anation of the source of pl atemovement rel ies on the
requirement of thermomechanical equil ibrium of the earth's material s.The upper
portion of the mantl e is in contact with the rel ativel y cool crust whil e thel ower
portion is in contact with the hot outer core. Obviousl y, a temperature gradient
mustexist within the mantl e (see Figure 2.4). The variation of mantl e density with
temperatureproduces the unstabl e situation of denser (cool er) material resting on
top of l ess dense(warmer) material . Eventual l y, the cool er, denser material begins
to sink under the action ofgravity and the warmer, l ess dense material begins to
rise. The sinking material gradual l ywarms and becomes l ess dense; eventual l y, it
wil l move l ateral l y and begin to rise again assubsequentl y cool ed material begins
to sink. This process is the famil iar one of convection.Convection currents in the
semimol ten rock of the mantl e, il l ustrated schematical l y inFigure 2.11, impose
shear stresses on the bottom of the pl ates, thus "dragging" them in various
directions across the surface of the earth. Other phenomena, such as ridge push or
sl abpul l , may al so contribute to the movement of pl ates (Hager, 1978).Tectonic
pl ate/Spreading ridge/ boundaryCOLDFigure 2.11 Convection currents in mantl e. Near
the bottom of the crust, horizontal component of convection currents impose shear
stresses on bottom of crust, causingmovement of pl ates on earth's surface. The
movement causes the pl ates to move apart insome pl aces and to converge in others.
(After Noson et al ., 1988.)Sec. 23 Continental Drift and Pl ate Tectonics 292.3.2
Pl ate BoundariesThree distinct types of pl ate boundaries have been identified, and
understanding the movementassociated with each wil l aid in the understanding of
pl ate tectonics. The characteristicsof the pl ate boundaries al so infl uence the
nature of the earthquakes that occur al ong them.2.3.2.1 Spreading Ridge Boundaries
In certain areas the pl ates move apart from each other (Figure 2.12) at boundaries
known as spreading ridges or spreading rifts. Mol ten rock from the underl ying
mantl e risesto the surface where it cool s and becomes part of the spreading pl ates.
In this way, the pl ates"grow" at the spreading ridge. Spreading rates range from
approximatel y 2 to 18 cm/yr (l to7 infyr); the highest rates are found in the
Pacific Ocean ridges and the l owest al ong theMid-Atl antic Ridge. It is estimated
(Garfunkel , 1975) that new oceanic crust is currentl yformed at a rate of about 3.1
kmvyr (1.2 mil esvyr) worl dwide. The crust, mainl y young,fresh basal t, is thin in
the vicinity of the spreading ridges. It may be formed by rel ativel ysl ow upward
movement of magma, or it may be ejected quickl y during seismic activity.Underwater
photographs have shown formations of pil l ow l ava and have even recordedl ava
eruptions in progress. Vol canic activity, much of which occurs beneath the ocean
surface,is common in the vicinity of spreading-ridge boundaries. Spreading ridges
can protrudeabove the ocean; the isl and of Icel and, where vol canic activity is
nearl y continuous(there are 150 active vol canos), is such an exampl e.The mantl e
material cool s after it reaches the surface in the gap between the spreadingpl ates.
As it cool s, it becomes magnetized (remnant magnetism) with a pol arity that depends
on the direction of the earth's magnetic fiel d at that time. The magnetic fiel d of
the earth is notconstant on a geol ogical time scal e; it has fl uctuated and reversed
at irregul ar historical interval s,thus imposing magnetic anomal ies (reversal s of
pol arity) in the rock that forms at the... Increasing age Increasing age ..Figure
2.12 Spreading ridge boundary. Magma rises to surface and cool s in gapformed by
spreading pl ates. Magnetic anomal ies are shown as stripes of normal andreversed
magnetic pol arity. (After Foster, 1971.)30 Seismol ogy and Earthquakes Chap. 2
spreading ridge boundaries. Measurement of the magnetic fiel d in a direction
perpendicul arto a spreading ridge pl ate boundary reveal s a fl uctuating pattern of
magnetic intensity, asil l ustrated for the eastern Pacific Ocean region in Figure
2.13. These magnetic anomal ieshave al l owed l arge portions of the major pl ates to be
dated. Comparison of the ages of variousmaterial s al l ows identification of the
geometry and movement of various pl ates and hasproven inval uabl e in the
verification and acceptance of the theory of pl ate tectonics.2.3.2.2 Subduction
Zone BoundariesSince the size of the earth remains constant, the creation of new
pl ate material atspreading ridges must be bal anced by the consumption of pl ate
material at other l ocations.This occurs at subduction zone boundaries where the
rel ative movement of two pl ates istoward each other. At the point of contact, one
pl ate pl unges, or subducts, beneath the other,34110°20120°- - - Fail ed rift
Propagating rift Age (Mal Chron ..features: 2 .... ---' Abandoned 3spreading center
10 4- - Wandering offset 5- Pseudofauit150° 140° 130°100 ., : : : : : : : " : : : :
: , , , , , : " : : : : : : : : : : : : : i : : : : i : : 170° 160°40°50°30°Figure
2.13 Magnetic anomal ies in the eastern Pacific Ocean. The dark l ines representbands
of common magnetic pol arity. (After Atwater and Sveringhaus, 1989.)Sec. 2.3
Continental Drift and Pl ate Tectonics 31Col umbiaPl ateauNot to scal ePuget Active
vol canoesSound (e.g. Mt. Rainier)Cascade IRange___X"'o-....--r<,North Americanpl ate
~~Juan de Fuca pl ateDeformed sediments Ol ympic jMountainsPacific Ocean(\,Juan de
FucaRidgeRising magmaFigure 2.14 Cascadia subduction zone off the coasts of
Washington and Oregon. TheJuan de Fuca pl ate originates at the Juan de Fuca
spreading ridge and subducts beneaththe North American pl ate. Magma rising from the
deeper part of the subduction zone hasformed a series of vol canos that run roughl y
paral l el to the subduction zone. One ofthese, Mt. St. Hel ens, erupted expl osivel y
in 1980. (After Noson et aI., 1988.)as shown in Figure 2.14. Subduction zone pl ate
boundaries exist off the western coasts ofMexico and Chil e, south of the Al eutian
Isl and chain of Al aska, and off the eastern coast ofJapan. The Cascadia subduction
zone off the coast of Washington and British Col umbia isshown in Figure 2.14.
Subduction zone boundaries are often found near the edges of continents. Becausethe
oceanic crust is general l y col d and dense, it sinks under its own weight beneath
thel ighter continental crust. When the rate of pl ate convergence is high, a trench
is formed atthe boundary between pl ates. In fact, subduction zone boundaries are
sometimes cal l edtrench boundaries. Earthquakes are generated in the sl oping Benioff
zone at the interfacebetween the subducting and overiding pl ates. When the rate of
convergence is sl ow, sediments.accumul ate in an accretionary wedge on top of the
crustal rock, thus obscuring thetrench.The subducting pl ate warms and becomes l ess
brittl e as it sinks. Eventual l y, itbecomes so ductil e as to be incapabl e of
producing earthquakes; the greatest recorded earthquakedepth of approximatel y 700
km (435 mil es) supports this hypothesis. Portions of thesubducting pl ate mel t,
producing magma that can rise to the surface to form a l ine of vol canosroughl y
paral l el to the subduction zone on the overriding pl ate.When pl ates carrying
continents move toward each other, continental col l isions canl ead to the formation
of mountain ranges. The Himal ayas consist of two crustal l ayers thathave formed as
the Austral ia-Indian pl ate has col l ided with the Eurasian pl ate. Continental
col l ision of the pl ates carrying Africa and Europe are currentl y reducing the size
of the MediterraneanSea and wil l eventual l y l ead to the formation of a col l ision-
type mountain range(McKenzie, 1970).32 Seismol ogy and Earthquakes Chap. 22.3.2.3
Transform Faul t BoundariesTransformfaul ts occur where pl ates move past each other
without creating new crustor consuming ol d crust. They are usual l y found offsetting
spreading ridges as il l ustrated inFigure 2.15. These transform faul ts are
identified by offsets in magnetic anomal ies and,where preserved, scarps on the
surface of the crust. Magnetic anomal y offsets definingfracturezones may be
observed over thousands of kil ometers; however, it is onl y the segmentof the
fracture zone between the spreading ridges that is referred to as the transform
faul t.As il l ustrated in Figure 2.15, the motion on the portions of the fracture
zone that extendbeyond the transform faul t is in the same direction on either side
of the fracture zone; hencethere is general l y no rel ative motion. These inactive
portions of the fracture zone can beviewed as fossil faul ts that are not producing
earthquakes.(a)JSpreading ridges~~~~ ~ ~ ~Inactive Active I Inactive- fracture -1--
transform --1- fracture -zone faul t zone(b)Figure 2.15 (a) Obl ique and (b) pl an
views of transform faul t and adjacent inactivefracture zones.The San Andreas faul t,
for exampl e, has been characterized as a transform faul t (Wil son,1965) connecting
the East Pacific ridge off the coast of Mexico with the Juan de Fucaridge off the
coast of Washington state. In real ity, the geometry of transform faul ts is usual l y
quite compl ex with many bends and kinks, and they are often divided into a number
offaul tsegments. Their depth is typical l y l imited but can extend horizontal l y over
very l ong distances.Other important transform faul ts incl ude the Motagua faul t
(which separates theNorth American and Caribbean pl ates), the Al pine faul t of New
Zeal and, and the Dead Seafaul t system that connects the Red Sea to the Bitl is
Mountains of Turkey (Kearey and Vine,1990).Pl ate
tectonics provides a very useful framework for understanding and expl aining
movements on the earth's surface and the l ocations of earthquakes and vol canoes.
Pl ate tectonicsaccounts for the formation of new and consumption of ol d crustal
material s in termsof the three types of pl ate movement il l ustrated in Figure 2.16.
It does not, however, expl ainal l observed tectonic seismicity. For exampl e, it is
known that intrapl ate earthquakes(earthquakes that occur within a pl ate, away from
its edges) have occurred on most continents.Wel l -known North American exampl es are
the series of midpl ate earthquakes thatoccurred in the vicinity of New Madrid,
Missouri, in 1811-1812, and the 1886 Charl eston(South Carol ina) earthquake. The
1976 Tangshan (China) and 1993 Marathawada (India)earthquakes are more recent
exampl es of damaging intrapl ate earthquakes.Sec. 2.4 Faul ts 33Subductionzone
boundary----.,.-----....... ~~-Spreadingridgeboundary~Upwel l ingmantl erocksFracture
zoneSpreading- ridge --~--._------boundary~..-Figure 2.16 Interrel ationships among
spreading ridge, subduction zone, and transformfaul t pl ate boundaries.2.4 FAULTS
Whil e the theory of pl ate tectonics general l y assigns the rel ative movement of
pl ates to oneof the three preceding types of pl ate boundaries, examination on a
smal l er scal e reveal s thatthe movement at a particul ar l ocation can be quite
compl icated. In some regions, pl ateboundaries are distinct and easy to identify,
whil e in others they may be spread out with theedges of the pl ates broken to form
smal l er pl atel ets or micropl ates trapped between thel arger pl ates. Local l y, the
movement between two portions of the crust wil l occur on new orpreexisting offsets
in the geol ogic structure of the crust known as faul ts.Faul ts may range in l ength
from several meters to hundreds of kil ometers and extendfrom the ground surface to
depths of several tens of kil ometers. Their presence may be obviOil S,as refl ected
in surficial topography, or they may be very difficul t to detect. The presenceof a
faul t does not necessaril y mean that earthquakes can be expected; movement canoccur
aseismical l y, or the faul t may be inactive. The l ack of observabl e surficial
faul ting, onthe other hand, does not impl y that earthquakes cannot occur; in fact,
faul t rupture does notreach the earth's surface in most earthquakes. The activity
of faul ts is discussed in moredetail in Chapter 4.2.4.1 Faul t GeometryStandard
geol ogic notation is used to describe the orientation of a faul t in space. Whil e
thesurface of a l arge faul t may be irregul ar, it can usual l y be approximated, at
l east over shortdistances, as a pl ane. The orientation of the faul t pl ane is
described by its strike and dip. The34 Seismol ogy and Earthquakes Chap. 2strike of
a faul t is the horizontal l ine produced by the intersection of the faul t pl ane and
ahorizontal pl ane as shown in Figure 2.17. The azimuth of the strike (e.g., N600E)
is used todescribe the orientation of the faul t with respect to due north. The
downward sl ope of thefaul t pl ane is described by the dip angl e, which is the angl e
between the faul t pl ane and thehorizontal pl ane measured perpendicul ar to the
strike. A vertical faul t woul d have a dipangl e of 900• Faul t pl aneDip angl eStrike
vectorHorizontal pl aneDip vectorFigure 2.17 Geometric notation for description of
faul t pl ane orientation.2.4.2 Faul t MovementThe type of movement occurring on a
faul t is usual l y reduced to components in the directionsof the strike and dip.
Whil e some movement in both directions is inevitabl e, movementin one direction or
the other wil l usual l y predominate.2.4.2.1 Dip Sl ip MovementFaul t movement that
occurs primaril y in the direction of the dip (or perpendicul ar tothe strike) is
referred to as dip sl ip movement. There are different types of dip sl ip movements,
cl assified according to the direction of movement and the dip angl e of the faul t.
Normal faul ts, il l ustrated in Figure 2.18, occur when the horizontal component of
dip sl ipmovement is extensional and when the material above the incl ined faul t
(sometimesreferred to as the hanging wal l ) moves downward rel ative to the material
bel ow the faul t(the foot wal l ). Normal faul ting is general l y associated with
tensil e stresses in the crust andresul ts in a horizontal l engthening of the crust.
When the horizontal component of dip sl ipmovement is compressional and the material
above the faul t moves upward rel ative to thematerial bel ow the faul t, reverse
faul ting is said to have occurred. Movement on reversefaul ts, il l ustrated in Figure
2.19, resul ts in a horizontal shortening of the crust. A special type of reverse
faul t is a thrust faul t, which occurs when the faul t pl ane has a smal l dipangl e.
Very l arge movements can be produced by thrust faul ting; the European Al ps are an
excel l ent exampl e of thrust structure.2.4.2.2 Strike-Sl ip MovementFaul t movement
occurring paral l el to the strike is cal l ed strike-sl ip movement. Strikesl ipfaul ts
are usual l y nearl y vertical and can produce l arge movements. Strike-sl ip faul ts are
further categorized by the rel ative direction of movement of the material s on
either side of thefaul t. An observer standing near a right l ateral strike-sl ip
faul t woul d observe the ground onSec. 2.4 Faul ts(a)(b)Figure 2.18 (a) Normal
faul ting (after Noson et aI., 1988); (b) scarp of the normal faul tthat produced the
1954 Dixie-Fairview earthquake in Nevada (K. Steinbrugge col l ection;courtesy of
EERC, Univ. of Cal ifornia).3536 Seismol ogy and EarthquakesFigure 2.19 Reverse
faul ting. Because the dip angl e is so smal l , this reverse faul twoul d probabl y be
cl assified as a thrust faul t. (After Noson et a!., 1988.)Chap. 2the other side of
the faul t moving to the right. Simil arl y, an observer adjacent to a l eft l ateral
strike-sl ip faul t woul d observe the material on the other side moving to the l eft.
The strikesl ipfaul t shown in Figure 2.20a woul d be characterized as a l eft l ateral
strike-sl ip faul t. TheSan Andreas faul t in Cal ifornia is an excel l ent exampl e of
right l ateral strike-sl ip faul ting; inthe 1906 San Francisco earthquake, several
roads and fences north of San Francisco wereoffset by nearl y 6 m (20 ft) (Figure
2.20b).Obl ique faul t movement (i.e., movement with both dip-sl ip and strike-sl ip
components)often occurs. The 1971 San Fernando earthquake ruptured the ground
surface over al ength of 15 km (9 mil es). The maximum vertical displ acement
(produced by reverse faul tmovement) was 1.46 m (4.8 ft), and the maximum horizontal
displ acement (from l eft l ateral strike-sl ip movement) was 2.13 m (7.0 ft) (Berl in,
1980).2.5 ELASTIC REBOUND THEORYThe pl ates of the earth are in constant motion, and
pl ate tectonics indicates that the majorityof their rel ative movement occurs near
their boundaries. The l ong-term effects ofthis movementcan be observed in the
geol ogic record, which refl ects deformations that haveoccurred over very l ong
periods of time. With the advent of modern el ectronic distancemeasurement
equipment, however, movements can al so be observed over much shortertime scal es.
Figure 2.21 shows a set of survey l ines establ ished across the San Andreas and
Cal averas faul ts by the Cal ifornia Department of Water Resources and Division of
Minesand Geol ogy. The shortening of chords 17 and 19, and the l engthening of20 and
23, indicatethat faul t movement is occurring. Chord 21, which l ies entirel y east of
the Cal averas faul t,shows very l ittl e change in l ength.As rel ative movement of the
pl ates occurs, el astic strain energy is stored in the material snear the boundary as
shear stresses increase on the faul t pl anes that separate the pl ates.When the shear
stress reaches the shear strength of the rock al ong the faul t, the rock fail s and
the accumul ated strain energy is rel eased. The effects of the fail ure depend on the
nature ofthe rock al ong the faul t. If it is weak and ductil e, what l ittl e strain
energy that coul d be storedSec. 2.5 El astic Rebound Theory(a)(b)Figure 2.20 (a)
Left l ateral strike-sl ip faul ting (from Noson et aI., 1988); (b) treesoffset by
strike-sl ip faul ting through citrus grove in 1940 Imperial Val l ey earthquake
(courtesy of U.S. Geol ogical Survey).37wil l be rel eased rel ativel y sl owl y and the
movement wil l occur aseismical l y. If, on the otherhand, the rock is strong and
brittl e, the fail ure wil l be rapid. Rupture of the rock wil l rel easethe stored
energy expl osivel y, partl y in the form of heat and partl y in the form of the stress
waves that are fel t as earthquakes. The theory of el astic rebound (Reid, 1911)
describes this38 Seismol ogy and Earthquakes Chap. 2MontereyBayYearI-1: l : ~-~hi ---I
m 0 _ ----~-~ ~~f-----~--I+1: I: : : : ,: : : : i: : 1: ~t?tiSsJd4>1Figure 2.21 (a) Survey
l ines across San Andreas and Cal averas faul ts in Cal ifornia;(b) change in chord
l ength (extension positive). (From Earthquakes by Bol t. Copyright© 1993 by W.H.
Freeman and Company. Used with permission.)process of the successive buil dup and
rel ease of strain energy in the rock adjacent to faul ts.It is often il l ustrated as
shown in Figure 2.22.The nature of the buil dup and rel ease of stress is of
interest. Faul ts are not uniform,either geometrical l y or in terms of material
properties-both strong and weak zones canexist over the surface of a faul t. The
stronger zones, referred to as asperities by some (Kanamoriand Stewart, 1978) and
barriers by others (Aki, 1979), are particul arl y important. Theasperity model of
faul t rupture assumes that the shear stresses prior to an earthquake are notuniform
across the faul t because of stress rel ease in the weaker zones by creep
orforeshocks.Rel ease of the remaining stresses hel d by the asperities produces the
main earthquake thatl eaves the rupture surface in a state of uniform stress. In the
barrier model , the pre-earthquakestresses on the faul t
are assumed to be uniform. When the main earthquake occurs,stresses are rel eased
from al l parts of the faul t except for the stronger barriers; aftershocksthen occur
as the rock adjusts to the new uniform stress fiel d. Since both foreshocks and
aftershocks are commonl y observed, it appears that some strong zones behave as
asperitiesand others as barriers (Aki, 1984). The engineering significance of
asperities and barriersl ies in their infl uence on ground-shaking characteristics
cl ose to the faul t. A site l ocatedcl ose to one of these strong zones may experience
stronger shaking than a site equal l y cl oseSec. 2.5 El astic Rebound Theory
Deformation of ductil e rockFracture of brittl e rock(a)Deformation of a ductil e
stickFracture of a brittl e stick39Figure 2.22 El astic rebound theory of
earthquakes. In (a) the sl ow deformation of rockin the vicinity of a pl ate boundary
resul ts in a buil dup of strain energy in the rock in thesame way that strain energy
buil ds up in a ductil e stick deformed as shown on the right.If the strength of the
rock is exceeded, it wil l rupture, rel easing strain energy in the formof
vibrations, much as the energy in the stick woul d be rel eased when the stick
breaks.After the earthquake, the rock is displ aced from its original position. The
total rel ativedispl acement of the pl ates is the sum of the sl ip displ acement at the
faul t and possibl edispl acements due to warping distortion of the edges of the
pl ates near the faul t. (AfterFoster, R.J., General Geol ogy, 5/e, © 1988. Adapted by
permission of Prentice Hal l ,Upper Sadde River, New Jersey.to the faul t but farther
from a strong zone. At l arger distances from the faul t the effects offaul t
nonuniformity decrease. Unfortunatel y, methods for l ocating these strong zones
priorto rupture have not yet been devel oped.Rupture general l y progresses across a
faul t as a series of disl ocations (some mul tipl eeventearthquakes can be thought of
as a series of smal l earthquakes that occur in cl ose spatial and temporal
proximity). Smal l earthquakes can be model ed as point processes sincetheir rupture
surfaces usual l y span onl y a few kil ometers. Large earthquakes, however, canrupture
over distances of tens, or even hundreds, of kil ometers, and the nature of ground
shaking can be infl uenced by the characteristics of the rupture process. For
exampl e, wavesemanate from the faul t with different strengths in different
directions; such directivityeffects can produce azimuthal differences in ground
motion characteristics (Benioff, 1955;Ben-Menachem, 1961). Constructive
interference of waves produced by successive disl ocationscan produce strong pul ses
ofl arge displ acement cal l edjl ing (Figure 2.23) at nearbysites toward which the
rupture is progressing (Benioff, 1955; Singh, 1985).2.5.1 Rel ationship to
Earthquake RecurrenceThe theory of el astic rebound impl ies that the occurrence of
earthquakes wil l rel ievestresses al ong the portion of a faul t on which rupture
occurs, and that subsequent rupture40 Seismol ogy and Earthquakes Chap. 2Direction
of Rupture PropagationTimeo__-------Irl t.-----------1 ...Jrl L. _2 Jrl 1.. _3
----'rl L. _4 In'-- _5 ....Jnt.--- _Pul sesResul tant in Direction AwayFrom Direction
of Faul t RuptureTime"~O ...Jrl L. _1 ----'rl 1.. _2 ---Jr"L. _3 ----'r"1.. _4
--Ir"1.. _5 ---In'-- _Pul ses_____....In.... _Resul tant in Directionof Faul t Rupture
Figure 2.23 Schematic il l ustration of directivity effect on ground motions at sites
toward and away from direction of faul t rupture. Overl apping of pul ses can l ead to
strong fl ing pul se at sites toward which the faul t ruptures. (After Singh, 1985;
used bypermission of EERI).wil l not occur on that segment until the stresses have
had time to buil d up again. Thechances of an earthquake occurring on a particul ar
faul t segment shoul d therefore be rel atedin some way to the time that has el apsed
since the l ast earthquake and, perhaps, to theamount of energy that was rel eased.
In a probabil istic sense, then, individual earthquakes ona particul ar faul t segment
shoul d not be considered as random, independent events. Thischaracteristic is
important in seismic hazard anal ysis.Because earthquakes rel ieve the strain energy
that buil ds up on faul ts, they shoul d bemore l ikel y to occur in areas where l ittl e
or no seismic activity has been observed for sometime. By pl otting faul t movement
and historical earthquake activity al ong a faul t, it is possibl eto identify gaps in
seismic activity at certain l ocations al ong faul ts. According to el asticrebound
theory, either the movement is occurring aseismical l y or strain energy isbuil ding
in the vicinity of these seismic gaps. In areas where the l atter is known to be the
case, seismic gaps shoul d represent the most l ikel y l ocations for future
earthquakes. A numberof seismic gaps have been identified around the worl d and
l arge earthquakes have subsequentl ybeen observed on several of them. The 1989 Loma
Prieta earthquake occurred ona segment of the San Andreas faul t that had previousl y
been identified as a gap, as shown inFigure 2.24. The use of seismic gaps offers
promise for improvement in earthquake predictioncapabil ities and seismic risk
eval uation.Sec. 2.5 El astic Rebound Theory 41Lorna PrietagapI~~~~"~I '..f100 150
200DISTANCE (mi)50.i'1ILi.!....,t ·"'~i: : .. : 'f • • • . t 'Ai; -",... L": ''': .. •
·1· ..... • : t.: .;...... ..: -.,,¥~• • .~.. ~fI.o~~~~~~~III"(ia) ~ 16I 15
+---------,-------,--------..: -,----------'--.....---+~ 01 (b) 0 1615
f-.------,..----'--"''--'''--''---.--------,----------,----Figure 2.24 Cross
section of the San Andreas faul t from north of San Francisco tosouth of Parkfiel d:
(a) seismicity in the 20 years prior to the 1989 Lorna Prietaearthquake is shown
with the Lorna Prieta gap highl ighted; (b) main shock (open circl e)and aftershocks
of the Lorna Prieta earthquake. Note the remaining gaps between SanFrancisco and
Portol a Val l ey and south of Parkfiel d. (After Housner et aI., 1990.)2.5.2
Rel ationship to Tectonic EnvironmentEl astic rebound al so impl ies that tectonic
environments capabl e of storing differentamounts of energy wil l produce earthquakes
of different size. Consider, for exampl e, thetectonic environment in the vicinity
of a spreading ridge pl ate boundary. First, the crust isthin; hence the vol ume of
rock in which strain energy can buil d up is smal l . Second, the horizontal component
of the rel ative pl ate movement is extensional ; hence the normal stress onthe faul t
pl ane, and with it the rupture strength, is l ow. Third, the rock is rel ativel y warm
andductil e, so it wil l not rel ease strain energy suddenl y. Taken together, these
factors l imit thetotal strain energy that can buil d up and be suddenl y rel eased at
a spreading ridge boundary.These factors expl ain the observed absence of very l arge
earthquakes at spreading ridgeboundaries.By the time the oceanic crust has moved
from a spreading ridge to a subduction zone,it has cool ed and become much thicker
and stronger. Rel ative movement of the pl ates istoward each other, so high
compressive normal stresses increase the rupture strength on thefaul t pl ane.
Because subduction zone pl ate boundaries are incl ined, the potential rupturearea is
l arge. Al l of these factors support the potential buil dup of very l arge amounts of
strain energy that, when suddenl y rel eased, can produce great earthquakes. In fact,
the l argestrecorded earthquakes have been produced by subduction zones.At transform
faul ts, the rock is general l y cool and brittl e, but l arge compressivestresses do
not usual l y devel op because the faul ts are often nearl y vertical and movement is
typical l y of a strike-sl ip nature. Because the depth of transform faul ting is
l imited, the total 42 Seismol ogy and Earthquakes Chap. 2amount of strain energy that
can be stored is control l ed by the l ength of rupture. Very l argeearthquakes
invol ving rupture l engths of hundreds of kil ometers have been observed ontransform
faul ts, but trul y "great" earthquakes may not be possibl e.2.5.3 Seismic MomentThe
concept of el astic rebound theory can be used to devel op a useful measure of the
sizeof an earthquake. The seismic moment of an earthquake is given byMO = /l AD
(2.1)where l -l is the rupture strength of the material al ong the faul t, A the
rupture area, and 15 theaverage amount of sl ip. The seismic moment is named for its
units of force times l ength;however, it is more a measure of the work done by the
earthquake. As such, the seismicmoment correl ates wel l with the energy rel eased
during an earthquake. The seismic momentcan be estimated from geol ogic records for
historical earthquakes, or obtained from thel ong-period components of a seismogram
(Bul l en and Bol t, 1985).2.6 OTHER SOURCES OF SEISMIC ACTIVITYThe sudden rel ease of
strain energy by rupture of the rock at pl ate boundaries is the primarycause of
seismic activity around the worl d. There are, however, other sources of seismic
activity that produce smal l er earthquakes that may be important in l ocal ized
regions.Earthquakes have been associated with vol canic activity. Shal l ow vol canic
earthquakesmay resul t from sudden shifting or movement of magma. In 1975, a
magnitude 7.2earthquake on the big isl and of Hawaii produced significant damage and
was fol l owedshortl y by an eruption of the Kil auea vol cano. The 1980 eruption ofMt.
St. Hel ens in southernWashington was actual l y triggered by a smal l (M, = 5.1),
shal l ow, vol canic earthquakethat triggered a massive l andsl ide on the north sl ope
of the vol cano. The unl oading of thenorth sl ope al l owed the main eruption to occur
approximatel y 30 sec l ater. Vol canic eruptionsthemsel ves can rel ease tremendous
amounts of energy essential l y at the earth's surfaceand may produce significant
ground motion.Seismic waves may be produced by underground detonation of chemical
expl osivesor
nucl ear devices (Bol t, 1975). Many significant devel opments in seismol ogy during
theCol d War years stemmed from the need to monitor nucl ear weapons testing
activities ofother countries. Col l apse of mine or cavern roofs, or mine bursts, can
cause smal l l ocal earthquakes, as can l arge l andsl ides. A 1974 l andsl ide invol ving
1.6 x 109 m3(2.1 x 109 yd')of material al ong the Montaro River in Peru produced
seismic waves equival ent to those ofa magnitude 4.5 earthquake (Bol t, 1989).
Reservoir-induced earthquakes have been the subject of considerabl e study and some
controversy. Local seismicity increased significantl y after the fil l ing of Lake
Mead behindHoover Dam on the Nevada-Arizona border in 1935. When the Koyna Dam
(India) reservoirwas fil l ed, l ocal shal l ow earthquakes became common in an area
previousl y thought tohave been virtual l y aseismic. In 1967, five years after
fil l ing of the Koyna reservoir hadSec. 2.7 Geometric Notation 43begun, a magnitude
6.5 earthquake kil l ed 177 persons and injured more than 2000 more.Local seismicity
has been observed to increase seasonal l y with seasonal increases in reservoirl evel .
In 1975, seven years after the fil l ing of Orovil l e Dam in an area ofl ow historical
seismicity in northern Cal ifornia, a swarm of earthquakes cul minated in a magnitude
5.7main shock. After construction of the High Dam, a magnitude 5.6 earthquake
occurred inAswan, Egypt where very l ittl e significant seismic activity had been
observed in the 3000yearhistory of the area. In these cases, seismic activity
appears to have been triggered by thepresence of the reservoir. Whil e the effect of
the weight of the impounded water is l ikel y tobe negl igibl e at the depths of the
induced seismic activity, an increase in porewater pressurethat migrates as a
"pul se" away from the reservoir after fil l ing may have been sufficient toreduce the
strength of the rock to the point where rupture coul d occur.2.7 GEOIMETRIC NOTATION
To describe the l ocation of an earthquake, it is necessary to use accepted
descriptive terminol ogy.Earthquakes resul t from rupture of the rock al ong a faul t,
and even though the rupturemay invol ve thousands of square kil ometers of faul t
pl ane surface, it must beginsomewhere. The point at which rupture begins and the
first seismic waves originate is cal l edthe focus, or hypocenter, of the earthquake
(Figure 2.25). From the focus, the rupturespreads across the faul t at vel ocities of
2 to 3 km/sec (1.2 to 1.9 mil es/sec) (Bol t, 1989).Al though faul t rupture can extend
to the ground surface, the focus is l ocated at some focal depth (or hypocentral
depth) bel ow the ground surface. The point on the ground surfacedirectl y above the
focus is cal l ed the epicenter. The distance on the ground surface betweenan
observer or site and the epicenter is known as the epicentral distance, and the
distancebetween the observer and the focus is cal l ed the focal distance or
hypocentral distance.Ground surfaceEpicentral distanceSite or observerHypocentral
distance-, Y'Focus or hypocenter - -/1"-Figure 2.25 Notation for description of
earthquake l ocation.442.8 LOCATION OF EARTHQUAKESSeismol ogy and Earthquakes Chap. 2
The l ocation of an earthquake is often initial l y specified in terms of the l ocation
of its epicenter.Prel iminary epicentral l ocation is a simpl e and straightforward
matter, but refinementof the final l ocation can be considerabl y more compl ex.
Prel iminary l ocation is based on therel ative arrival times of p- and s-waves at a
set of at l east three seismographs.Since p-waves travel faster than s-waves, they
wil l arrive first at a given seismograph.The difference in arrival times wil l
depend on the difference between the p- and s-wavevel ocities, and on the distance
between the seismograph and the focus of the earthquake,according toAtp_ s d= ~~
l /vs-l /vpwhere Atp_s is the difference in time between the first p- and s-wave
arrival s, and vp and Vsare the p- and s-wave vel ocities, respectivel y. In bedrock,
p-wave vel ocities are general l y 3to 8 km/sec (1.9 to 5 mil es/sec) and s-wave
vel ocities range from 2 to 5 km/sec (1.2 to 3.1mil es/sec). At any singl e
seismograph it is possibl e to determine the epicentral distance butnot the
direction of the epicenter. This l imited knowl edge is expressed graphical l y by
pl ottinga circl e of radius equal to the epicentral distance. When the epicentral
distance from asecond seismograph is pl otted as a circl e about its l ocation, the
possibl e l ocation of the epicenteris narrowed to the two points of intersection of
the circl es. Obviousl y, a third seismographis necessary to identify the most l ikel y
l ocation of the epicenter as il l ustrated inFigure 2.26. More refined estimates of
the epicentral , or hypocentral , l ocation are madeusing mul tipl e seismographs, a
three-dimensional seismic vel ocity model of the earth, andnumerical optimization
techniques. The accuracy of these techniques depends on thenumber, qual ity, and
geographic distribution of the seismographs and on the accuracy of theseismic
vel ocity model (Dewey, 1979).p-s time shows that earthquake occurredat this
distance from station AEpicenterFigure 2.26 Prel iminary l ocation of epicenter from
differential wave-arrival -timemeasurements at seismographs A, B, and C. Most l ikel y
epicentral l ocation is at theintersection of the three circl es. (After Foster, R.J.,
General Geol ogy, 5/e, © 1988.Adapted by permission of Prentice Hal l , Upper Saddl e
River, New Jersey.)Sec. 2.9 Size of Earthquakes 452.9 SIZE: : OF EARTHQUAKESThe
"size" of an earthquake is obviousl y a very important parameter, and it has been
described in different ways. Prior to the devel opment of modern instrumentation,
methodsfor characterizing the size of earthquakes were based on crude and
qual itative descriptionsof the effects of the earthquakes. More recentl y, modern
seismographs have al l owed thedevel opment of a number of quantitative measures of
earthquake size. Since several ofthese measures are commonl y used in both
seismol ogy and earthquake engineering, the distinguishingfeatures of each shoul d be
understood.2.9.1 Earthqua'(e 'ntensityThe ol dest measure of earthquake size is the
earthquake intensity. The intensity is a qual itativedescription of the effects of
the earthquake at a particul ar l ocation, as evidenced byobserved damage and human
reactions at that l ocation. Because qual itative descriptions ofthe effects of
earthquakes are avail abl e throughout recorded history, the concept of intensitycan
be appl ied to historical accounts to estimate the l ocations and sizes of
earthquakes thatoccurred prior to the devel opment of modern seismic instruments
(preinstrumental earthquakes).This appl ication has been very useful in
characterizing the rates of recurrence ofearthquakes of different sizes in various
l ocations, a critical step in eval uation of the l ikel ihoodof seismic hazards.
Intensities can al so be used to estimate strong ground motion l evel s(Section
3.3.1.1), for comparison of earthquake effects in different geographic regions,and
for earthquake l oss estimation.The Rossi-Forel (RF) scal e of intensity, describing
intensities with val ues rangingfrom I to X, was devel oped in the 1880s and used for
many years. It has l argel y beenrepl aced in Engl ish-speaking countries by the
modified Mercal l i intensity (MMI) scal eoriginal l y devel oped by the Ital ian
seismol ogist Mercal l i and modified in 1931 to better representconditions in
Cal ifornia (Richter, 1958). The MMI scal e is il l ustrated in Tabl e 2- I.The
qual itative nature of the MMI scal e is apparent from the descriptions of each
intensityl evel .The Japanese Meteorol ogical Agency (JMA) has its own intensity
scal e, and theMedvedev-Spoonheuer-Karnik (MSK) scal e is used in central and eastern
Europe. A comparisonof the RF, MMI, JMA, and MSK scal es is shown in Figure 2.27.
Earthquake intensities are usual l y obtained from interviews of observers after the
event. The interviews are often done by mail , but in some seismical l y active areas,
permanentobservers are organized and trained to produce rational and unemotional
accounts ofground shaking. Since human observers and structures are scattered more
widel y than anyseismol ogical observatory coul d reasonabl y hope to scatter
instruments, intensity observationsprovide information that hel ps characterize the
distribution of ground shaking in aregion. A pl ot of reported intensities at
different l ocations on a map al l ows contours of equal intensity, or isoseisms, to be
pl otted. Such a map is cal l ed an isoseismal map (Figure 2.28).The intensity is
general l y greatest in the vicinity of the epicenter of the earthquake, andthe term
epicentral intensity is often used as a crude description of earthquake size.
Isoseismal maps show how the intensity decreases, or attenuates, with increasing
epicentral distance.46Tabl e 2-1 Modified Mercal l i Intensity Scal e of 1931Seismol ogy
and Earthquakes Chap. 2IIIIIIVVVIVIIVIIIIXXXIXIINot fel t except by a very few under
especial l y favorabl e circumstancesFel t by onl y a few persons at rest, especial l y on
upper fl oors of buil dings; del icatel ysuspended objects may swingFel t quite
noticeabl y indoors, especial l y on upper fl oors of buil dings, but many peopl e do not
recognize it as an earthquake; standing motor cars may rock sl ightl y; vibration
l ike passingof truck; duration estimatedDuring the day fel t indoors by many,
outdoors by few; at night some awakened; dishes,windows, doors disturbed; wal l s
make cracking sound; sensation l ike heavy truck strikingbuil ding; standing motor
cars rocked noticeabl yFel t by nearl y everyone, many awakened; some dishes, windows,
etc., broken; a few instancesof cracked pl aster; unstabl e objects overturned;
disturbances of trees, pil es, and other tal l objects sometimes noticed; pendul um
cl ocks may stopFel t by al l , many frightened
and run outdoors; some heavy furniture moved; a few instances offal l en pl aster or
damaged chimneys; damage sl ightEverybody runs outdoors; damage negl igibl e in
buil dings of good design and construction,sl ight to moderate in wel l -buil t ordinary
structures, considerabl e in poorl y buil t or badl ydesigned structures; some chimneys
broken; noticed by persons driving motor carsDamage sl ight in special l y designed
structures, considerabl e in ordinary substantial buil dings, with partial col l apse,
great in poorl y buil t structures; panel wal l s thrown out of framestructures; fal l
of chimneys, factory stacks, col umns, monuments, wal l s; heavy furniture overturned;
sand and mud ejected in smal l amounts; changes in wel l water; persons driving motor
cars disturbedDamage considerabl e in special l y designed structures; wel l -designed
frame structures thrownout of pl umb; great in substantial buil dings, with partial
col l apse; buil dings shifted off foundations;ground cracked conspicuousl y;
underground pipes brokenSome wel l -buil t wooden structures destroyed; most masonry
and frame structures destroyedwith foundations; ground badl y cracked; rail s bent;
l andsl ides considerabl e from river banksand steep sl opes; shifted sand and mud;
water spl ashed over banksFew, if any (masonry) structures remain standing; bridges
destroyed; broad fissures in ground;underground pipel ines compl etel y out of
service; earth sl umps and l and sl ips in soft ground;rail s bent greatl yDamage total ;
practical l y al l works of contruction are damaged greatl y or destroyed; waves seenon
ground surface; l ines of sight and l evel are distorted; objects thrown into the air
2.9.2 Earthquake MagnitudeThe possibil ity of obtaining a more objective,
quantitative measure of the size of an earthquakecame about with the devel opment of
modem instrumentation for measuring groundmotion during earthquakes. In the past 60
years, the devel opment of seismic instruments,and our understanding of the
quantities they measure, have increased dramatical l y. Seismicinstruments al l ow an
objective, quantitative measurement of earthquake size cal l ed earthquakemagnitude
to be made. Most measurements of earthquake magnitude are instrumental (i.e., based
on some measured characteristic of ground shaking).Sec. 2.9MMISize of EarthquakesI
II III IV V VI VII VIII IX X XI XII47RF I EE: EEI: : E VIII IX XJMA I [il l V VI VIIMSK
I II III IV V VI VII VIII IX X XI XIIFigure 2.27 Comparison of intensity val ues
from modified Mercal l i (MMI), RossiForel (RF), Japanese Meteorol ogical Agency (JMA),
and Medvedev-SpoonheuerKarnik(MSK) scal es. (After Richter (1958) and Murphy and
O'Brien (1977).)6VINew Zeal and200 400Kil ometers6Cal ifornia6666(a) (b)Figure 2.28
Isoseismal maps from (a) the 1968 Inangahua earthquake iu New Zeal and(After Eiby,
1980) and (b) the 1989 Lorna Prieta earthquake in northern Cal ifornia(Modified
Mercal i intensities). (After Housner, 1990.)48 Seismol ogy and Earthquakes Chap_ 2
2.9.2.1 Richter Local MagnitudeIn 1935, Charl es Richter used a Wood-Anderson
seismometer to define a magnitudescal e for shal l ow, l ocal (epicentral distances
l ess than about 600 km (375 mil es)) earthquakesin southern Cal ifornia (Richter,
1935). Richter defined what is now known as thel ocal magnitude as the l ogarithm
(base 10) of the maximum trace ampl itude (in micrometers)recorded on a Wood-
Anderson seismometer l ocated 100 km (62 mil es) from theepicenter of the earthquake.
The Richter l ocal magnitude (ML) is the best known magnitudescal e, but it is not
al ways the most appropriate scal e for description of earthquake size.2.9.2.2
Surface Wave MagnitudeThe Richter l ocal magnitude does not distinguish between
different types of waves.Other magnitude scal es that base the magnitude on the
ampl itude of a particul ar wave havebeen devel oped. At l arge epicentral distances,
body waves have usual l y been attenuated andscattered sufficientl y that the
resul ting motion is dominated by surface waves. The surfacewave magnitude
(Gutenberg and Richter, 1936) is a worl dwide magnitude scal e based onthe ampl itude
of Rayl eigh waves with a period of about 20 sec. The surface wave magnitudeis
obtained fromM, = 10gA + 1.66 10gLl + 2.0 (2.3)where A is the maximum ground
displ acement in micrometers and II is the epicentral distanceof the seismometer
measured in degrees (360° corresponding to the circumference ofthe earth). Note
that the surface wave magnitude is based on the maximum ground displ acement
ampl itude (rather than the maximum trace ampl itude of a particul ar seismograph);
therefore, it can be determined from any type of seismograph. The surface wave
magnitudeis most commonl y used to describe the size of shal l ow (l ess than about 70
km (44 mil es)focal depth), distant (farther than about 1000km (622 mil es)) moderate
to l arge earthquakes.2.9.2.3 Body Wave MagnitudeFor deep-focus earthquakes, surface
waves are often too smal l to permit rel iabl e eval uationof the surface wave
magnitude. The body wave magnitude (Gutenberg, 1945) is aworl dwide magnitude scal e
based on the ampl itude of the first few cycl es of p-waves whichare not strongl y
infl uenced by the focal depth (Bol t, 1989). The body wave magnitude canbe expressed
asmb = 10gA -l ogT + O.Ol Ll + 5.9 (2.4)where A is the p-wave ampl itude in
micrometers and T is the period of the p-wave (usual l yabout one sec). Body wave
magnitude can al so be estimated from the ampl itude of one-second-period, higher-
mode Rayl eigh waves (Nuttl i, 1973); the resul ting magnitude, mbLg' iscommonl y used
to describe intrapl ate earthquakes.2.9.2.4 Other Instrumental Magnitude Scal es
Magnitude scal es using different parts of the instrumental record have al so been
proposed.The coda of an earthquake motion are the backscattered waves (Aki, 1969)
that fol l owpassage of the primary (unrefl ected) body and surface waves. Aki (1969),
showing thatcertain characteristics of the coda are independent of the travel path,
devel oped a codamagnitude, Me> that coul d be obtained from those characteristics.
The duration magnitude,Sec. 2.9 Size of Earthquakes 49(2.5)MD, which is based on
the total duration of the earthquake, can be used to describe smal l earthquakes that
are often of more interest to seismol ogists than engineers (Real and Teng,1973).
The Japanese Meteorol ogical Agency uses l ong-period waves to determine a l ocal
magnitude scal e, MJMA, for Japanese earthquakes.2.9.2.5 Moment MagnitudeIt is
important to real ize that the previousl y described magnitude scal es are empirical
quantities based on various instrumental measurements of ground-shaking
characteristics.As the total amount of energy rel eased during an earthquake
increases, however, theground-shaking characteristics do not necessaril y increase
at the same rate. For strongearthquakes, the measured ground-shaking
characteristics become l ess sensitive to the sizeof the earthquake than for smal l er
earthquakes. This phenomenon is referred to as saturation;the body wave and Richter
l ocal magnitudes saturate at magnitudes of 6 to 7 and thesurface wave magnitude
saturates at about M, =8. To describe the size of very l arge earthquakes,a
magnitude scal e that does not depend on ground-shaking l evel s, and consequentl ydoes
not saturate, woul d be desirabl e. The onl y magnitude scal e that is not subject to
saturationis the moment magnitude (Kanamori, 1977; Hanks and Kanamori, 1979) since
it isbased on the seismic moment, which is a direct measure of the factors that
produce ruptureal ong the faul t. The moment magnitude is given byM = 10gMo- 107w 1.5
.where Mo is the seismic moment (Section 2.5.3) in dyne-ern,The rel ationship
between the various magnitude scal es can be seen in Figure 2.29.Saturation of the
instrumental scal es is indicated by their fl attening at higher magnitude9r--r-
r-,--,----,--,--,----.----.----.----,--,--,---,,---,------,------mb-87433 4 5 6 7 8
Moment Magnitude9 10Figure 2.29 Saturation of various magnitude scal es: Mw (moment
magnitude), ML(Richter l ocal magnitude), M, (surface wave magnitude), mi; (short-
period body wavemagnitude), mB (l ong-period body wave magnitude), and MJMA
(JapaneseMeteorol ogical Agency magnitude). (After Idriss, 1985.)50 Seismol ogy and
Earthquakes Chap. 2val ues. As an exampl e of the effects of magnitude saturation,
both the 1906 San Franciscoand 1960 Chil e earthquakes produced ground shaking that
l ed to surface wave magnitudesof 8.3, even though the sizes of their rupture
surfaces, il l ustrated by the shaded areas in Figure2.30, were vastl y different. The
great disparity in energy rel ease was, however, refl ectedin the moment magnitudes
of the earthquakes: 7.9 for San Francisco and 9.5 for Chil e.San Francisco, 1906
Ms=8.3Mw=7.9Chil e, 1960Ms=8.3Mw = 9.5Figure 2.30 Comparison of rel ative areas of
faul t rupture (shaded) and magnitudes for1906 San Francisco and 1960 Chil e
earthquakes. Al though the shaking of bothearthquakes produced surface wave
magnitudes of 8.3, the amounts of energy rel easedwere very different, as refl ected
in their moment magnitudes. (After Boore, 1977. Themotion of the ground during
earthquakes, Scientific American, Vol . 237, No.6, usedwith permission.)Bol t (1989)
suggests thatML or mb be used for shal l ow earthquakes of magnitude 3 to7, M, for
magnitudes 5 to 7.5, and Mw for magnitudes greater than 7.5.2.9.3 Earthquake Energy
The total seismic energy rel eased during an earthquake is often estimated from the
rel ationship(Gutenberg and Richter, 1956)l ogE = 11.8 + 1.5Ms (2.6)where E is
expressed in ergs. This rel ationship was l ater shown (Kanamori, 1983) to be
appl icabl e to moment magnitude as wel l . It impl ies that a unit change in magnitude
correspondsto a 101.5 or 32-fol d increase in seismic energy. A magnitude 5
earthquake thereforewoul d rel ease onl y about 0.001 times the energy of a magnitude
7 earthquake,
thereby il l ustratingthe ineffectiveness of smal l earthquakes in rel ieving the
buil dup of strain energy thatcauses very l arge earthquakes. Combining equations
(2.5) and (2.6) (using Mw) shows thatthe amount of energy rel eased during an
earthquake is proportional to the seismic moment.The amount of energy rel eased by
earthquakes is often difficul t to comprehend;al though a singl e erg is smal l (1 erg
= 7.5 x 10-8 ft-l b), the energy rel eased in an atomicbomb of the size used at
Hiroshima (20,000-ton TNT equival ent) woul d correspond to amagnitude 6.0
earthquake. On that basis, the 1960 Chil e earthquake (Mw =9.5) rel eased asmuch
energy as 178,000 such atomic bombs (Figure 2.31).Sec. 2.10 Summary 51Meterorite
Impact(10 km diameter, 20 km/sec vel ocity)Earth's dail y receipt of sol ar energy
Chil e, 1960 (M =9.5)Al aska,1964 (M = 9.2)Annual gl obal seismicityMt. St. Hel ens
TornadoEl ectrical energy oftypical thunderstormAverage tornado(kinetic energy)
Earth's annual heat fl ow 1.X~?~~,U.S. Annual energy consumption ....Average
hurricane (10 day l ifetime) ,//Chil e 1960 (M9.5) //Al aska 1964 (M9.2) 7.5 x 1025
Worl d's l argest Average annual nucl ~ar expl osion "gl obal seismicNew Madnd 1811-12U
'energy rel ease(M8.0-8.3) -,Hurricane Mt. st. Hel ens(kinetic energy) ,5x 1025
Average annual stabl e 1 Megaton \,continental seismicity nucl ear expl osib~,,,,
Hiroshima atomic bomb 2.5'~ 1025,,,,,,0.5 X 1025'0~==~~~~: J5.0 6.0 7.0 8.0 9.0 10.0
Moment magnitude (M)Lightning Bol t2.0 3.0 4.0 5.0 6.0 7.0 8.0 9.0 10.0 11.0 12.0
13.0Equival ent Moment Magnitude (M)Figure 2.31 Rel ative energy of various natural
and human-made phenomena. (AfterJohnston, 1990. Reprinted by permission of USGS.)
2.10 SUMMARY1. The earth has a l ayered structure-the surficial crust is underl ain
in turn by the mantl e,the outer core, and the inner core. The temperature of each
l ayer increases withdepth. The temperature gradient in the mantl e causes the
semimol ten rock to movesl owl y by convection.2. The crust is broken into a number of
l arge pl ates and smal l er pl atel ets. Shear stresseson the bottoms of the pl ates,
caused by l ateral movement of the convecting mantl e,and gravitational forces cause
the pl ates to move with respect to each other.3. Rel ative movement of the pl ates
causes stresses to buil d up on their boundaries. Asmovement occurs, strain energy
accumul ates in the vicinity of the boundaries. This52 Seismol ogy and Earthquakes
Chap. 2energy is eventual l y dissipated: either smoothl y and continuousl y or in a
stick-sl ipmanner that produces earthquakes. The size of an earthquake depends on
the amountof energy rel eased.4. There are three different types of pl ate boundaries
and their characteristics infl uencethe amount of strain energy that can buil d up
near them. As a resul t, the different typesof boundaries have different earthquake
capabil ities: subduction zone boundaries canproduce the l argest earthquakes,
fol l owed by transform faul t boundaries and thenspreading ridge boundaries.5. The
surfaces on which rel ative movements occur are cal l ed faul ts. At a particul ar
l ocation, a faul t is assumed to be pl anar with an orientation described by its
strike anddip. Faul t movement is divided into dip-sl ip components (normal and
reverse faul ting)and strike-sl ip components (l eft l ateral and right l ateral
faul ting).6. The energy-rel easing function of earthquakes suggests that a period of
time for strainenergy accumul ation shoul d be expected between l arge earthquakes at
the same l ocation.It al so suggests that earthquakes shoul d be most l ikel y to occur
al ong portions ofa faul t for which l ittl e seismic activity has been observed-unl ess
the pl ate movementhas occurred aseismical l y.7. Earthquake intensity is a
qual itative measure of the effects of an earthquake at a particul arl ocation. It is
rel ated to the size of the earthquake but is al so infl uenced byother factors.
Isoseismal maps can be used to describe the spatial variation of intensityfor a
given earthquake. Because no instrumental measurements are required, historical
accounts can be used to estimate intensity val ues for preinstrumental earthquakes.
8. Earthquake magnitude is a quantitative measure of the size of an earthquake.
Mostmagnitude scal es are based on measured ground motion characteristics. The l ocal
magnitude is based on the trace ampl itude of a particul ar seismometer, the surface
wave magnitude on the ampl itude of Rayl eigh waves, and the body wave magnitudeon
the ampl itude of p-waves. Because these ampl itudes tend to reach l imiting val ues,
these magnitude scal es may not accuratel y refl ect the size of very l arge
earthquakes.The moment magnitude, which is not obtained from ground motion
characteristics, isabl e to describe the size of any earthquake.9. Earthquake
magnitude scal es are l ogarithmic, hence a unit change in magnitude correspondsto a
l O-fol d change in the magnitude parameter (ground motion characteristicor seismic
moment). The energy rel eased by an earthquake is rel ated tomagnitude in such a way
that a unit change in magnitude corresponds to a 32-fol dchange in energy.HOMEWORK
PROBLEMS2.1 Convection causedby thermal gradients in the uppermantl eis thoughtto be
a primarycauseofcontinental drift. Estimate the average thermal gradientin the
upper mantl e.2.2 The coefficient of thermal expansion of the upper mantl e is
about2.5 x 1O-5/°K. Estimatetheratio of the densityat the top of the uppermantl eto
that at the bottomon the uppermantl e.Chap. 2 Homework Probl ems 53.2.3 Using the
data from Figure 2.21, determine whether the San Andreas and Cal averas faul ts are
undergoing right l ateral or l eft l ateral strike sl ip faul ting.2.4 Using the data
from Figure 2.21, estimate the average rate of rel ative movement al ong the San
Andreas and Cal averas faul ts during the period from 1959 to 1970..2.5 Assuming p-
and s-waves travel ed through the crust at 6 km/sec and 3 km/sec, respectivel y,
estimate the epicentral l ocation (l atitude and l ongitude) of the hypothetical
earthquake whosecharacteristics are given bel ow: SeismographLatitude Longitude p-
wave arrival time s-wave arrival time37°22'30" 121°52'30" 06: 11: 18.93 06: 11: 26.90
37°45'00" 122°20'00" 06: 11: 14.84 06: 11: 18.7137°52'33" 121°43'38" 06: 11: 17.26 06:
11: 23.532.6 Using a map of Cal ifornia, determine which faul t the hypothetical
earthquake of Probl em 2.5woul d most l ikel y have occurred on?.2.7 An earthquake
causes an average of 2.5 m strike-sl ip displ acement over an 80 km l ong, 23 kmdeep
portion of a transform faul t. Assuming that the rock al ong the faul t had an average
rupturestrength of 175 kPa, estimate the seismic moment and moment magnitude of the
earthquake.3Strong Ground Motion3.1 INTRODUCTIONThe earth is far from quiet-it
vibrates al most continuousl y at periods ranging from mil l isecondsto days and
ampl itudes ranging from nanometers to meters. The great majority of these
vibrationsare so weak that they cannot be fel t or even detected without special ized
measurementequipment. Such microseismic activity is of greater importance to
seismol ogists than engineers.Earthquake engineers are interested primaril y in
strong ground motion (i.e., motion of sufficientstrength to affect peopl e and their
environment). Eval uation of the effects of earthquakes at aparticul ar site requires
objective, quantitative ways of describing strong ground motion.The ground motions
produced by earthquakes can be quite compl icated. At a givenpoint, they can be
compl etel y described by three components of transl ation and three componentsof
rotation. In practice, the rotational components are usual l y negl ected; three
orthogonal components of transl ational motion are most commonl y measured. Typical
ground motion records, such as the accel eration-time histories shown in Figure 3.1,
containa tremendous amount of information. To express al l of this information
precisel y (i.e., toreproduce each of the three time histories exactl y), every twist
and turn in each pl ot must bedescribed. The motions shown in Figure 3.1, for
exampl e, were determined from 2000accel eration val ues measured at time increments
of 0.02 sec. This l arge amount of informationmakes precise description of a ground
motion rather cumbersome.54[: rF: ·: ~' ~-w,,, Io 10 20 30 40Time (sec)Sec. 3. I
Introduction 55~~0-5~8 ~ 0Q)~o« -0.5 : : : o 10 20 30 40Time (sec)0.5 § 0.5 §c North
- South c North - South0 .Q~ 0 ~ O~ .UII"I,d. .,1Q) Q) 1'1" l Q5 Q5o oo o« -0.5 «
-0.50 10 20 30 40 0 10 20 30 40Time (sec) Time (sec)~ 0.5 El ~05E,c Vertical
Vertical 0~0Q) Io~~... Q5oo-c -0.5 « -0.5 : 0 10 20 30 40 0 10 20 30 40Time (sec)
Time (sec)Figure 3.1 Accel eration time histories recorded at two sites in Gil roy,
Cal iforniaduring the 1989 Loma Prieta earthquake. The Gil roy No.1 instrument was
l ocated onan outcrop of Franciscan sandstone, whil e the Gil roy No.2 instrument was
underl ain by165m (540 ft) of stiff, al l uvial soil s. The Gil roy No.1 (rock) and
Gil roy No.2 (soil )sites were l ocated at epicentral distances of21.8 km (13.5 mil es)
and 22.8 km (14.2mil es), respectivel y.Fortunatel y, it is not necessary to reproduce
each time history exactl y to describe theground motion adequatel y for engineering
purposes. It is necessary, however, to be abl e todescribe the characteristics of
the ground motion that are of engineering significance and toidentify a number of
ground motion parameters that refl ect those characteristics. For engineering
purposes, three characteristics of earthquake motion are of primary significance:
(1)the ampl itude, (2) frequency content, and (3) duration of the motion. A number
of differentground motion parameters have been proposed, each of which provides
information aboutone or more of these characteristics. In practice, it is usual l y
necessary to use more than oneof these parameters to characterize a particul ar
ground motion adequatel y.This chapter describes the instruments and techniques
used to measure strong groundmotion, and the procedures by which measured motions
are corrected. It then presents a varietyof parameters that can be used to
characterize the ampl itude, frequency content, and durationof strong ground
motions. Rel ationships that can be used to predict these parameters areal so
presented. The chapter concl udes with a brief description of the spatial
variabil ity of56 Strong Ground Motion Chap. 3ground motions. Before proceeding
further, the reader shoul d review the topics discussed inAppendices A and B-
famil iarity with the concepts presented in those appendices isassumed in this
chapter and the remainder of the book.3.2 STRONG-MOTION MEASUREMENTThe
identification and eval uation of ground motion parameters requires access to
measurementsof strong ground motions in actual earthquakes. Accurate, quantitative
measurement of strongground motion is critical for both seismol ogical and
earthquake engineering appl ications. Asstated by the National Research Council
Committee on Earthquake Engineering Research(Housner, 1982): ''The recording of
strong ground motion provides the basic data for earthquakeengineering. Without a
knowl edge of the ground shaking generated by earthquakes, it is not possibl eto
assess hazards rational l y or to devel op appropriate methods of seismic design."
3.2.1 SeismographsAl though written descriptions of earthquakes date back as far as
780 B.C., the first accuratemeasurements of destructive ground motions were made
during the 1933 Long Beach, Cal iforniaearthquake (Hudson, 1984). Measurement of
ground motion has advanced considerabl ysince then, most rapidl y in the past 20
years or so.Various instruments are avail abl e for ground motion measurement.
Seismographs areused to measure rel ativel y weak ground motion; the records they
produce are cal l ed seismograms.Strong ground motions are usual l y measured by
accel erographs and expressed inthe form of accel erograms. The simpl est type of
seismograph can be il l ustrated by a massspring-damper singl e-degree-of-freedom
(SDOF) system, as shown in Figure 3.2. A rotatingdrum is connected to the
seismograph housing with a styl us attached to a mass. Themass is connected to the
housing by a spring and dashpot arranged in paral l el , and the housingis connected
to the ground. Since the spring and dashpot are not rigid, the motion of themass
wil l not be identical to the motion of the ground during an earthquake. The
rel ativemovement of the mass and the ground wil l be indicated by the trace made by
the styl us onHousing,,I Styl us~ ~ mrr'7LL Drumc ( ") ( ")
I~~~~~~~~~~~~~~~~~~~~~~~~~~~~.
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.• • Figure 3.2 Simpl e mass-
spring-dashpot type of seismograph. The housing is firml yconuected to the ground.
When the ground shakes, the styl us marks a trace on therotating drum that shows the
rel ative displ acement between the mass and the ground.Most modern instruments are
more compl icated than the one shown here.Sec. 3.2 Strong-Motion Measurement 57the
rotating drum. A typical seismograph station may have three seismographs oriented
torecord motion in the vertical and two perpendicul ar horizontal directions.
Seismographs can be designed to measure various ground motion characteristics. To
understand how this can be done it is necessary to consider the dynamic response of
a simpl eseismograph such as the one shown in Figure 3.2. This seismograph is a SDOF
oscil l atorwhose response to shaking is given by the equation of motion (Appendix B)
mii + cu + ku = -mug (3.1)where u is the seismograph trace displ acement (the
rel ative displ acement between the seismographand the ground) and ug is the ground
displ acement.If the ground displ acement is simpl e harmonic at a circul ar frequency
OOg, the displ acementresponse ratio (the ratio of trace displ acement ampl itude to
ground displ acementampl itude) wil l bep2J( 1 - P2)2 + (2~P)2(3.2)where p(= 00/000)
is the tuning ratio, roo (= ,Jk/m) is the undamped natural circul ar frequency,and ~
(= c/2 JkI1i) is the damping ratio. Figure 3.3a shows how the displ acementresponse
ratio varies with frequency and damping. For ground motion frequencies wel l above
the natural frequency of the seismograph (i.e., l arge val ues of P), the trace
ampl itude is equal to the ground motion ampl itude. The l owest frequency for which
this equal ity hol ds (withina given range of accuracy) depends on the damping ratio.
Because the frequency response isfl at and phase angl es are preserved at damping
ratios of 60%, SDOF displ acement seismographsare usual l y designed with damping
ratios in that range (Richart et aI., 1970).Simil arl y, the accel eration response
ratio (the ratio of trace displ acement ampl itude toground accel eration ampl itude)
is given byl uiIugl (3.3)The variation of accel eration response ratio with frequency
and damping is shown inFigure 3.3b. The trace ampl itude is proportional to the
ground accel eration ampl itude forfrequencies wel l bel ow the natural frequency of
the seismograph (i.e., l ow val ues of P). Aseismograph with 60% damping wil l
accuratel y measure accel erations at frequencies up toabout 55%of its natural
frequency. Most seismographs of this type have natural frequenciesof about 25 Hz
with damping ratios near 60%, with desirabl e fl at response (constant accel eration
response ratio) at frequencies up to about 13 Hz.The preceding paragraphs show how
the same physical system can act as both a displ acementseismograph and an
accel erograph. It measures displ acements at frequencieswel l above and accel erations
at frequencies wel l bel ow its natural frequency. The WoodAndersonseismograph, used
by Richter to devel op the first earthquake magnitude scal e,used a smal l mass
suspended eccentrical l y on a thin tungsten torsion wire. A mirror attachedto the
wire al l owed optical recording with a ground motion magnification of about 3000.
Damping was provided el ectromagnetical l y at 80% of critical ; the damped natural
periodwas about 0.8 sec.Chap. 34 5 6 7 8 910Strong Ground Motion1.5 2 3A ./"V
~Vb ... ,..: .../ /i/I/'l iI~: --~~C-------+t---------'----'----'----: --+-i---i~----
·-;------'----;---/: : .J~.60+;L- y.-r- iV /"1 Vf-------------: -----,UJ'-
N,'1JJj /" /"fil l ! l / ./ 2~------.c .....'.7. jj J-M.)/.I} //iiI ;/VIA!///0.1 IP!/
0.3 0.4 0.5 0.6 0.8 1582.52.01.51.00.90.80.7l Ui 0.6IUgl 0.50.40.30.20.15~(a)• /0 i i
0\~ ~. ~~....~ V~: 10.: .. : 0.707~ : - : Tj~i+f: : : : : : ~ . ...... ......<,!--. : 2
x: \\'\!'o.. r-, \. \~<; '-5 " ~ ~-, .<;= 1(1"- '\ ~t-, "' r-, "' -, \',......-,'\
i" -, \\ ,"\ : 1.52.52.00.10.040.060.08 0.1 0.15 0.2 0.3 0.4 0.50.6 0.81.0 1.5 2 3 4
1.00.90.80.7l Ui 06- 0.5 IUgl 0.40.30.20.15~(b)Figure 3.3 (a) Displ acement response
ratio and (b) accel eration response ratio (roo = Irad/sec) for SDOF system
subjected to simpl e harmonic base motion.In most modern seismographs, an el ectronic
transducer often referred to as a seismometersenses the motion and produces an
anal og (continuous) el ectrical signal that isrecorded for subsequent processing.
Most accel erographs currentl y in use are accel erometers,el ectronic transducers that
produce an output vol tage proportional to accel eration. Anumber of different types
of accel erometers are avail abl e. Servo (orforce bal ance) accel erometersuse a
suspended mass to which a displ acement transducer is attached. When thehousing is
accel erated, the signal produced by the rel ative displ acement between theSec. 3.2
Strong-Motion Measurement 59housing and mass is used to generate a restoring force
that pushes the mass back toward itsequil ibrium position. The restoring force is
proportional to the accel eration and can be measuredel ectronical l y. Servo
accel erometers can provide very good accuracy over the rangeof frequencies of
greatest interest in earthquake engineering. Piezoel ectric accel erometersuse a mass
attached to a piezoel ectric material (usual l y quartz, tourmal ine, or a
ferroel ectricceramic) to sense accel erations. The piezoel ectric material acts as
the spring in the diagramof Figure 3.2; damping is general l y negl igibl e. When
accel erated, the inertial force of themass strains the piezoel ectric material ,
which devel ops an el ectrical charge on its surfaces.The resul ting vol tage is (if
the diel ectric constant does not vary with charge) proportional toaccel eration.
Because piezoel ectric material s are quite stiff, their natural frequencies arevery
high, so they are particul arl y useful for high-frequency measurements. Their
responseat l ow frequencies, however, can be strongl y infl uenced by signal
conditioning system characteristics.Triaxial accel erometers, in which three
orthogonal components of accel erationare measured with a common time base, are
commonl y used. From the three components,the accel eration in any direction can be
computed. Some seismographs use vel ocity transducers,or geophones, in addition or
as an al ternative to accel erometers.Seismographs, accel erographs, and ancil l ary
equipment are protected by an instrumentshel ter (Figure 3.4). An important
component of a seismograph or accel erograph is anaccurate cl ock, particul arl y when
more than one component of motion is measured or whenthe ground motion at one
l ocation is being compared with that at another. Most modeminstruments maintain
time accuracy by synchronizing on a dail y basis with radio time signal stransmitted
by a standard time service or by recording such signal s al ong with theground-motion
data. Universal Coordinated Time (the scientific equival ent of GreenwichMean Time)
is used as a common worl dwide time basis.A seismoscope (Hudson, 1958) is a
rel ativel y inexpensive ground motion instrument.Seismoscopes are conical pendul ums
(Figure 3.5a) in which a metal styl us
attached to a suspendedmass inscribes a record of ground motion on a smoked gl ass
pl ate, producing a twodimensional record of the type shown in Figure 3.5b. Scott
(1972) found that smal l oscil l ationsof the trace were rel ated to the instrument
rather than the earthquake and that theycoul d be used to provide a time scal e to
the seismoscope trace. The time scal e al l ows accel erogramsto be computed from the
seismoscope trace.': 1.2.2 Data Acquisition and DigitizationEarl y ground motion
instruments transformed the motion of the ground to the motion of aphysical
mechanism. The mechanism, perhaps in the form of a pen or styl us or refl ective
mirror, caused the motion to be recorded in anal og form on paper or photographic
fil mattached to a rotating drum. Later-generation instruments recorded motions
el ectronical l yin anal og form on magnetic tape. Rather than record continuousl y,
instruments of thesetypes l ay dormant until triggered by the exceedance of a smal l
threshol d accel eration at thebeginning of the earthquake motion. As a resul t, any
vibrations that may have preceded triggeringwere not recorded, thereby introducing
a basel ine error into the accel eration record.To use the recorded ground motions
for engineering computations, the anal og groundmotion records must be digitized.
Original l y, digitization was performed manual l y with paper,pencil , and an
engineering scal e. Semiautomatic digitizers, with which a user moved a l ens with
crosshairs across an accel erogram mounted on a digitizing tabl e, were commonl y used
in the l ate60(a)(b)Figure 3.4 (a) Modern digital strong motion instrument (sol ar-
powered, l 6-bitresol ution, 250 sampl es/sec, GPS timing, and cel l ul ar modem) mounted
in cast-in-pl acereinforced concrete vaul t, and (b) compl eted instal l ation with
insul ated cover and sol arpanel (courtesy of Terra Technol ogy Corp, Redmond,
Washington).Sec. 3.2 Strong-Motion Measurement 61I\\""'.(b)~NISuspensionwireSmoked
gl asspl ate~DamPingdiskt(a)IIStyl us -----TDampingmagnetsFigure 3.5 (a) Typical
Wil mot-type seismoscope in which a fixed styl us scribes arecord of rel ative motion
on a smoked gl ass pl ate mounted on the suspendedseismoscope; (b) typical
seismoscope record. (After Newmark and Rosenbl ueth, 1971.)1970s. By pressing a
foot-operated switch, the coordinates of the crosshairs were recorded.These forms
of digitizing invol ved exacting and tiring work; operator accuracy and fatigue were
important considerations (Hudson, 1979). Ful l y automatic computer-based
digitization, typical l yat sampl ing rates of 200 or more sampl es per second, is now
commonpl ace.In recent years, digital seismographs have become much more commonl y
used.Al though they use anal og transducers, digital instruments convert the anal og
signal to digital form in the fiel d. They record ground motions continuousl y at
rates of 200 to 1000 sampl es/sec with 12- to 16-bit resol ution, saving the recorded
data onl y if a triggering accel erationis exceeded. Their on-board memories can
typical l y save 4 to 6 Mb worth of data, from beforean earthquake begins until after
it ends, thereby preserving the initial portion of the record thatis l ost with
triggered anal og systems. Because digital systems are more compl ex, more expensive,
and more difficul t to maintain in the fiel d, they have not yet repl aced anal og
systems.3.2.3 Strong-Motion ProcessingThe raw data obtained from a strong-motion
instrument may incl ude errors from several possibl esources, each of which must be
careful l y eval uated and corrected to produce an accuraterecord of the actual ground
motion. Raw data often incl ude background noise from differentsources. Microseisms
from ocean waves can be detected by sensitive instruments. Other noisemay be caused
by traffic, construction activity, wind (transmitted to the ground by vibration of
trees, buil dings, etc.), and even atmospheric pressure changes. Obviousl y, this
range of sourcescan produce nonseismic noise at both l ow and high frequencies. To
isol ate the motion actual l yproduced by the earthquake, background noise must be
removed or at l east suppressed.62 Strong Ground Motion Chap. 3Al l accel erographs
have their own dynamic response characteristics, or instrumentresponse, that can
infl uence the motions they measure. Consequentl y, instrument responsemust be
corrected for in strong-motion processing. Instrument response corrections are
usual l yperformed by model ing the instrument itsel f as a SDOF system and using the
SDOFmodel to decoupl e the response of the instrument from the actual ground motion.
For mostmodern accel erographs with fl at frequency response up to about 12 to 13 Hz,
the instrumentcorrection is onl y important for frequencies above the usual range of
engineering interest.However, some accel erographs are l ocated in buil dings (usual l y
on the ground fl oor or in thebasement) or near the abutments of dams or bridges.
The motions they record can be affectedat frequencies of interest by the response
of the structure in or near which they are l ocated.Even the motions recorded by
strong motion instruments l ocated in the free fiel d (away fromthe infl uence ofl arge
structures) may be infl uenced by the response of their instrument shel ter(Bycroft,
1978; Crouse et aI., 1984), al though these effects are usual l y important onl y at
rel ativel y high frequencies (Crouse and Hushmand, 1989) for typical instrument
shel ters.Another correction is required to reduce the effects of errors in ground
motion measurement,such as those associated with the triggering of anal og
seismographs. If a seismographdoes not start until some triggering l evel of motion
is reached, the entire accel erogramis in error by the l evel of motion at the time
of triggering. Integration of an uncorrectedaccel eration time history, for exampl e,
wil l produce a l inear error in vel ocity and a quadraticerror in displ acement. An
accel eration error as smal l as O.OOl g at the beginning of a 30-secl ongaccel erogram
woul d erroneousl y predict a permanent displ acement of 441 em at theend of the
motion. Correction of such errors, termed basel ine correction, was original l y
accompl ished by subtracting a best-fit parabol a from the accel erogram before
integrating tovel ocity and displ acement but is now performed using high-pass
fil ters and modern dataprocessing techniques (Joyner and Boore, 1988). The motions
shown in Figure 3.1, forexampl e, were bandpass fil tered to remove frequencies bel ow
0.08 Hz and frequenciesabove 23 Hz. Computer software for processing strong-ground-
motion records (Converse,1992) is avail abl e from the U.S. Geol ogical Survey (USGS).
3.2.4 Strong-Motion Instrument ArraysLarge earthquakes produce ground motions with
different characteristics at different pointson the ground surface. The spatial
variation of ground motion, whether on worl dwide,regional , or l ocal scal es, is
important in both seismol ogy and earthquake engineering.Arrays and networks of
strong motion instruments have proven useful in determining thespatial variation of
strong ground motion.3.2.4.1 Worl dwide and Regional ArraysUnderstanding of
earthquake and tectonic processes improved dramatical l y with theestabl ishment of
the Worl dwide Standard Seismograph Network (WWSSN) in 1961. TheWWSSN was original l y
devel oped, in l arge part, to monitor compl iance with nucl ear weaponstesting bans.
Before that time, worl dwide seismicity data were obtained from a widevariety of
very different types of instruments operated by many different organizations.
Differencesin instruments and operating procedures made comparison of resul ts
difficul t.WWSSN stations use standardized instruments; each station has at l east
two three-componentanal og seismographs to monitor both short- and l ong-period
motions. The capabil itiesof the WWSSN instruments, however, are l imited by modern
standards (Aki and Richards,Sec. 3.2 Strong-Motion Measurement 631980), and they
are being repl aced by digital instruments such as those of the Gl obal Digital
Seismometer Network (GDSN) and the Gl obal Seismographic Network (GSN). The
IncorporatedResearch Institutions for Seismol ogy (IRIS), a consortium of U.S. and
foreignresearch institutions, oversees operation of the GSN and a set of portabl e
instruments thatcan be depl oyed to monitor aftershocks fol l owing l arge earthquakes.
Regional arrays of seismographs are now operating in most seismical l y active
countries.In the United States, for exampl e, the USGS operates regional arrays in
different partsof the country. In the Cal ifornia Strong Motion Instrumentation
Program (CSMIP), the Cal iforniaDivision of Mines and Geol ogy operates an extensive
array of free-fiel d seismographs(Figure 3.6) as wel l as seismographs in buil dings
and bridges.LOS ANGELES AREA 37" .. • • • • Figure 3.6 Locations of strong motion
instruments operated by the U.S. Geol ogical Survery in cooperation with other
agencies as of April , 1990. Boxes in northern andsouthern Cal ifornia indicate areas
with high instrument density. (After Joyner and Boore,Geotechnical News, March,
1991, p, 24. Used by permission of BiTech Publ ishers, Ltd..)3.2.4.2 Local and Dense
ArraysWhil e widel y spaced regional and worl dwide arrays are useful for studying
earthquakemechanisms and the spatial distribution of many important earthquake
parameters, geotechnical earthquake engineering often requires spatial distribution
information on a smal l er areal scal e and bel ow the ground surface. In recent years a
number of l ocal and dense arrays, somewith downhol e instrumentation, have been
instal l ed at various l ocations around the worl d.Japan has been very active in the
instal l ation ofl ocal strong-motion instrument arrays.The three-dimensional dense
accel erometer array at Chiba (Katayama and Sato, 1982), forexampl e, incl udes 44
three-component accel erometers, 15 of which are at the ground surfaceand
the remainder at depths of up to 40 m (130 ft.). In Taiwan, the SMART-l dense
accel erometer array near Lotung (Figure 3.7) consists of a central accel erometer
surrounded64 Strong Ground Motion Chap. 3o115°30WIBrawl ey 010 kmArray.No 1No 11. "*
f;;. 32°45'N• No 121940 Epicenter "''l ~ . "'",No13 ~USL------- .~ Surface
accel erometer ------MEXICO 1979 E ic*enter8IIII ~ eNo2No 61 · No 3Imperial \ • • No
4 o No?. NoSEI Centro NO: • No10 No9Figure 3.7 Original configuration ofSMART-l
array in Lotung, Taiwan.Figure 3.8 The El Centro Array in southern Cal ifornia.The
El Centro Differential Array is l ocated near Station 9.by three rings of 12
accel erometers each at radii of 200 m (650 ft), 1 km (0.6 mi), and 2 krn(1.2 mi). A
few years after the SMART-l array was instal l ed, 1- and 1- -scal e model s of4 12
nucl ear containment structures were constructed in its midst (Tang, 1987). The
structureswere instrumented to record their response during earthquakes, and
additional surface anddownhol e (to depths of 47 m (154 ft)) ground motion
instruments were instal l ed adjacent tothe ~ -scal e model and in the free-fiel d.In
the United States, one of the most important l ocal arrays has been the EI Centro
Array, a 45-km-Iong (28 mi) array of 13 stations that crosses the Imperial and
Brawl eyfaul ts in southern Cal ifornia (Figure 3.8). It al so contains the EI Centro
Differential Array,a dense array consisting of six three-component accel erometers
al ong a 305-m (1000 ft)l ine. Shortl y after instal l ation, the arrays recorded the
1979 Imperial Val l ey earthquake(M,. =6.9), which occurred onl y 5.6 km away and
produced very useful information onnear-fiel d ground motions. Near Anza,
Cal ifornia, an array of ten three-component stationsal ong a 30-km (19 mi) stretch
of the San Jacinto faul t was instal l ed to study various earthquakecharacteristics
(Berger et al ., 1984). Data are tel emetered by digital VHF radio to anearby
mountain peak station and then on to another station in La Jol l a, Cal ifornia.These
are but a few of the many strong-motion arrays that have been instal l ed in
seismical l yactive countries around the worl d. The rapid prol iferation of l ocal ,
regional , andworl dwide seismograph arrays in recent years has come hand in hand
with technol ogical advances in data acquisition, storage, and communication. The
abil ity to acquire and storel arge quantities of digital seismic data at high
speeds, and to retrieve the data from remotel ocations by tel emetry, has and wil l
continue to make such data more pl entiful .3.2.5 Strong-Motion RecordsStrong-motion
records can now be easil y obtained from a number of sources. The U.S. Geol ogical
Survey, for exampl e, publ ished a compact disk (Seekins et al ., 1992) that contained
uncorrectedstrong motion records from North American earthquakes between 1933 and
1986; morethan 4000 records were incl uded. A number of strong-motion databases can
be accessed over theInternet, with individual records downl oaded by anonymous ftp
(fil e transfer protocol )Sec. 3.3 Ground Motion Parameters 65procedures. The Gil roy
records shown in Figure 3.1 were obtained from the database maintainedby the
Lamont-Doherty Earth Observatory at Col umbia University in conjunction with the
National Center for Earthquake Engineering Research at SUNY Buffal o. An exampl e of
theinformation provided with such records is shown in Figure 3.9. A useful Worl d
Wide Web sitewith l inks to many sources of ground motion and earthquake information
is maintained at theUniversity of Washington
(http: //www.geophys.washington.edul seismosutfing.html ). A varietyof geotechnical
earthquake engineering information can be found at a web site maintained at the
University of Southern Cal ifornia (http: //rccgOl .usc.edul eqdatal home.html ).3.155.54
-5.71-3.92-5.4512.52-25.70-42.300.0-0.192.78-15.56-13.34-12.18-43.07-46.07-32.89-
9.81-15.90-20.33-2.1615.48-26.0315.761.26-1.3811. 683.8229.66-16.0437.65-25.5467.54
9.493.4415.80-3.87-23.51-14.0915.818.83-6.74-20.14-8.69-10.6412.96-23.2439.13-4.54-
3.90-4.94-4.98-10.0714.4341.53-20.3024.915.467.40-3.88-9.950.25-15.52-17.72-89.33-
3.95-5.76-17.185.180.31-37.29-6.89-31. 37NCEER ASCII STRONG-MOTION DATA FORMATEVENT
PARAMETERS: DATE: y~ar=1989 month=10 day=18TIME: hourl minute (24hr)=0004 second=
2.200 time code=UTCLOCATIo~r: l atitude= 37.03700 l ongitude= -121.80300 depth (km)=
18.0NAME: S: ANTA CRUZ MTNS (LOMA PRIETA) EARTHQUAKESITE PARJl .METERS: LOCATIO~r:
l atitude= 36.97300 l ongitude= -121.57200 el evation (m)=SITE ID: 47379 - GILROY #1 -
GAVILAN COLLEGE, WATER TANKCODE: 47379RECORD/TRACE PARAMETERS: START D1~E: year=1989
month=10 day=18START Tl : ME: hourl minute (24hr)=0004 second=23.900 time code=UTC
EPICENTF~ DISTANCE: distance (km)= 21.8 azimuth (deg)=N289.1SPECS: s: ampl ing rate
(sec)=0.020~Lumber of points= 2000Il Lnits=CM/SEC/SECt: ype of data=ACCELERATIONdata
format=10F8.2 INTERPOLATEDCOMPO~IT OF MOTION: azimuth=NO (HORIZONTAL)COMMENT:
POLARITY OF MOTION RELATIVE TO GROUNDDATA: -15.68-3.67-8.23-0.33-22.9719.2326.86
74.64Figure 3.9 Event, size, and record information preceding the digitized
accel erationdata for the Gil roy No. I (rock) strong-motion record.3.3 GRC)UND
MOTION PARAMETERSGround motion parameters are essential for describing the
important characteristics ofstrong ground motion in compact, quantitative form.
Many parameters have been proposed66 Strong Ground Motion Chap. 3to characterize
the ampl itude, frequency content, and duration of strong ground motions;some
describe onl y one of these characteristics, whil e others may refl ect two or three.
Because of the compl exity of earthquake ground motions, identification of a singl e
parameterthat accuratel y describes al l important ground motion characteristics is
regarded asimpossibl e (Jennings, 1985; Joyner and Boore, 1988).3.3.1 Ampl itude
ParametersThe most common way of describing a ground motion is with a time history.
The motionparameter may be accel eration, vel ocity, or displ acement, or al l three
may be displ ayed asshown in Figure 3.10. Typical l y, onl y one of these quantities is
measured directl y with theothers computed from it by integration and/or
differentiation. Note the different predominantfrequencies in the accel eration,
vel ocity, and displ acement time histories. The accel erationtime history shows a
significant proportion of rel ativel y high frequencies. Integrationproduces a
smoothing or fil tering effect [in the frequency domain, v(ro) = a(ro)/ro andGil roy
NO.1 East - West SO'S ~co~0wQjoo« -0.5 : : : o 10 20 30 40Time (sec)20 30 40Time
(sec)§ 0.5co~0 ~ ~~".L~._,,~_~ IQjoo« -0.5 L-__-'----__'---_----"'----_---l o 10
~40EJ' ~ ~ -- : : -E~t-~ . . . • . > -40 -------'-------o 10 20 30 40Time (sec)
K20EJ j10~ ~w 0 - ~o~-10 -- --(J)is -20 : : : o 10 20 30 40Time (sec)~40Ej' ! 2~ : :
-- ---~'g -20 .. --~ -40 : : : o 10 20 30 40Time (sec) 120E3 E 10 - -w~o· ..j-10~o
-20 . . .o 10 20 30 40Time (sec)Figure 3.10 Accel eration, vel ocity, and
displ acement time histories for the E-Wcomponents of the Gil roy No. I (rock) and
Gil roy NO.2 (soil ) strong motion records.The vel ocities and displ acements were
obtained by integrating the accel eration recordsshown in Figure 3.1 using the
trapezoidal rul e. Note that the Gil roy No. I (rock) siteexperienced higher
accel erations, but the Gil roy No.2 (soil ) site experienced highervel ocities and
displ acements.Sec. 3.3 Ground Motion Parameters 67I: i(oo) = v(00)/00 , where it, V,
and aare the transformed displ acement, vel ocity,andaccel eration,respectivel y].
Therefore, the vel ocity time history shows substantial l y l ess high-frequencymotion
than the accel eration time history. The displ acement time history, obtainedby
another round of integration, is dominated by rel ativel y l ow frequency motion.
3.3.1.1 Peak Accel erationThe most commonl y used measure of the ampl itude of a
particul ar ground motion isthe peak horizontal accel eration (PHA). The PHA for a
given component of motion is simpl ythe l argest (absol ute) val ue of horizontal
accel eration obtained from the accel erogram ofthat component. By taking the vector
sum of two orthogonal components, the maximumresul tant PHA (the direction of which
wil l usual l y not coincide with either of the measuredcomponents) can be obtained.
Horizontal accel erations have commonl y been used to describe ground motionsbecause
of their natural rel ationship to inertial forces; indeed, the l argest dynamic
forcesinduced in certain types of structures (i.e., very stiff structures) are
cl osel y rel ated to thePHA. The PHA can al so be correl ated to earthquake intensity
(e.g., Trifunac and Brady,1975a; Murphy and O'Brien, 1977; Krinitzsky and Chang,
1987). Al though this correl ationis far from precise, it can be very useful for
estimation of PHA when onl y intensity informationis avail abl e, as in the cases of
earthquakes that occurred before strong motion instrumentswere avail abl e
(preinstrumental earthquakes). A number of intensity-accel erationrel ationships have
been proposed, several of which are shown in Figure 3.11. The use ofintensity-
attenuation rel ationships al so al l ows estimation of the spatial variabil ity of peak
accel eration from the isoseismal maps of historical earthquakes.100"'0Q)-!!!EoI510~
Q)Qi o~,Trifunac and Brady ,','(1975a) - horizontal ------,/Trifunac and Brady / "
(1975a) - vertical ~ /.a.~J-.t'------ Medvedevand----..... Sponheuer (1969)
Savarensky andKimos (1955)Kawasumi (1951)Ishimoto (1932)II IV VI VIII x XII
Equival ent modified Mercal l i intensityFigure 3.11 Proposed rel ationships between
PHA and MMI. (After Trifunac andBrady, 1975a. Used by permission of the
Seismol ogical Society of America.)68 Strong Ground Motion Chap. 3Vertical
accel erations have received l ess attention in earthquake
engineering than horizontal accel erations, primaril y because the margins of safety
against gravity-induced staticvertical forces in constructed works usual l y provide
adequate resistance to dynamic forcesinduced by vertical accel erations during
earthquakes. For engineering purposes, the peakvertical accel eration (PVA) is often
assumed to be two-thirds of the PHA (Newmark andHal l , 1982). The ratio ofPVA to
PHA, however, has more recentl y been observed to be quitevariabl e but general l y to
be greater than two-thirds near the source of moderate to l argeearthquakes and l ess
than two-thirds at l arge distances (Campbel l , 1985; Abrahamson andLitehiser, 1989).
Peak vertical accel erations can be quite l arge; a PVA of 1.74g was measuredbetween
the Imperial and Brawl ey faul ts in the 1979 Imperial Val l ey earthquake.Ground
motions with high peak accel erations are usual l y, but not al ways, more destructive
than motions with l ower peak accel erations. Very high peak accel erations that l ast
for onl ya very short period of time may cause l ittl e damage to many types of
structures. A number ofearthquakes have produced peak accel erations in excess of
0.5gbut caused no significant damageto structures because the peak accel erations
occurred at very high frequencies and the durationof the earthquake was not l ong.
Al though peak accel eration is a very useful parameter, itprovides no information on
the frequency content or duration of the motion; consequentl y, itmust be
suppl emented by additional information to characterize a ground motion accuratel y.
3.3.1.2 Peak Vel ocityThe peak horizontal vel ocity (PHV) is another useful parameter
for characterizationof ground motion ampl itude. Since the vel ocity is l ess
sensitive to the higher-frequencycomponents of the ground motion, as il l ustrated in
Figure 3.10, the PHV is more l ikel y thanthe PHA to characterize ground motion
ampl itude accuratel y at intermediate frequencies.For structures or facil ities that
are sensitive to l oading in this intermediate-frequency range(e.g., tal l or
fl exibl e buil dings, bridges, etc.), the PHV may provide a much more accurate
indication of the potential for damage than the PHA. PHV has al so been correl ated
to earthquakeintensity (e.g., Trifunac and Brady, 1975a; Krinitzsky and Chang,
1987).3.3.1.3 Peal ' Displ acementPeak displ acements are general l y associated with
the l ower-frequency componentsof an earthquake motion. They are, however, often
difficul t to determine accuratel y(Campbel l , 1985; Joyner and Boore, 1988), due to
signal processing errors in the fil tering andintegration of accel erograms and due
to l ong-period noise. As a resul t, peak displ acement isl ess commonl y used as a
measure of ground motion than is peak accel eration or peak vel ocity.Exampl e 3.1
Determine the peak accel erations, vel ocities, and displ acements for the E-
Wcomponents of theGil roy No. I (rock) and Gil roy No.2 (soil ) ground motions.
Sol ution The peak ampl itude val ues can be estimated graphical l y from Figure 3.9.
The actual peak val ues, based on the data from which Figure 3.9 was pl otted, are:
ParameterPeak accel erationPeak vel ocity (em/sec)Peak displ acement (em)Gil roy No. I
(Rock)0.442g33.78.5Gil roy No.2 (Soil )O.332g39.213.3Sec. 3.3 Ground Motion
Parameters 693.3.1.4 Other Ampl itude ParametersAl though the parameters discussed
previousl y are easil y determined, they describeonl y the peak ampl itudes of singl e
cycl es within the ground motion time history. In somecases, damage may be cl osel y
rel ated to the peak ampl itude, but in others it may require several repeated cycl es
of high ampl itude to devel op. Newmark and Hal l (1982) described theconcept of an
effective accel eration as "that accel eration which is most cl osel y rel ated to
structural response and to damage potential of an earthquake. It differs from and
is l ess thanthe peak free-fiel d ground accel eration. It is a function of the size
of the l oaded area, the frequencycontent of the excitation, which in turn depends
on the cl oseness to the source of theearthquake, and to the weight, embedment,
damping characteristic, and stiffness bf thestructure and its foundation." ISome
time histories are characterized by singl e-cycl e peak ampl itudes that are !much
greater than the ampl itudes of other cycl es. An exampl e of such a case is the Stone
cl nyonrecord shown in Figure 3.12a. These singl e cycl es often occur at high
frequencies and eonsequentl yhave l ittl e effect on structures with l ower natural
frequencies. In other time his ories,such as the Koyna record of Figure 3.12b, a
number of peaks of simil ar ampl itude are observed.(a) (b)Figure 3.12 Accel erograms
from (a) the N29W Mel endy Ranch record of the 1972Stone Canyon (M =4.6) earthquake
and (b) the l ongitudinal record from the 1967 Koyna(M =6.5) earthquake. The time
and accel eration scal es are identical for both records. Peakaccel erations are very
cl ose, il l ustrating the l imitations of using peak ampl itude as a sol emeasure of
strong ground motion. (After Hudson, 1979; used by permission OfEERL)Sustained
Maximum Accel eration and Vel ocity. Nuttl i (1979) usedl ower peaks of the
accel erogram to characterize strong motion by defining the sustainedmaximum
accel eration for three (or five) cycl es as the third (or fifth) highest (absol ute)
val ue of accel eration in the time history. The sustained maximum vel ocity was
defined simil arl y.Al though the PHA val ues for the 1972 Stone Canyon earthquake and
1967 Koynaearthquake records (Figure 3.12) were nearl y the same, a quick visual
inspection indicatesthat their sustained maximum accel erations (three- or five-
cycl e) were very different. For astructure that required several repeated cycl es of
strong motion to devel op damage, the70 Strong Ground Motion Chap. 3Koyna motion
woul d be much more damaging than the Stone Canyon motion, even thoughthey had
nearl y the same PHA. For these motions, the sustained maximum accel erationwoul d be
a better indicator of damage potential than the PHA.Exampl e 3.2Determine the three-
and five-cycl e sustained maximum accel erations and vel ocities for theE-W components
of the Gil roy No. I (rock) and Gil roy No.2 (soil ) ground motions.Sol ution The
sustained maximum accel eration and vel ocity val ues can be obtained graphical l yfrom
Figure 3.10. The actual val ues, based on the data from which Figure 3.10 was
pl otted, are: ParameterSustained maximum accel erationThree-cycl eFive-cycl eSustained
maximum vel ocity (em/sec)Three-cycl eFive-cycl eGil roy No. I (Rock)O.434gO.418g31.6
29.9Gil roy No.2 (Soil )O.312gO.289g38.438.2Effective Design Accel eration. The notion
of an effective design accel eration,with different definitions, has been proposed
by at l east two researchers. Since pul sesof high accel eration at high frequencies
induce l ittl e response in most structures, Benjaminand Associates (1988) proposed
that an effective design accel eration be taken as the peakaccel eration that remains
after fil tering out accel erations above 8 to 9 Hz. Kennedy (1980)proposed that the
effective design accel eration be 25% greater than the third highest (absol ute)peak
accel eration obtained from a fil tered time history.3.3.2 Frequency Content
ParametersOnl y the simpl est of anal yses (see Section B.5.3 of Appendix B) are
required to show thatthe dynamic response of compl iant objects, be they buil dings,
bridges, sl opes, or soil deposits,is very sensitive to the frequency at which they
are l oaded. Earthquakes produce compl icatedl oading with components of motion that
span a broad range of frequencies. Thefrequency content describes how the ampl itude
of a ground motion is distributed among differentfrequencies. Since the frequency
content of an earthquake motion wil l strongl y infl uencethe effects of that motion,
characterization of the motion cannot be compl ete withoutconsideration of its
frequency content.3.3.2.1 Ground Motion SpectraAny periodic function (i.e., any
function that repeats itsel f exactl y at a constant interval )can be expressed using
Fourier anal ysis as the sum of a series of simpl e harmonic termsof different
frequency, ampl itude, and phase. Using the Fourier series (see Section A.3 of
Appendix A), a periodic function, x(t), can be written asx (t) = Co + I. c.; sin
(OJnt + cjJn)n ~ 1(3.4)Sec. 3.3 Ground Motion Parameters 71In this form, Cnand <l >n
are the ampl itude and phase angl e, respectivel y, of the nth harmonicof the Fourier
series [see equations (A. 10) and (A. 11) for their definitions]. The Fourierseries
provides a compl ete description of the ground motion since the motion can be
compJl etel yrecovered by the inverse Fourier transform.Fourier Spectra. A pl ot of
Fourier ampl itude versus frequency [en versus wnfrom equation (3.4)] is known as a
Fourier ampl itude spectrum; a pl ot of Fourier phaseangl e (<I>n versus wn) gives the
Fourier phase spectrum. The Fourier ampl itude spectrum ofa strong ground motion
shows how the ampl itude of the motion is distributed with respectto frequency (or
period). It expresses the frequency content of a motion very cl earl y.The Fourier
ampl itude spectrum may be narrow or broad. A narrow spectrum impl iesthat the motion
has !l dominant frequency (or period), which can produce a smooth, al mostsinusoidal
time history. A broad spectrum corresponds to a motion that contains a variety of
frequencies that produce a more jagged, irregul ar time history. The Fourier
ampl itude spectrafor the E-W components of the Gil roy No.1 (rock) and Gil roy No.2
(soil ) motionsshown in Figure 3.13. The jagged shapes of the spectra are typical of
those observed for6 7 8I Gil roy No.1 (rock) I4 5Period - sec2 3U
3o,---------,-------,-----,--------,-----------------,wI/)IOl --: 20 ...W-0.~Q.~ 10
--W-c: : J~oLl !~~~---.l : =: : : : ~--..l .-~: : : r: ===: : : : t=====±=: ====: t====---Jo6 7 8I
Gil roy NO.2 (soil ) I4 5Period - sec2 3U 30,--------,-------,-----,--------,--------
_--_-__-,wI/)IOl --:
20 --W-0: : J: t=Q.~ 10 ._.m'C: : J~oLl !~~l ..------.Jl ..--~=--.J----.J--
======: : : : j=====±=====--JoFigure 3.13 Fourier ampl itude spectra for the E-W
components of the Gil roy No. I(rock) and Gil roy No.2 (soil ) strong motion records.
Fourier spectra were obtained bydiscrete Fourier transform (Section A.3.3 of
Appendix A) and consequentl y have units ofvel ocity. Fourier ampl itude spectra can
al so be pl otted as functions of frequency (seeFigure E3.3).72 Strong Ground Motion
Chap. 3individual ground motions. The shapes of the spectra are quite different:
the Gil roy No.1(rock) spectrum is strongest at l ow periods (or high frequencies)
whil e the reverse isobserved for the Gil roy No.2 (soil ) record. A difference in
frequency content can bedetected by cl osel y examining the motions in the time
domain (Figure 3.1), but the differenceis expl icitl y il l ustrated by the Fourier
ampl itude spectra.When the Fourier ampl itude spectra of actual earthquake motions
are smoothed andpl otted on l ogarithmic scal es, their characteristic shapes can be
seen more easil y. As il l ustratedin Figure 3.14, Fourier accel eration ampl itudes
tend to be l argest over an intermediaterange of frequencies bounded by the
cornerfrequency t: on the l ow side and the cutofffrequencyfrnaxon the high side.
The comer frequency can be shown theoretical l y (Brune, 1970,1971) to be inversel y
proportional to the cube root of the seismic moment. This resul t indicatesthat
l arge earthquakes produce greater l ow-frequency motions than do smal l er
earthquakes.The cutoff frequency is not wel l understood; it has been characterized
both as anear-site effect (Hanks, 1982) and as a source effect (Papageorgiou and
Aki, 1983) and isusual l y assumed to be constant for a given geographic region.Q)~Cl
g(])"0.~Q.EctICD'C: : J aLL.Frequency (l og scal e)Figure 3.14 Ideal ized shape of
smoothedFourier ampl itude spectrum il l ustrating thecorner frequency, fe' and cutoff
frequency,fm.x·Since phase angl es control the times at which the peaks of harmonic
motions occur(Appendix A), the Fourier phase spectrum infl uences the variation of
ground motion withtime. In contrast to Fourier ampl itude spectra, Fourier phase
spectra from actual earthquakerecords do not displ ay characteristic shapes.Power
Spectra. The frequency content of a ground motion can al so be describedby a power
spectrum or power spectral density function. The power spectral density functioncan
al so be used to estimate the statistical properties of a ground motion and to
compute stochasticresponse using random vibration techniques (Cl ough and Penzien,
1975; Vanmarcke,1976; Yang, 1986).The total intensity of a ground motion of
duration Td is given in the time domain bythe area under the time history of
squared accel eration: (3.5)Sec. 3.3 Ground Motion Parameters 73(3.10)Using Parseval '
s theorem, the total intensity can al so be expressed in the frequency domain, asI =
.!.fc 2 o 7T n di» (3.6)owhere mN = 1t /~t is the Nyquist frequency (the highest
frequency in the Fourier series). Theaverage intensity, An, can be obtained by
dividing equations (3.5) and (3.6) by the duration.Td OJNx, = .If [a (t)] 2dt = -L.
Tfc~ di» (3.7)Td 1t do 0Notice that the average intensity is equal to the mean-
squared accel eration. The power spectral density, G(m), is defined such thatOJNAO =
fG(ro) di» (3.8)ofrom which we can easil y see, by comparing equations (3.7) and
(3.8), thatG(OJ) = IT c~ (3.9)7T dThe cl ose rel ationship between the power spectral
density function and the Fourier ampl itudespectrum is apparent from equation (3.9).
The power spectral density is often normal izedby dividing its val ues by the area
beneath itGn(m) = ioG(m)where Ao, as before, is the mean-squared accel eration.The
power spectral density function is useful in characterizing the earthquake as a
random process. The power spectral density function by itsel f can describe a
stationary randomprocess (i.e., one whose statistical parameters do not vary with
time). Actual strongmotion accel erograms, however, frequentl y show that the
intensity buil ds up to a maximumval ue in the earl y part of the motion, then remains
approximatel y constant for a period oftime, and final l y decreases near the end of
the motion. Such nonstationary random processbehavior is often model ed by
mul tipl ying a stationary time history by a deterministic intensityfunction (e.g.,
Hou, 1968; Shinozuka, 1973; Saragoni and Hart, 1983). Changes in frequencycontent
during the motion have been described using an evol utionary powerspectrum approach
(Priestl ey, 1965, 1967; Liu, 1970).Response Spectra. A third type of spectrum is
used extensivel y in earthquakeengineering practice. The response spectrum describes
the maximum response of a singl edegree-of-freedom (SDOF) system to a particul ar
input motion as a function of the natural frequency (or natural period) and damping
ratio of the SDOF system (Section B.7 ofAppendix B). Computed response spectra for
the Gil roy No.1 (rock) and Gil roy No.2 (soil )records are il l ustrated in Figure
3.15.74 Strong Ground Motion Chap. 3Gil roy NO.2 (soil )2Gil roy No.1 (rock)---~; 0:
K==: : : : : : : : : =JI ~; 0: E;: Jo 1 3 0 2 3,: ,: I: : ;d ,: ,: : IL: Jo 1 2 3 0 1 2 3Period,
T(sec) Period, T(sec)Figure 3.15 Response spectra (5% damping) for Gil roy No. I
(rock) and Gil roy No.2(soil ) strong motion records. The frequency contents of the
two motions are refl ected inthe response spectra. The Gil roy I (rock) motion, for
exampl e, produced higher spectral accel erations at l ow periods than did the Gil roy 2
(soil ) motion, and l ower spectral accel erations at higher periods. The higher l ong-
period content of the Gil roy 2 (soil )motion produced spectral vel ocities and
displ acements much higher than those of theGil roy I (rock) motion.Response spectra
may be pl otted individual l y to arithmetic scal es, or may be combined,by virtue of
the rel ationships of equation (3.11), in tripartite pl ots (Section A.2.2).The
tripartite pl ot displ ays spectral vel ocity on the vertical axis, natural frequency
(orperiod) on the horizontal axis, and accel eration and displ acement on incl ined
axes. Theaccel eration and displ acement axes are reversed when the spectral val ues
are pl otted againstnatural period rather than natural frequency. The shapes of
typical response spectra indicatethat peak spectral accel eration, vel ocity, and
displ acement val ues are associated with differentfrequencies (or periods). At l ow
frequencies the average spectral displ acement isnearl y constant; at high
frequencies the average spectral accel eration is fairl y constant. Inbetween l ies a
range of nearl y constant spectral vel ocity. Because of this behavior, response
spectra are often divided into accel eration-control l ed (high-frequency), vel ocity-
control l ed(intermediate-frequency), and displ acement-control l ed (l ow-frequency)
portions.El astic response spectra assume l inear structural force-displ acement
behavior. Formany real structures, however, inel astic behavior may be induced by
earthquake groundSec. 3.3 Ground Motion Parameters 75motions. An inel astic response
spectrum (i.e., one that corresponds to a nonl inear forcedispl acementrel ationship,
can be used to account for the effects of inel astic behavior. Figure3.16 shows
inel astic response spectra for accel eration and yiel d displ acement for various
val ues of the ductil ity factor 11 = umaJuy , where Umax is the maximum al l owabl e
displ acement and u y is the yiel d displ acement. A separate inel astic spectrum must
be pl ottedto show total (el astic pl us pl astic) displ acement. Spectral accel erations
decrease withincreasing ductil ity, but total displ acements increase.---------L----
t0.1320.50.20.050.030.03 0.05 0.1 0.2 0.5 1 2 5 10 2030Undamped natural frequency
(HZ)Uy umaxDispl acement,J) o: 5LL.Figure 3.16 Inel astic response spectra for the EI
Centro N-S component of the 1940Imperial Val l ey earthquake. Onl y the el astic
component of the displ acement is pl otted.Spectral accel erations are correct, but
spectral vel ocities are not. (After Newmark andHal l , 1982; used by permission
ofEERI.)Response spectra refl ect strong ground motion characteristics indirectl y,
since theyare "fil tered" by the response of a SDOF structure. The ampl itude,
frequency content, andto a l esser extent, duration of the input motion al l
infl uence spectral val ues. The differentfrequency contents of the Gil roy No. I
(rock) and Gil roy No.2 (soil ) ground motions arecl earl y il l ustrated by the
different shapes of their respective response spectra (Figure 3.15).It is important
to remember that response spectra represent onl y the maximumresponses of a number
of different structures. However, the response of structures is of greatimportance
in earthquake engineering, and the response spectrum has proven to be animportant
and useful tool for characterization of strong ground motion.3.3.2.2 Spectra'
ParametersSection 3.3.2.1 described three types of spectra that can be used to
characterize strongground motion. The Fourier ampl itude spectrum and the cl osel y
rel ated power spectral density,combined with the phase spectrum, can describe a
ground motion compl etel y. Theresponse spectrum does not describe the actual ground
motion, but it does provide val uabl eadditional information on its potential effects
on structures. Each of these spectra is a compl icatedfunction and, as with time
histories, a great many data are required to describe themcompl etel y. A number of
spectral parameters have been proposed to extract importantpieces of information
from each spectrum.76 Strong Ground Motion Chap. 3Predominant Period. A singl e
parameter that provides a useful , al thoughsomewhat crude representation of the
frequency content of a ground motion is the predominantperiod, Tp- The predominant
period is defined as the period of vibration correspondingto
the maximum val ue of the Fourier ampl itude spectrum. To avoid undue infl uence of
individual spikes of the Fourier ampl itude spectrum, the predominant period is often
obtainedfrom a smoothed spectrum. Whil e the predominant period provides some
informationregarding the frequency content, it is easy to see (Figure 3.17) that
motions with radical l ydifferent frequency contents can have the same predominant
period.PeriodTpFigure 3.17 Two hypothetical Fourier ampl itude spectra with the same
predominantperiod but very different frequency contents. The upper curve describes
a widebandmotion and the l ower a narrowband motion.Exampl e 3.3Determine the
predominant periods for the E-W components of the Gil roy No.1 (rock) and Gil royNo.2
(soil ) ground motions.Sol ution The Fourier ampl itude spectra of most ground motions
are quite jagged in the vicinityof their peaks, so some smoothing is required to
identify the predominant period. Thesmoothing and predominant period identification
is most easil y accompl ished by pl otting theFourier ampl itude spectrum as a function
of frequency. By numerical l y smoothing their Fourierampl itude spectra (Figure
E3.3), the predominant periods areGil roy NO.1 (rock): Tp = 0.39 secGil roy No.2
(soil ): T; = 0.53 secNote that the predominant period of the Gil roy No.2 (soil )
motion is greater than that of the Gil royNo. I (rock) motion, thereby il l ustrating
the rel ative strength of the l onger period (l owerfrequency)components of the Gil roy
No. I (soil ) motion.Bandwidth. The predominant period can be used to l ocate the
peak of the Fourierampl itude spectrum; however, it provides no information on the
dispersion of spectral ampl itudes about the predominant period. The bandwidth of the
Fourier ampl itude spectrumis the range of frequency over which some l evel of
Fourier ampl itude is exceeded.Bandwidth is usual l y measured at the l evel where the
power of the spectrum is hal f its maximumval ue; this corresponds to a l evel of 1/J2
times the maximum Fourier ampl itude.The irregul ar shape of individual Fourier
ampl itude spectra often renders bandwidth difficul tto eval uate. It is determined
more easil y for smoothed spectra.Central Frequency. The power spectral density
function can be used to estimatestatistical properties of the ground motion.
Defining the nth spectral moment of aground motion bySec. 3.3 Ground Motion
Parameters 7710 15 20 25Frequency - Hz5~
30,------.,--------.,--------.,--------.,------.,--------,o(J) enIen~ 20"0(a) ,g0..
Eal CD"§~10 15 20 25Frequency - Hz5~
30,------,---------,---------,---------,----------,<.>(J) enIen~ 20 ...-0(b) ,g0..~
10 ..CD~ ol JL~Jl ~~~~~~f!1~~~_~~~.....,""""~"""'~~"""""""''''''''."JoFigure E3.3 Raw
and smoothed FAS for Gil roy No.1 (rock) and Gil roy No.2 (soil )motions.An = fmnG(m)
dmothe central frequency Q (Vanmarcke, 1976) is given by(3.11)Q= r: ; (3.12)~~The
central frequency is a measure of the frequency where the power spectral density is
concentrated.It can al so be used, al ong with the average intensity and duration, to
cal cul ate thetheoretical median peak accel erationUmax = noIn( 2.8 ~: d) (3.13)Shape
Factor. The shapefactor (Vanmarcke, 1976) indicates the dispersion ofthe power
spectral density function about the central frequency: 8= )1- AT (3.14),1,0,1,2The
shape factor al ways l ies between 0 and 1, with higher val ues corresponding tol arger
bandwidths.78 Strong Ground Motion Chap. 3Kanai-Tajimi Parameters. Al though
individual power spectral density functionsmay have highl y irregul ar shapes,
averaging a number of normal ized power spectral density functions for simil ar strong
ground motions reveal s a smooth characteristic shape.Kanai (1957) and Tajimi (1960)
used a l imited number of strong motion records to proposethe fol l owing three-
parameter model for power spectral density: 1 + [2~g(co/COg)] 2G(co) = Go (3.15)[1-
(CO/COg)2]2 + [2~g(CO/COg)]2where the parameters Go, ~g , and COg determine the
shape of the function (Figure 3.18).Figure 3.18 Shape of the Kanai-Tajimipower
spectral density function.The displ acement response of a SDOF system with natural
frequency COg and dampingratio ~g to white noise base motion woul d be described by
a Kanai-Tajimi power spectral density function. As such, high-frequency components
of the input motion wil l be attenuated,and frequency components in the vicinity of
COg wil l be ampl ified. Typical val ues ofKanai-Tajimi parameters for various site
conditions are shown in Tabl e 3-1.Cl ough and Penzien (1975) proposed a correction
to the Kanai-Tajimi power spectral density function to prevent excessive vel ocities
and displ acements at very l ow frequencies.The corrected Kanai-Tajimi power spectral
density function requires two additional parametersto describe the power spectral
density.3.3.2.3 vmaxl "amaxBecause peak vel ocities and peak accel erations are
usual l y associated with motions ofdifferent frequency, the ratio vrnaJarnax shoul d
be rel ated to the frequency content of themotion (Newmark, 1973; Seed et al ., 1976;
McGuire, 1978). For a simpl e harmonic motionTabl e 3-1 Ground Intensity, Ground
Frequency, and Ground Damping for Various SiteConditionsGround Ground GroundGround
Site Number of Intensity, Frequency, Damping,Motion Conditions Records Go Ol g ~g
Horizontal Al l uvium 161 0.102 18.4 0.34Al l uvium on rock 60 0.078 22.9 0.30Rock 26
0.070 27.0 0.34Vertical Al l uvium 78 0.080 26.2 0.46Al l uvium on rock 29 0.072 29.1
0.46Rock 13 0.053 38.8 0.46Source: El ghadamsi et al . (1988).Sec. 3.3 Ground Motion
Parameters 79of period T, for exampl e, vmax/amax =T/2rc. For earthquake motions
that incl ude manyfrequencies, the quantity 2rc(vmax/amax) can be interpreted as the
period of vibration of anequival ent harmonic wave, thus providing an indication of
which periods of the groundmotion are most significant. Seed and Idriss (1982)
suggested the fol l owing representativeaverage val ues for different site conditions
l ess than 50 km from the source: SiteConditionRockStiff soil s« 200 ft)
Deepstiffsoil s(>200 ft)55 crn/sec/g110 crn/sec/g135cm/sec/g0.056sec0.112 sec
0.138secThe corresponding periods of equival ent harmonic waves for the rock, stiff
soil , and deepstiff soil site conditions are 0.35 sec, 0.70 sec, and 0.87 sec,
respectivel y, which indicates ashift toward l onger-period (l ower-frequency) motion
on softer soil deposits.Exampl e 3.4Determine the ratio vmm/amax for the N-S
components of the Gil roy NO.1 (rock) and Gil royNo.2 (soil ) ground motions. Compare
the quantities 2n(vma/amaJ with the predominant periodsof the motions.Sol ution
Based on the vmax and a max val ues from Exampl e 3.1,Gil roy No.1 (rock): Vmax 33.7
crn/sec = 0.078 secamax 0.442 (981 ern/sec")Gil N 2 ('1) Vmax I roy o. SOl : 39.2
em/sec -_ 0.124 secamax 0.322 (981 em/sec")The quantity 2n(vmax/amax) is equal to
the predominant period of a simpl e harmonic motion. Tosee how wel l it corresponds
to the predominant period of the Gil roy NO.1 (rock) and Gil royNo.2 (soil ) ground
motions,Gil roy NO.1 (rock): Gil roy No.2 (soil ): 2nVmax = 0.49 seca max2nVmax = 0.78
secamaxTp = 0.39secTp = 0.53secThough the ratio vmax/amax certainl y indicates that
the Gil roy NO.1 (rock) motion has a higherfrequency content than the Gil roy No.2
(soil ) motion, it overestimates the predominant periodof both the Gil roy NO.1
(rock) motion and Gil roy No.2 (soil ) motions. Due to the approximatenature of the
predominant period and the stochastic nature of both Vmax and amax, cl ose agreement
between vma/amax and predominant period shoul d not be expected.3.3.3 DurationThe
duration of strong ground motion can have a strong infl uence on earthquake damage.
Many physical processes, such as the degradation of stiffness and strength of
certain types ofstructures and the buil dup of porewater pressures in l oose,
saturated sands, are sensitive tothe number of l oad or stress reversal s that occur
during an earthquake. A motion of shortduration may not produce enough l oad
reversal s for damaging response to buil d up in a structure,even if the ampl itude of
the motion is high. On the other hand, a motion with moderateampl itude but l ong
duration can produce enough l oad reversal s to cause substantial damage.80 Strong
Ground Motion Chap. 3The duration of a strong ground motion is rel ated to the time
required for rel ease ofaccumul ated strain energy by rupture al ong the faul t. As the
l ength, or area, of faul t ruptureincreases, the time required for rupture
increases. As a resul t, the duration of strong motionincreases with increasing
earthquake magnitude. Whil e this rel ationship has been supportedby empirical
evidence for many years, advances in source mechanism model ing (Hanks andMcGuire,
1981) have provided theoretical support, indicating that the duration shoul d be
proportional to the cube root of the seismic moment. When bil ateral rupture [i.e.,
rupturethat propagates in opposite directions from the focus (as in the case of the
1989 Lorna Prietaearthquake)] occurs, the strong motion duration may be
considerabl y l ower.An earthquake accel erogram general l y contains al l accel erations
from the time theearthquake begins until the time the motion has returned to the
l evel of background noise.For engineering purposes, onl y the strong-motion portion
of the accel erogram is of interest.Different approaches have been taken to the
probl em of eval uating the duration of strongmotion in an accel erogram. The
bracketed duration (Bol t, 1969) is defined as the timebetween the first and l ast
exceedances of a threshol d accel eration (usual l y O.05g). Anotherdefinition of
duration (Trifunac and Brady, 1975b) is based on the time interval between the
points at which 5% and 95% of the total energy has been recorded. Boore (1983) has
takenthe duration to be equal to the corner period (i.e., the inverse ofthe comer
frequency).
Therate of change of cumul ative root-mean-square (rms) accel eration has al so been
used as thebasis for eval uation of strong-motion duration (McCann and Shah, 1979).
Power spectral density concepts can al so be used to define a strong-motion duration
(Vanmarcke and Lai,1977). Other definitions of strong-motion duration have been
proposed (Perez, 1974; Trifunacand Westermo, 1977). Because it impl icitl y refl ects
the strength of shaking, thebracketed duration is most commonl y used for earthquake
engineering purposes.The duration of strong motion has been investigated by
interpretation of accel erogramsfrom earthquakes of different magnitudes. Using a
O.05g threshol d accel eration,Chang and Krinitszky (1977) estimated the bracketed
durations for soil and rock sites atshort (l ess than 10 km) epicentral distances
shown in Tabl e 3-2.Duration has al so been expressed in terms of equival ent cycl es
of ground motion. Onesuch approach was devel oped in conjunction with an earl y
procedure for eval uation of l iquefactionpotential (Seed et al ., 1975). To represent
an irregul ar time history of shear stressTabl e 3-2 Typical Earthquake Durations at
Epicentral Distances Less Than 10 kmDuration (sec)Magnitude5.05.56.06.57.07.58.08.5
Rock Sites4681116223143Soil Sites812162332456286Source: Chang and Krinitzsky
(1977).Sec. 3.3 Ground Motion Parameters 81by a uniform series of harmonic stress
cycl es, the concept of an equival ent number of significantstress cycl es was
devel oped. The equival ent number of uniform stress cycl es,shown in Tabl e 3-3, was
sel ected to cause pore pressure buil dup equival ent to that of anactual shear
stress-time history at a harmonic stress ampl itude of 65% of the maximumactual
shear stress.Tabl e 3-3 Equival ent Number ofUniform Stress Cycl esEarthquakeMagnitude
Number of SignificantStress Cycl es2-35101526Exampl e 3.5Determine the bracketed
durations of the E-W components of the Gil roy No. I (rock) and Gil royNo.2 (soil )
ground motions.Sol ution Based on a threshol d accel eration ofO.05g, the bracketed
durations can be obtainedgraphical l y from the accel erograms shown in Figure E3.5.
0.5 ~---------------------------~: §co(a) ~ 0CDa>o~Firstexceedance(+0.059~-0.059Last
exceedance(+0.059,~ IA~ ~R ~i\I!~ii~ r ~-0.059~ Firstexceedance: -0.5o0.5: §co(b) ~ 0
CDa>~-0.5o55101015 29Time (sec)15 20Time (sec)FigureE3.5252530303535404082Gil roy
No.1 (rock): Gil roy No.2 (soil ): 3.3.4 Other Ground Motion ParametersStrong Ground
MotionTd = 9.8 secTd = 14.7 secChap. 3The preceding parameters rel ated primaril y to
the ampl itude, frequency content, or durationof a ground motion. Since al l of these
characteristics are important, ground motion parametersthat refl ect more than one
are very useful . The fol l owing paragraphs present a numberof parameters that
refl ect two or three important ground motion characteristics.A singl e parameter
that incl udes the effects of the ampl itude and frequency content ofa strong motion
record is the rms accel eration, defined asTdarms idS[a(t)]2dt = JIO (3.16)owhere Td
is the duration of the strong motion and AD is the average intensity (or
meansquaredaccel eration). Because the integral in equation (3.16) is not strongl y
infl uenced byl arge, high-frequency accel erations (which occur onl y over a very
short period of time) andbecause it is infl uenced by the duration of the motion,
the rms accel eration can be very useful for engineering purposes. Its val ue,
however, can be sensitive to the method used todefine strong motion duration.A
parameter cl osel y rel ated to the rms accel eration is the Arias intensity (Arias,
1970), defined ast, = 2~S[a(t)]2dt (3.17)oThe Arias intensity has units of vel ocity
and is usual l y expressed in meters per second.Since it is obtained by integration
over the entire duration rather than over the duration ofstrong motion, its val ue
is independent of the method used to define the duration of strongmotion.Exampl e
3.6Determine the rms accel erations and Arias intensities of the E-W components of
the Gil royNo.1 (rock) and Gil roy No.2 (soil ) ground motions.Sol ution By integrating
the accel erograms of the Gil roy No.1 (rock) and Gil roy No.2 (soil )ground motions
numerical l y, the rms accel erations and Arias intensities areGil roy NO.1 (rock):
Gil roy No.2 (soil ): arms = O.112garms = O.072gfa = 1.667 m/secfa = 1.228 m/secThe
characteristic intensity, defined ast, = a~sT~·5 (3.18)is rel ated l inearl y to an
index of structural damage due to maximum deformations andabsorbed hysteretic
energy (Ang, 1990).Sec. 3.3 Ground Motion Parameters 83(3.20)The cumul ative
absol ute vel ocity is simpl y the area under the absol ute accel erogram: r,CAY =
f1a(t)l dt (3.19)aThe cumul ative absol ute vel ocity has been found to correl ate wel l
with structural damagepotential . For exampl e, a CAV of 0.30g-sec (obtained after
fil tering out frequencies above10 Hz) corresponds to the l ower l imit for MMIVII
shaking (Benjamin and Associates,1988).Since many structures have fundamental
periods between 0.1 and 2.5 sec, theresponse spectrum ordinates in this period
range shoul d provide an indication of the potential response of these structures.
The response spectrum intensity (Housner, 1959) wastherefore defined as2.5SI(~) =
fPSV (~, T) dT0.1(i.e., the area under the pseudovel ocity response spectrum between
periods of 0.1 sec and2.5 sec. The response spectrum intensity, as indicated in
equation (3.20), can be computedfor any structural damping ratio. It captures
important aspects of the ampl itude and frequencycontent (in the range of primary
importance for structures) in a singl e parameter.Von Thun et al . (1988) referred to
the response spectrum intensity for 5% damping asthe vel ocity spectrum intensity.
The vel ocity spectrum intensity was suggested as being useful for eval uation of the
response of earth and rockfil l dams, which typical l y have fundamental periods
between 0.6 and 2.0 sec (Makdisi and Seed, 1978). To characterize strongground
motion for anal ysis of concrete dams, which general l y have fundamental periods of
l ess than 0.5 sec, Von Thun et al . (1988) introduced the accel eration spectrum
intensity,defined as0.5ASI = fSa (~=0.05,T) dT (3.21)0.1(i.e., the area under the
accel eration response spectrum between periods of 0.1 sec and0.5 sec).The Appl ied
Technol ogy Council (1978) defined two factors by which standardresponse spectra
coul d be normal ized. The effective peak accel eration (EPA) was definedas the
average spectral accel eration over the period range 0.1 to 0.5 sec divided by 2.5
(thestandard ampl ification factor for a 5% damping spectrum). The effective peak
vel ocity(EPV) was defined as the average spectral vel ocity at a period of 1 sec
divided by 2.5.Determination of EPA and EPV is shown schematical l y in Figure
3.19.The process of averagingthe spectral accel erations and vel ocities over a range
of periods minimizes the infl uenceofl ocal spikes in the response spectrum on the
EPA and EPV. The EPA and EPV havebeen used in the specification of smoothed design
response spectra in buil ding codes(Chapter 8).84100.1Strong Ground Motion
-----------------.---- : : IEe~ = ;51jEev = ~IChap. 30.1 0.5Period (sec)5 10 50
Figure 3.19 Determination of effective peak accel eration and effective peak
vel ocityfrom response spectra. (After Appl ied Technol ogy Council , 1978.)3.3.5
DiscussionA wide variety of strong motion parameters have been presented. Some
describe onl y theampl itude of the motion, others onl y the frequency content or
duration. Some of theseparameters are infl uenced by two or three of these important
ground motion characteristics.Tabl e 3-4 indicates which ground motion
characteristics strongl y infl uence the variousground motion parameters.Seismic
hazard anal yses (Chapter 4) and the devel opment of design ground motions(Chapter 8)
rel y heavil y on the characterization of strong ground motion by ground motion
parameters. Characterization by a singl e parameter is onl y rarel y appropriate; the
use ofseveral parameters is usual l y required to describe adequatel y the important
characteristicsof a particul ar ground motion. Since different engineering probl ems
are infl uenced by differentground motion characteristics, the significance of
different parameters depends onthe types of probl ems for which they are used.3.4
ESTIMATION OF GROUND MOTION PARAMETERSProper design of earthquake-resistant
structures and facil ities requires estimation of thel evel of ground shaking to
which they wil l be subjected. Since the l evel of shaking is mostSec. 3.4 Estimation
of Ground Motion Parameters 85Tabl e 3-4 Ground Motion Characteristics That Are
Strongl y Refl ected in Various GroundMotion ParametersGround Motion Characteristic
Ground Motion Parameter Ampl itude Frequency Content DurationPeak accel eration, PHA
and PHV xPeak vel ocity, PHV xSustained maximum accel eration, SMA xEffective design
accel eration, EDA xPredominant period, Tp xBandwidth xCentral frequency, Q xShape
factor, Ii xPower spectrum intensity, Go x x xGround frequency, wg xGround damping,
~g xvma/amax XDuration, TdXrms accel eration, arms X XCharacteristic intensity, Ie X
X xArias intensity, Ia x x xCumul ative absol ute vel ocity, CAY x x xResponse
spectrum intensity, SI(~) x xVel ocity spectrum intensity, VSI x xAccel eration
spectrum intensity, ASI x xEffective peak accl eration, EPA x xEffective peak
vel ocity, EPV x xconvenientl y described in terms of ground motion parameters,
methods for estimatingground motion parameters are required. Predictive
rel ationships, which express a particul arground motion parameter in terms of the
quantities that affect it most strongl y, are used toestimate ground motion
parameters. Predictive rel ationships pl ay an important rol e in seismichazard
anal yses
(Chapter 4) used for seismic design.3: .4.1 Magnitude and Distance EffectsMuch of
the energy rel eased by rupture al ong a faul t takes the form of stress waves. Since
the amount of energy rel eased in an earthquake is strongl y rel ated to its
magnitude, thecharacteristics of the stress waves wil l al so be strongl y rel ated to
magnitude. Figure 3.20il l ustrates the infl uence of earthquake magnitude on actual
ground motion characteristics inthe time domain. Each earthquake came from
essential l y the same source, and each accel erogramwas measured at about the same
distance from the source. The variations in ampl itude,frequency content, and
duration with magnitude are apparent.As stress waves travel away from the source of
an earthquake, they spread out and arepartial l y absorbed by the material s they
travel through. As a resul t, the specific energy(energy per unit vol ume) decreases
with increasing distance from the source. Since thecharacteristics of stress waves
are strongl y rel ated to specific energy, they wil l al so bestrongl y rel ated to
distance. The distance between the source of an earthquake and a86 Strong Ground
Motion Chap. 3Figure 3.20 Accel erograms from sixearthquakes off the Pacific coast
of Mexico.Each accel erogram was measured at nearl ythe same epicentraJ distance. The
recordfrom the M = 8. J (1985 Michoacan)earthquake continues for another 25 sec.
(After Anderson, 1991, Geotechnical News,Vol . 9, No. J, p. 35. Used by permission
ofBiTech Publ ishers, Ltd.)40 50Accel erationscal e: J100S~~zM=7.0M=5.1M= 3.1M=4.120
30Time (sec)10"',oU -+'1j ~ M=5.5t--~~~~~~,~ "-"r'l n'~,!II'If"'I'~TC-O""'YOJCD o~
particul ar site can be interpreted in different ways. Figure 3.21 il l ustrates some
of the mostcommonl y used measures of distance. Rl and R2 are the hypocentral and
epicentral distances,which are the easiest distances to determine after an
earthquake. If the l ength of faul trupture is a significant fraction of the distance
between the faul t and the site, however,energy may be rel eased cl oser to the site,
and Rl and R2 may not accuratel y represent the"effective distance." R3 is the
distance to the zone of highest energy rel ease. Since ruptureof this zone is l ikel y
to produce the peak ground motion ampl itudes, it represents the bestdistance
measure for peak ampl itude predictive rel ationships. Unfortunatel y, its l ocation is
difficul t to determine after an earthquake and nearl y impossibl e to predict before
an earthquake.R4 is the cl osest distance to the zone of rupture (not incl uding
sediments overl yingbasement rock) and R5 is the cl osest distance to the surface
projection of the faul t rupture.R4 and R5 have both been used extensivel y in
predictive rel ationships.Surface projectionEpicenter ~Hi9h-stres~zoneFaul t rupture
surfaceHypocenterFigure 3.21 Various measures of distanceused in strong-motion
predictive rel ationships.(After Shakal and Bernreuter, 1981.)3.4.2 Devel opment of
Predictive Rel ationshipsPredictive rel ationships usual l y express ground motion
parameters as functions of magnitude,distance, and in some cases, other variabl es,
for exampl e,(3.22)Sec. 3.4 Estimation of Ground Motion Parameters 87where Y is the
ground motion parameter of interest, M the magnitude of the earthquake, Ra measure
of the distance from the source to the site being considered, and the Pi are other
parameters (which may be used to characterize the earthquake source, wave
propagationpath, and/or l ocal site conditions). Predictive rel ationships are
devel oped by regressionanal yses of recorded strong motion dat~bases. As such, they
change with time as additional strong motion data become avail abl e. Most predictive
rel ationships are updated in the l iteratureevery 3 to 5 years or shortl y after the
occurrence of l arge earthquakes in wel l -instrumentedregions.The functional form of
the predictive rel ationship is usual l y sel ected to refl ectthe mechanics of the
ground motion process as cl osel y as possibl e. This minimizes thenumber of empirical
coefficients and al l ows greater confidence in appl ication of the predictive
rel ationship to conditions (magnitudes and distances) that are poorl y representedin
the database. Common forms for predictive rel ationships are based on the fol l owing
observations: 1. Peak val ues of strong motion parameters are approximatel y
l ognormal l y distributed(i.e., the l ogarithms of the parameters are approximatel y
normal l y distributed). As aresul t, the regression is usual l y performed on the
l ogarithm of Yrather than on Yitsel f.2. Earthquake magnitude is typical l y defined
as the l ogarithm of some peak motionparameter (Section 2.9.2). Consequentl y, In Y
shoul d be approximatel y proportional toM.3. The spreading of stress waves as they
travel away from the source of an earthquakecauses body wave [p- and s-wave
(Section 5.2.2.5)] ampl itudes to decrease accordingto l /R and surface wave
[primaril y Rayl eigh wave (Section 5.3.1)] ampl itudes todecrease according to 1/JR .
4. The area over which faul t rupture occurs increases with increasing earthquake
magnitude(Section 4.2.1.2). As a resul t, some of the waves that produce strong
motion ata site arrive from a distance, R, and some arrive from greater distances.
The effectivedistance, therefore, is greater than R by an amount that increases
with increasingmagnitude.5. Some of the energy carried by stress waves is absorbed
by the material s they travel through [material damping (Section 5.5.1)]. This
material damping causes groundmotion ampl itudes to decrease exponential l y with R.6.
Ground motion parameters may be infl uenced by source characteristics (e.g.
strikesl ip,normal , or reverse faul ting) or site characteristics (e.g. hard rock,
soft rock, al l uvium,etc.).Combining these observations, a typical predictive
rel ationship may have the formIn Y = C1 + C2M + C3Mc, + Csl n[R + C6 exp(C7M)]'-y-i
v '-----v-----' '----v-----'CD ® @ @+ CsR + f(source) + f(site)'-y--J v ® ®88
Strong Ground Motion Chap. 3(3.24)where the circl ed numbers indicate the
observations associated with each term. Some predictiverel ationships util ize al l
these terms (and some have even more) and others do not.The crl nY term describes
uncertainty in the val ue of the ground motion parameter givenby the predictive
rel ationship. Statistical l y, it represents an estimate of the standard deviationof
In Yat the magnitude and distance of interest. Historical l y, most crl nY val ues have
beenconstants, but several recent predictive rel ationships indicate crl riYval ues
that vary with magnitude.At a given magnitude, therefore, the probabil ity that the
ground motion parameterwil l exceed a val ue y* woul d be 1- Fz(z*) where Fz(z*) is
the val ue of the standard cumul ativedistribution function (see Section C7.2 of
Appendix C) at z* =(In y* ---' In Y)/cr InY.When using anypredictive rel ationship,
it is very important to know how parameterssuch as M and R are defined and to use
them in a consistent manner. It is al so important torecognize that different
predictive rel ationships are usual l y obtained from different datasets. To make
reasonabl e predictions of ground motion parameters, a predictive rel ationshipbased
on data that are consistent with the conditions rel evant to the prediction is
required.3.4.3 Estimation of Ampl itude ParametersPredictive rel ationships for
parameters that decrease with increasing distance (such aspeak accel eration and
peak vel ocity) are often referred to as attenuation rel ationships. Afew of a l arge
number of useful attenuation rel ationships for different geographic and tectonic
environments are described in the fol l owing sections.3.4.3.1 Peak Accel erationSince
peak accel eration is the most commonl y used ground motion parameter, manypeak
accel eration attenuation rel ationships have been devel oped. Al l are best suited to
conditionssimil ar to those in the databases from which they were devel oped. As
additional strong motion data have become avail abl e, attenuation rel ationships have
become morerefined. Consider, for exampl e, two attenuation rel ationships devel oped
some 13 years apart.In 1981,Campbel l (1981) used worl dwide data to devel op an
attenuation rel ationship forthe mean PHA for sites within 50 km of the faul t
rupture in magnitude 5.0 to 7.7 earthquakes: InPHA(g) = -4.141 + 0.868M -1.09In[R +
0.0606 exp(0.7M)]crl nPHA = 0.37where M is the l ocal magnitude or surface wave
magnitude for magnitudes l ess than orgreater than 6, respectivel y, and R is the
cl osest distance to the faul t rupture in kil ometers.In this rel ativel y simpl e
attenuation rel ationship, which represented the state of the art in1981, the peak
accel eration was taken as a function of M and R onl y and crl n PHA was constant.In
1994, Campbel l and Bozorgnia (1994) used worl dwide accel erograms from earthquakesof
moment magnitude ranging from 4.7 to 8.1 to devel op the attenuation rel ationshipIn
PHA(gal s) = - 3.512 + 0.904Mw- 1.328 InJR2 +[0.14gexp(0.647Mw)]2+ (1.125 - 0.1121nR
- 0.0957 Mw)F + (0.440 - 0.1711nR)SsR+ (0.405 - 0.222l nR)SHR (3.25)__ {0.889 -
0.0691M M ~ 7.4crl n PHA0.38 M> 7.4Sec. 3.4 Estimation of Ground Motion Parameters
89where R is the cl osest distance (~60 km) to seismic rupture in kil ometers (with
minimumval ues on.3, 5.8, 3.5, and 3.0 km for magnitudes of 5.0,5.5,6.0, and 6.5,
respectivel y); thesource term, F, takes on val ues of 0 for strike-sl ip and normal
faul ting, and 1 for reverse,reverse-obl ique, and thrust faul ting; SSR = 1for soft-
rock sites (sedimentary deposits of Tertiaryage), SHR =1 for hard-rock sites
(primaril y ol der sedimentary deposits, metamorphicrock, and crystal l ine rock), and
SSR = SHR = 0 for al l uvium sites. The 1994 rel ationship,which is based on more
data, is cl earl y more specific (and more compl icated) than the 1981rel ationship.
The incorporation of additional terms refl ecting source and site characteristicsare
typical of the refinement of predictive rel ationships that has taken pl ace in
recent years.Boore et al . (1993) used data from western North America earthquakes
of magnitude5.0 to 7.7 at distances within 100 km (62 mi) of the surface projection
of the faul t to devel opthe predictive rel ationshipl og PHA(g) = b, + b2(Mw - 6) +
b3(Mw - 6)2 + b4R + bsl ogR + b6GB + b7GC (3.26)whereR = Jd2 + h2 , d is the cl osest
distance to the surface projection of the faul t in kil ometers,and10 for site cl ass
AGB = 1 for site cl ass Bo for site cl ass C 10 for site cl ass AGc = 0 for site cl ass
B1 for site cl ass CNote that the Boore et al . (1993) attenuation rel ationship is
expressed in terms of the common(base 10) l ogarithm rather than the natural
l ogarithm. The site cl asses are defined onthe basis of the average shear wave
vel ocity in the upper 30 m (100 ft) (Tabl e 3-5). Coefficientsfor the Boore et al .
(1993) attenuation rel ationship were devel oped for two measuresof peak
accel eration: the randoml y oriented component and the l arger horizontal component
(the former considers two orthogonal horizontal records at a particul ar site as
separate events and the l atter considers onl y the l arger of the two). The
coefficients aregiven in Tabl e 3-6.Tabl e 3-5 Definitions of Site Cl asses forBoore
et al . (1993) AttenuationRel ationshipSite Cl ass Vs in Upper 30 m (100 ft)A > 750
m/sec (2500 ft/sec)B 360-750 m/sec (1200-2500 ft/sec)C 180-360 m/sec (600-1200
ft/sec)Tabl e 3-6 Coefficients for Boore et al . (1993) Attenuation Rel ationship
ComponentRandomLargerh- 0.105 0.229 0.0 0.0 - 0.778 0.162 0.251 5.57-0.038 0.216
0.0 0.0 -0.777 0.158 0.254 5.48O"]ogPHA0.2300.20590 Strong Ground Motion Chap. 3
(3.27)for R < 5 kmfor 5 km ~ R ~ 20 kmfor R > 20 km(3.28)Since the continental
crust in eastern North America is stronger and more intact thanthe crust in western
North America, peak accel erations tend to be higher. For the midcontinental portion
of eastern North America, Toro et al . (1994) devel oped an attenuationrel ationship
for peak horizontal rock accel eration: In PHA(g) = 2.20 + 0.81 (Mw - 6) - 1.271nRm+
o.l l max( In~~, 0) -0.0021Rmwhere R; = JR2 + 9.32, R is the cl osest horizontal
distance to the earthquake rupture (inkm), (JM = 0.36 + 0.07 (Mw - 6) , and{0.54(JR
= 0.54 - 0.0227(R - 5)0.20Subduction zone earthquakes general l y occur at greater
hypocentral depths than earthquakesthat occur on transform faul ts. Consequentl y,
the seismic waves that emanate from subductionzone earthquakes fol l ow different
paths from those of transform faul ts. Youngs et al .(1988) used strong-motion
measurements obtained on rock from 60 earthquakes and numerical simul ations of
Mw : : : ': 8 earthquakes to devel op a subduction zone attenuation rel ationship: In
PHA(g) = 19.16 + 1.045Mw-4.738 In [R + 205.5exp(0.0968Mw ) ]+ 0.54Z{ (Jl nPHA = 1.55
- 0.125Mwwhere R is the cl osest distance to the zone of rupture in kil ometers and
Z{ is afor interfaceevents and 1 for intrasl ab events.The four preceding
attenuation rel ationships are shown graphical l y for earthquakemagnitudes 5.5, 6.5,
and 7.5 in Figure 3.22. The shapes of the attenuation rel ationships aresimil ar,
despite the fact that they represent different geographic regions and different
sourcemechanisms and use different measures of distance.Peal ( Vel ocity. Regression
anal ysis of PHV data has provided a number ofuseful rel ationships describing the
attenuation ofthat parameter. Joyner and Boore (1988),for exampl e, used strong-
motion records from earthquakes of moment magnitude between5.0 and 7.7 to devel op
the attenuation rel ationshipl ogPHV(cml sec) = i, + l : (M - 6) + 13 (M - 6) 2 + i,
l ogR + j-R + i6 (3.29)where PHV can be sel ected as the randoml y oriented or l arger
horizontal component,R = JT6 + i~, and To is the shortest distance (in kil ometers)
from the site to the vertical projection of the earthquake faul t rupture on the
surface of the earth. The coefficients for theJoyner and Boore (1988) attenuation
rel ationship are given in Tabl e 3-7.Tabl e 3-7 Coeffecients for Joyner and Boore
(1988) Peak Horizontal Vel ocityAttenuation Rel ationshipComponentRandomLarger2.09
2.170.490.490.00.0-1.0-1.0-0.0026-0.00260.170.17h4.04.0O"JogPHV0.330.33Sec. 3.4
Estimation of Ground Motion Parameters 9110001000(d)(b)Campbel l and Bozorgnia
(1994) Boore et al . (1993)1.0 1.0(a)M=7.5§ M=6.5 § M=7.5c c~05Me;;~ 0~0.510.0
.......• ....(])(j)oo « 0.01 10 100 1000 1 10 100Distance (km) Distance (km)Toro et
al . (1994) Youngs et al . (1988)2.0 1.0M=7.5 (c)§ §c c0 0~ 1.0 ~ 0.5(]) (])(j) (j)o
o M=7.5o o « « M=6.5M=5.50.0 0.01 10 100 1000 1 10 100Distance (km) Distance (km)
Figure 3.22 Variation of peak horizontal accel eration with distance for M =5.5, M =
6.5, and M =7.5 earthquakes according to various attenuation rel ationships: (a)
Campbel l and Bozorgnia (1994), soft rock sites and strike-sl ip faul ting; (b) Boore
et al . (1993), sitecl ass B; (c) Toro et al . (1994); and (d) Youngs et al . (1988),
intrasl ab event.': ~.4.4 Estimation of Frequency Content ParametersLarge earthquakes
produce l arger and l onger-period ground motions than do smal l er earthquakes;
consequentl y, the frequency content of a ground motion is rel ated to the earthquake
magnitude. As seismic waves travel away from a faul t, their higher-frequency
componentsare scattered and absorbed more rapidl y than are their l ower-frequency
components. As aresul t, the frequency content al so changes with distance.3.4.4.1
Predominant PeriodOne aspect of the change in frequency content with distance
invol ves the shiftingof the peak of the Fourier ampl itude spectrum to l ower
frequencies (or higher periods). Asa resul t, the predominant period increases with
increasing distance, as il l ustrated inFigure 3.23.M=8M=7.5921.2U1.0Ql -!!!-""0 0.80
.~0.. 0.6"Ectl c'E 0.40""0~a. 0.2Strong Ground Motion Chap. 380 120 160 200 240 280
320 360Distance from causative faul t - km400.0 '----- ----.JoFigure 3.23 Variation
of predominant period at rock outcrops with magnitude anddistance. (After Seed et
al ., 1969.)Exampl e 3.7Determine the predominant period that woul d have been
expected for the recorded motion atthe Gil roy No. 1 (rock) site.Sol ution The Gil roy
No.1 (rock) motion was recorded at a site l ocated at an epicentral distanceof 21.8
km from the Lorna Prieta (M =7.1) earthquake. From Figure 3.23, the expected
predominant period woul d be 0.33 sec. As determined in Exampl e 3.3, the actual
predominantperiod of the Gil roy No. 1 (rock) motion was 0.39 sec.(3.30)3.4.4.2
Fourier Ampl itude SpectraOrdinates of the Fourier ampl itude spectrum can be
estimated empirical l y by regressionon the Fourier spectral ordinates of actual
strong-motion data (e.g., Trifunac, 1976;McGuire et al ., 1984; Trifunac and Lee,
1987; Castro et al ., 1990). Al ternativel y, a physical l ybased model of source,
travel path, and site behavior may be cal ibrated to predict Fourierampl itude
spectra.Based on Brune's (1970, 1971) sol ution for instantaneous sl ip of a circul ar
rupturesurface, the Fourier ampl itudes for a far-fiel d event at distance R can be
expressed(McGuire and Hanks, 1980; Boore, 1983) asIA(f)1 = [CMo f2 1 Je-7rf R/Q(f )
V,1- (fl fe)2Jl + (fl fmax) 8 R(3.31)where fe is the comer frequency (see Figure
3.14),fmax the cutoff frequency (Figure 3.14),Q(j) is the frequency-dependent
qual ity factor (inversel y proportional to the damping ratioof the rock; see Section
5.5.1), and C is a constant given byC = Rs.pFV4Jrpv;Sec. 3.4 Estimation of Ground
Motion Parameters 93(3.32)where RS!JI ('" 0.55) accounts for the radiation pattern,
F (= 2) accounts for the free-surfaceeffect, V (= Ji12) accounts for partitioning
the energy into two horizontal components,p is the density of the rock al ong the
rupture surface, and Vs is the shear wave vel ocity ofthe rock.If/max is assumed
constant for a given geographic region (15 Hz and 40 Hz are typical val ues for
western and eastern North America, respectivel y), the spectra for different
earthquakesare functions of the seismic moment, Mo, and/o which can be rel ated
(Brune, 1970,1971) byI, = 4.9 6 (L1Ci)1/3 X 10 Vs Mowhere V s is in km/sec, M; is
in dyne-ern, and L1Ci is referred to as the stress parameter orstress drop in bars.
Stress parameters of 50 bars and 100 bars are commonl y used forsources in western
and eastern North America, respectivel y. Figure 3.24a shows how theFourier
ampl itude spectra predicted by equation (3.30) vary with magnitude. Note thestrong
infl uence of magnitude on both the ampl itude and frequency content of the motion.As
the magnitude increases, the bandwidth increases and the corner frequency
decreases,impl ying that more l ow-frequency (l ong-period) motion wil l occur. Figure
3.24b showstime histories of accel eration generated from the spectra of equation
(3.30) for magnitude4 and magnitude 7 events. The stress parameter and seismic
moment are commonl y used tospecify the source spectrum, given by the expression in
brackets in equation (3.30). Thefinal expression is the travel path operator, which
describes attenuation of the Fourierampl itudes as the energy travel s away from the
site. An expression that describes theeffects of soil response (a site operator)
can be added to equation (3.31), if necessary, toaccount for near-surface effects.
The response of soil deposits during earthquakes is discussedin detail in Chapter
7.Since it is based on the mechanics of source rupture and wave propagation,
equation(3.30) offers significant advantages over purel y empirical methods for
magnitudes and distancesfor which few or no data are avail abl e.M=70- 100'" M=4.0
M=7.0 6J!!10 1 5 0.6g'" ~ -0: E0- 0.1 E'"" o§ 0.01tl : > 0.0010.1 1 10Frequency (Hz)
(a)100(b)Fiigure 3.24 (a) Variation of Fourier ampl itude spectra
at R = 10 km for different moment magnitudes (!'.a =100 bars); (b) accel erograms
generated from the M = 4 and M = 7 spectra. (After Boore, 1983. Used bypermission
of the Seismol ogical Society of America.)94 Strong Ground Motion Chap. 33.4.4.3
Ratio vmax/amaxAs a measure of the frequency content of a ground motion, the ratio
vmax/amax is al sorel ated to earthquake magnitude and distance. This dependence has
been studied by several investigators, with a summary of their resul ts provided by
McGuire (1978), who proposedthe magnitude and distance dependencies shown in Tabl e
3-8. The tabl e indicates that, asexpected, the vma/amax ratio increases with
increasing earthquake magnitude and increasingsource-to-site distance.Tabl e 3-8
Magnitude and Distance Dependence of Vmax/8max aSite ConditionsRock sitesSoil sites
Magnitude DependenceOAOM eeO.15MDistance DependenceSource: After McGuire (1978).
aThe ratio vma/amax is proportional to these dependence rel ationships.Exampl e 3.8
Estimate the val ues of vmax/amax that woul d be observed at rock and soil sites 40
km from thesource of a M = 6 earthquake l ocated near the 1989 Lorna Prieta
earthquake.Sol ution Using the val ues from the Gil roy No.1 (rock) and Gil roy No.2
(soil ) sites, andrecal l ing that those sites were l ocated 21.8 km and 22.8 km,
respectivel y, from the M = 7.1Lorna Prieta earthquake,Rock site: Soil site: Vmax e
(0.40) (6) 40°. 12amax'" 0.078sec e(OAO) (7.1) 21.8°.12vmax e (0.15) (6) 400.23amax
'" 0.124sec e (0.15) (7.1) 22.80.23= 0.054 sec= 0.120 sec3.4.4.4 Response Spectrum
OrdinatesThe importance of response spectra in earthquake engineering has l ed to
the devel opmentof methods for predicting them directl y. For many years, the shapes
of al l responsespectra were, for a given cl ass of soil conditions, assumed to be
identical . Design spectrawere devel oped by scal ing standard spectral shapes by some
ground motion parameter, usual l ythe PHA. As more recorded motions became avail abl e,
the magnitude dependence ofspectral shapes was recognized increasingl y. For
exampl e, Figure 3.25 shows the responsespectra computed from the accel erograms
shown in Figure 3.21. The difference in spectral shapes at different magnitudes,
particul arl y in the l ong-period range, are apparent. Thisshape dependence was l ater
accounted for, at l east approximatel y, by using PHA, PHV, andpeak displ acement to
scal e design spectra in different frequency ranges (Newmark and Hal l ,1978, 1982),
as discussed in Section 8.3.2. More recentl y, regression anal yses have beenused to
devel op predictive rel ationships for spectral ordinates at various oscil l ator
periods(e.g., Joyner and Boore, 1982, 1988; Crouse, 1991; Boore et aI., 1993).For
exampl e, Boore et al . (1993) determined val ues of coefficients that when usedwith
equation (3.26), predict pseudospectral vel ocities for oscil l ators of different
natural periods. These attenuation rel ationships, the coefficients of which are
presented inSec. 3.4 Estimation of Ground Motion Parameters 95Figure 3.25 Response
spectra computedfrom the accel erograms of Figure 3.20.Epicentral distances from
eachaccel erogram were nearl y equal . Note thatthe shapes, as wel l as the val ues, of
thespectra vary with earthquake magnitude.(After Anderson, 1991. Geotechnical News,
Vol . 9, No.1, p.35. Used by permission ofBiTech Publ ishers, Ltd.)10Period (sec)0.1
'----__-'-----"_-=-'----'---"'-''----_-''----'--....J0.011000 ",.---...,..,..---
~------,.--,,-,100c: .;-al ~..s 10>en0...Tabl es 3-9 and 3-10, are appl icabl e to the
same conditions as the attenuation rel ationship ofequation (3.26). These
coefficients produce smooth response spectra, as il l ustrated inFigure 3.26.
Al though the smooth response spectra are l imited to periods of 2.0 sec, their
general shapes are simil ar to those of actual spectra.~: I.4.5 Estimation of
DurationThe duration of strong ground motion increases with increasing earthquake
magnitude.However, the manner in which strong motion duration varies with distance
depends on howit is defined. Since accel eration ampl itudes decrease with distance,
durations based on absol uteaccel eration l evel s, such as the bracketed duration,
woul d be expected to decrease withdistance; at some distance al l accel erations wil l
drop bel ow the threshol d accel eration andthe bracketed duration wil l be zero. Page
et al . (1972) and Chang and Krinitzsky (1977) confirmedthis, as shown in Figure
3.27. Durations based on rel ative accel eration l evel s (e.g.,Trifunac and Brady,
1975b; Dobry et al ., 1978) increase with increasing distance and mayhave l ong
durations even when the accel eration ampl itudes are very l ow. For engineering
purposes, the bracketed duration appears to provide the most reasonabl e indication
of theinfl uence of duration on potential damage.3.4.6 Estimation of Other
ParametersParameters that refl ect more than one important characteristic of strong
ground motion arel ikel y to see increasing use. For most, however, onl y l imited data
for the devel opment ofpredictive rel ationships are currentl y avail abl e.96 Strong
Ground Motion Chap, 3Tabl e 3-9 Smoothed Coeffecients of Predictive Rel ationships
for the Larger Horizontal Component of 5% Damped PSVT bl bz b-; b4 bs b6 b7 h (Jl ogY
0,10 1.700 0.321 -0.104 0.0 -0.921 0.039 0.128 6.18 0.1940,11 1.777 0.320 -0.110
0,0 -0,929 0.065 0.150 6.57 0.1940.12 1.837 0.320 -0,113 0.0 -0.934 0,087 0.169
6.82 0.1930,13 1,886 0.321 -0.116 0,0 -0.938 0.106 0.187 6.99 0.1930.14 1.925 0.322
-0.117 0.0 -0.939 0.123 0.203 7.09 0.1930.15 1.956 0.323 -0.117 0.0 -0,939 0.137
0,217 7.13 0,1940.16 1.982 0,325 -0.117 0.0 -0.939 0.149 0.230 7,13 0.1940.17 2,002
0.326 -0.117 0.0 -0.938 0.159 0.242 7,10 0,1950,18 2.019 0.328 -0.115 0.0 -0,936
0.169 0,254 7.05 0,1950.19 2,032 0.330 -0,114 0.0 -0.934 0,177 0.264 6.98 0.1960.20
2.042 0.332 -0.112 0,0 -0.931 0.185 0.274 6.90 0.1960,22 2.056 0,336 -0.109 0.0
-0.926 0.198 0.291 6.70 0.1980.24 2.064 0.341 -0,105 0.0 -0.920 0.208 0.306 6.48
0.1990.26 2.067 0.345 -0.101 0,0 -0.914 0.217 0.320 6.25 0.2010.28 2,066 0,349
-0.096 0.0 -0.908 0,224 0.333 6.02 0.2020,30 2.063 0.354 -0.092 0.0 -0,902 0.231
0,344 5,79 0.2040,32 2.058 0.358 -0.088 0.0 -0,897 0.236 0.354 5.57 0.2050.34 2.052
0.362 -0,083 0.0 -0.891 0,241 0.363 5.35 0.2060.36 2,045 0.366 -0.079 0,0 -0.886
0.245 0.372 5.14 0.2080.38 2.038 0,369 -0.076 0.0 -0.881 0.249 0.380 4,94 0.2090.40
2.029 0.373 -0,072 0.0 -0,876 0.252 0.388 4.75 0,2110.42 2.021 0.377 -0.068 0,0
-0.871 0.255 0.395 4.58 0,2130.44 2.013 0,380 -0.065 0.0 -0.867 0,258 0.401 4.41
0.2130.46 2,004 0.383 -0.061 0.0 -0,863 0.261 0.407 4,26 0,2150.48 1.996 0.386
-0.058 0.0 -0.859 0.263 0.413 4.11 0.2160,50 1.988 0.390 -0.055 0.0 -0.856 0.265
0.418 3.97 0.2170.55 1.968 0.397 -0.048 0,0 -0.848 0.270 0.430 3.67 0.2210.60 1.949
0.404 -0.042 0.0 -0.842 0.275 0.441 3.43 0.2230,65 1.932 0.410 -0.037 0.0 -0.837
0.279 0.451 3,23 0.2260,70 ',917 0.416 -0,033 0.0 -0,833 0.283 0.459 3,08 0,2290.75
1.903 0.422 -0.029 0,0 -0.830 0.287 0.467 2.97 0.2320.80 1.891 0.427 -0.025 0.0
-0.827 0.290 0.474 2.89 0.2340.85 1.881 0.432 -0.022 0.0 -0.826 0.294 0.481 2.85
0,2370.90 1.872 0.436 -0.020 0.0 -0,825 0.297 0.486 2.83 0.2400.95 1.864 0.440
-0.0l 8 0.0 -0.825 0.301 0.492 2.84 0.2421.00 1.858 0.444 -0,016 0.0 -0.825 0.305
0.497 2.87 0.2451.10 1,849 0.452 -0.014 0,0 -0.828 0.312 0.506 3.00 0.2491.20 1,844
0.458 -0.013 0.0 -0.832 0.319 0.514 3.19 0.2541.30 1.842 0.464 -0,012 0.0 -0,837
0.326 0.521 3.44 0.2581.40 1.844 0.469 -0,013 0,0 -0.843 0.334 0.527 3.74 0,2621.50
1.849 0.474 -0.014 0.0 -0.851 0.341 0.533 4.08 0.2671.60 1,857 0.478 -0.016 0.0
-0.859 0.349 0.538 4.46 0.2701.70 1.866 0.482 -0.019 0.0 -0,868 0.357 0,543 4.86
0,2741.80 ',878 0.485 -0.022 0.0 -0.878 0.365 0.547 5.29 0.2791.90 1.891 0.488
-0.025 0.0 -0.888 0.373 0.551 5.74 0.2832.00 1.905 0.491 -0,028 0.0 -0.898 0.381
0.554 6.21 0.287Source: After Boore et a!., (1993),Sec. 3.4 Estimation of Ground
Motion Parameters 97Tabl e 3-10 Smoothed Coefficients of Predictive Rel ationships
for the Random Horizontal Component of 5% Damped PSVT bl b2 b3 b4 bs b6 b7 h O"l ogY
0.10 1.653 0.327 -0.098 0.0 -0.934 0.046 0.136 6.27 0.208CI.11 1.725 0.318 -0.100
0.0 -0.937 0.071 0.156 6.65 0.2080.12 1.782 0.313 -0.101 0.0 -0.939 0.093 0.174
6.91 0.2080.13 1.828 0.309 -0.101 0.0 -0.939 0.111 0.191 7.08 0.2090.14 1.864 0.307
-0.100 0.0 -0.938 0.127 0.206 7.18 0.2090.15 1.892 0.305 -0.099 0.0 -0.937 0.140
0.221 7.23 0.2110.16 1.915 0.305 -0.098 0.0 -0.935 0.153 0.234 7.24 0.2110.17 1.933
0.305 -0.096 0.0 -0.933 0.163 0.246 7.21 0.2120.18 1.948 0.306 -0.094 0.0 -0.930
0.173 0.258 7.16 0.2130.19 1.959 0.308 -0.092 0.0 -0.927 0.182 0.269 7.10 0.2150.20
1.967 0.309 -0.090 0.0 -0.924 0.190 0.279 7.02 0.2150.22 1.978 0.313 -0.086 0.0
-0.918 0.203 0.297 6.83 0.2180.24 1.982 0.318 -0.082 0.0 -0.912 0.214 0.314 6.62
0.2200.26 1.982 0.323 -0.078 0.0 -0.906 0.224 0.329 6.39 0.2220.28 1.979 0.329
-0.073 0.0 -0.899 0.232 0.343 6.17 0.2250.30 1.974 0.334 -0.070 0.0 -0.893 0.239
0.356 5.94 0.2260.32 1.967 0.340 -0.066 0.0 -0.888 0.245 0.367 5.72 0.2280.34 1.959
0.345 -0.062 0.0 -0.882 0.251 0.378 5.50 0.2300.36 1.950 0.350 -0.059 0.0 -0.877
0.256 0.387 5.30 0.2320.38 1.940 0.356 -0.055 0.0 -0.872 0.260 0.396 5.10 0.2350.40
1.930 0.361 -0.052 0.0 -0.867 0.264 0.405 4.91 0.2360.42 1.920 0.365 -0.049 0.0
-0.862 0.267 0.413 4.74 0.2380.44 1.910 0.370 -0.047 0.0 -0.858 0.271 0.420 4.57
0.2390.46 1.900 0.375 -0.044 0.0 -0.854 0.273 0.427 4.41 0.2410.48 1.890 0.379
-0.042 0.0 -0.850 0.276 0.433 4.26 0.2430.50 1.881 0.384 -0.039 0.0 -0.846 0.279
0.439 4.13 0.2440.55 1.857 0.394 -0.034 0.0 -0.837 0.284 0.452 3.82 0.2480.60 1.835
0.403 -0.030 0.0 -0.830 0.289 0.464 3.57 0.2510.65 1.815 0.411 -0.026 0.0 -0.823
0.293 0.474 3.36 0.2540.70 1.797 0.418 -0.023 0.0 -0.818 0.297
0.483 3.20 0.2570.75 1.781 0.425 -0.020 0.0 -0.813 0.300 0.490 3.07 0.2590.80
1.766 0.431 -0.018 0.0 -0.809 0.303 0.497 2.98 0.2610.85 1.753 0.437 -0.016 0.0
-0.805 0.306 0.503 2.92 0.2640.90 1.742 0.442 -0.015 0.0 -0.802 0.309 0.508 2.89
0.2660.95 1.732 0.446 -0.014 0.0 -0.800 0.312 0.513 2.88 0.2681.00 1.724 0.450
-0.014 0.0 -0.798 0.314 0.517 2.90 0.2701.10 1.710 0.457 -0.013 0.0 -0.795 0.319
0.523 2.99 0.2741.20 1.701 0.462 -0.014 0.0 -0.794 0.324 0.528 3.14 0.2771.30 1.696
0.466 -0.015 0.0 -0.793 0.328 0.532 3.36 0.2801.40 1.695 0.469 -0.017 0.0 -0.794
0.333 0.535 3.62 0.2831.50 1.696 0.471 -0.019 0.0 -0.796 0.338 0.537 3.92 0.2851.60
1.700 0.472 -0.022 0.0 -0.798 0.342 0.538 4.26 0.2861.70 1.706 0.473 -0.025 0.0
-0.801 0.347 0.539 4.62 0.2891.80 1.715 0.472 -0.029 0.0 -0.804 0.351 0.539 5.01
0.2901.90 1.725 0.472 -0.032 0.0 -0.808 0.356 0.538 5.42 0.2922.00 1.737 0.471
-0.037 0.0 -0.812 0.360 0.537 5.85 0.293Source: After Boore et al ., (1993).98
Strong Ground Motion Chap. 3100 0.9 r----,----~--___._--___,U 0.5 CD.!!: .10 §E .s
Cf)'" ,.Cf)10 0.5 1.5 2O'-------------------'---------.Jo1'-------~-~----------'-'
0.1Period (sec) Period (sec)Figure 3.26 Response spectra for the random component
of ground motion at a site cl ass B site at R= 10 km according to the predictive
rel ationship of Boore et al . (1993): (a) pseudospectral vel ocitycomputed directl y
from equation (3.27) and Tabl e 3-10, and (b) pseudospectral accel eration computed
from the pseudospectral vel ocities.(3.33)3.4.6.1 RMS Accel erationHanks and McGuire
(1981) used a database of Cal ifornia earthquakes of l ocal magnitude4.0 to 7.0 to
devel op an attenuation rel ationship for rms accel eration for hypocentral distances
between 10 and 100 km (6.2 and 62 mi): = 0 119 J f max/ f carms' Rwherej, is the
corner frequency,fmax is the cutoff frequency, and R is in kil ometers.Kavazanjian
et al . (1985) used the definition of duration proposed by Vanmarcke andLai (1980)
with a database of 83 strong motion records from 18 different earthquakes toobtain(
0.966 0.255)(l rms = 0.472 + 0.268Mw + 0.129 l og 7 +~ - 0.1167 R (3.34)Sec. 3.4
Estimation of Ground Motion Parameters 99uOJ.!!2- 70co~60: : : : l ""0 50""0OJ ]1 40o~
30 co20M=6.5WM=6-M=5.50<, A.~l z.. ""-.... t...... ~ i2...... ...........r-;
~ ......................~~: ~~~~;~?!~~• .: -;;: : '-' , • • !-....o 10 20 30 40 50 60
70 80 90 100 110(a) Epicentral distance (km)9080UOJ .!!2- 70c 60 0~: : : : l 50""0""0
40 -*o'" 30~co 20100(b)\..-,-, <,<, '" -, <, r-,JW","<, <, '-<, ~~ c: ~"'. "~~ r----
.~ ~~~~'.O ...• -.-~, . '"""-1 : J 1 1o 10 20 30 40 50 60 70 80 90 100 110 120
Epicentral distance (km)Figure 3.27 Variation of bracketedduration (0.05g
threshol d) with magnitudeand epicentral distance: (a) rock sites; (b)soil sites.
(After Chang and Krinitzsky,1977.)(3.35)for basement rockwhere R is the distance to
the cl osest point of rupture on the faul t. The database wasrestricted to Mw > 5, R
< 110 km (68 mi), rupture depths l ess than 30 km (19 mi), and soil thicknesses
greater than 10 m (33 ft).3.4.6.2 Arias IntensityCampbel l and Duke (1974) used data
from Cal ifornia earthquakes to predict the variationof Arias intensity within 15 to
110 km (9 to 68 mi) of magnitude 4.5 to 8.5 events.eM, (O.33M, - 1.47)fa (m/sec) =
313 R3.79 S10.57 R°.461.02Ro.51 for sedimentary rockwhere S = 0.37RO.81 for
al l uvium s; 60ft thick0.65Ro.74 for al l uvium> 60ft thickand R is the distance from
the center of energy rel ease in kil ometers.100 Strong Ground Motion Chap. 3Wil son
(1993) anal yzed strong motion records from Cal ifornia to devel op an attenuation
rel ationship which, using the Arias intensity definition of equation (3.17), can be
expressed as10gIa (m/sec) = Mw - 210gR - kR - 3.990 + 0.365 (1 - P) (3.36)where R =
JD2 + h2, D is the minimum horizontal distance to the vertical projection of the
faul t pl ane, h is a correction factor (with a defaul t val ue on.5 km (4.7 mi)), k is
a coefficient ofanel astic absorption (with a defaul t val ue of zero), and P is the
exceedance probabil ity.Accel eration and Vel ocity Spectrum Intensities. Von Thun et
al .(1988) used 30 strong motion records, primaril y from rock outcrops in the
western UnitedStates and Ital y, to devel op the attenuation rel ationships for
accel eration spectrum intensityand vel ocity spectrum intensity shown graphical l y in
Figure 3.28. Large earthquakes inthese areas are general l y accompanied by surface
faul ting. The use of these attenuation rel ationshipsis recommended onl y for areas
with simil ar tectonic conditions.600U(])-!!! 500E~~·iii c 400(])~E2 30013(])Q.(f)c
2000~(])OJ 100 oo<l : ~-,I~ -,r-, -.1""'- -, <, M=7.5 -, ~ ~r-, <, <,~=;7.0r-, r-, <,
<,i"<, M=6.5<,<, r-............ i M=6.0o "• M-=-5-.5 -.....;o 5 10 15 20 25 30 35
Causative faul t distance (km)250(b)E 200~~·iiic.~ 150E213~ 100(f)~.g~ 50r-,......"
r-, ......,-, ...... -, -, -, -,r-, r-, <, M=7.5'" -, r-, ......."'" '" r-, ~="~: ~"
<, <, <, : M=6.5<, <,<, ~6.0~o rvi~o 5 10 15 20 25 30 35(a) Causative faul t distance
(km)Figure 3.28 Attenuation of (a) accel eration spectrum intensity and (b) vel ocity
spectrum intensity. (After Von Thun et aI., 1988. Earthquake ground motions for
designand anal ysis of dams, Earthquake Engineering and Soil Dynamics II. Reprinted
bypermission of ASCE.)3.5 SPATIAL VARIABILITY OF GROUND MOTIONSThe preceding
sections considered the spatial variation of ground motions on a regional scal e.
Ground motions al so vary spatial l y on l ocal scal es, and this l ocal variation can be
important for certain types of structures. The l ongest dimension of most structures
is usual l ysmal l enough that the ground motion at one end is virtual l y the same as
that at the other end.For structures such as bridges and pipel ines that extend over
considerabl e distances, differentground motions may occur beneath different parts
of the structure. In such cases theSec. 3.5 Spatial Variabil ity of Ground Motions
101l ocal spatial variation (or incoherence) of the ground motion may exert an
important infl uenceon the response of the structure.Spatial incoherence can be
caused by a number of factors. One is the travel ing-waveor wave-passage effect, in
which nonvertical waves reach different points on the groundsurface at different
times, producing a time shift between the motions at those points (Figure3.29a). A
cause of incoherence in the nearfiel d is the extended source effect, in which
differencesin the rel ative geometry of the source and sites produce different time
shifts, andconsequentl y different motions, at the sites (Figure 3.29b). Final l y,
ray-path effects causedby scattering (refl ection, refraction, etc.) of waves by
inhomogeneities al ong the travel path(Figure 3.30c) can cause incoherence.(a) (b)
(c)2 3 2 3 2 3---<>--------0-~efrOntFaul tPl an viewFigure 3.29 Sources of incoherent
ground motions: (a) wave-passage effect causesincl ined wavefront to reach l ocations
1,2, and 3 at different times; (b) extended sourceeffect causes waves due to
rupture at A and B to reach points I and 2 at different times;(c) scattering of
waves by heterogeneity causes different waves to arrive at differentl ocations at
different times. (After Abrahamson, 1991.)The simil arity between ground motions at
different l ocations can be described in thetime domain or the frequency domain.
Consider two points j and k at which accel erogramsal t) and ak(t) are recorded. The
simil arity of the motions can be described in the timedomain by the cross
covarianceNLaj(ti)ak(ti + r )i = 1(3.37)(3.38)where .. is a time increment and N is
the number of time sampl es. The autocovariance, Cjj(or Ckk), is obtained by
anal yzing the covariance of an accel erogram against itsel f. The maximumval ue of the
autocovariance wil l , obviousl y, correspond to a val ue of .. = O. The simil arityof
the motions in the frequency domain can be described by the coherencyYk ( ro) -_
-=.=SJi!=k(~ro=)=J JSj/ro)Skk(ro)where the smoothed cross-spectrum, SjkCro), is the
Fourier transform of the cross-covarianceand the autospectra, Siro) and Skk(ro),
are the Fourier transforms of the autocovariances,eji") and Ckk( .. ). The
coherency describes the degree of positive or negative correl atiortbetween the
ampl itudes and phase angl es of two time histories at each of their component
frequencies. A val ue of 1 indicates ful l coherence (or perfect correl ation), whil e
a val ue of102 Strong Ground Motion Chap. 3zero indicates ful l incoherence (or no
correl ation). The modul us of the coherency (thesquare root of the sum of the
squares of the real and imaginary parts) is cal l ed the l aggedcoherency. Because the
wave passage effect from a point source simpl y introduces a phaseshift at each
frequency, it does not infl uence the l agged coherency.Ground motions recorded by
dense arrays show that coherency decreases with increasingdistance between
measuring points and with increasing frequency, as shown inFigure 3.30. Measured
coherency functions from dense arrays in Cal ifornia and Japan are simil arto those
from the SMART-l array, suggesting that they may be appl icabl e to other areas as
wel l , al though research in this area is continuing (Abrahamson, 1991). Smooth
anal ytical coherency functions (coherency as a function of separation distance and
frequency) that matchthe most significant trends in measured coherency functions
have been proposed (Haricharanand Vanmarcke, 1986; Luco and Wong, 1986, Hao et al .,
1989; Abrahamson et al ., 1991).3.6 SUMMARY1. Compl ete description of a strong
ground motion invol ves three components of transl ationand three components of
rotation. In practice, onl y the transl ational componentsare usual l y measured, and
they are usual l y measured in orthogonal directions.2. A number of different
instruments
can be used for strong-motion measurement. Eachhas its own dynamic response
characteristics that determine the conditions for whichit is best suited. Ol der
strong-motion instruments are l ikel y to acquire data in anal ogform, whil e newer
instruments often acquire data digital l y.3. Raw strong-motion data may incl ude
errors from several sources that require correctionto produce accurate strong
motion records. Strong motion processing is oftenrequired to minimize background
noise, correct for the dynamic response of the transducer,and to correct for
measurement errors.4. Strong ground motions can be quite compl icated, and their
compl ete descriptioninvol ves a l arge amount of data. For engineering purposes, the
essential characteristicsof a strong ground motion can be described in much more
compact farm usingground motion parameters.5. From an earthquake engineering
standpoint, the most important characteristics of astrong ground motion are the
ampl itude, frequency content, and duration. Al l of thesecharacteristics can
significantl y infl uence earthquake damage. Consequentl y, knowl edgeof the ampl itude,
frequency content, or duration al one may not be sufficient todescribe accuratel y
the damage potential of a ground motion.6. A variety of parameters are avail abl e
for description of strong ground motions. Someof these parameters describe ground
motion ampl itude, some describe frequency content,and others describe duration.
Other parameters refl ect two or more of theseimportant characteristics. More than
one parameter is general l y required to characterizea strong ground motion.7.
Commonl y used ampl itude parameters incl ude peak accel eration, peak vel ocity, and
peakdispl acement. The peak accel eration provides a good indication of the high-
frequencycomponent of a ground motion. The peak vel ocity and peak displ acement
describe theampl itudes of the intermediate- and l ow-frequency components,
respectivel y.Sec. 3.6 Summary 103d 100 md=200m - - -d=400 m ----d=800m ------d=
1000 m .1.00.90.8 c-,oc~ 0.7(])s: 0o 0.6"6(]) 0.5: : l co > 0.4 (])"50 0.3 C/)..0<t:
0.20.1/ ......."\......\"<, \ , .').: '.r....\ \ / ""\..... ""\... "' \. \""\ /'-.-
..: ........ \\ " \ 1/\ /' \ L' ... ...-......L.. _' »r: \'....... ,. '" .
............. ~\"~ /I ....."..\..'..-.J' ,\.V·..·\·.~/...jt","> 1/ ~\ /s-, \: ' I l \
,,,,,,,~\ J' ,. -, ....... /\'<... ',,/...... ~ , ,.... , ~ ~ : .;Frequency (Hz)(a)
1.00.90.8 c-,oc~ 0.7(])s: o0 0.6"6(]) 0.5: : l co > 0.4 (])"50 0.3 C/)..0<t: 0.20.1f=
1.0 Hzf= 2.0 Hzf= 4.0 Hzf= 6.0 Hzf= 8.0 Hz0.0
L-,--,--,---,------,------,------,------,------,------,------,------,------,------,
------,------,------,------,------'------.Jo 100 200 300 400 500 600 700 800 900
1000Separation distance (m)(b)Figure 3.30 Measured decay of coherency with
increasing frequency and separationdistance for M =6.9 event at hypocentral depth
of 30.6 km and epicentraJ distance of116.6 km from SMART-I dense array at Lotung,
Taiwan. (After Haricharan andVanmarcke, 1986. Stochastic variation of earthquake
ground motion in space and time,Journal ofEngineering Mechanics, Vol . 112, No.2.
Reprinted by permission of ASCE.)104 Strong Ground Motion Chap. 38. The frequency
content of a strong ground motion is general l y described through theuse of
different types of spectra. Fourier spectra and power spectra directl y il l ustrate
the frequency content of the motion itsel f. Response spectra refl ect the infl uence
ofthe ground motion on structures of different natural periods. A variety of
spectral parameters are avail abl e to describe the frequency content of a strong
ground motion.9. Strong-motion durations can be described in absol ute or rel ative
terms. The bracketedduration, defined as the time between the first and l ast
exceedances of a threshol daccel eration, is based on an absol ute measure of
accel eration (the threshol d accel eration).Measures of duration based on rel ative
ground motion l evel s can define veryl ong durations for weak ground motions. For
engineering purposes, the bracketedduration is most commonl y used.10. Some
parameters refl ect the ampl itude, frequency content, and duration of a strongground
motion. Al though these parameters, such as rms accel eration, Arias intensity,and
response spectrum intensity, may be more difficul t to cal cul ate than parametersmore
commonl y used, they often refl ect the potential of the motion to produce damagemore
accuratel y.11. The characteristics of a ground motion at a particul ar site depend
on earthquake magnitudeand on the distance between the source of the earthquake and
the site. Consequentl y,ground motion parameters al so vary with earthquake magnitude
and sourceto-site distance.12. Measured ground motion data have been used to
devel op rel ationships that predictval ues of ground motion parameters as functions
of earthquake magnitude and sourceto-site distance. Predictive rel ationships are
general l y empirical ; each is obtained byregression on a specific set of data.
Consequentl y, each predictive rel ationship isappropriate for conditions that are
consistent with the conditions of the database.13. Predictive rel ationships are not
precise; they typical l y express the mean val ue of aground motion parameter and
incl ude a measure of the distribution of val ues about themean. The standard
deviation of the parameter (or the natural l ogarithm of the parameter)is usual l y
estimated in the devel opment of the predictive rel ationship.14. Predictive
rel ationships for variabl es that decrease with increasing source-to-site distance
are frequentl y referred to as attenuation rel ationships. Many attenuation
rel ationshipshave been reported in the l iterature, and the most commonl y used
rel ationshipsare updated every few years.15. Ground motions vary on l ocal as wel l
as regional scal es. Local variations may causedifferential movements of the
supports ofl ong structures such as bridges and pipel ines.Design and anal ysis of
such structures may require consideration of l ocal variations.16. The l ocal
variabil ity of ground motions is usual l y expressed in terms of coherency.The
coherency of two ground motions can be computed-s-it is a measure of thecorrel ation
of the ampl itudes and phase angl es of the motions at different frequencies.The
coherency of two cl osel y spaced ground motions is higher than that of two distant
ground motions. Al so, the coherency of the l ow-frequency (l ong-wavel ength)
componentsof a pair of motions is higher than that of the high-frequency (short-
wavel ength)components.Chap. 3 Homework Probl emsHOMEWORK PROBLEMS105Strong motion
records can be obtained from a variety of sources over the Internet, often by
anonymousftp. Downl oad the strong motion record indicated by your instructor and
use it to sol ve Probl ems 3.1 3.6.The use of a mathematical anal ysis program such as
MATLAB is highl y recommended; it wil l greatl y simpl ify the required computations.3.1
Pl ot the time history of accel eration and determine: (a) The peak accel eration.(b)
The sustained maximum accel eration (3rd cycl e and 5th cycl e).(c) The bracketed
duration.3.2 Integrate the time history of accel eration to produce time histories
of vel ocity and displ acement.Pl ot the time histories of vel ocity and displ acement
and determine the peak vel ocity andpeak displ acement.3.3 Compute and pl ot the
Fourier ampl itude spectrum of the strong motion record.3.4 Determine the
predominant period of the strong motion record.3.5 Compute the rms accel eration for
the strong motion record.3.6 Compute the Arias intensity for the strong motion
record.3.7 Determine and pl ot the variations of peak horizontal accel eration with
distance for a Mw = 6.5earthquake using the attenuation rel ationship of Campbel l
(1981).3.8 Determine and pl ot the variations of peak horizontal accel eration with
distance for a Mw = 6.5earthquake at soft rock, hard rock, and al l uvium sites using
the attenuation rel ationship ofCampbel l and Bozorgnia (1994). Which of these
conditions agrees best with the attenuationrel ationship of Campbel l (1981)?3.9
Using the attenuation rel ationship of Toro et al . (1994), determine the probabil ity
that a Mw =7earthquake in mid-continental eastern North America woul d produce a
peak accel erationgreater than 0.30 g at a point l ocated 50 km from the cl osest
point of rupture.: UO Determine the peak horizontal vel ocity that woul d have a 10%
probabil ity of being exceededby a Mw = 7.5 earthquake occurring at a distance of 40
km. Use the Joyner and Boore (1988)attenuation rel ationship.3.11 Using the Boore et
al . (1994) attenuation rel ationship, determine and pl ot the mean and mean± one
standard deviation response spectra for aM = 6.75 earthquake that occurs at a
distanceof70 km. w: U2 Determine the val ues of Arias intensity that have
10%,25%,50%,75%, and 90% probabil itiesof being exceeded by aM =7.25 earthquake at a
distance of 45 km. Use the attenuation rel ationshipof Wil son (1993rwith zero
anel astic absorption.4Seismic Hazard Anal ysis4.1 INTRODUCTIONIn many areas of the
worl d, the threat to human activities from earthquakes is sufficient torequire
their careful consideration in the design of structures and facil ities. The goal of
earthquake-resistant design is to produce a structure or facil ity that can
withstand a certainl evel of shaking without excessive damage. That l evel of shaking
is described by a designground motion, which can be characterized by design ground
motion parameters. The specificationof design ground motion parameters is one of
the most difficul t and most importantprobl ems in geotechnical earthquake
engineering.Much of the difficul ty in design ground motion specification resul ts
from its unavoidabl erel iance on subjective decisions that must be made with
incompl ete or uncertain information.These decisions l argel y revol ve around
the definition of the boundary betweenacceptabl e and excessive damage, and
uncertainty in the size, time, and l ocation of futureearthquakes. If very l ittl e
damage is acceptabl e, a rel ativel y strong l evel of shaking must bedesigned for, and
the measures required to resist that shaking can be quite expensive. Ifgreater
l evel s of damage are tol erabl e, l ower design l evel s of shaking may be consideredand
the resul ting design wil l be l ess expensive. Obviousl y, there are trade-offs
between theshort-term cost of providing an earthquake-resistant design and the
potential l ong-term cost(which, for many structures, may never be real ized) of
earthquake-induced damage.106Sec. 4.2 Identification and Eval uation of Earthquake
Sources 107Seismic hazard anal yses invol ve the quantitative estimation of ground-
shaking hazardsat a particul ar site. Seismic hazards may be anal yzed
deterministical l y, as when a particul arearthquake scenario is assumed, or
probabil istical l y, in which uncertainties inearthquake size, l ocation, and time of
occurrence are expl icitl y considered. Al though seismichazard anal ysis is a critical
part of the devel opment of design ground motions, it is notthe onl y part. This
chapter presents different methods for anal ysis of seismic hazards; thebroader
probl em of design ground motions is addressed in Chapter 8.4.2 IDEI'ITIFICATION AND
EVALUATIONOF E~RTHQUAI{ESOURCESTo eval uate seismic hazards for a particul ar site or
region, al l possibl e sources of seismicactivity must be identified and their
potential for generating future strong ground motioneval uated. Identification of
seismic sources requires some detective work; nature's cl ues,some of which are
obvious and others quite obscure, must be observed and interpreted.The avail abil ity
of modern seismographs and seismographic networks has madeobservation and
interpretation of current earthquakes rather convenient. The occurrence ofa l arge
earthquake is now recorded by hundreds of seismographs around the worl d. Within
hours, seismol ogists are abl e to determine its magnitude, l ocate its rupture
surface, and eveneval uate source parameters. In the 1990s, it is virtual l y
impossibl e for a significant earthquakeanywhere in the worl d to go undetected.The
current abil ity to identify and l ocate al l earthquake sources is a rel ativel y
recentdevel opment, particul arl y when compared with the time scal es on which l arge
earthquakesusual l y occur. The fact that no strong motions have been instrumental l y
recorded in a particul ararea does not guarantee that they have not occurred in the
past or that they wil l notoccur in the future. In the absence of an instrumental
seismic record, other cl ues of earthquakeactivity must be uncovered. These may take
the form of geol ogic and tectonic evidence,or historical (preinstrumental )
seismicity.4.2.1 Geol ogic EvidenceThe theory of pl ate tectonics assures us that the
occurrence of earthquakes is written in thegeol ogic record, primaril y in the form
of offsets, or rel ative displ acements, of various strata.Study of the geol ogic
record of past earthquake activity is cal l ed pal eoseismol ogy (Wal l ace,1981). In
some parts of the worl d, this geol ogic record is easil y accessibl e and rel ativel y
easil y interpreted by the trained seismic geol ogist. In other l ocations, however,
the geol ogicrecord may be very compl ex or it may be hidden by thick l ayers of
recent sediments thathave not been displ aced by seismic activity. The
identification of seismic sources from geol ogicevidence is a vital , though often
difficul t part of a seismic hazard anal ysis.The search for geol ogic evidence of
earthquake sources centers on the identificationof faul ts. A variety of tool s and
techniques are avail abl e to the geol ogist, incl uding thereview of publ ished
l iterature; interpretation of air photos and remote sensing (e.g., infrared
photograph) imagery; fiel d reconnaissance incl uding l ogging of trenches (Figure
4.1); testpits and borings; and geophysical techniques. Criteria for identification
of faul ts aredescribed in numerous textbooks on structural geol ogy, fiel d geol ogy
and geomorphol ogy(Adair, 1979). The fol l owing l ist of features that suggest
faul ting is that of Reiter (1990): 246Distance from main faul t (m)8 10m -: Lake
deposit--- -@Ql <: : Ql soICorrel ation of Lithol ogic and Soil UnitsSag fil l and
associatedcol l uvium Soil si{I'~; ~ ~s~ _ ,(S @] I .- (': ;\ , '0 '~ ~ (~@~@@ \:
@[§i@ \;@~ Al l uvial -fan~ deposit2: Ol E~Ol : J aZone of sl umping and backfil l o,\
\,,,,,,,,,,,,,,,,,,,,,,"<,',---------------.. Unit descriptions Lithol ogic units0
CD Bonnevil l e Lake deposits CIl l ® Post-provo al l uvial fan depositSag fil l derived
from the northand associated col l uvium@ Col l uvium8@ Transitional deposit@ Pond
depositLocal l y derived SAG fil l and associated col l uvium@ Col l uvium-basal facies
6.-----@ Col l uvium@ Pond deposit Q)Cii o@ Channel (?) deposit C/)~ 4@ Mudfl ow
deposit c: 0.~® Pond deposit 0.I: : Young scarp col l uviumIIQ)(§ Col l uvium-basal facies
Cii oC/) 2 ,----@ Col l uvium/sl opewash deposit Cii~Pre-settl ement deposit ~CD Pond
deposit soil Historical deposit® Al l uvium and pond deposits 0'-----Soil units® Soil
devel oped on post-Provo al l uvial -fan deposits@ Pal eoentisol soil devel oped on 4B -2
@ Pal eoentisol soil devel oped on 3B@ Topsoil -2Figure 4.1 Trench l og across the
Wasatch faul t near Kaysvil l e, Utah. Col l uvial units 3A, 4A/4B/S2, and 6A are three
separate deposits.Each resul ted from erosion ofthe scarp produced by a surface-
faul ting eathquake. (After Swan et al ., 1980; Schwartz, 1988.)Sec. 4.2
Identification and Eval uation of Earthquake Sources 1091. Directl y observabl e
fracture surfaces and indicators of fracturing. These incl ude disruptionof the
ground surface and evidence of the movement and grinding of the twosides of the
faul t (sl ickensides.faul t gouge, and faul t breccia).2. Geol ogical l y mappabl e
indicators. These incl ude the juxtaposition of dissimil armaterial s, missing or
repeated strata and the truncation of strata or structures.3. Topographic and
geomorphic (surface l andform) indicators [Figure 4.2]. These incl udetopographic
scarps or triangul ar facets on ridges, offset streams or drainage, til ting or
changes in el evation of terraces or shorel ines, sag ponds (water ponded by
depressionsnear strike-sl ip faul ts) and anomal ous stream gradients.4. Secondary
geol ogic features. These incl ude abrupt changes in groundwater l evel s,gradients,
and chemical composition, al ignment of springs or vol canic vents and thepresence of
hot springs.5. Lineaments on remote sensing imagery. These may be caused by
topography, vegetation,or tonal contrasts.6. Geophysical indicators of subsurface
faul ting. These incl ude steep l inear gravity ormagnetic gradients, differences in
seismic wave vel ocities, and offset of seismicrefl ection horizons.7. Geodetic
indicators. These incl ude faul t movement appearing in geodetic surveys astil ting
and changes in the distance between fixed points.4.2.1 .1 Faul t ActivityThe mere
presence of a faul t, however, does not indicate the l ikel ihood of future
earthquakes. The notion of faul t activity is important and has been a topic of
considerabl ediscussion and controversy over the years. Al though there is general
agreement concerningthe use of the terms active faul t to describe a faul t that
poses a current earthquake threat andinactive faul t to describe one on which past
earthquake activity is unl ikel y to be repeated,there is no consensus as to how
faul t activity shoul d be eval uated.Formal definitions of faul t activity are
important because they often trigger l egal requirements for special investigations
or special design provisions. However, there arewide variations in the criteria for
faul t activity in the commonl y used definitions. Sl emmonsand McKinney (1977), for
exampl e, found 31 different definitions of the term active faul t.Most were based on
the el apsed period of time since the most recent faul t movement. TheCal ifornia
Division of Mines and Geol ogy defines an active faul t as one that has produced
surface displ acement within Hol ocene time (approximatel y the past 10,000 years).
Fordams, the U.S. Army Corps of Engineers has used a time period of 35,000 years,
and theU.S. Bureau of Recl amation has used 100,000 years (Idriss, 1985). The U.S.
Nucl ear Regul atoryCommission (Code ofFederal Regul ations, 1978), on the other hand,
has used theterm capabl e faul t (rather than active faul t) for those that exhibit1.
movement at or near the ground surface at l east once within the past 35,000 years
ormovement of a recurring nature within the past 500,000 years;2. macroseismicity
instrumental l y determined with records of sufficient precision todemonstrate a
direct rel ationship with the faul t; or110Offset drainage channel Linear ridge(a)
Spring(b)Figure 4.2 Typical terrain in the vicinity of a faul t (a) showing
topographic andgeomorphic indicators offaul ting (After Wesson et aI., 1975.) and
(b) an aerial view ofsuch terrain al ong the San Andreas faul t in the Carrizo Pl ain
(photo by Robert Wal l ace;used by permission of U.S. Geol ogical Survey.)Sec 4.2
Identification and Eval uation of Earthquake Sources 1113. a structural rel ationship
to a capabl e faul t according to characteristics (1) or (2) above,such that movement
on one coul d reasonabl y be expected to be accompanied bymovement on the other.
Actual l y, the specification of faul t activity by specific time interval s is not
very real istic(Cl uff et al ., 1972; Cl uff and Cl uff, 1984); faul ts do not suddenl y
become inactive onthe 1O,OOOth or 35,000th anniversary of their l ast movement.
Rather, faul t activity is rel ativeand can change as faul ts move from active to
inactive states over geol ogic time. Cl uff andCl uff (1984) suggested six cl asses
(and five subcl asses) of faul t activity based on characteristicssuch
as sl ip rate, sl ip per event, rupture l ength, earthquake size, and recurrence
interval . Approaches of this type offer a more satisfying framework for
characterization offaul t activity but can be difficul t to impl ement in the
pol itical and economic environment inwhich many seismic hazard anal yses are
conducted.4.2.1.2 MagnitUde IndicatorsGeol ogic evidence can al so be used to
estimate the magnitude of past earthquakes bycorrel ating observed deformation
characteristics with the known magnitudes of recordedearthquakes. Studies of
worl dwide earthquakes have shown that faul ts do not rupture overtheir entire
l engths or areas during individual events. Instead, individual faul t segments with
physical l y control l ed boundaries (Schwartz and Coppersmith, 1986; Schwartz, 1988)
rupturerepeatedl y. Rupture l ength, rupture area, and faul t displ acement can be
eval uated bypostearthquake, fiel d geol ogical investigations. Correl ation of
magnitude with such quantitiesinvol ves regression on l imited data sets and,
consequentl y, produces an estimate of theexpected val ue of the magnitude. The
uncertainty in these estimates, which can be considerabl e,must be recognized when
appl ying them.Faul t rupture l ength has often been used to estimate earthquake
magnitude. A numberof studies (e.g., Tocher, 1958; Bonil l a and Buchanan, 1970; Mark
and Bonil l a, 1977; Sl emmons,1977, 1982; Acharaya, 1979; Chen, 1984; Bonil l a et al .,
1984; Wel l s and Coppersmith,1994) have il l ustrated the general nature of the
rel ationship between faul t rupturel ength and magnitude. Estimation of magnitude
based on faul t rupture l ength does notaccount for variations in the width of the
rupture surface; rupture l ength methods are bestsuited to cases in which the
rupture surface is fairl y narrow, typical l y l ess than about 20 km(12.4 mi) (Bonil l a
et al ., 1984). Obviousl y, they are not useful for cases in which rupturedoes not
extend to the ground surface. Faul t rupture area, by virtue of its rel ationship to
seismicmoment, woul d appear to be more fundamental l y rel ated to magnitude than
faul t rupturel ength al one. Indeed, for faul ts of width greater than about 20 km
(12.4 mi), magnitudesare more cl osel y correl ated to faul t rupture area than are any
other parameter (Wyss, 1979;Wel l s and Coppersmith, 1994). Al though the average
faul t displ acement is used to eval uatethe seismic moment, the unavail abil ity of
faul t displ acement measurements over an entirerupture surface renders its
determination impossibl e. Instead, maximum surface displ acements(Sl emmons, 1982;
Wel l s and Coppersmith, 1994) have been correl ated to magnitude.Empirical
rel ationships based on statistical anal yses of worl dwide historical earthquakedata
are presented in Tabl e 4-1 and Figure 4.3.The rel ationships in Tabl e 4-1 can be
used to predict mean val ues of the dependentvariabl es (Mw' l og L, l og A, and l og
D); the standard deviations of the dependent variabl escan be used (see Appendix C)
to compute val ues other than the mean.112 Seismic Hazard Anal ysis Chap. 4Tabl e 4-1
Empirical Rel ationships between Moment Magnitude, Mw' Surface Rupture Length,L
(km), Rupture Area, A (krrr'). and Maximum Surface Displ acement, D(m)Faul tMovement
Strike sl ipReverseNormal Al l Strike Sl ipReverseNormal Al l Strike sl ipReverse"Normal Al l
Numberof Events4319157783432214843211680Rel ationshipM; = 5.16 + 1.12l og LMw = 5.00
+ 1.22 l og LMw = 4.86 + 1.32 l og LMw = 5.08 + 1.16l og LMw = 3.98 + 1.02 l og AMw =
4.33 + 0.90 l og AMw = 3.93 + 1.02 l og AMw = 4.07 + 0.98 l og AMw = 6.81 + 0.78 l og D
Mw = 6.52 + 0.44 l og DMw = 6.61 + 0.71 l og DMw = 6.69 + 0.74 l og D0.280.280.340.28
0.230.250.250.240.290.520.340.40Rel ationshipl og L = 0.74Mw - 3.55l og L = 0.63Mw -
2.86l og L = 0.50Mw - 2.01l og L = 0.69Mw - 3.22l og A = 0.90Mw - 3.42l og A = 0.98Mw -
3.99l og A = 0.82Mw - 2.87l og A = 0.91Mw - 3.49l ogD= 1.03Mw-7.03l og D = 0.29Mw -
1.84l og D = 0.89Mw - 5.90l og D = 0.82Mw - 5.460.230.200.210.220.220.260.220.240.34
0.420.380.42Source: Wel l s and Coppersmith (1994).aRegression rel ationships are not
statistical l y significant at a 95% probabil ity l evel (note inconsistency of
regression coefficients and standard deviations).9 9 90 0 Strike sl ip 0 Strike sl ip
~8 0 ~" 8 0 Reverse8 0 Reverse.a "- o~...e...'O "- Normal "- Normal '2 77 ,- 148 EQs
80 EQsg'7 7 7E'E6 ~9q 0 6 6(I)E ~ ......~0~5 ,,'t: ..5 "-4 4 41 10 100 103 1 10 100
103 104 10-2 10-1 1 10Surface rupture l ength (km) Rupture area (km) Maximum
displ acement (m)Fl gure 4.3 Scatter inherent in databases from which correl ations of
Tabl e 4-1 weredevel oped. (After Wel l s and Coppersmith, 1994. Used by permission of
theSeismol ogical Society of America.)Exampl e 4.1Compute the probabil ity that a
moment magnitude 7.0 earthquake on the San Andreas faul twoul d cause a surface
rupture l onger than 100 km.Sol ution The San Andreas faul t is known to produce
strike-sl ip movement (Section 2.4.2.2).From Tabl e 4-1, the mean surface rupture
l ength for aMw = 7.0 earthquake woul d be computed as10gL = 0.74Mw - 3.55 = 0.74
(7.0) - 3.55 = 1.63Then the mean, or expected, val ue of L is given byL = 101 63 =
42.7 kmThe standard normal variate (Section C.7.2 of Appendix C) for a 100-km-Iong
surface rupturewoul d bez = l og 100 -l og42.7 1.320.28Sec. 4.2 Identification and
Eval uation of Earthquake Sources 113From Tabl e C-l , the probabil ity that the
surface rupture l ength woul d exceed 100 km is 0.0934or 9.34%.4.2.2 Tectonic
EvidencePil atetectonics and el astic rebound theory tel l us that earthquakes occur
to rel ieve the strainenergy that accumul ates as pl ates move rel ative to each other.
The rate of movement, therefore,shoul d be rel ated to the rate of strain energy
accumul ation and al so to the rate of strainenergy rel ease (Smith, 1976; Woodward-
Cl yde Consul tants, 1979; Idriss, 1985). For majorsubduction zones, Ruff and
Kanamori (1980) rel ated maximum magnitude to both the rateof convergence and the
age of the subducted sl ab according toMw = -0.0089T+0.134V+7.96 (4.1)where T is the
age in mil l ions of years and Vis the rate of convergence in ern/yr. Heaton and
Kanamori (1984) used this rel ationship to suggest that the Cascadia subduction zone
off thecoasts of Oregon, Washington, and British Col umbia coul d be capabl e of
generating greatearthquakes of magnitude wel l above 8 (Figure 4.4). Subsequentl y,
geol ogic evidence ofhistorical great earthquakes was discovered (e.g., Atwater,
1987; Atwater et al ., 1987) al ongthe coasts of Washington and Oregon.
12 ,----,----",----,--,-----...------,-------,------,Figure 4.4 Rel ationship
betweenearthquake magnitude, age, and rate ofconvergence in subduction zone
environments. Diagonal l ines correspond toequation (4.1). Data points represent
actual earthquakes. (After Heaton and Kanamori,1984. Used by permission of the
Seismol ogical Society of America.)8.5.C. Chil e • 9.590 9.2 ~.o~· Chil eKam~hatka
Peru : .1 ~~.1 C.A~~rica 8.0,.Al eutians • Col ombIa7.9·, • Sumatra ........92• 80 .
Ryukyus Al aska • • 8.0140 120 100 80 60 40 20 0Age (mil l ion years before present)
0'----'---'----'''---'---'--------'---''''-------'-----'----'180'i: ' 10~~B 8eQICl ~
6s: : ou 4'0~a: 24.2.3 Historical SeismicityEarthquake sources may al so be
identified from records of historical (preinstrumental ) seismicity.The written
historical record extends back onl y a few hundred years or l ess in theUnited
States; in Japan and the Middl e East it may extend about 2000 years and up to 3000
years or so in China (Ambraseys, 1971, 1978; Al l en, 1975; Bol t, 1988).Historical
accounts of ground-shaking effects can be used to confirm the occurrenceof past
earthquakes and to estimate their geographic distributions of intensity. When
sufficientdata are avail abl e, the maximum intensity can be determined and used to
estimate thel ocation of the earthquake epicenter and the magnitude of the event.
Al though the accuracyof l ocations determined in this way depends strongl y on
popul ation density and the rate of114 Seismic Hazard Anal ysis Chap. 4earthquake
recurrence, a geographic pattern of historic epicenters provides strong evidencefor
the existence of earthquake source zones. Since historical records are dated, they
can al sobe used to eval uate the rate of recurrence of earthquakes, or seismicity,
in particul ar areas.4.2.4 Instrumental SeismicityOver the past 80 or 90 years,
about 10 earthquakes of M, > 7 have occurred somewhere inthe worl d each year
(Kanamori, 1988). Instrumental records from l arge earthquakes havebeen avail abl e
since about 1900, al though many from before 1960 are incompl ete or ofuneven
qual ity. Neverthel ess, instrumental recordings represent the best avail abl e
informationfor the identification and eval uation of earthquake sources. Their most
significant l imitationis the short period of time, compared with the average period
of time between l argeearthquakes, for which they have been avail abl e. Again, the
al ignment of instrumental l yl ocated epicenters or hypocenters indicates the
existence of earthquake sources. Anal ysis ofaftershocks can al so aid in the
del ineation of earthquake source zones.4.3 DETERMINISTIC SEISMIC HAZARD ANALYSISIn
the earl y years of geotechnical earthquake engineering, the use of deterministic
seismichazard anal ysis (DSHA) was preval ent. A DSHA invol ves the devel opment of a
particul arseismic scenario upon which a ground motion hazard eval uation is based.
The scenario consistsof the postul ated occurrence of an earthquake of a specified
size occurring at a specifiedl ocation. Atypical DSHA can be described as a four-
step process (Reiter, 1990) consisting of: 1. Identification and characterization of
al l earthquake sources capabl e of producingsignificant ground motion at the site.
Source characterization incl udes definition ofeach source's geometry (the
source zone) and earthquake potential .2. Sel ection of a source-to-site distance
parameter for each source zone. In mostDSHAs, the shortest distance between the
source zone and the site of interest issel ected. The distance may be expressed as
an epicentral distance or hypocentral distance,depending on the measure of distance
of the predictive rel ationship(s) used inthe fol l owing step.3. Sel ection of the
control l ing earthquake (i.e., the earthquake that is expected to producethe
strongest l evel of shaking), general l y expressed in terms of some groundmotion
parameter, at the site. The sel ection is made by comparing the l evel s of shaking
produced by earthquakes (identified in step 1) assumed to occur at the distances
identified in step 2. The control l ing earthquake is described in terms of its size
(usual l yexpressed as magnitude) and distance from the site.4. The hazard at the
site is formal l y defined, usual l y in terms of the ground motions producedat the
site by the control l ing earthquake. Its characteristics are usual l ydescribed by one
or more ground motion parameters obtained from predictive rel ationshipsof the types
presented in Chapter 3. Peak accel eration, peak vel ocity, andresponse spectrum
ordinates are commonl y used to characterize the seismic hazard.The DSHA procedure
is shown schematical l y in Figure 4.5. Expressed in these fourcompact steps, DSHA
appears to be a very simpl e procedure, and in many respects it is.Sec. 4.3
Deterministic Seismic Hazard Anal ysis 115~so u rce 3Siteo M3STEP 1 STEP 2,M '
2,, ,, ~ , , ,DistanceSTEP 4STEP 3Figure 4.5 Four steps of a deterministic seismic
hazard anal ysis.When appl ied to structures for which fail ure coul d have
catastrophic consequences, such asnucl ear power pl ants and l arge dams, DSHA
provides a straightforward framework foreval uation of worst-case ground motions.
However, it provides no information on thel ikel ihood of occurrence of the
control l ing earthquake, the l ikel ihood of it occurring whereit is assumed to occur,
the l evel of shaking that might be expected during a finite period oftime (such as
the useful l ifetime of a particul ar structure or facil ity), or the effects of
uncertaintiesin the various steps required to compute the resul ting ground motion
characteristics.Perhaps most important, DSHA invol ves subjective decisions,
particul arl y regardingearthquake potential (step 1), that can require the combined
expertise and opinions of seismol ogists,seismic geol ogists, engineers, risk
anal ysts, economists, social scientists, andgovernment official s. The broad range
of backgrounds and often divergent goal s of such professional scan cause difficul ty
in reaching a consensus on earthquake potential . Over theyears there have been many
terms used to describe earthquake potential ; among them themaximum credibl e
earthquake (MCE), design basis earthquake (DBE), safe shutdownearthquake (SSE),
maximum probabl e earthquake (MPE), operating basis earthquake(aBE), and seismic
safety eval uation earthquake. The MCE, for exampl e, is usual l y definedas the
maximum earthquake that appears capabl e of occurring under the known tectonic
framework. The DBE and SSE are usual l y defined in essential l y the same way. The MPE
hasbeen defined as the maximum historical earthquake and al so as the maximum
earthquake116 Seismic Hazard Anal ysis Chap. 4l ikel y to occur in a l OO-yearinterval .
Many DSHAs have used the two-pronged approach ofeval uating hazards for both the MCE
and MPE (or SSE and OBE). Disagreements over thedefinition and use of these terms
have forced the del ay, and even cancel l ation, of a numberof l arge construction
projects. The Committee on Seismic Risk of the Earthquake EngineeringResearch
Institute (EERI) has stated that terms such as MCE and MPE "are misl eading... and
their use is discouraged" (Committee on Seismic Risk, 1984).Exampl e 4.2The site
shown in Figure E4.2 is l ocated in the vicinity of three independent seismic
sourcesrepresented by source zones 1, 2, and 3. Using a deterministic seismic
hazard anal ysis, computethe peak accel eration.Mmax = 7.3(-50,75)Source 1Mmax = 5.0
(20, 78)Source 3 o(0,60)(0,0)o (20,18)Site(-15, -30)(100,78)Mmax =7.7(100,18)Figure
E4.2Sol ution Taking the site as the center of a l ocal x-y coordinate system, the
coordinates of thesource zone boundaries (in kil ometers) are given in parentheses.
Source zone 1 is a l l l -kml ongl inear source zone that can produce a maximum
magnitude of 7.3 at any point al ong itsl ength. Source zone 2 is an areal source
zone of 4800 km2 capabl e of generating a magnitude 7.7earthquake anywhere within
its boundaries. Source zone 3 is a point source that can produce amaximum magnitude
of 5.0. Fol l owing the four-step procedure described earl ier: 1. The probl em
statement provides the l ocation and maximum magnitude of each sourcezone. In real
DSHAs, this is often an extremel y compl ex and difficul t task.2. The source-to-site
distance can be represented by the minimum between the site and anypart of each
source zone. On that basis, the distances are: Distance, RSource Zone (krn)1 23.72
25.03 60.03. If the l evel of shaking is assumed to be adequatel y characterized by
the peak horizontal accel eration, an appropriate attenuation rel ationship can be
used to sel ect the control l ingearthquake. Using the rel ationship of Cornel l et al .
(1979), devel oped with data fromM = 3.0 to 7.7 earthquakes at distances of 20 to
200 km in the western United States,InPHA (gal s) = 6.74 + 0.859M - 1.80l n (R + 25)
the PHA val ues generated by each of the source zones woul d be: Sec. 4.4
Probabil istic Seismic Hazard Anal ysis 117Source Zone123M7.37.75.0R(km)23.725.060.0
PHA0.42g0.57g0.02gOn this basis, the source zone 2 event woul d be sel ected as the
control l ing earthquake.(Note: Though currentl y out of date, the Cornel l et al .
rel ationship is used here becauseof its simpl icity which wil l make a subsequent
exampl e on probabil istic seismic hazardanal ysis much easier to understand.)4. The
hazard woul d be taken as that which woul d resul t from a magnitude 7.7 earthquake
occurring at a distance of 25 km. This motion woul d produce a peak accel eration of
0.57g; other ground motion parameters coul d be obtained from the predictive
rel ationshipsdescribed in Chapter 3.4.4 PROBABILISTIC SEISMIC HAZARD ANALYSISIn the
past 20 to 30 years the use of probabil istic concepts has al l owed uncertainties in
thesize, l ocation, and rate of recurrence of earthquakes and in the variation of
ground motioncharacteristics with earthquake size and l ocation to be expl icitl y
considered in the eval uationof seismic hazards. Probabil istic seismic hazard
anal ysis (PSHA) provides a frameworkin which these uncertainties can be identified,
quantified, and combined in a rational manner to provide a more compl ete picture of
the seismic hazard.Understanding the concepts and mechanics of PSHA requires
famil iarity with someof the terminol ogy and basic concepts of probabil ity theory.
Such background informationcan be found in Appendix C. The PSHA methodol ogy
described in this section is simil ar inmany respects to the wel l -establ ished
methods devel oped by Cornel l (1968), and Al germissenet al . (1982).The PSHA can al so
be described as a procedure of four steps (Reiter, 1990), each ofwhich bear some
degree of simil arity to the steps of the DSHA procedure, as il l ustrated inFigure
4.6.1. The first step, identification and characterization of earthquake sources,
is identical tothe first step of the DSHA, except that the probabil ity distribution
of potential rupturel ocations within the source must al so be characterized. In most
cases, uniform probabil itydistributions are assigned to each source zone, impl ying
that earthquakes areequal l y l ikel y to occur at any point within the source zone.
These distributions arethen combined with the source geometry to obtain the
corresponding probabil ity distributionof source-to-site distance. The DSHA, on the
other hand, impl icitl y assumesthat the probabil ity of occurrence is 1 at the points
in each source zone cl osest to thesite, and zero el sewhere.2. Next, the seismicity
or temporal distribution of earthquake recurrence must be characterized.A
recurrence rel ationship, which specifies the average rate at which anearthquake of
some size wil l be exceeded, is used to characterize the seismicity ofeach source
zone. The recurrence rel ationship may accommodate the maximum sizeearthquake, but
it does not l imit consideration to that earthquake, as DSHAs often do.118 Seismic
Hazard Anal ysis Chap. 4Source 1 Source 3R12Magnitude, x>-STEP 1Distance, RSTEP 3
": >0.,A>-Ii: : 'STEP 2Parameter val ue, y*STEP 4Figure 4.6 Four steps of a
probabil istic seismic hazard anal ysis.3. The ground motion produced at the site by
earthquakes of any possibl e size occurringat any possibl e point in each source zone
must be determined with the use of predictiverel ationships. The uncertainty
inherent in the predictive rel ationship is al so consideredin a PSHA.4. Final l y, the
uncertainties in earthquake l ocation, earthquake size, and ground motionparameter
prediction are combined to obtain the probabil ity that the ground motionparameter
wil l be exceeded during a particul ar time period.The proper performance of a PSHA
requires careful attention to the probl ems ofsource characterization and ground
motion parameter prediction and to the mechanics of theprobabil ity computations.
4.4.1 Earthquake Source CharacterizationCharacterization of an earthquake source
requires consideration of the spatial characteristicsof the source and of the
distribution of earthquakes within that source, of the distributionof earthquake
size for each source, and of the distribution of earthquakes with time. Each of
these characteristics invol ves some degree of uncertainty.4.4.1.1 Spatial
UncertaintyThe geometries of earthquake sources depend on the tectonic processes
invol ved intheir formul ation, Earthquakes associated with vol canic activity, for
exampl e, general l ySec. 4.4 Probabil istic Seismic Hazard Anal ysis 119originate in
zones near the vol canoes that are smal l enough to al l ow them to be characterizedas
point sources. Wel l -defined faul t pl anes, on which earthquakes can occur at many
different l ocations, can be considered as two-dimensional areal sources. Areas
where earthquakemechanisms are poorl y defined, or where faul ting is so extensive as
to precl ude distinctionbetween individual faul ts, can be treated as three-
dimensional vol umetric sources.For the purposes of a seismic hazard anal ysis, the
source zones may be simil ar to orsomewhat different than the actual source,
depending on the rel ative geometry of the sourceand site of interest and on the
qual ity of information about the sources. For exampl e, the rel ativel yshort faul t in
Figure 4.7a can be model ed as a point source since the distancebetween any point
al ong its l ength and the site is nearl y constant. Simil arl y, the depth of thefaul t
pl ane shown in Figure 4.7b is sufficientl y smal l that variations in hypocentral
depthhave l ittl e infl uence on hypocentral distance. In such a case the hazard
anal ysis can be simpl ifiedwith negl igibl e l oss of accuracy by approximating the
pl anar source as a l inearsource zone. In Figure 4.7c, the avail abl e data are
insufficient to determine accuratel y theactual geometry of the source, so it is
represented as a vol umetric source.Earthquakes are usual l y assumed to be uniforml y
distributed within a particul arsource zone (i.e., earthquakes are considered
equal l y l ikel y to occur at any l ocation). Theassumption of uniformity is by no
means required; nonuniform distributions may be usedwhen sufficient information to
justify them exists. A uniform distribution within the sourcezone does not,
however, often transl ate into a uniform distribution of source-to-site distance.
Since predictive rel ationships express ground motion parameters in terms of some
measure of source-to-site distance, the spatial uncertainty must be described with
respect tothe appropriate distance parameter. The uncertainty in source-to-site
distance can bedescribed by a probabil ity density function.For the point source of
Figure 4.8a, the distance, R, is known to be rs ; consequentl y,the probabil ity that
R = rs is assumed to be 1 and the probabil ity that R "* rs , zero. Other casesare
not as simpl e. For the l inear source of Figure 4.8b, the probabil ity that an
earthquakeoccurs on the smal l segment of the faul t between L =I and L =1+ dl is the
same as the probabil itythat it occurs between R =rand R =r + dr; that is,h(l )dl =
h(r)dr (4.2)where/L(l ) and!R(r) are the probabil ity density functions for the
variabl es Land R, respectivel y.Consequentl y,(a) (b) (c)Figure 4.7 Exampl es of
different source zone geometries: (a) short faul t that can bemodel ed as a point
source; (b) shal l ow faul t that can be model ed as a l inear source; (c)three-
dimensional source zone.120 Seismic Hazard Anal ysis Chap. 4SourceSite ~~rsfR(r)~~dr
dr • rs R(a)~Lf-----------. , ~L de Irmin' e rSite' r+ dr \L~sou rce~(r)• rmin R(b)
s;~o SourceR(c)(4.3)Figure 4.8 Exampl es of variations of source-to-site distance
for different source zonegeometries. The shape of the probabil ity distribution can
be visual ized by consideringthe rel ative portions of the source zone that woul d
fal l between each of a series of circl es(or spheres for three-dimensional probl ems)
with equal differences in radius.d[ !R (r) = fdl ) drIf earthquakes are assumed to
uniforml y distributed over the l ength of the faul t,fL(l ) =l il .;Since [2 = r2 - r~n
the probabil ity density function of R is given by!R (r) = r (4.4)Lf Jr2- r~inFor
source zones with more compl ex geometries, it is easier to eval uatefR(r) by
numerical rather than anal ytical methods. For exampl e, dividing the irregul ar source
zone of Figure 4.8cinto a l arge number of discrete el ements of equal area, a
histogram that approximates fR(r) canbe constructed by tabul ating the val ues of R
that correspond to the center of each el ement.The preceeding discussion assumes
that al l the energy is rel eased at the hypocenter ofthe earthquake. However, energy
is rel eased over the entire faul t rupture surface, parts ofwhich may be much cl oser
to the site than the hypocenter. Der-Kiureghian and Ang (1977)noted that the
rupture surface of a l arge earthquake with a distant hypocenter coul d rel easeenergy
much cl oser to the site, and devel oped methods to account for rupture surface
dimensions in PSHA.4.4.1.2 Size UncertaintyOnce an earthquake source is identified
and its corresponding source zone characterized,the seismic hazard anal yst's
attention is turned toward eval uation of the sizes of earthquakesthat the source
zone can be expected to produce. Al l source zones have a maximumearthquake
magnitude that cannot be exceeded; it can be l arge for some and smal l for others.In
general , the source zone wil l produce earthquakes of different sizes up to the
maximumearthquake, with smal l er earthquakes occurring more frequentl y than l arger
ones. The strainenergy may be rel eased aseismical l y, or in the form of earthquakes.
Assuming that al l strainenergy is rel eased by earthquakes of magnitude 5.5 to 9.0
and that the average faul t displ acementis one-hal f the maximum surface
displ acement, Sl emmons (1982) showed howthe rate of movement was rel ated to
earthquake magnitude and recurrence interval (Figure 4.9). The distribution of
earthquake sizes in a given period of time is described bySec. 4.4 Probabil istic
Seismic Hazard Anal ysis 121107106105C.2: : ~... 104 Gl a.c.a 103Gl I%: 10210110°6IIII
-----------r----------IIIII7 8Magnitude9Figure 4.9 Effect of faul t sl ip rate and
earthquake magnitude on return period.(After Sl emmons, 1982.)a recurrence l aw. A
basic assumption ofPSHA is that the recurrence l aw obtained from pastseismicity is
appropriate for the prediction of future seismicity.Gutenberg-Richter Recurrence
Law. Gutenberg and Richter (1944)gathered data for southern Cal ifornia earthquakes
over a period of many years and organizedthe data according to the number of
earthquakes that exceeded different magnitudesduring that time period. They divided
the number of exceedances of each magnitude by thel ength of the time period to
define a mean annual rate of exceedance, Am of an earthquakeof magnitude m. As
woul d be expected, the mean annual rate of exceedance of smal l earthquakesis
greater than that of l arge earthquakes. The reciprocal of the annual rate of
exceedancefor a particul ar magnitude is commonl y referred to as the return period
of earthquakesexceeding that magnitude. When the l ogarithm of the annual rate of
exceedance of southernCal ifornia earthquakes was pl otted against earthquake
magnitude, a l inear rel ationship wasobserved. The resul ting Gutenberg-Richter l aw
for earthquake recurrence was expressed asl ogAm = a - bm (4.5)where Am is the mean
annual rate of exceedance of magnitude m, IO" is the mean yearl ynumber of
earthquakes of magnitude greater than or equal to zero, and b (the b val ue)
describes the rel ative l ikel ihood of l arge and smal l earthquakes. The Gutenberg-
Richterl aw is il l ustrated schematical l y in Figure 4. l Oa. As the b val ue increases,
the number ofl arger magnitude earthquakes decreases compared to those of smal l er
magnitudes. TheGutenberg-Richter l aw is not restricted to the use of magnitude as a
descriptor of earthquakesize; epicentral intensity has al so been used. Worl dwide
recurrence data are shown inFigure 4. l Ob.122 Seismic Hazard Anal ysis Chap. 49 m 10
xxx,,,,__1. _ ,,,,,,,,8(b)6 710 -1 -A!pide bel t0.1 --~------;-------",,,,,,__ L
_ ,,,,,500 ,-,-----.,---,-----,----0.01100Am(a)MagnitudeoFigure 4.10 (a) Gutenberg-
Richter recurrence l aw, showing meaning of a and bparameters; and (b) appl ication
of Gutenberg-Richter l aw to worl dwide seismicity data.(After Esteva, 1970.)Exampl e
4.3Using Figure 4. l Ob, compute the return period of M = 8 earthquakes on the
Circumpacific andAl pide bel ts.Sol ution At a magnitude of 8, Figure 4.1Obindicates
that the Circumpacific and Al pide bel tshave mean annual rates of exceedance of 1.76
per year and 0.31 per year, respectivel y. Therefore,the corresponding return
periods areCircumpacific: TR = I 1Am 1.76/yr = 0.6 yearAl pide: TR = I 1Am 0.31/yr =
3.2 yearsThe a and b parameters are general l y obtained by regression on a database
of seismicityfrom the source zone of interest. Unl ess the source zone is extremel y
active, the databaseis l ikel y to be rel ativel y sparse. Since the use of both
instrumental and historical events isusual l y required, the database may contain
both magnitude (possibl y based on differentscal es) and intensity data,
necessitating the conversion of one measure of size to the other.Insome areas, the
record of seismicity may be distorted by the presence of dependent eventssuch as
aftershocks and foreshocks (Merz and Cornel l , 1973). Al though such dependentevents
can cause significant damage, a PSHA is intended to eval uate the hazard from
discrete,independent rel eases of seismic energy. Therefore, dependent events must
be removedfrom the seismicity database and their effects accounted for in separate
anal yses. Compl etenessof the database must al so be considered. The historical
record is usual l y more compl etefor l arge earthquakes than for smal l earthquakes;
smal l earthquakes can go undetected for aSec. 4.4 Probabil istic Seismic Hazard
Anal ysis 123variety of physical and demographic reasons. Fitting a straight l ine
such as that impl ied bythe Gutenberg-Richter l aw through recurrence data in which
the mean rate of exceedance ofsmal l earthquakes is underestimated
wil l tend to fl atten the l ine. As a resul t, the actual meanrate of smal l
earthquakes wil l be underpredicted and the mean rate of l arge earthquakes wil l be
overpredicted. Different methods have been proposed (Stepp, 1972; Weichert, 1980;
EPRI, 1986) to correct incompl ete records.Bounded Gutenberg-Richter Recurrence
Laws. The standard Gutenberg-Richter recurrence l aw of equation (4.5) may al so be
expressed asAm = l Oa - bm = exp (ex- ~m) (4.6)where ex =2.303a and ~ =2.303b.
Equation (4.6) shows that the Gutenberg-Richter l awimpl ies that earthquake
magnitudes are exponential l y distributed. The standard GutenbergRichterl aw covers
an infinite range of magnitudes, from -00 to +00. For engineeringpurposes, the
effects of very smal l earthquakes are of l ittl e interest and it is common to
disregardthose that are not capabl e of causing significant damage. If earthquakes
smal l er thana l ower threshol d magnitude mo are el iminated, the mean annual rate of
exceedance can bewritten (McGuire and Arabasz, 1990) asAm = vexp [-~ (m - mo)] m >
mo (4.7)where v = exp (ex - ~mo)' In most PSHAs, the l ower threshol d magnitude is
set at val uesfrom about 4.0 to 5.0 since magnitudes smal l er than that sel dom cause
significant damage.The resul ting probabil ity distribution of magnitude for the
Gutenberg-Richter l aw withl ower bound can be expressed in terms of the cumul ative
distribution function (CDF): FM(m) = P[M<mIM>mo] =Am -Am -,.R,(m-111{)o = 1- eAmo
(4.8)or the probabil ity density function (PDF): f () d F ( ) = A e-~ (m - 111{)M m =
dm M m fJ (4.9)(4.10)At the other end of the magnitude scal e, the standard
Gutenberg-Richter l aw predictsnonzero mean rates of exceedance for magnitudes up to
infinity. This impl ies, for exampl e,that the Circumpacific bel t (Figure 4.l Ob),
woul d produce a magnitude 10 earthquake at amean annual exceedance rate of about
0.02 per year (a return period of onl y 50 years), eventhough earthquakes of that
size have never been observed. Some maximum magnitude, mmax'is associated with al l
source zones. If it is known or can be estimated, the mean annual rateof exceedance
can be expressed (McGuire and Arabasz, 1990) asA = v exp [-~ (m - mo)] - exp [-~
(mmax - mo) ]m 1 - exp [-~ (mmax - mo) ]The bounded recurrence l aw of equation
(4.10) is shown in Figure 4.11a for conditions of constantrate of seismicity (i.e.,
constant mean annual rate of exceedance of mO)' An al ternativeinterpretation, based
on a constant rate of seismic moment (hence energy) rel ease, producesthe recurrence
curves of Figure 4.11b. In the constant moment rate model , increasing the maximum
magnitude requires a substantial decrease in the mean annual rate of exceedance of
10-1124""'"-, ,\,,,\\6 7 8. 6,'" ,,,7 8\\Seismic Hazard Anal ysis Chap. 4(b)
Magnitude, m Magnitude, m(a)5 6 7 8Figure 4.11 Bounded Gutenberg-Richterrecurrence
l aws for ma=4 and mmax =6, 7,and 8 constrained by (a) constant seismicityrate and
(b) constant moment rate. (AfterYoungs and Coppersmith, 1985.)(4.11)l ower magnitude
events to account for the extra energy rel eased in l arge earthquakes. Sincethe
seismic moment is proportional to the amount of sl ip (displ acement) that occurs in
anearthquake, the moment rate is proportional to the sl ip rate. Hence the constant-
moment-ratemodel is equival ent to a constant-sl ip-rate model and can be used when
the sl ip rate is knownto be constant. The extent to which actual sl ip rates vary
with time, however, appears to be differentfor different faul ts and can even
fl uctuate with time al ong the same faul t.The CDF and PDF for the Gutenberg-Richter
l aw with upper and l ower bounds canbe expressed as1 - exp [-~ (m - mo)]FM(m) = P [M
< ml mo ~ m ~ mmax] = -,------": : ---: ;: -'-: ------"--: -~1- exp [-~ (mmax -mo)]fM(m) =
~exp[-~(m-mo)] (4.12)1 - exp [-~ (mmax- mo) ]Characteristic Earthquake Recurrence
Laws. The Gutenberg-Richterl aw was devel oped from a set of regional data that
incl uded many different seismic sources.Since PSHAs are usual l y conducted for
specific sites rather than l arge regions, the earthquake-generating characteristics
of individual faul ts is important. Inrecent years the abil ityof the Gutenberg-
Richter l aw to represent the behavior of a singl e source has been cal l edinto
question (Schwartz and Coppersmith, 1984; Schwartz, 1988).Pal eoseismic studies
indicate that individual points on faul ts and faul t segments tendto move by
approximatel y the same distance in each earthquake. This has been interpretedto
suggest that individual faul ts repeatedl y generate earthquakes of simil ar (within
aboutone-hal f magnitude unit) size, known as characteristic earthquakes, at or near
their maximummagnitude. Al ternativel y, the apparentl y repetitive nature of faul t
movement at individual points may be control l ed by l ocal ized geol ogic constraints
and, consequentl y, notrefl ect earthquake magnitude very accuratel y. Resol ution of
these al ternative interpretationsawaits further pal eoseismic research.Sec. 4.4
Probabil istic Seismic Hazard Anal ysis 125By dating these characteristic
earthquakes, their historical rate of recurrence can beestimated. Geol ogic evidence
indicates the characteristic earthquakes occur more frequentl ythan woul d be impl ied
by extrapol ation of the Gutenberg-Richter l aw from highexceedance rates (l ow
magnitude) to l ow exceedance rates (high magnitude). The resul t isa more compl ex
recurrence l aw that is governed by seismicity data at l ow magnitudes andgeol ogic
data at high magnitudes, as shown in Figure 4.12.100
,=_--,--------,---,--.----..,..----,---,----=1.0 cr-~E Seismicitydata, . 10 : : 0.1-
0.01-• 2 3Control l e: by tCO: Ol l ed byseismicity data geol ogic dataMagnitude, m9Figure
4.12 Inconsistency of mean annual rate of exceedance as determined fromseismicity
data and geol ogic data. (AfterYoungs and Coppersmith, 1985.)Youngs and Coppersmith
(1985) devel oped a general ized magnitude-frequency densityfunction that combined an
exponential magnitude distribution at l ower magnitudes witha uniform distribution
in the vicinity of the characteristic earthquake. Recurrence rel ationshipsderived
from the Youngs and Coppersmith model and the bounded Gutenberg-Richtermodel ,
assuming the same mmax, b val ue, and sl ip rate, are shown in Figure 4.13. The
characteristic earthquake model predicts higher rates of exceedance at magnitudes
near thecharacteristic earthquake magnitude and l ower rates at l ower magnitudes.
Other model s thataccount for characteristic earthquakes have been devel oped by
Wesnorsky et al . (1984) andWu et al . (1995).Other Recurrence Laws. A number of
other recurrence l aws have been proposed.Merz and Cornel l (1973a) used a quadratic
expression to describe the mean annual rate at which earthquakes of magnitude
greater than ma and l ess than mmax are exceeded.Shah et al . (1975) used a bil inear
recurrence l aw in an eval uation of seismic risk for Nicaragua.In another approach,
the Gutenberg-Richter l aw was modified on the basis of seismicmoment and faul t sl ip
(Lomnitz-Adl er and Lomnitz, 1979).Discussion. Avail abl e evidence is insufficient to
determine whether the Gutenberg-Richter, characteristic earthquake, or some other
recurrence l aw is correct. Eval uationof which model is most appropriate for a given
source is hampered by the brevity of historical 12610 r-r-r-r-r-r-r-,.-,BoundedG-R
model " ""/'-Characteristicearthquakemodel 10-4Seismic Hazard Anal ysis Chap. 4
Magnitude, mFigure 4.13 Comparison of recurrencel aws from bounded Gutenberg-Richter
andcharacteristic earthquake model s. (AfterYoungs and Coppersmith, 1985.)and/or
instrumental records. Based ort five decades of seismicity records for the major
seismicsources of southern Cal ifornia, Wesnousky (1994) concl uded that whil e the
avail abl edata were not sufficient to disprove the Gutenberg-Richter recurrence l aw,
the characteristicearthquake model better represented the observed distribution of
earthquake magnitudes.Additional research in this area is in progress and wil l
undoubtedl y be an active topic of discussionin the forthcoming seismol ogy
l iterature.4.4.2 Predictive Rel ationshipsPredictive rel ationships are nearl y al ways
obtained empirical l y by l east-squares regressionon a particul ar set of strong
motion parameter data. Despite attempts to remove questionabl edata and the use of
qual ity-based weighting schemes, some amount of scatter in thedata is inevitabl e.
The scatter resul ts from randomness in the mechanics of rupture and fromvariabil ity
and heterogeneity of the source, travel path, and site conditions. Scatter in the
data can be quantified by confidence l imits (Campbel l , 1985) or by the standard
deviationof the predicted parameter. Refl ecting the form of most predictive
rel ationships, the standarddeviation of the (natural ) l ogarithm of the predicted
parameter is usual l y computed.This considerabl e uncertainty must be accounted for
in computation of seismic hazards.The probabil ity that a particul ar ground motion
parameter Yexceeds a certain val ue, y*, foran earthquake of a given magnitude, m,
occurring at a given distance, r, is il l ustrated graphical l yin Figure 4.14. In
probabil istic terms, it is given byp[Y>Y*l m,r] = I-Fy(y*) (4.13)where Fy (y) is the
val ue of the CDF of Yat m and r. The val ue of Fy (y) depends on the probabil ity
distribution used to represent Y. In general , ground motion parameters are usual l y
assumed to be l ognormal l y distributed (the l ogarithm ofthe parameter is normal l y
distributed);however, the unbounded characteristics of that distribution can
attribute a nonzero probabil ityto unreal istic val ues of the ground motion
parameter. For exampl e, a hypothetical PHA attenuationrel ationship that predicts a
mean PHA of 0.5g with crin y = 0.5 woul d impl y a 0.06%probabil ity that the PHA
woul d exceed 2.5g. The use of distributions that impose
an upperl imit on Yhave been studied by Kul karni et al . (1979), Bender (1984), and
Zemel l (1984).Sec. 4.4 Probabil istic Seismic Hazard Anal ysisYP[Y>y* I m, r]127r l og
RFigure 4.14 Schematic il l ustration of conditional probabil ity of exceeding a
particul arval ue of a ground motion parameter for a given magnitude and distance.
Exampl e 4.4Using the predictive rel ationship of Campbel l and Bozorgnia (1994) given
as equation (3.25),compute the probabil ity that aMw = 7 earthquake on a strike-sl ip
faul t woul d cause a peak horizontal accel eration greater than OAg at a soft-rock
site 15km from the cl osest point on the rupturesurface.Sol ution. For equation
(3.26), the parameters F = 0 and SHR = 0 and SSR = 1, so the meanval ue of the
natural l ogarithm of peak horizontal accel eration is given byl nPHA (gal ) = 3.512+
0.904Mw-1.328JR2 + [0. 149exp(0.647Mw) ] 2+ 0.904 - 0.171 In R= 3.512 + (0.904) (7)
-1.328J152 + [0.149 exp(0.647 x 7)] 2+ 0.904 - 0.171 In (15)= 6.31from which PHA=
e6.31 = 552 gal . Then, from Section C.7.2 of Appendix C, the standard normal variate
isl nPHA-l nPHA z =From Tabl e C-1,In [(OAOg) (981 gal /g)] -In (552 gal ) = -0.84300405
Fz (-0.843) = P Iz< 0.843] P [a max < 0.4g] = 0.200so the desired probabil ityP [PHA
> OAOgl M = 7.0, R = 15 km] 1 - 0.200 = 0.8004.4.3 Temporal UncertaintyTo cal cul ate
the probabil ities of various hazards occurring in a given time period, the
distributionof earthquake occurrence with respect to time must be considered.
Earthquakeshave l ong been assumed to occur randoml y with time, and in fact,
examination of avail abl eseismicity records has reveal ed l ittl e evidence (when
aftershocks are removed) of temporal patterns in earthquake recurrence. The
assumption of random occurrence al l ows the use ofsimpl e probabil ity model s, but is
inconsistent with the impl ications of el astic rebound theory(Section 2.5.1).128
Seismic Hazard Anal ysis Chap. 44.4.3.1 Poisson Model The temporal occurrence of
earthquakes is most commonl y described by a Poissonmodel . The Poisson model
provides a simpl e framework for eval uating probabil ities ofevents that fol l ow a
Poisson process, one that yiel ds val ues of a random variabl e describingthe number
of occurrences of a particul ar event during it given time interval or in a
specifiedspatial region. Since PSHAs deal with temporal uncertainty, the spatial
appl ications of thePoisson model wil l not be considered further. Poisson processes
possess the fol l owingproperties: 1. The number of occurrences in one time interval
are independent of the number thatoccur in any other time interval .2. The
probabil ity of occurrence during a very short time interval is proportional to the
l ength of the time interval .3. The probabil ity of more than one occurrence during a
very short time interval isnegl igibl e.These properties indicate that the events of
a Poisson process occur randoml y, with no"memory" of the time, size, or l ocation of
any preceding event.For a Poisson process, the probabil ity of a random variabl e N,
representing the numberof occurrences of a particul ar event during a given time
interval is given byn -f!P[N=n] = ~n!(4.14)(4.15)(4.16)where I..l is the average
number of occurrences of the event in that time interval . The timebetween events in
a Poisson process can be shown to be exponential l y distributed. To characterizethe
temporal distribution of earthquake recurrence for PSHA purposes, the Poisson
probabil ity is usual l y expressed asP [N = n] = (Atte-Atn!where Ais the average rate
of occurrence of the event and t is the time period of interest.Note that the
probabil ity of occurrence of at l east one event in a period of time t is given by
P[N~l ] = P[N= 1] +P[N=2] +P[N=3] + ...+ P [N =00] = 1 - P [N =0] = 1 _ e-AtWhen the
event of interest is the exceedance of a particul ar earthquake magnitude, the
Poissonmodel can be combined with a suitabl e recurrence l aw to predict the
probabil ity of atl east one exceedance in a period of t years by the expression
(4.17)4.4.3.2 Other Model sEl astic rebound theory suggests that the occurrence of
earthquakes on a particul ar faul tor faul t segment shoul d not be independent of past
seismicity. If earthquakes occur to rel easestrain energy that buil ds up over
extended periods of time, the occurrence of a l arge earthquakeshoul d substantial l y
reduce the chances of another independent, l arge earthquakeSec. 4.4 Probabil istic
Seismic Hazard Anal ysis 129(from the same source) occurring shortl y thereafter. If
earthquakes are triggered when thestress on a faul t reaches some l imiting val ue,
the chances of occurrence shoul d depend on thetimes, sizes, and l ocations of
preceding events.A number of model s that account for prior seismicity have been
proposed (Anagnosand Kiremidjian, 1988). Nonhomogeneous Poisson model s (e.g., Vere-
Jones and Ozaki,1982) al l ow the annual rate of exceedance to vary with time.
Renewal model s (Esteva, 1970;Hagiwara, 1974; Savy et al ., 1980; Kiremidjian and
Anagnos, 1984; Cornel l and Winterstein,1986) use arrival -time distributions other
than exponential (impl ied by the homogeneousPoisson model ) to al l ow the hazard rate
to increase with time since the l ast event;gamma and Weibul l distributions are most
common. Time-predictabl e model s specify a distributionof the time to the next
earthquake that depends on the magnitude of the most recentearthquake; sl ip-
predictabl e model s consider the distribution of earthquake magnitude todepend on
the time since the most recent earthquake. Markov model s incorporate a type of
memory that describes the chances that a process moves from some past "state" to a
particul arfuture state. The time for which the process stays in a particul ar state
before moving toanother state is exponential l y distributed; semi-Markov model s are
not restricted to the exponential distribution. Both Markov model s (Vere-Jones,
1966; Vagl iente, 1973; Venezianoand Cornel l , 1974; Nishioka and Shah, 1980) and
semi-Markov model s (Patwardhan et al .,1980; Cl uff et al ., 1980; Coppersmith, 1981;
Guagenti-Grandori and Mol ina, 1984) havebeen used in seismic hazard anal ysis. The
semi-Markov model s of Patwardhan et a1. (1980)and Cl uff et a1. (1980), for exampl e,
rel ate the probabil ity of future earthquakes of varioussizes to the size of the
most recent event and the el apsed time since its occurrence. Triggermodel s (Vere-
Jones and Davies, 1966; Shl ien and Tokosz, 1970; Merz and Cornel l , 1973b;Lai, 1977)
can account for cl usters of events (aftershocks) that occur after triggering
events.4.4.3.3 Model Appl icabil ityInvestigations of the appl icabil ity of Poisson
and non-Poissonian model s (Cornel l and Winterstein, 1986) have shown that the
Poisson model is useful for practical seismicrisk anal ysis except when the seismic
hazard is dominated by a singl e source for which thetime interval since the
previous significant event is greater than the average interevent timeand when the
source displ ays strong "characteristic-time" behavior. For this and other reasons
rel ated to simpl icity, ease of use, and l ack of sufficient data to support more
sophisticatedmodel s, the Poisson model is the most widel y used in contemporary
PSHA.Each of the more sophisticated model s uses a "pattern" of earthquake
occurrence toreconcil e their computed probabil ities with the mechanics of the
el astic rebound process ofearthquake generations. As a resul t, each requires
additional parameters whose val ues mustbe eval uated from historical and
instrumental seismicity records that are, in most cases, toosparse to permit
accurate eval uation. As time passes and additional data becomes avail abl e,the use
of these model s wil l undoubtedl y increase..tl ~.4.4 Probabil ity ComputationsThe
resul ts of a PSHA can be expressed in many different ways. Al l invol ve some l evel
ofprobabil istic computations to combine the uncertainties in earthquake size,
l ocation, frequency,and effects to estimate seismic hazards. A common approach
invol ves the devel opmentof seismic hazard curves, which indicate the annual
probabil ity of exceedance of130 Seismic Hazard Anal ysis Chap. 4(4.18)(4.21)
different val ues of a sel ected ground motion parameter. The seismic hazard curves
can thenbe used to compute the probabil ity of exceeding the sel ected ground motion
parameter in aspecified period of time.4.4.4.1 Seismic Hazard CurvesSeismic hazard
curves can be obtained for individual source zones and combined toexpress the
aggregate hazard at a particul ar site. The basic concept of the computations
required for devel opment of seismic hazard curves is fairl y simpl e. The probabil ity
ofexceeding a particul ar val ue, y*, of a ground motion parameter, Y, is cal cul ated
for one possibl eearthquake at one possibl e source l ocation and then mul tipl ied by
the probabil ity thatthat particul ar magnitude earthquake woul d occur at that
particul ar l ocation. The process isthen repeated for al l possibl e magnitudes and
l ocations with the probabil ities of eachsummed. The required cal cul ations are
described in the fol l owing paragraphs.For a given earthquake occurrence, the
probabil ity that a ground motion parameter Ywil l exceed a particul ar val ue y* can
be computed using the total probabil ity theorem, that is,P[Y>/] = p[Y>/IX]p[X] =
fp[y>/rX]l x(X)dXwhere X is a vector of random variabl es that infl uence Y. In most
cases the quantities in Xare l imited to the magnitude, M, and distance, R. Assuming
that M and R are independent,the probabil ity of exceedance can be written asP[Y>/]
= ffp[Y>/jm,r]IM(m)!R(r)dmdr (4.19)where P [Y> y* 1m, r] is obtained from the
predictive rel ationship andIM(m) andIR(r) are theprobabil ity density functions for
magnitude and distance, respectivel y.If the site of interest is in a region of Ns
potential earthquake sources, each of whichhas an average rate of threshol d
magnitude exceedance, Vi
[= expro, - ~imO)], the total averageexceedance rate for the region wil l be given
byNsAy* = L vJf P [Y> /Im, r] IMi (m) IRi (r) dm dr (4.20)i= IThe individual
components of equation (4.20) are, for virtual l y al l real istic PSHAs,sufficientl y
compl icated that the integral s cannot be eval uated anal ytical l y. Numerical
integration,which can be performed by a variety of different techniques, is
therefore required.One approach, used here for simpl icity rather than efficiency,
is to divide the possibl eranges of magnitude and distance into NM and NR segments,
respectivel y. The averageexceedance rate can then be estimated byNs NM NRi; = L L L
viP [Y> /Imj, rk]IMi (m) IRi (rk) Am Sri=Ij=l k=]where mj= mo+ (j - 0.5)(mmax-
mo)/NM, rk= rmin + (k - 0.5)(rmax - rmin)/NR,Am = (mmax- mo)/NI11, and Ar= (rmax-
rmin)/NR. This is equival ent to assuming that each source is capabl eof generating
onl y NM different earthquakes of magnitude, mj, at onl y NR differentsource-to-site
distances, rk' Equation (4.21) is then equival ent toSec. 4.4 Probabil istic Seismic
Hazard Anal ysis 131Ns NM NRAy * ""L L LviP [Y> /Imj' rk] P [M = mj] P [R = rk]
(4.22)i=l j=l k=l The accuracy of the crude numerical integration procedure described
above increases withincreasing NM and NR' More refined methods of numerical
integration wil l provide greateraccuracy at the same val ues of NM and NR• Exampl e
4.5The basic procedures of a typical PSHA can be il l ustrated for the site shown in
Figure 4-6 if therecurrence rel ationships for each of the source zones is known.
Assuming that the seismicity ofthe respective source zones are described bySource
zone I: l og Am 4.4 - l .OMSource zone 2: l ogAm 3.5 -0.8MSource zone 3: l og Am 2.7
-l .2Mthe PSHA can be performed in the four previousl y described steps: 1. The
probl em statement provides the l ocation, geometry, and maximum magnitude of each
source zone. The distribution of source-to-site distance must al so be
characterized. To l imitthe number of computations invol ved in this simpl e exampl e,
we wil l characterize the distributionof source-to-site distance by a rel ativel y
coarse histogram. Consider first sourcezone I. It is a simpl e matter to show that
the shortest possibl e source-to-site distance wil l be23.72 km and that the l ongest
wil l be 90.12 km. We can divide this total range into 10 distanceinterval s ofl ength
(90.12 km - 23.72 km)/IO = 6.64 km. If we divide the source zoneinto a l arge number
of segments of equal l ength, we can characterize the distribution ofsource-to-site
distance by determining how many of the segments fal l within each distanceinterval .
For 1000 segments, the normal ized histogram of source-to-site distance is shownin
Figure E4.5a. The ordinates of the normal ized histogram represent the rel ative
frequencythat woul d be equal to the probabil ity if an infinite number of segments
were used, butwhich is an approximation to the probabil ity in this case. The
probabil ity that the source-tositedistance is between 23.72 and 30.36 km (or about
equal to the midpoint of that range,27.04 km) is approximatel y 0.336. For source
zone 2, the source-to-site range 25 to 125 kmcan be divided into 10 interval s of
IO-km l ength; dividing the areal source zone into 2500el ements of equal area, the
normal ized histogram of Figure E4.5b is obtained. Since thereis onl y one possibl e
source-to-site distance, obtaining the normal ized histogram of FigureE4.5c for
source zone 3 is trivial matter.0000000000T"""C\JC": )VLO(of'-.COO>~0.0 YI--I--+-+-
+'''''I--+-I--+----10000000000(f)Vl !)COI""'--OOO'l ~;: : : ~0.0 0.00.4 0.21.0';: : "
';: : " ';: : "0": 0": 0": Q Q QEpicentral distance, r{km) Epicentral distance, r{km)
Epicentral distance, r{km)00 ~ ~Figure E4.5a--c Approximations to source-to-site
probabil ity distributions for sourcezones (a)l , (b)2, and (c)3.132 Seismic Hazard
Anal ysis Chap. 42. The temporal distribution of earthquake recurrence can be
characterized using the recurrencerel ationships provided in the probl em statement.
Assuming that earthquakes ofmagnitude l ess than 4.0 do not contribute to the
seismic hazard, the mean rates of exceedanceof magnitude 4.0 events from each of
the source zones areSource zone I: VI104.4 - 1.0 (4.0) = 2.512Source zone 2: V2
103.5 - 0.8 (4.0) = 1.995Source zone 3: V3 = 102.7 - 1.2 (4.0) = 0.008giving Vtol al
= 4.515. For each source zone, the probabil ity that the magnitude wil l bewithin an
interval between a l ower bound m/ and an upper bound mu is given bywhere!M(m) is
given in equation (4.12). IfNM= 10, the l owest magnitude interval forsource zone I
wil l be from M =4.0 to M =4.33. The probabil ity that the magnitudewoul d fal l within
that interval woul d be2.303e-2.303 (4.165-4.0)P[4.0<M<4.33] '" l _e-2.303(7.3-4.0)
(4.33-4.0) = 0.522The probabil ities of various magnitudes for each source zone are
as shown in FigureE4.5d-f.3. To compare the resul ts of this PSHA with those from
the DSHA exampl e, we wil l use thesame predictive rel ationship: that is, the Cornel l
et al . (1979) rel ationshipInPHA (gal s) = 6.74 + 0.859M - 1.80l n (R + 25)Uncertainty
in this rel ationship is expressed by the standard deviation crl ny = 0.57.4.
Final l y, we compute the total seismic hazard as the sum of the contributions from
each possibl ecombination of source-to-site distance and earthquake magnitude on
each of the threesource zones. First, we consider source zone I. For the l owest
magnitude interval (j =I),P [M =ml ] = P [M =4.165] = 0.522as computed in step 2.
For the l owest distance interval (k = I),P [R = rd P [R = 27.04 km] = 0.336
0.8.,---------,0.70.6"E 0.5~ 0.4 cr 0.30.20.10.0J"-.,orntOO)C\ILOCOT'""..;t...-
LOCO...-vOO ..... o: : : tco...-...t~~L6u;l l icDccicor-..: Magnitude, m(d)
0.8.,----------,0.70.6"E 0.5~ 0.4 cr 0.30.20.10.0O')(Q(Y')OJ"-.,v..-COLOC\Ir-LQ 0'>
(I') co o-.: r t---...-LO..,f~~triu1« : i<6cOr--: r-..: Magnitude, m(e)0.8
-,-----------,0.70.6"E 0.5~ 0.4 cr 0.30.20.10.0LOI.OLOLOLOLOLOLOLOI.O
O.,..C\J(I")">t"LOCOI'--COm~ ..¢ ..¢ ...; ..¢ ~ ..¢ ..¢ ..¢ ..¢Magnitude, m(1)
Figure E4.5d-f Approximations to magnitude probabil ity distributions for source
zones(d)l , (e)2, and (f)3.Sec. 4.4 Probabil istic Seismic Hazard Anal ysis 133as
computed in step I. This combination of magnitude and distance indicates an
expectedval ue of In PHA ofInPHA = 3.204Now we can cal cul ate the probabil ities that
various target peak accel eration l evel s wil l be exceeded. For a* = O.Ol g (9.81
gal s), the corresponding standard normal variabl e isz" = Ina* - Ii.l PHA = In (9.81)
- 3.204 = -1.63Giny 0.57Then the probabil ity that the peak accel eration is greater
than a.01g, using Tabl e 4-2, isP [PHA > O.Ol gIM = 4.165, R = 27.04 km] P l z" >
-1.63]1- P l z"< -1.63]1- Fz (-1.63)= 0.9484Annual rate of exceedance of a peak
accel eration of O.Ol gby an earthquake of magnitude 4.165at a distance of 27.04 km
on source zone I (given that an earthquake ofM>maoccurs on sourcezone 1) wil l be
Aaal g = VjP [PHA > O.Ol gIM =4.165, R =27.04 km]x P [M = 4.165] P [R = 27.04]2.512
(0.9484) (0.522) (0.336)0.4181If the preceding cal cul ations are repeated for the 99
other possibl e combinations of magnitudeand distance for source zone I, the
contributions of each wil l beDistance (km)Magnitude 27.04 33.68 40.32 46.96 53.60
60.24 66.88 73.52 80.16 86.804.165 0.4181 0.1501 0.0805 0.0676 0.0559 0.0456 0.0369
0.0293 0.0231 0.01794.495 0.2027 0.0753 0.0424 0.0376 0.0331 0.0289 0.0250 0.0213
0.0181 0.01494.825 0.0957 0.0363 0.0209 0.1092 0.0177 0.0162 0.0148 0.0134 0.0120
0.01065.155 0.0452 0.0172 0.0100 0.0093 0.0088 0.0084 0.0079 0.0074 0.0070 0.0065
5.485 0.0212 0.0081 0.0047 0.0045 0.0042 0.0041 0.0040 0.0038 0.0037 0.00355.815
0.0099 0.0037 0.0022 0.0021 0.0020 0.0019 0.0019 0.0018 0.0018 0.00186.145 0.0045
0.0018 0.0010 0.0010 0.0010 0.0010 0.0009 0.0009 0.0009 0.00096.475 0.0023 0.0009
0.0005 0.0005 0.0005 0.0005 0.0005 0.0004 0,0004 0.00046.805 0.0011 0.0004 0.0003
0.0002 0.0002 0.0002 0.0002 0.0002 0.0002 0.00027.135 0.0005 0.0002 0.0001 0.0001
0.0001 0.0001 0.0001 0.0001 0.0001 0.0001Summing al l of these contributions
indicates that the mean annual rate at which an accel erationof O.Ol g wil l be
exceeded by an earthquake on source zone 1 wil l be 1.923. Repeating al l ofthese
cal cul ations for the other source zones yiel ds equival ent exceedance rates of 1.016
forsource zone 2 and 0.005 for source zone 3. Consequentl y, the probabil ity that a
target accel erationofO.Ol g wil l be exceeded by an earthquake ofM> rna on any of the
three source zones wil l be 1.923 + 1.016 + 0.005 =2.944. This impl ies a return
period of 0.34 year for this l ow accel eration.By repeating this process for
different target accel erations, the seismic hazard curves ofFigure E4.5g can be
devel oped.Seismic Hazard Anal ysisSource zone 2Source zone 3Chap. 40.7 0.8Source
zone 10.5 0.6Al l source zones0.1 0.2 0.3 0.4134101Cl : 100 : r-D. 0 10-1IIIUc: Cl l 10-
2't: l IIIIIIU 10-3 ><III'0 10-4III'.l .i.iCO 10-5: : : l c: c: 10-6 Cl l c: Cl l III 10-7 : a:
10-80.0Peak horizontal accel eration (PHA)Figure E4.Sg Seismichazardcurvesfor
sourcezones 1, 2, and 3 and total seismichazardcurvefor al l sources.4.4.4.2 Finite
Time PeriodsThe seismic hazard curve can easil y be combined with the Poisson model
to estimateprobabil ities of exceedance in finite time interval s. From equation
(4.17), the probabil ity ofexceedance of y* in a time period Tis(4.23)Exampl e 4.6
Returning to Exampl e 4.5, the probabil ity that an accel eration of O.l Og woul d be
exceeded in a30-year period woul d beP[PHA>0.l Ogin30years] = I_e-A,T = l _e-co.o822)
(30) = 0.915 = 91.5%It is often necessary to compute the val ue of a ground motion
parameter corresponding to a particul arprobabil ity of exceedance in a given time
period. For exampl e, the accel eration l evel thathas a 10% probabil ity of exceedance
in a 50-year
period woul d be that with an annual rate ofexceedance, obtained by rearranging
equation (4.23), ofIn(I-P[YT>y*])TIn(l 5~0.1) = 0.00211From the total seismic hazard
curve of Figure 4.18, that accel eration l evel woul d be approximatel y0.63g.These
types of anal yses have been performed for a variety of seismical l y active areas
within the United States. As the exposure time, T, increases, the probabil ity of
exceeding aparticul ar ground motion parameter val ue al so increases. Simil arl y, the
val ue of a groundmotion parameter with a particul ar probabil ity of exceedance
increases with increasingexposure time. Figure 4.15 il l ustrates the peak
accel eration with a 10% probabil ity ofSec. 4.4 Probabil istic Seismic Hazard
Anal ysis 1351.0.------------------,San Francisco0.8sl : 0 : ;: : : 0.6.C..l l GIGel o«
0.4~Cl l GID.0.20.0010--n St. Louis_ - --0.05\0 - - - - - - : : - - - -~.r-v,--_- - -
- - AIl JuquerqueEI P~Q.. - - - - -- - - - - White Sands, NM50 100 150 200 250
Exposure time (yrs)Figure 4.15 Peak horizontal bedrockaccel erations with 10%
probabil ity ofexceedance over various exposure timesfor 14 areas in North America.
(AfterNEHRP,1993.)(4.24)exceedance for a number of metropol itan areas within the
United States. Seismic hazardmaps, such as that shown in Figure 4.16, have been
devel oped to express the rel ative seismicityof different regions in buil ding codes
(Chapter 8).4.4.4.3 D~aggregationThe PSHA procedures described in the preceding
sections al l ow computation of themean annual rate of exceedance at a particul ar
site based on the aggregate risk from potential earthquakes of many different
magnitudes occurring at many different source-site distances.The rate of exceedance
computed in a PSHA, therefore, is not associated with anyparticul ar earthquake
magnitude or source-site distance.In some cases, however, it may be useful to
estimate the most l ikel y earthquake magnitudeand/or the most l ikel y source-site
distance. These quantities may be used, for exampl e,to sel ect existing ground
motion records (recorded in earthquakes of simil ar magnitudeat simil ar source-site
distance) for response anal yses. This process of deaggregationrequires that the
mean annual rate of exceedance be expressed as a function of magnitudeand/or
distance. Computational l y, this simpl y invol ves the removal of terms from the
summationsof Equation 4.22. For exampl e, the mean annual rate of exceedance can be
expressed as a function of magnitude byNs NRAy* (m) "" P [M =mj] L L ViP [Y> l l mj'
rd P [R = rk]i = l k = ISimil arl y, the mean annual rate of exceedance can be
expressed as a function of source-sitedistance byNs NMAy* (rk) "" P [R =rk] L LViP
[Y> l l mj' rk] P [M =mj]i = 1j = 1(4.25).. 10\1Cl 'Figure 4.16 Contours of mean
horizontal accel eration on rock (expressed as a percentage of gravity) with I0%
probabil ity of exceedance in 50 years.(After Al germissen et aI., 1990.)Sec. 4.4
Probabil istic Seismic Hazard Anal ysis 137Final l y, it is possibl e to compute the
mean annual rate of exceedance as functions of bothearthquake magnitude and source-
site distance, i.e.Ns\* (m;, rk) '" P [M = m;J P [R = r k] LV,P [Y> l im;, rk]i = 1
(4.26)4.4.4.4 Logic Tree MethodsThe probabil ity computations described previousl y
al l ow systematic consideration ofuncertainty in the val ues of the parameters of a
particul ar seismic hazard model . In some cases,however, the best choices for
el ements of the seismic hazard model itsel f may not be cl ear. Theuse of l ogic trees
(Power et al ., 1981; Kul kami et al ., 1984; Coppersmith and Youngs, 1986)provides a
convenient framework for the expl icit treatment of model uncertainty.The l ogic tree
approach al l ows the use of al ternative model s, each of which isassigned a weighting
factor that is interpreted as the rel ative l ikel ihood of that model beingcorrect.
It consists of a series of nodes, representing points at which model s are specified
and branches that represent the different model s specified at each node. The sum of
theprobabil ities of al l branches connected to a given node must be 1. The simpl e
l ogic treeshown in Figure 4.17 al l ows uncertainty in sel ection of model s for
attenuation, magnitudedistribution, and maximum magnitude to be considered. In this
l ogic tree, attenuationaccording to the model s of Campbel l and Bozorgnia (1994) and
Boore et al . (1993) are consideredequal l y l ikel y to be correct, hence each is
assigned a rel ative l ikel ihood of 0.5. Proceedingto the next l evel of nodes, the
Gutenberg-Richter magnitude distribution isconsidered to be 50% more l ikel y to be
correct than the characteristic earthquake distribution.At the final l evel of
nodes, different rel ative l ikel ihoods are assigned to the maximummagnitude. This
l ogic tree terminates with a total of 2 x 2 x 3 = 12 (no. of attenuationAttenuation
model MagnitudedistributionMaximummagnitude7 ~ (0.3)7 ~ (0.6)8 (0.1)7 ~ (0.3)7 ~
(0.6)8 (0.1)7 ~ (0.3)7 ~ (0.6)8 (0.1)7 ~ (0.3)7 ~ (0.6)8 (0.1)Figure 4.17 Simpl e
l ogic tree for incorporation of model uncertainty.138 Seismic Hazard Anal ysis Chap.
4model s x no. of magnitude distributions x no. of maximum magnitudes) branches. The
rel ativel ikel ihood of the combination of model s and/or parameters impl ied by each
terminal branch is given by the product of the rel ative l ikel ihood of the terminal
branch and al l priorbranches l eading to it. Hence the rel ative l ikel ihood of the
combination of the Campbel l attenuation model , Gutenberg-Richter magnitude
distribution, and maximum magnitude of7.5 is 0.5 x 0.6 x 0.3 = 0.09. The sum of the
rel ative l ikel ihoods of the terminal branches,or of those at any prior l evel , is
equal to 1.To use the l ogic tree, a seismic hazard anal ysis is carried out for the
combination ofmodel s and/or parameters associated with each terminal branch. The
resul t of each anal ysisis weighted by the rel ative l ikel ihood of its combination of
branches, with the final resul ttaken as the sum of the weighted individual resul ts.
It is easy to see that the required computational effort increases quickl y with
increasingnumbers of nodes and branches. Parameters best characterized by
continuous distributions(e.g., the maximum magnitude in the exampl e of Figure 4.17)
are difficul t to treat inthe l ogic tree without resorting to l arge numbers of
branches. Neverthel ess, the l ogic tree isa very useful tool for the anal ysis of
seismic hazards.4.5 SUMMARY1. Earthquake-resistant design seeks to produce
structures that can withstand a certainl evel of shaking without excessive damage.
That l evel of shaking is described by adesign ground motion which is usual l y
determined with the aid of a seismic hazardanal ysis.2. Seismic hazard anal yses
invol ve the quantitative estimation of ground motioncharacteristics at a particul ar
site. They may be conducted deterministical l y orprobabil istical l y.3. Seismic hazard
anal yses require the identification and characterization of al l potential sources of
seismic activity that coul d produce significant ground motions at the site of
interest. Earthquake sources may be identified on the basis of geol ogic, tectonic,
historical ,and instrumental evidence.4. Deterministic seismic hazard anal yses
invol ve the assumption of some scenario-theoccurrence of an earthquake of a
particul ar size at a particul ar l ocation-for whichground motion characteristics are
determined. In practice, DSHAs often assume thatearthquakes of the l argest possibl e
magnitude occur at the shortest possibl e distanceto the site within each source
zone. The earthquake that produces the most severe sitemotion is then used to
compute site-specific ground motion parameters.S. When appl ied to structures for
which fail ure coul d have catastrophic consequences,such as nucl ear power pl ants and
l arge dams, DSHA provides a straightforwardframework for eval uation of "worst-case"
ground motions. However, it provides noinformation on the l ikel ihood of occurrence
of the control l ing earthquake, the l ikel ihoodof it occurring where it is assumed to
occur, the l evel of shaking that might beexpected during a finite period of time
(such as the useful l ifetime of a particul arstructure or facil ity), or the effects
of uncertainties in the various steps required tocompute the resul ting ground
motion characteristics.Chap. 4 Homework Probl ems 1396. Probabil istic seismic hazard
anal yses al l ow uncertainties in the size, l ocation, rate ofrecurrence, and effects
of earthquakes to be expl icitl y considered in the eval uation ofseismic hazards. A
PSHA requires that uncertainties in earthquake l ocation, size,recurrence, and
ground shaking effects be quantified.7. For each source zone, uncertainty in
earthquake l ocation is characterized by a probabil itydensity function of source-to-
site distance. Eval uation of the probabil ity densityfunction requires estimation of
the geometry of the source zone and of thedistribution of earthquakes within it.8.
Uncertainty in the sizes of earthquakes produced by each source zone can be
described by various recurrence l aws. The Gutenberg-Richter recurrence l aw, which
assumes an exponential distribution of magnitude, is commonl y used with
modificationsto account for minimum and maximum magnitudes. A more recent
interpretationof faul t activity has produced the characteristic earthquake
recurrence l aw.Which of these two recurrence l aws is more correct has not yet been
determined; it isnot uncommon to incorporate both into a PSHA by means of a l ogic
tree.9. The l evel of shaking produced by an earthquake of a given size occurring at
a givensource-to-site distance is determined from predictive rel ationships. The
uncertainty inthese ground motions is a function of the scatter in the databases
from which the predictiverel ationships were devel oped.10. The probabil ities of
earthquakes of various sizes occurring in finite
periods of time areusual l y computed assuming that earthquakes occur as Poisson
processes. Al though thePoisson model assumes an independence of events that is not
consistent with el asticrebound theory, it remains the most commonl y used model in
contemporary PSHA.11. Standard methods of probabil ity anal ysis can be used to
combine the quantified uncertaintiesin earthquake size, l ocation, recurrence, and
effects to compute ground motionl evel s with various probabil ities of exceedance in
different periods of time. Because ofthe compl ex and empirical nature of the
probabil ity density functions, exceedanceprobabil ities are usual l y computed by
numerical , rather than anal ytical , methods.12. The accuracy of a PSHA depends on
the accuracy with which uncertainty in earthquakesize, l ocation, recurrence, and
effects can be characterized. Al though model sand procedures for characterization of
uncertainty of these parameters are avail abl e,they may be based on data col l ected
over periods of time that, geol ogical l y, are veryshort. Engineering judgment must
be appl ied to the interpretation of PSHA resul ts.13. Model uncertainties can be
incorporated into a PSHA by means of a l ogic tree. A l ogictree al l ows the use of
al ternative model s, each of which is assigned a weighting factorrel ated to the
l ikel ihood of that model being correct. The weighting factors are usual l yassigned
subjectivel y, often using expert opinion.HOMEWORK PROBLEMS4.1 Estimate the surface
rupture l engths, rupture areas, and maximum surface displ acements forearthquakes of
Mw =6 and Mw =8.140 Seismic Hazard Anal ysis Chap. 4(0,40)(0,20)(20,30) Faul t B
(50,30)Faul t C(50,20)---(62,20)Site • (60,15)(70,50)(70,40)(10,10)(40,0)Figure P4.2
4.2 The site shown in Figure P4.2 is l ocated near four active faul ts in Cal ifornia.
Faul t A is anormal faul t. Faul ts Band C are strike-sl ip faul ts, and Faul t D is a
reverse faul t. The coordinatesof the site and faul ts shown above are in krn.
Assuming that onl y l inear segments can rupturein an individual event, perform a
deterministic seismic hazard anal ysis to: (a) Determine the anticipated peak
accel eration at the site. Use the attenuation rel ationship(s)from Chapter 3 you
bel ieve are most appropriate and briefl y justify your sel ection.(b) Determine and
pl ot the anticipated response spectrum at the site.(c) Assuming that the site
consists of a bedrock outcrop, determine the anticipated duration ofstrong motion
at the site.(d) Estimate the anticipated Arias Intensity at the site.Site B- - - -
- - - - - - - - ~12 km30 km30 kmFigure P4.3+NIC11ap.4 Homework Probl ems 1414.3 The
hypothetical vertical faul t segment shown in Figure P4.3 is represented a s a
quarter-circl e.On the same graph, pl ot histograms of expected epicentral distance
for motions at Site A andSite B assuming: (a) Earthquakes are equal l y l ikel y to
occur at any point on the faul t segment.(b) Earthquakes are twice as l ikel y to
occur at the midpoint of the faul t segment as at either endand the l ikel ihood is
l inearl y distributed between the midpoint and the ends.4.4 In a hypothetical
seismical l y active region, earthquakes have been recorded over an 80-yearperiod.
Part of the record is instrumental , but part is not. Combining al l avail abl e data,
itappears that the earthquakes have been distributed as fol l ows: MOMENT MAGNITUDE3-4
4-55-6>6NUMBER OF EARTHQUAKES-1800-ISOII1(a) Estimate the Gutenberg-Richter
parameters for the region.(b) Negl ecting earthquakes of magnitude l ess than 3,
compute the probabil ity that an earthquakein the region wil l have a moment
magnitude between 5.5 and 6.5.(c) Repeat Part(b) assuming that pal eoseismic
evidence indicates that the region is notcapabl e of producing earthquakes of moment
magnitude greater that 6.5.4.5 The seismicity of a particul ar region is described
by the Gutenberg-Richter recurrence l aw: l ogAm = 4.0 - 0.7M(a) What is the
probabil ity that at l east one earthquake of magnitude greater than 7.0 wil l occur in
a l O-year period? In a 50-year period? In a 250-year period?(b) What is the
probabil ity that exactl y one earthquake of magnitude greater than 7.0 wil l occur in
a l O-year period? In a 50-year period? In a 250-year period?(c) Determine the
earthquake magnitude that woul d have a 10% probabil ity of beingexceeded at l east
once in a 50-year period.4.6 A hypothetical site is l ocated at the end of a
hypothetical 300 km l ong l inear feature known asHol tz's Faul t. Historical data
indicates that the seismicity of Hol tz's Faul t can be described bythe rel ationship
l ogAm = 3.0 - 0.75MPeak accel erations in the region of Hol tz's Faul t can be
described by the simpl e attenuationrel ationship of Campbel l (1981), even for
distances greater than 50 km.Given the above information: (a) Devel op a seismic
hazard curve for peak accel eration at the site.(b) Determine the probabil ity that
an accel eration of 0.25 g wil l be exceeded at l east once ina 100-year period.(c)
Determine the peak accel eration that woul d have a 10% probabil ity of being exceeded
atl east once in a 50-year period.142 Seismic Hazard Anal ysis Chap. 40.8 0,8~ 0.6
0.50 &' 0.6: 0 0.4 : 0 <1l <1l 0.4.0 .0 e 0.2 0.10 0.10 e 0. 0. 0.2 0.075 0.0250.0
0.010 20 30 40 4 5 6 7Distance (km) MagnitudeFigureP4.74.7 A new seismic source is
identified in the vicinity of the site described in Probl em 6. The geometryof the
source is such that the source-site distance is distributed as indicated by the
simpl ehistogram above. The source is expected to produce an earthquake of magnitude
4 or moreevery 5 years, on average, and the magnitudes are expected to be
distributed as indicated in thesimpl e histogram above. Assuming that earthquakes of
magnitude 4, 5, 6, or 7 can occur at distancesof 10, 20, 30, or 40 krn, and using
the attenuation rel ationship given in Probl em 6,(a) Devel op and pl ot a seismic
hazard curve for peak accel eration at the site that considers bothHol tz's Faul t and
the new seismic source. Pl ot the individual seismic hazard curves on thesame graph.
(b) Considering both sources, determine the probabil ity that an accel eration of
0.25 g wil l beexceeded at l east once in a 100-year period.(c) Considering both
sources, determine the peak accel eration that woul d have a 10% probabil ityof being
exceeded at l east once in a 50-year period.5Wave Propagation5.1 INT"~ODUCTIONIt is
the continuous nature of geol ogic material s that causes soil dynamics and
geotechnical earthquake engineering to diverge from their structural counterparts.
Whil e most structurescan readil y be ideal ized as assembl ages of discrete masses
with discrete sources of stiffness,geol ogic material s cannot. They must be treated
as continua, and their response to dynamicdisturbances must be described in the
context of wave propagation.Some basic concepts of wave propagation have been
al l uded to in previous chapters;a more fundamental treatment of the basic concepts
is presented in this chapter. The presentationfol l ows a repeated pattern of simpl e-
to-compl ex appl ications. The rel ativel y simpl eprobl em of waves in unbounded media
is fol l owed by the more compl icated probl em ofwaves in bounded and l ayered media.
Within each, the concepts are presented first for thesimpl e case of one-dimensional
wave propagation, and then for the more general threedimensional case. The careful
reader wil l note that the basic techniques and principl es usedto sol ve the more
compl icated cases are general l y the same as those used for the simpl ecases; the
additional compl exity simpl y resul ts from the need to consider more dimensions.143
1445.2 WAVES IN UNBOUNDED MEDIAWave Propagation Chap. 5(5.1)(5.2)The propagation of
stress waves is most easil y understood by first considering anunbounded, or
"infinite," medium [i.e., one that extends infinitel y in the direction(s) ofwave
propagation). A simpl e, one-dimensional ideal ization of an unbounded medium isthat
of an infinitel y l ong rod or bar. Using the basic requirements of equil ibrium of
forcesand compatibil ity of displ acements, and using strain-displ acement and stress-
strain rel ationships,a one-dimensional wave equation can be derived and sol ved. The
process can berepeated, using the same requirements and rel ationships, for the more
general case of wavepropagation in a medium that extends infinitel y in three
orthogonal directions.5.2.1 One-Dimensional Wave PropagationThree different types
of vibration can occur in a thin rod: l ongitudinal vibration duringwhich the axis
of the rod extends and contracts without l ateral displ acement; torsional vibration
in which the rod rotates about its axis without l ateral displ acement of the axis;
andfl exural vibration during which the axis itsel f moves l ateral l y. The fl exural
vibration probl emhas l ittl e appl ication in soil dynamics and wil l not be considered
further. For the firsttwo cases, however, the operative wave equations are easil y
derived and sol ved.5.2.1.1 Longitudinal Waves in an Infinitel y Long RodConsider the
free vibration of an infinitel y l ong, l inear el astic, constrained rod with cross-
sectional area A, Young's modul us E, Poisson's ratio v, and density p, as shown in
Figure 5.1. Ifthe rod is constrained against radial straining, then particl e
displ acements caused by a l ongitudinal wave must be paral l el to the axis of the rod.
Assume that cross-sectional pl anes wil l remain pl anar and that stresses wil l be
distributed uniforml y over each cross section. As astress wave travel s al ong the
rod and passes through the smal l el ement shown in Figure 5.2, theaxial stress at
the l eft end of the el ement (x = xo) is O"x . At the right end (x =Xo +dx), the
axial ostress is O"x + (dO"x Idx) dx. Then dynamic equil ibrium of the el ement
requires thato(dO"x) d2uo, + -=: ;-dx A - o, A = pA dxooX 0 at2where u is the
displ acement in the x-direction. This simpl y
states that the unbal anced external forces acting on the ends of the el ement [the
l eft side of Equation (5.1)] must equal theinertial force induced by accel eration
of the mass of the el ement (the right side). Simpl ifyingyiel ds the one-dimensional
equation ofmotiondO"x d2u~ = Pdt2pEvAFigure 5.1 Constrained, infinite rod for one-
dimensional wave propagation. Constraintagainst radial straining schematical l y
represented by rol l ers.Sec. 5.2 Waves in Unbounded Media 145Figure 5.2 Stresses and
displ acements at ends of el ement of l ength dx and crosssectional area, A.In this
form, the equation of motion is val id for any stress-strain behavior but cannot be
sol ved directl y because it mixes stresses [on the l eft side of equation (5.2)] with
displ acements(on the right side). To simpl ify the equation of motion, the l eft side
can be expressedin terms of displ acement by using the stress-strain rel ationship,
O'x =MEx, where the constrainedmodul us M = {(l -v)/[(l +v)(l -2v)]}E, and the strain-
displ acementrel ationship, Ex = au/ax. These substitutions al l ow the one-dimensional
equation of motionto be written in the famil iar form of the one-dimensional
l ongitudinal wave equation for aconstrained rod: (5.3)The one-dimensional wave
equation can be written in the al ternative forma2u = v2 a2uat2 p ax2(5.4)(5.5)where
vp is the wave propagation vel ocity; for this case, the wave travel s at vp = JM/ p.
Note that the wave propagation vel ocity depends onl y on the properties of the rod
material (its stiffness and density) and is independent of the ampl itude of the
stress wave. The wavepropagation vel ocity increases with increasing stiffness and
with decreasing density. Thewave propagation vel ocity is an extremel y important
material property that is rel ied uponheavil y in soil dynamics and geotechnical
earthquake engineering.The wave propagation vel ocity is the vel ocity at which a
stress wave woul d travel al ong the rod. It is not the same as the particl e vel ocity,
which is the vel ocity at which a singl l epoint within the rod woul d move as the wave
passes through it. Knowing that au = ExaX(from the strain-displ acement
rel ationship), Ex =O'x/M(from the stress-strain rel ationship),and ax=Vp at (from
the definition of wave propagation vel ocity), the particl e vel ocity u canbe shown
to beu = au = ExaX = O'xvpat = O'x v = ~v = ~at at M at M p pv~ p pvpEquation (5.5)
shows that the particl e vel ocity is proportional to the axial stress in the rod.The
coefficient of proportional ity, pVp, is cal l ed the specific impedance of the
material . Thespecific impedance is another important property that infl uences the
behavior of waves atboundaries (Section 5.4).Exampl eS.1Compute vp for steel ,
vul canized rubber, and water.146 Wave Propagation Chap. 5Sol ution The constrained
modul i and specific gravities of steel , rubber, and water can befound in a number
of reference books; typical val ues are summarized bel ow: MATERIALSteel Vul canized
RubberWaterSPECIFIC GRAVITY, SG7.851.21.0MIPSI)40.4 x 106167 X 1060.34 X 106With
this information, the vp-val ues can be cal cul ated from: For steel ,(40.4 X 106 psi)
( 144 in2fft2)(32.2 ft/sec2)(7.85)(62.4 pcf)For vul canized rubber,(167 X 106 psi)
( 144 in2fftz)(32.2 ft/sec2)( 1.20)(62.4 pcf)For water,19556 ft/sec101691 ft/sec
(5.6)(0.34 X 106 psi)( 144 in2fft2)(32.2 ft/sec2) = 5026 ft/sec( 1.00)( 62.4 pcf)
5.2.1.2 Torsional Waves in an Infinitel y Long RodTorsional waves invol ve rotation
of the rod about its own axis. In the case of the l ongitudinal wave, the direction
of particl e motion was paral l el to the direction of wave propagation.For torsional
waves, particl e motion is constrained to pl anes perpendicul ar to thedirection of
wave propagation. Devel opment of a wave equation for torsional vibrations,however,
fol l ows exactl y the same steps as for l ongitudinal vibration. Consider the short
segment of a cyl indrical rod shown in Figure 5.3 as a torsional wave of torque
ampl itude Ttravel s al ong the rod. Dynamic torsional equil ibrium requires that the
unbal anced external torque [l eft side of equation (5.6)] is equal to the inertial
torque (right side): ( aT ) aZe T; + -=: ;-dx - T, = pI dx- o ox 0 z atwhere I is the
pol ar moment of inertia of the rod about its axis. This equil ibrium equationcan be
simpl ified to produce the equation of motionaTaxNow, incorporating the torque-
rotation rel ationshipT = GIaeax(5.7)(5.8)where G is the shear modul us of the rod,
the torsional wave equation can be written asSec. 5.2 Waves in Unbounded Media
Figure 5.3 Torque and rotation at ends of el ement of l ength dx and cross-sectional
area, A.147Gl 2a = QGl 2a = v; Gl 2a (5.9)Gl t2 P Gl x2 Gl x2where Vs = -l CIP is the
vel ocity of propagation of the torsional wave. Note that the form ofthe wave
equation for torsional waves [equation (5.9)] is identical to that for l ongitudinal
waves [equation (5.3)], but the wave propagation vel ocities are different. The wave
propagationvel ocity depends both on the stiffness of the rod in the mode of
deformation induced bythe wave and on the material density but is independent of
the ampl itude of the stress wave.Exampl e 5.2Compute Vs for steel , vul canized
rubber, and water.Sol ution The shear modul i of steel , rubber, and water can be
found in a number of referencebooks; typical val ues are summarized bel ow: MATERIAL
Steel Vul canized RubberWaterGIPSI)11.5 x 1060.167 X 106oWith this information, the
vs-val ues can be cal cul ated fromFor steel ,(11.5 X 106 psi) (144 in2/ft2) (32.2
ft/sec 2)(7.85) (62.4 pef)For vul canized rubber,(0.167 X 106 psi) (144 in2/ft2)
(32.2 ft/sec 2)(1.20) (62.4 pef)For water,10434 ft/sec3216 ft/secVs = 0The l ast
resul t is obvious-as an inviscid fl uid, water can produce no resistance to shear
stressesand consequentl y cannot transmit torsional waves.148 Wave Propagation Chap.
55.2.1.3 Sol ution of the One-Dimensional Equation of MotionThe one-dimensional wave
equation is a partial differential equation of the formcPu cPu - = v2 - (5.10) at2
ax2where v represents the wave propagation vel ocity corresponding to the type of
stress waveof interest. The sol ution of such an equation can be written in the form
U(x,t) = j(vt-x)+g(vt+x) (5.11)wherejand g can be any arbitrary functions of (vt -
x) and (vt + x) that satisfy equation(5.10). Note that the argument ofjremains
constant when x increases with time (at vel ocityv), and the argument of g remains
constant when x decreases with time. Therefore, the sol utionof equation (5.l l )
describes a displ acement wave [fl : vt- x)] travel ing at vel ocity v in thepositive x-
direction and another [g(vt + x)] travel ing at the same speed in the negative
xdirection.It al so impl ies that the shapes of the waves do not change with position
or time.If the rod is subjected to some steady-state harmonic stress aCt) =ao cos
rotwhere aDis the stress wave ampl itude and Wis the circul ar frequency of the
appl ied l oading, the sol utioncan be expressed using the wave number, k = wiv , in
the formU(x,t) = Acos(rot-kx) + Bcos(rot+ kx) (5.12)Here the first and second terms
describe harmonic waves propagating in the positive andnegative x-directions,
respectivel y. The wave number is rel ated to the wavel ength, A, of themotion byA =
vI' = vj2n-=-vto2nk(5.13)where T is the period of the appl ied l oadingjnote that
wave number is to wavel ength ascircul ar frequency is to period) and 1 = 11 T . Note
that at a given frequency, the wavel engthincreases with increasing wave propagation
vel ocity. Equation (5.12) indicates thatthe displ acement varies harmonical l y with
respect to both time and position as il l ustrated inFigure 5.4. Equation 5.13 and
Figure 5.4 show that the wave number is to the wavel ength asthe circul ar frequency
is to the period of vibration. For a wave propagating in the positivex-direction
onl y (B =0), differentiating u(x, t) twice with respect to x and twice with respect
to t and substituting into the wave equation [equation (5.10)] givesuI_T=~_I(a)ux
Figure 5.4 Particl e displ acement (a) as function of time, and (b) as function of
positional ong the rod.Sec. 5.2 Waves in Unbounded Media 149_ro2A cos (rot - kx) =
_v2k2A cos (rot - kx) (5.14)which reduces to the identity ro = kv, thereby
verifying equation (5.12) as a sol ution to thewave equation.Using compl ex notation
(Appendix A), the equival ent form of the sol ution can bewritten asu(x, t) =
Cei(ffit-kx) + Dei(ffit+kx) (5.15)This form of the sol ution can be verified in the
same way as the trigonometric form.Exampl e 5.3Cal cul ate the wavel engths of harmonic
l ongitudinal and torsional waves travel ing al ong constrainedsteel and vul canized
rubber rods. Assume that the waves are harmonic at a frequencyof 10 Hz.Sol ution
Using equation 5.13 and the wave propagation vel ocities computed in Exampl es5.1 and
5.2,Longitudinal wavesTorsional wavesSteel : Vul canized rubber: Steel : Vul canized
rubber: A = ~ = 19556 ft/sec = 1956 ftf 10 sec!A=~ = 101691 fUsec = 10169 ftf 10
sec!A = ~ = 10434 fUsec = 1043 ftf 10 sec!A=~ =3216 fUsec =322 ftf 10 sec"!5: .2.2
Three-Dimensional Wave PropagationThe preceding discussion of wave propagation in
rods il l ustrates some of the basic principl esof wave propagation, but an infinite
rod is hardl y an adequate model for describing thepropagation of seismic waves
through the earth. Since the earth is three-dimensional andsources of seismic
energy are three-dimensional , seismic waves must be described in termsof three-
dimensional wave propagation.Derivations of three-dimensional equations of motion
fol l ow the same steps as thoseused for one-dimensional propagation; the equations
of motion are formul ated from equil ibriumconsiderations, stress-strain
rel ationships, and strain-displ acement rel ationships. Inthethree-dimensional case,
however, the various rel ationships are more
compl ex and the derivationmore cumbersome. Brief reviews of three-dimensional
stress and strain notation andthree-dimensional stress-strain behavior wil l precede
derivation of the equations of motion,5.2.2.1 Review of Stress NotationThe stress
at a point on some pl ane passing through a sol id does not usual l y act normal to that
pl ane but has both normal and shear components. Considering a smal l el ement withone
corner at the center of an x-y-z Cartesian coordinate system (Figure 5.5), a total
of ninecomponents of stress wil l act on its faces. These stresses are denoted by
axx, axy, axZ, and150 Wave Propagation Chap. 5dzxdyFigure 5.5 Stress notation for
el ement of dimensions dx by dy by dz: so on, where the first and second l etters in
the subscript describe the direction of the stressitsel f and the axis perpendicul ar
to the pl ane in which it acts. Thus (Jxx, (Jyy, and (Jzz, are normal stresses, whil e
the other six components represent shear stresses. Moment equil ibriumof the el ement
requires that(5.16)which means that onl y six independent components of stress are
required to define the stateof stress of the el ement compl etel y. In some
references, the notation (Jx, (Jy, (Jz, 'txy , 'tyZ' and'txz is used to describe
(Jxx, (Jyy, (Jzz, (Jxy, (Jyz' and (Jxz, respectivel y.5.2.2.2 Review of Strain
NotationComponents of strain are easil y visual ized by considering the two-
dimensional strainin the x-y pl ane shown in Figure 5.6.The point P, at coordinates
(xo, Yo), is at one comer ofthe infinitesimal el ement PQRS which has a square shape
before deformation. After deformation,the infinitesimal el ement has been displ aced,
distorted, and rotated into the shapeP/Q'R/S~ From Figure 5.6, tan <Xl = dvl dx and
tan <X2 = dul dy, where u and v represent displ acementsin the x- and y-directions,
respectivel y. The shear strain in the x-y pl ane is givenby Cxy = <Xl + <X2' For
smal l deformations, the angl es may be taken equal to their tangents sothat the
rel ationship between the shear strain and the displ acements is Cxy =dvl dx + dul dy.
The rotation of the el ement about the z-axis is given by Qz=(<X1 - « 2)12. Anal ogous
definitionscan be devel oped for the x-z and y-z pl anes. For the three-dimensional
case, thestrain-displ acement rel ationships are defined byc du dv dw xx dx Cyy =dy
czz dzdv + du dw +dv du+dw(5.17)Cxy = cYZ czx =dx dy dy dz dz dxSec. 5.2 Waves in
Unbounded Mediadu ..Q.----+---I--------,dydxFigure 5.6 Square el ement subjected to
pl ane strain deformation.R'151Rigid-body rotation about the X-, y-, and z-axes are
given by the rotation-displ acementrel ationshipsQ = !(dV _dU)z 2 dx dy(5.18)The
first three quantities, f xx, f yy, and fzz, represent the extensional and
compressional strain paral l el to the X-, y-, and z-axes, and are cal l ed normal
strains. The second three quantities,fxy, fyZ' and f zx' represent the components
of shear strain in the pl anes correspondingto their suffixes. These six quantities
are the components of strain that correspond to thedeformation at P. In some
references, the notation fx, f y, fz, '[xy. Yyz' and Yxz is used todescribe f xx, f
yy, f w fxy, f yZ' and f xz' respectivel y.5.2.2.3 Review .of Stress-Strain
Rel ationshipsStresses and strains are proportional in a l inear el astic body. The
stress-strain rel ationshipcan be described by Hooke's l aw, which can be written in
general ized form asO'xx Cl 1 f xx + C12fyy + Cl 3fzz + CI4fxy + Cl sfyz + CI6fzxO'yy =
C21f xx + C22 fyy + c23fzz + C24fxy + C2S fyz + C26 fzxO'zz C31f xx + C32 fyy +
C33fzz + C34fl y + C3S fyz + C36 fzx(5.19)O'xy = C41f xx + C42 fyy + C43fzz + C44
fxy + C4S fyz + C46fzxO'yz = CSI f xx + CS2 fyy + CS3fzz + CS4fxy + Cssfy Z+ CS6fzx
O'zx C61f xx + C62fyy + C63fzz + C64 fxy + C6S fyz + C66fzx152 Wave Propagation
Chap. 5where the 36 coefficients represent the el astic constants of the material .
The requirementthat the el astic strain energy must be a unique function of the
strain (which requires that cij =Cjifor al l i and j) reduces the number of
independent coefficients to 21. If the material is isotropic,the coefficients must
be independent of direction, so thatCl 2 C21 Cl 3 C3l = C23 = C32 = A~ ~ ~ ~ ~.~CII
C22 C33 A+2~and al l other constants are zero. Therefore, Hooke's l aw for an
isotropic, l inear, el asticmaterial al l ows al l components of stress and strain to be
expressed in terms of the two Lameconstants, Aand u: 0 xx AE + 2~£xx 0 xy = ~£xy0 yy
AE + 2~£yy 0 yz ~£yz (5.21)0 zz AE + 2~£zz 0 zx ~£zxwhere the vol umetric strain £
=£xx + £yy + £zz. Note that the symbol Ais used universal l yfor both Lame's constant
and for wavel ength; the context in which it is used shoul d make itsmeaning obvious.
For convenience, several other parameters are often used to describe the stress-
strainbehavior of isotropic, l inear, el astic material s, each of which can be
expressed in terms ofLame's constants. Some of the more common of these areYoung's
modul us: E=~(3A + 2~) (5.22a)A+~Bul k modul us: K A+ 2~ (5.22b)3Shear modul us: G=~
(5.22c)Poisson's ratio: A v = (5.22d)2 (A +~)Hooke's l aw for an isotropic, l inear,
el astic material can be expressed using any combinationof two of these parameters
and/or Lame's constants.5.2.2.4 Equations of Motion for a Three-Dimensional El astic
Sol idThe three-dimensional equations of motion for an el astic sol id are obtained
fromequil ibrium requirements in much the same way as for the one-dimensional rod,
except thatequil ibrium must be ensured in three perpendicul ar directions. Consider
the variation instress across an infinitesimal cube al igned with its sides paral l el
to the x-y-z axes shown inFigure 5.7. Assuming that the average stress on each face
of the cube is represented by thestress shown at the center of the face, the
resul tant forces acting in the x-, y-, and z-directionscan be eval uated. In the x-
direction, the unbal anced external forces must be bal anced by aninertial force in
that direction, so thatSec. 5.2 Waves in Unbounded Media 153dcrxz do xz + -----az z
dxxI + d_~_;_y dy• • • • ~.• .• • • • • • • • .............• • • • • • • ............• • • • ': ?.J------J~
ydxdcrxxa xx + ---ax dxFigure 5.7 Stresses in x-direction on infinitesimal cube.P
dx dy dZ~: ~ (O"xx + d~;XdX)dy dz - O"xx dy dz(dO"XY ) + O"xy+ dy dy dxdz-O"xydxdz+
( O"xz + d~;ZdZ)dX dy - O"xz dx dywhich simpl ifies tod2u dO"xx dO"xy dO"xzPdt2 = ~+
dY +azRepeating this operation in the y- and z-directions givesd2U dO"yx dO"yy
dO"yzPdt2 = dX + dY +az(5.23)(5.24)(5.24b)(5.24c)Equations (5.24) represent the
three-dimensional equations of motion of an el astic sol id.Note that these equations
of motion were derived sol el y on the basis of equil ibrium considerationsand thus
appl y to sol ids of any stress-strain behavior. To express these equations ofmotion
in terms of displ acements, it is again necessary to use a stress-strain
rel ationship anda strain-displ acement rel ationship. Using Hooke's l aw as devel oped
in Section 5.2.2.3, thefirst of the equations of motion [equation (5.24a)] can be
written in terms of strains as.154 Wave Propagation Chap. 5(5.25)(5.26a)
Substituting the strain-displ acement rel ationships()u ()v ()u ()w ()ut xx = ()x t
xy = ()x +()y t x z = ()x +()zinto equation (5.25) produces the desired equation of
motion in terms of displ acements: ()2u ()£ 2 P-2 = (A+J..L): \+J..LV u()t oxwhere the
Lapl acian operator V2 representsV2 = £+£+~()x 2 ()y2 ()Z2Repeating this process in
the y and z directions givesP~: ~ = (A+ J..L)~~ + J..LV 2v()2w ()-Parz = (A+J..L)
()~+J..LV2w(5.26b)(5.26c)5.2.2.5 Sol utions of the Three-Dimensional Equations of
MotionTogether, equations (5.26) represent the three-dimensional equations of
motion for anisotropic, l inear, el astic sol id. It turns out that these equations
can be manipul ated to producetwo wave equations. Consequentl y, onl y two types of
waves can travel through suchan unbounded sol id. The characteristics of each type
of wave wil l be reveal ed by theirrespective wave equations.The sol ution for the
first type of wave can be obtained by differentiating each of equations(5.26) with
respect to x, y, and z and adding the resul ts together to giveor()2-P~ =
(A+J..L)V2£+J..LV2£()tRearranging yiel ds the wave equation: : 12;: ; ~ 2u v = ~V2£()t
2 P(5.27)(5.28)Sec. 5.2 Waves in Unbounded Media 155(5.31)(5.32)Recal l ing that E is
the vol umetric strain (which describes deformations that invol ve noshearing or
rotation), this wave equation describes an irrotational , or dil atational , wave. It
indicates that a dil atational wave wil l propagate through the body at a vel ocityvp
= j"- +P2fJ. (5.29)This type of wave is commonl y known as a p-wave (or primary
wave) and vp is referred toas the p-wave vel ocity of the material . The general
nature of p-wave motion was il l ustratedin Figure 2.l a. Note that particl e
displ acements are paral l el to the direction of wave propagation,just as they were
in the constrained rod of Section 5.2.1.1. The l ongitudinal wavein the constrained
rod is actual l y a p-wave. Using equations (5.22c) and (5.22d), vp can bewritten in
terms of the shear modul us and Poisson's ratio asv = G(2 - 2v) (5.30)p p(l -2v)As v
approaches 0.5 (at which point the body becomes incompressibl e, i.e., infinitel y
stiffwith respect to dil atational deformations), vp approaches infinity.To obtain
the sol ution for the second type of wave, E is el iminated by differentiating
equation (5.26b) with respect to z and equation (5.26c) with respect to y, and
subtracting onefrom the other: l ...(dW _dV) = VZ(dW _ dV)PdtZ dy dZ fJ. dy dZ
Recal l ing the definition of rotation [equation (5.19)], equation (5.31) can be
written in theform of the wave equationdznx = I: !: vzndt Z P xwhich describes an
equivol uminal , or distortional wave, of rotation about the x-axis. Simil ar
expressions can be obtained by the same process for rotation
about the y- and z-axes.Equation (5.32) shows that a distortional wave wil l
propagate through the sol id at a vel ocity(5.33)This type of wave is commonl y known
as a s-wave (or shear wave) and vs is referred to asthe shear wave vel ocity of the
material . Note that the particl e motion is constrained to apl ane perpendicul ar to
the direction of wave propagation, just as it was in the case of the torsional wave
of Section 5.2.1.2. Consequentl y, the torsional wave represented a form of ans-
wave, The cl ose rel ationship between s-wave vel ocity and shear modul us is used to
advantage in many of the fiel d and l aboratory tests discussed in Chapter 6. The
general nature of s-wave motion was il l ustrated in Figure 2.1b.S-waves are often
divided into two types, or resol ved into two perpendicul ar components.SH-waves are
s-waves in which particl e motion occurs onl y in a horizontal pl ane.SV-waves are s-
waves whose particl e motion l ies in a vertical pl ane. A given s-wave witharbitrary
particl e motion can be represented as the vector sum of its SH and SV components.
156 Wave Propagation Chap. 5In summary, onl y two types of waves, known as body
waves, can exist in an unbounded(infinite) el astic sol id. P-waves invol ve no
rotation of the material they pass through andtravel at vel ocity, vp . S-waves
invol ve no vol ume change and travel at vel ocity, Vs' The vel ocitiesof p- and s-
waves depend on the stiffnesses of the sol id with respect to the types of
deformationinduced by each wave. Comparing the vel ocities [equations (5.30) and
(5.33)]~ = J2-2VVS1 _ 2v (5.34)the p-wave vel ocity can be seen to exceed the s-wave
vel ocity by an amount that depends onthe compressibil ity (as refl ected in Poisson's
ratio) of the body. For a typical Poisson's ratioof 0.3 for geol ogic material s, the
ratio vpl vs = 1.87.5.3 WAVES IN A SEMI-INFINITE BODYThe earth is obviousl y not an
infinite body-it is a very l arge sphere with an outer surfaceon which stresses
cannot exist. For near-surface earthquake engineering probl ems, the earthis often
ideal ized as a semi-infinite body with a pl anar free surface (the effects of the
earth'scurvature are negl ected). The boundary conditions associated with the free
surface al l owadditional sol utions to the equations of motion to be obtained. These
sol utions describewaves whose motion is concentrated in a shal l ow zone near the
free surface (i.e., surfacewaves). Since earthquake engineering is concerned with
the effects of earthquakes onhumans and their environment, which are l ocated on or
very near the earth's surface, andsince they attenuate with distance more sl owl y
than body waves, surface waves are veryimportant.Two types of surface waves are of
primary importance in earthquake engineering.One, the Rayl eigh wave, can be shown
to exist in a homogeneous, el astic hal f-space. Theother surface wave, the Love
wave, requires a surficial l ayer of l ower s-wave vel ocity thanthe underl ying hal f-
space. Other types of surface waves exist but are much l ess significantfrom an
earthquake engineering standpoint.5.3.1 Rayl eigh WavesWaves that exist near the
surface of a homogeneous el astic hal f-space were first investigatedby Rayl eigh
(1885) and are known to this date as Rayl eigh waves. To describe Rayl eighwaves,
consider a pl ane wave (Figure 5.8) that travel s in the x-direction with zero
particl e displ acement in the y-direction (v = 0). The z-direction is taken as
positive downward,so al l particl e motion occurs in the x-z pl ane. Two potential
functions, <t> and \{I, canbe defined to describe the displ acements in the x- and
z-directions: u =w=d<t> + d\{l dX dZd<t> d\{l az- dxThe vol umetric strain, or
dil atation, E, of the wave is given by E = Exx + EZl ' or(5.35a)(5.35b)Sec. 5.3
Waves in a Semi-Infinite Body 157yPl ane wavezFigure 5.8 Motion induced by a typical
pl ane wave that propagates in the x-direction.Wave motion does not vary in the y-
direction.(5.36)(5.38b)(5.39a)(5.38a)E = au + aW = ~(a<l > + a'l l ) + ~(a<l > _a'l l ) =
a2<l >+ a2<l >= V2<l > ax az ax ax az az az ax ax2 az2The rotation in the x-z pl ane is
given [equation (5.19)] by2Q - au _aw _ ~(a<l > + a'l l ) _~(a<l > _a'l l ) = a2'P+ a2'P
= V2'P (5.37)Y - az ax - az ax az az az ax az2 ax2Use of the potential functions
al l ows separation of the effects of dil atation and rotation [i.e.,equations (5.36)
and (5.37) indicate that <l > and P are associated with dil atation and rotation,
respectivel y]. Therefore, Rayl eigh waves can be thought of as combinations ofP: and
s-waves(SV waves for this case, since the x-z pl ane is vertical ) that satisfy
certain boundary conditions.Substitution of the expressions for u and w into the
equations of motion as written inequations (5.26a) and (5.26c) givesp~(a2<l » +
p~(a2'P) = (Ie + 21l )~(V2<l » + 1l ~(V2'P)ax at2 az at2 ax azp~(a2<l » _ p~(a2'P) =
(1e+21l )~(V2<l » _ 1l ~(V2'P)az at2 ax at2 az axSol ving equations (5.38)
simul taneousl y for a2<l >/aF and a2'P/aF showsa2<l >= Ie + 21l V2<l > = v2V2<l > at2 p p
(5.39b)If the wave is harmonic with frequency co and wave number kR, so that it
propagates withRayl eigh wave vel ocity VR = co/kR, the potential functions can be
expressed as<l > = F(z)ei(Ol t-kRx) (5.40a)158 Wave Propagation Chap. 5(5.41a)qt =
G(z)ei(ffit-kRx) (5.40b)where F and G are functions that describe the manner in
which the ampl itude of the dil atational and rotational components of the Rayl eigh
wave vary with depth. Substituting theseexpressions for <I> and qt into equations
(5.39) gives_ 0)2F(z) = _ k'kF(z) + d2F(z)v2 dz2P_ 0)2G(z) = _ k'kG(z) + d2G(z ) v;
dz2which can be rearranged to give the second-order differential equationsd2F_ (k'k
_ 0)2)F = 0dz2 v~d2G _ (k'k _ 0)2)G = 0dz2 v;The general sol ution to these
equations can be written in the formF(z) = Al e-qz+Bl eqZwhere(5.41b)(5.42a)(5.42b)
(5.43a)(5.43b)The second term of equations (5.43) corresponds to a disturbance
whose displ acementampl itude approaches infinity with increasing depth. Since this
type of behavior is not real istic,Bl and B2 must be zero, and the potential
functions can final l y be written as(5.44a)(5.44b)Since neither shear nor normal
stresses can exist at the free surface of the hal f-space, 0'xz=oand O'zz = 0 when z
= O. Therefore,~ - ~ - dw 0O'zz = l I.e + 21J,£zz = 11.£ + 21J, dz =(dW dU)O'xz =
l J,exz = IJ, dx + dz = 0(5.45a)(5.45b)Sec. 5.3 Waves in a Semi-Infinite Body 159
(5.47a)(5.46b)(5.46a)Using the potential function definitions of u and w [equations
(5.35)] and the sol ution for thepotential functions [equation (5.44)], the free
surface boundary conditions can be rewritten asazzCz = 0) = Ad(A + 21l )q2 - Akk] -
2iA21l kRS = 0axz(z = 0) = 2iAtkRq + A2(s2 + kk) = 0which can be rearranged to yiel d
At(A+21l )q2- Akk_ l = 0A2 2il l kRsAt 2iqkR + 1 = 0 (5.47b)A2 s2+ kkWith these
resul ts, the vel ocities and displ acement patterns of Rayl eigh waves can be
determined.5.3.1.1 Rayl eigh Wave Vel ocityThe vel ocity at which Rayl eigh waves
travel is of interest in geotechnical earthquakeengineering. As discussed in
Chapter 6, Rayl eigh waves are often mechanical l y generatedand their vel ocities
measured in the fiel d to investigate the stiffness of surficial soil s. Adding
equations (5.47) and cross-mul tipl ying gives(5.48)which, upon introducing the
definitions of q and s and factoring out a C2 k~ term, yiel ds( 2)( 2) ( ~ 2
2)2( 2)2 16 1-~ 1-~ = 2-~~ 2-~v~kk v;kk 11 v~kk v;kkDefining KRsas the ratio of the
Rayl eigh wave vel ocity to the s-wave vel ocitythen(5.49)VR _ 0) 0) - - -- = = aKR vp
vpkR vskRJ(A + 211)/11 swhere a = ~'-~---: /---: (A'---+-: : 2: -1l ---: ) = J(l - 2v)/(2 -
2v) . Then equation (5.49) can be rewrittenas16 (l - a2K~s)( 1 - K~s) = (2 -
~2a2KksJ(2 - Kks)2which can be expanded and rearranged into the equationKt - 8Kt +
(24 - 16(2)Kks + 16( a 2 - 1) = 0(5.50)(5.51)160 Wave Propagation Chap. 5This
equation is cubic in K~s ' and real sol utions for KRscan be found for various
val ues ofPoisson's ratio. These al l ow eval uation of the ratios of the Rayl eigh wave
vel ocity to boths- and p-wave vel ocities as functions ofv. The sol ution shown in
Figure 5.9 shows that Rayl eighwaves travel sl ightl y sl ower than s-waves for al l
val ues of Poisson's ratio except 0.5.5r--~--~---,---,----,-,4 --32 -s-wavesRayl eigh
waveso 0.1 0.2 0.3 0.4Poisson's ratio, vFigure 5.9 Variation of Rayl eigh wave0.5
and body wave propagation vel ocities withPoisson's ratio.5.3.1.2 Rayl eigh Wave
Displ acement Ampl itudeSection 5.3.1.1 showed how the vel ocity of a Rayl eigh wave
compares with that ofpands-waves. Some of the intermediate resul ts of that section
can be used to il l ustrate thenature of particl e motion during the passage of
Rayl eigh waves. Substituting the sol utionsfor the potential functions <I> and \}'
[equations (5.44)] into the expressions for u and w[equation 5.35] and carrying out
the necessary partial differentiations yiel ds(5.52a)(5.52b)(5.53a)From equation
(5.47b),2qikR A2 = ---AIs2 + k~which, substituting into equations (5.52), gives_ A
( 'k -qz + 2iqskR -sz) i(ffi( - kRx) U - I -I ee --- e es2 + k~w = AIC;: k: ~ c sz -
qe-qZ)ei(ffi(-kRX) (5.53b)where the terms in parentheses describe the variation of
the ampl itudes of u and w with depth.These horizontal and vertical displ acement
ampl itudes are il l ustrated for several val ues ofSec. 5.3 Waves in a Semi-Infinite
Body 161o0.40.21.0 1.2Vertical 0.6Component zARv =0.25 0.8v =0.33v =0.40v =0.50 1.0
1.21.40.8Ampl itude at Depth zAmpl itude at Surface-0.6 -0.4 -0.2 o 0.2 0.4 0.6Figure
5.10 Horizontal and vertical motion of Rayl eigh waves. A negative ampl ituderatio
indicates that the displ acement is in the opposite direction of the surface
displ acement. (After Richart et aI., 1970.)Poisson's ratio in Figure 5.10.
Examination of equations (5.53) indicates that the horizontal and vertical
displ acements are out of phase by 90°. Hence the horizontal displ acement wil l be
zero when the vertical displ acement reaches its maximum (or minimum), and vice
versa. Themotion of a particl e near the surface of the hal f-space is in the form of
a retrograde el l ipse (asopposed to the prograde el l ipse particl e motion observed at
the surface of water waves). Thegeneral nature of Rayl eigh wave motion was
il l ustrated in Figure 2.2a.The Rayl eigh waves produced by earthquakes were once
thought to appear onl y atvery l arge epicentral distances (several hundred Ian). It
is now recognized, however, thatthey can be significant at much shorter distances
(a few tens of kil ometers). The ratio ofminimum epicentral distance, R, to focal
depth, h, at which Rayl eigh waves first appear ina homogeneous medium is given byRh
(5.54)where Vp and VR are the wave propagation vel ocities of p-waves and Rayl eigh
waves,respectivel y (Ewing et al ., 1957).162 Wave Propagation Chap. 55.3.2 Love
Waves(5.56a)for z: 2: H (5.56b)for 0 s. z s. HIn a homogeneous el astic hal f-space,
onl y p-waves, s-waves, and Rayl eigh waves can exist.If the hal f-space is overl ain
by a l ayer of material with l ower body wave vel ocity, however,Love waves can
devel op (Love, 1927). Love waves essential l y consist of SH-waves that aretrapped by
mul tipl e refl ections within the surficial l ayer. Consider the case of a homogeneous
surficial l ayer of thickness H overl ying a homogeneous hal f-space as shown in
Figure5.11. A Love wave travel ing in the +x-direction woul d invol ve particl e
displ acementsonl y in the y-direction (SH-wave motion), and coul d be described by
the equationv(x,z,t) = V(z)ei(kLx-Wf) (5.55)where v is the particl e displ acement in
the y-direction, V(z) describes the variation of v withdepth, and kL is the wave
number of the Love wave. The Love wave must satisfy the waveequations for s-waves
in both the surficial l ayer and the hal f-spaceI~,'(~: ~ + ~: ~)d2vdt2 = G2( d2v +
d2V)P2 dx2 dz2The ampl itude can be shown (Aki and Richards, 1980) to vary with
depth according to{A l e-VIZ+ B ,eVIZ forOs.zs.H (5.57a)V(z) =A2e- V2 Z + B2eV2 Z
for z: 2: H (5.57b)where the A and B coefficients describe the ampl itudes of
downgoing and upgoing waves,respectivel y, andkZ - (j)2 kZ _(j)2VI = V2 = (5.58)
G1/PI G2/P2Since the hal f-space extends to infinite depth, B2 must be zero (no
energy can be suppl iedor refl ected at infinite depth to produce an upgoing wave).
The requirement that al l stressesvanish at the ground surface is satisfied ifdv =
dV(Z)ei(kLx-Wt) = _Ajvl e-VIZ+v,BjeVIZ = (A1-Bj)v, (e-V1Z + eV1Z) = 0dZ dZin other
words, if A I =B I. The ampl itudes can now be rewritten in terms of the two
remainingunknown ampl itudes asx J..P1 G1P2 G2Surficial l ayerHal f-spaceFigure 5.11
Schematic il l ustration ofsofter surficial l ayer (Gj/PI < G2/P2)overl ying el astic
hal f-space, the simpl estconditions for which Love waves can exist.Sec. 5.3 Waves in
a Semi-Infinite Body 163forO: S;z: S; H1AI(e-VIZ + eVIl )V(z) =A2e-v2z for z;?: HAt
the z =H interface, continuity of stresses requires that(5.59a)(5.59b)2iGIvIAl
sin(ivIH) = G2v2A2e-V2Hand compatibil ity of displ acements requires that2A
Icos(ivIH) = A2e-V2HUsing equations (5.60) and (5.61), A2 can be expressed in terms
of Al byA - 2cos(ivIH)A2 - -v2H I e(5.60)(5.61)(5.62)Substituting equations (5.59)
and (5.60) into (5.55) gives( (5.64)tanmH 21 - 21 )112 =vsI VL1[ (1 1)112 J2A Icos
co 2 - 2 z ei(kLX-wt) for 0 : s; z: S; H (5.63a)VsI VLv( X, z, t) = 1 1 112 1 1 112
i(kLX - tor)2A Icos[m(2 - 2) HJ exp [-m(2 - 2) (z - H)J eVsI VL VL Vs2 (5.63b)for
z;?: Hwhere VSI and VS2 are the shear wave vel ocities of material s 1 and 2,
respectivel y, and VL isthe vel ocity of the Love wave. Equation (5.63) shows, as
il l ustrated in Figure 5.12, that theLove wave displ acement ampl itude varies
sinusoidal l y with depth in the surficial l ayer anddecays exponential l y with depth
in the underl ying hal f-space. Because of this, Love wavesare often described as SH-
waves that are trapped in the surficial l ayer. The general nature ofLove wave
displ acement was shown in Figure 2.2a.The Love wave vel ocity is given by the
sol ution ofG2Jl Iv1- l IV;2GI Jl Iv;1 -l Iv1v(z)Ht .zFigure 5.12 Variation of particl e
displ acement ampl itude with depth forLove waves.164 Wave PropagationFigure 5.13
Variation of Love wave vel ocity with frequency.Chap. 5which indicates, as
il l ustrated in Figure 5.13 that Love wave vel ocities range from the s-wavevel ocity
of the hal f-space (at very l ow frequencies) to the s-wave vel ocity of the surficial
l ayer(at very high frequencies). This frequency dependence indicates that Love
waves are dispersive(Section 5.3.4).5.3.3 Higher-Mode Surface WavesAny surface wave
must (1) satisfy the equation of motion, (2) produce zero stress at theground
surface, and (3) produce zero displ acement at infinite depth. Nontrivial sol utions
donot exist for arbitrary combinations of frequency and wave number; rather, a set
of discreteand unique wave numbers exist for a given frequency. Each wave number
describes a differentdispl acement pattern, or mode, of the surface wave. The
preceding derivations havebeen l imited to the fundamental modes of Rayl eigh and
Love waves, which are the mostimportant for earthquake engineering appl ications.
Detail ed treatment of higher-mode surfacewaves can be found in most advanced
seismol ogy texts.5.3.4 Dispersion of Surface WavesFor a homogeneous hal f-space, the
Rayl eigh wave vel ocity was shown to be rel ated to thebody wave vel ocities by
Poisson's ratio. Since the body wave vel ocities are constant withdepth, the
Rayl eigh wave vel ocity in a homogeneous hal f-space is independent of frequency.The
vel ocity of the Love wave, on the other hand, varies with frequency betweenan upper
and a l ower l imit.Dispersion is a phenomenon in which waves of different frequency
(and differentwavel ength) propagate at different vel ocities. Hence Love waves are
cl earl y dispersive, andRayl eigh waves in a homogeneous hal f-space are
nondispersive. Near the earth's surface,however, soil and rock stiffnesses usual l y
increase with depth. Since the depth to which aRayl eigh wave causes significant
displ acement increases with increasing wavel ength (Figure5.10), Rayl eigh waves of
l ong wavel ength (l ow frequency) can propagate faster thanRayl eigh waves of short
wavel ength (high frequency). Therefore, in the real worl d of heterogeneous
material s, Rayl eigh waves are al so dispersive. The dispersion of Rayl eighwaves can
be used to eval uate subsurface stiffness profil es by fiel d testing techniques
described in Chapter 6.Since the vel ocities of both Rayl eigh waves and Love waves
decrease with increasingfrequency, the l ow-frequency components of surface waves
produced by earthquakes canSec. 5.4 Waves in a Layered Body 165be expected to
arrive at a particul ar site before their high-frequency counterparts. This tendency
to spread the seismic energy Overtime is an important effect of dispersion.5.3.5
Phase and Group Vel ocitiesThe sol utions for Rayl eigh wave vel ocity, VR, and Love
wave vel ocity, Vv were based on theassumption of harmonic l oading which produces an
infinite wave train. These vel ocitiesdescribe the rate at which points of constant
phase (e.g., peaks, troughs, or zero points)travel through the medium and are
cal l ed phase vel ocities. A transient disturbance may producea packet of waves with
simil ar frequencies. This packet of waves travel s at the groupvel ocity, cg, given
bydccg = C + kdk (5.65)where C is the phase vel ocity (equal to VR or vi»depending
on which type of wave is being considered)and k is the wave number (equal to ro/VR
or ffi/VL)' In a nondispersive material , dc/dk=0, so the group vel ocity is equal to
the phase vel ocity. Since both VR and VL general l ydecrease with increasing
frequency in geol ogic material s, dc/dk is l ess than zero and the groupvel ocity is
l ower than the phase vel ocity. Consequentl y, a wave packet woul d appear to consist
of a series of individual peaks that appear at the back end of the packet, move
through thepacket to the front, and then disappear. The opposite behavior can be
observed by (for C < cg)dropping a rock into a cal m pond of water and watching the
resul ting rippl es careful l y.5.4 WAVES IN A LAYERED BODYThe model of a homogeneous
el astic hal f-space is useful for expl aining the existence ofbody waves and Rayl eigh
waves, and the addition ora softer surficial l ayer al l ows Lovewaves to be
described. In the earth, however, conditions are much more compl icated withmany
different material s of variabl e thickness occurring in many areas. To anal yze wave
propagation under such conditions, and to understand the justification for
ideal izations ofactual conditions when al l features cannot be expl icitl y anal yzed,
the general probl em ofwave behavior at interfaces must be investigated.5,,4.1 One-
Dimensional Case: Material Boundary in anInfinite RodConsider a harmonic stress
wave travel ing al ong a constrained rod in the +x direction aridapproaching an
interface between two different material s, as shown in Figure 5.14. Since thewave
is travel ing toward the interface, it wil l be referred to as the incident wave.
Since it istravel ing in material 1, its wavel ength wil l be Al =2rr/k l , and it can
therefore be described by(5.66a)When the incident wave reaches the interface, part
of its energy wil l be transmitted throughthe interface to continue travel ing in the
positive x-direction through material 2.This trans"mitted wave wil l have a
wavel ength A2 =2rrJk2. The remainder wil l be refl ected at the interfaceand wil l
travel back through material 1 in the negative x-direction as a refl ected wave.The
transmitted and refl ected waves can be described by166 Wave Propagation~---I ~
l nciderit : : : ~~TransmittedRefl ectedChap. 5Figure
5.14 One-dimensional wave propagation at material interface. Incident andrefl ected
waves travel in opposite directions in material I. The transmitted wave travel s
through material 2 in the same direction as the incident wave.(jT(X, t) = (j/e
i(ffit-k2x) (5.66b)(jR(X, t) = (jrei(ffi/+k,x) (5.66c)Assuming that the
displ acements associated with each of these waves are of the same harmonicform as
the stresses that cause them; that is,U/ (x, t) =Aiei(ffit-k,x)UR (x, t) =
Arei(ffit+k,x)(5.67a)(5067b)(5.68c)(5068b)(5.68a)UT(X, t) = A tei(ffit-k 2x)
(5.67c)Stress-strain and strain-displ acement rel ationships can be used to rel ate
the stress ampl itudesto the displ acement ampl itudes: au/ (x, t) _ ok M A i(ffit-k,x)
(j/ (x, t) = M] ax - -1] ] i eaUR (x, t) + ik: M A i(ffit+k,x)(JR (x, t) M 1 ax = 1
l IreaUT (x, t) _ ik M A i(ffit-k2x)(jT (x, t) = M 2 ax - -1 2 2 t eFrom these, the
stress ampl itudes are rel ated to the displ acement ampl itudes byo, = -ik]M]A;
(5.69a)o, = +ik]M]Ar (5.69b)c, = -ik2M2At (5.69c)At the interface, both
compatibil ty of displ acements and continuity of stresses must besatisfied. The
former requires thatand the l atter thatU/ (0, t) + UR (0, t) = UT (0, t) (5.70)
(5.71)Sec. 5.4 Waves in a Layered Body 167Substituting equations (5.67) and (5.66)
into equations (5.70) and (5.71), respectivel y, indicatesthat(5.72)(5.73)at the
interface. Substituting equations (5.69) into equation (5.73) and using the
rel ationshipkM = copv, gives(5.74)(5.75)(5.76)Equation (5.74) can be rearranged to
rel ate the displ acement ampl itude of the refl ectedwave to that of the incident
wave: PI VI - P2v2 1- P2v2/PI vIAr =. Ai = . AiPI vI + P2v2 1+ P2V2 / p,VIand
knowing Ai and Ar, equation (5.72) can be used to determine At asAt = 2Pl Vj A- = 2
AiPjvI+P2v2 I I+P2v2/Pl vl Remember that the product of the density and the wave
propagation vel ocity is the specificimpedance of the material . Equations (5.75) and
(5.76) indicate that the partitioning ofenergy at the interface depends onl y on the
ratio of the specific impedances of the material son either side of the interface.
Defining the impedance ratio as Uz = P2V2/pIVI, the displ acementampl itudes of the
refl ected and transmitted waves are(5.77)2 (5.78)At = -l +--Ai U zAfter eval uating
the effect of the interface on the displ acement ampl itudes of the refl ectedand
transmitted waves, its effect on stress ampl itudes can be investigated. From
equations(5.69)Aio,=--- (5.79a)ikl M,A - (Jr r ~ (5.79b) ik,MI(JtAt = - ik2M2(5.79c)
Substituting equations (5.79) into equations (5.77) and (5.78) and rearranging
gives168 Wave Propagation Chap. 5(5.80)(5.81)2a at=--za1 + az IThe importance of
the impedance ratio in determining the nature of refl ection and transmissionat
interfaces can cl earl y be seen. Equations (5.77), (5.78), (5.80), and (5.81)
indicatethat fundamental l y different types of behavior occur when the impedance
ratio is l ess thanor greater than 1. When the impedance ratio is l ess than 1, an
incident wave can be thoughtof as approaching a "softer" material . For this case,
the refl ected wave wil l have a smal l erstress ampl itude than the incident wave and
its sign wil l be reversed (an incident compressionpul se wil l be refl ected as a
tensil e pul se, and vice versa). If the impedance ratio isgreater than 1, the
incident wave is approaching a "stiffer" material in which the stressampl itude of
the transmitted wave wil l be greater than that of the incident wave and thestress
ampl itude of the refl ected wave wil l be l ess than, but of the same sign, as that of
theincident wave. The displ acement ampl itudes are al so affected by the impedance
ratio. Therel ative stress and displ acement ampl itudes of refl ected and transmitted
waves at boundarieswith several different impedance ratios are il l ustrated in Tabl e
5-1.The cases of az = 0 and az = 00 are of particul ar interest. An impedance ratio
of zeroimpl ies that the incident wave is approaching a "free end" across which no
stress can betransmitted (at =0). To satisfy this zero stress boundary condition,
the displ acement of theboundary (the transmitted displ acement) must be twice the
displ acement ampl itude of theincident wave (At =2Ai)' The refl ected wave has the
same ampl itude as the incident wavebut is of the opposite pol arity (a r = - ai)' In
other words, a free end wil l refl ect a compressionwave as a tension wave of
identical ampl itude and shape and a tension wave as an identical compression wave.
An infinite impedance ratio impl ies that the incident wave isapproaching a "fixed
end" at which no displ acement can occur (Ut =0). In that case thestress at the
boundary is twice that of the incident wave (at= 2ai) and the refl ected wave hasthe
same ampl itude and pol arity as the incident wave (Ar =- Ai)'The case of az =1, in
which the impedances on each side of the boundary are equal ,is al so of interest.
Equations (5.77), (5.78), (5.80), and (5.81) indicate that no refl ected waveTabl e
5-1 l nfl uence of Impedance Ratio on Displ acement and Stress Ampl itudes of Refl ected
and Transmitted WavesImpedance Displ acement Ampl itudes Stress Ampl itudesRatio,<Xz
Incident Refl ected Transmitted Incident Refl ected Transmitted0 Ai Ai 2Ai o, -cri 0I
Ai 3A/5 8A/5 a, -3a/5 2(J/5 : II Ai A/3 4A/3 c, -G/3 2a/3 21 Ai 0 Ai o, 0 a,2 Ai
-A/3 2A/3 c, a/3 4a/34 Ai -3A/5 2A/5 ai 3a/5 8a/5Ai -Ai 0 o, ai 2a iSec. 5.4 Waves
in a Layered Body 169is produced and that the transmitted wave has, as expected,
the same ampl itude and pol arityas the incident wave. In other words, al l of the
el astic energy of the wave crosses the boundaryunchanged and travel s away, never to
return. Another way of l ooking at a boundary withan impedance ratio of unity is as
a boundary between two identical , semi-infinite rods. Aharmonic wave travel ing in
the positive x-direction (Figure 5.15a) woul d impose an axial force [see equation
(5.5)] on the boundary: F = O"xA = pvmAuThis axial force is identical to that which
woul d exist if the semi-infinite rod on the right sideofthe boundary were repl aced
by a dashpot (Figure 5.15b) of coefficient c =pvmA.In otherwords, the dashpot woul d
absorb al l the el astic energy of the incident wave, so the responseof the rod on
the l eft woul d be identical for both cases il l ustrated in Figure 5.15. This resul t
has important impl ications for ground response and soil -structure interaction
anal yses(Chapter 7), where the repl acement of a semi-infinite domain by discrete
el ements such asdashpots can provide trememdous computational efficiencies.!Area, A
- I- (a) 00 -----fJ\L._p_vm _(b) 00 --6 PVm ~"----- -----'r: : LJ~Figure 5.15 (a)
Harmonic wave travel ing al ong two connected semi-infinite rods; (b)semi-infinite
rod attached to dashpot. With proper sel ection of dashpot coefficient,response in
semi-infinite rod on l eft wil l be identical for both cases.Exampl e 5.4A vertical l y
propagating shear wave travel s upward through a l ayered soil deposit. Computethe
ampl itudes of the refl ected and transmitted waves that devel op when the shear wave
reachesthe boundary shown in Figure E5.4.Sol ution Al though the transmission-
refl ection behavior in the preceding section was derivedfor constrained
l ongitudinal waves, extension to the case of shear waves is straightforward. The
(shear wave) impedance ratio for an upward-travel ing wave isl Xz= (1.76 Mg/m3) (400
m/sec) = 0.419(2.24 Mg/m3 )(750 m/sec)The stress ampl itude of the refl ected wave is
given by equation (5.79)l Xz - 1 0.419 - 1e, = 1 + l Xze. = 1 + 0.419 (100 kPa) =
-40.9 kPaFrom equation (5.80), the stress ampl itude of the transmitted wave is170
Incident waveStress ampl itude =100kPaFrequency =2 HzWave PropagationFigure ES.4
Chap. 5The displ acement ampl itude of the incident wave can be computed from the
shear wave equival entof equation 5.78aA _ Ui _ ici,i - - ikjG j - pjffiVs!_ i (100
kN/m 2) (1 Mg/9.807 kN) (9.807 m/sec 2) = 0.00477 m =4.77 mm- (2.24 Mg/m') 211: (2
Hz) (750 m/sec)The i term simpl y describes the 90° phase angl e between stresses and
displ acements. Then,using equations (5.76) and (5.77), the displ acement ampl itudes
of the refl ected and transmittedwaves are1 - 0.4191 + 0.419 (4.77 mm)21 + 0.419
(4.77 mm)= 1.95 mm= 6.72 mmIn this exampl e, the incident wave travel s from a
material of higher impedance to a material ofl ower impedance. As a resul t, the
displ acement ampl itude of the transmitted wave is greaterthan that of the incident
wave, but the stress ampl itude is smal l er.5.4.2 Three-Dimensional Case: Incl ined
WavesIn general , waves wil l not approach interfaces at 900 angl es as they did in
Section 504.1. Theorientation of an incl ined body wave can strongl y infl uence the
manner in which energy isrefl ected and transmitted across an interface. Fermat's
principl e defines the propagationtime of a seismic pul se between two arbitrary
points A and B as the minimum travel timeal ong any continuous path that connects A
and B. The path that produces the minimumtravel time is cal l ed a ray path, and its
direction is often represented by a vector cal l ed a ray.A wavefront is defined as a
surface of equal travel time, consequentl y, a ray path must (inan isotropic
material ) be perpendicul ar to the wavefront as il l ustrated in Figure 5.16. Snel l
considered the change of direction of ray paths at interfaces between material s
with differentwave propagation vel ocities. Using Fermat's principl e, Snel l showed
thatSec. 5.4 Waves in a Layered Body~ RayWavefrontRay pathWavefront171/(a)Ray paths
(b)Figure 5.16 Ray path, ray, and wavefront for (a) pl ane wave and (b) curved
wavefront.sini = constant (5.82)vwhere i is the angl e between the ray path and the
normal to the interface and v is the vel ocityof the wave (p- or s-wave) of
interest. This rel ationship hol ds for both refl ected and transmittedwaves.
It indicates that the transmitted wave wil l be refracted (except when i = 0)when
the wave propagation vel ocities are different on each side of the interface.
Consider the case of two hal f-spaces of different el astic material s in contact with
eachother. As for the previous case, the requirements of equil ibrium and
compatibil ity and thetheory of el asticity can be used to determine the nature of
and distribution of energy amongthe refl ected and transmitted waves for the cases
of an incident p-wave, an incident SVwave,and an incident SH-wave.The types of
waves produced by incident p-, SV-, and SH-waves are shown in Figure5.17. Since
incident p- and SV-waves invol ve particl e motion perpendicul ar to the pl ane ofthe
interface; they wil l each produce both refl ected and refracted p- and SV-waves. An
incidentSH-wave does not invol ve particl e motion perpendicul ar to the interface;
consequentl y,onl y SH-waves are refl ected and refracted. The directions and rel ative
ampl itudesof the waves produced at the interface depend on both the direction and
ampl itude of theRefracted SV Refracted SV Refracted SHRefracted P Refracted P-: /
Incident P Incident SHRefl ected SV Refl ected SV Refl ected SH(a) (b) (c)Figure 5.17
Refl ected and refracted rays resul ting from incident (a) p-wave, (b) SYwave,and (c)
SH-wave.172 Wave Propagation Chap. 5(5.83)incident wave. Using Snel l 's l aw and the
requirements of equil ibrium and compatibil ity,these directions and ampl itudes can
be determined. Using the notation of Richter (1958): Wave Type Vel ocity Ampl itude
Angl e with Normal Incident p U A aIncident s V B bRefl ected p U C cRefl ected s V D d
Refracted p Y E eRefracted s Z F fthe directions of al l waves are easil y rel ated to
the direction of the incident wave usingSnel l 's l aw: sin a _ sin b _ sin c _ sin d _
sin e _ sin f----rT - -y - ----rT - -y - ----y - ------: z-Since incident and
refl ected waves travel through the same material , a = c and b = d, whichshows that
the angl e ofincidence is equal to the angl e ofrefl ection for both p- and s-waves.
The angl e ofrefraction is uniquel y rel ated to the angl e of incidence by the ratio
of thewave vel ocities of the material s on each side of the interface. Snel l 's l aw
indicates thatwaves travel ing from higher-vel ocity material s into l ower-vel ocity
material s wil l berefracted cl oser to the normal to the interfaces. In other words,
waves propagating upwardthrough horizontal l ayers of successivel y l ower vel ocity
(as is common near the earth's surface)wil l be refracted cl oser and cl oser to a
vertical path (Figure 5.18). This phenomenonis rel ied upon heavil y by many of the
methods of ground response anal ysis presented inChapter 7.The critical angl e
ofincidence, i , is defined as that which produces a refracted wave c 0 that
travel s paral l el to the interface (e orf= 90). Therefore,i - sin-1Q" sin-I.!: : '
(5.84)c - y ZVs = 1,000 fpsVs =1,500 fpsVs = 2,000 fps Figure 5.18 Refraction of an
SH-waveray path through series of successivel ysofter (l ower Vs ) l ayers. Note that
orientation of ray path becomes cl oser tovertical as ground surface is approached.
Refl ected rays are not shown.Sec. 5.4 Waves in a Layered Body 173(5.87)The concept
of critical refraction is used in the interpretation of seismic refraction tests
(Section 6.3.1.1).Assuming that the incident wave is simpl e harmonic, satisfaction
of the requirementsof equil ibrium and compatibil ity at the interface give rise to
the fol l owing systems of simul taneousequations (Richter, 1958), which al l ows the
ampl itudes of the refl ected and refractedwaves (C, D, E, and F) to be expressed in
terms of the ampl itude ofthe incident p-wave (A).(A - C) sina + Dcosb - Esine +
Fcosf = 0(A + C) cosa + Dsinb - Ecose - Fsinf = 0- (A + C) sin2a + D~cos2b +
EK(~J~sin2e - FK( ~Y¥COS2f = 0 (5.85)- (A - C) cos2b + Dfsin2b + EK-Dcos2f - FK~sin
(21) = 0where K = Pl /P2 (the subscripts 1 and 2 refer to material s 1 and 2,
respectivel y). Note thatthe ampl itudes are functions of the angl e of incidence, the
vel ocity ratio, and the densityratio. Figure 5.19 shows the variation of ampl itude
with angl e of p-wave incidence for thefoil l owing conditions: U =8.000, Y =2.003, K
=0.606, and v=0.25. The sensitivity of therefl ected and refracted wave ampl itudes
to the angl e of incidence is apparent. SV-waves areneither refl ected nor refracted
at angl es of incidence of 0 and 90", but can carry the majorityof the wave energy
away from the interface at intermediate angl es.For an incident SV-wave, both SV-
and p-waves are refl ected and refracted. The equil ibrium/compatibil ity equations
rel ating the rel ative ampl itudes are(B + D) sinb + Ccosa - Ecose - Fsinf = 0(B - D)
cosb + Csina + Esine - Fcosf = 0(B+D)COS2b-cfsin2a+EK~~sin2e-FK~coS2f = 0 (5.86)-
(B - D) sin2b + C~COS2b + EK~cos2f + FK~sin2f = 0which produce the ampl itude
behavior shown in Figure 5.20. For angl es of incidencegreater than sin-1 (V/U),
about 36"in Figure 5.20(a), no p-wave can be refl ected, so the incidentwave energy
must be carried away by the remaining waves. A more detail ed discussionof this
phenomenon can be found in McCamy et a1. (1962).An incident Sl I-wave invol ves no
particl e motion perpendicul ar to the interface; consequentl y,it cannot produce p-
waves (C = E = 0) or SV-waves. The equil ibrium/compatibil ityequations are
considerabl y simpl ified and easil y sol ved as1_K~cosfD = Vcosb B1 + K~cosfVcosb[1 KZ
COSf]F = B 1 + Vcosb1 +K~cosfVcosb174 Wave Propagation Chap. 52.0 2.01.6 1.61.2 1.2
C 0A A0.8 '. 0.8 ..0.40 30 60 90 0 30 60 90Incident angl e (deg) Incident angl e
(deg)(a) (b)2.0 2.01.6 .. 1.61.2 1.2E FA A0.8 _.0.4 0.40 30 60 90 0 30 60 90
Incident angl e (deg) Incident angl e (deg)(c) (d)Figure 5.19 Ratio of ampl itudes of
(a) refl ected p-wave, (b) refl ected SV-wave, (c)refracted p-wave, and (d) refracted
SV-wave to ampl itude of incident p-wave versusangl e of incidence.The preceding
resul ts show that the interaction of stress waves with boundaries can bequite
compl icated. As seismic waves travel away from the source of an earthquake, they
invariabl y encounter heterogeneities and discontinuities in the earth's crust. The
creation ofnew waves and the refl ection and refraction of ray paths by these
heterogeneities cause seismicwaves to reach a site by many different paths. Since
the paths have different l engths, themotion at the site is spread out in time by
this scattering effect.5.5 ATTENUATION OF STRESS WAVESThe preceding sections have
considered onl y the propagation of waves in l inear el asticmaterial s. In a
homogeneous l inear el astic material , stress waves travel indefinitel y withoutchange
in ampl itude. This type of behavior cannot occur, however, in real material s. The
ampl itudes of stress waves in real material s, such as those that comprise the
earth, attenuatewith distance. This attenuation can be attributed to two sources,
one of which invol ves theSec. 5.5 Attenuation of Stress Waves 1752.0 2.01.6 1.61.2
1.2 --C 08 0.8I 8 0. \ 8\0.4 \ 0.4 --\ " .....\ / ,I ,0 30 60 90 0 30 60 90Incident
angl e (deg) Incident angl e (deg)(a) (b)2.0 2.01.6 -- 1.61.2 -- 1.2E F8 80.8 -- 0.8
--0.4 0.40 30 60 90 0 30 60 90Incident angl e (deg) Incident angl e (deg)(c) (d)
Figure 5.20 Ratio of ampl itudes of (a) refl ected p-wave, (b) refl ected SV-wave, (c)
refracted p-wave, and (d) refracted SV-wave to ampl itude of incident SV-wave versus
angl e of incidence.material s through which the waves travel and the other the
geometry of the wave propagationprobl em.5.~;.1 Material Damping(5.88)In real
material s, part of the el astic energy of a travel ing wave is al ways converted to
heat.The conversion is accompanied by a decrease in the ampl itude of the wave.
Viscous damping,by virtue of its mathematical convenience, is often used to
represent this dissipation ofel astic energy. For the purposes of viscoel astic wave
propagation, soil s are usual l y model l edas Kel vin-Voigt sol ids (i.e., material s
whose resistance to shearing deformation is thesum of an el astic part and a viscous
part). A thin el ement of a Kel vin-Voigt sol id can beil l ustrated as in Figure 5.21.
The stress-strain rel ationship for a Kel vin-Voigt sol id in shear can be expressed
as't = GY+l l *176 Wave Propagation Chap. 5du crxz---j~"'" O Tdz zL ...... 1xFigure
5.21 Thin el ement of a Kel vin-Voigt sol id subjected to horizontal shearing.Total
resistance to shearing deformation is given by the sum of an el astic (spring)
component and a viscous (dashpot) component.where t (= (Jxz) is the shear stress,
Y(= du/dz) is the shear strain, and 11 is the viscosity of thematerial . Thus the
shear stress is the sum of an el astic part (proportional to strain) and a viscous
part (proportional to strain rate). For a harmonic shear strain of the formthe
shear stress wil l beY = Yo sin cot (5.89)"[ = GYosincot+ co11Yocoscot (5.90)
Together, equations (5.89) and (5.90) show that the stress-strain l oop of a Kel vin-
Voigtsol id is el l iptical . The el astic energy dissipated in a singl e cycl e is given
by the area of theel l ipse, orftO+ 21t/ 0l d~W = "[Jdt = 1t11COY6dttowhich indicates
that the dissipated energy is proportional to the frequency of l oading. Real soil s,
however, dissipate el astic energy hysteretical l y, by the sl ippage of grains with
respectto each other. As a resul t, their energy dissipation characteristics are
insensitive to frequency.For discrete Kel vin-Voigt systems (Appendix B), the
damping ratio,~, was shownto be rel ated to the force-displ acement (or,
equival entl y, the stress-strain) l oop as shown inFigure 5.22. Since the peak energy
stored in the cycl e isPeak energyduring cycl e, Wu,yEnergy dissipatedin one cycl e,
~wFigure 5.22 Rel ationship between hysteresis l oop and damping ratio.Sec. 5.5then
Attenuation of Stress Waves 177(5.91)S- 2 1 1tT](OYo _ T](O-
41t !GY6 - 2GTo el iminate frequency dependence whil e maintaining the convenience
of the viscoel asticformul ation, equation (5.91) is often rearranged to produce an
equival ent viscosity that isinversel y proportional to frequency. The use of this
equival ent viscosity ensures that thedamping ratio is independent of frequency:
(5.92)(5.94)(5.96)A Kel vin-Voigt sol id for vertical l y propagating SH-waves may be
represented by astack of infinitesimal el ements of the type shown schematical l y in
Figure 5.21. The onedimensional equation of motion for vertical l y propagating SH-
waves can be written ascPu _ dcrxz 5 93Pdt2 - az (. )Substituting equation (5.88)
into (5.93) with r =crxz and Y=dU/dz, and differentiating theright side al l ows the
wave equation to be expressed asd2u d2u d3up- - G-+T]--dt2 - dz2 dz 2dtFor harmonic
waves, the displ acements can be written asu (z, t) = U (z) ei rot (5.95)which, when
substituted into the wave equation (5.94) yiel ds the ordinary differential equation
d2U(G + i(OT]) -2 = _p(02Udzord2UG*-2 = -p(02U (5.97)dzwhere G* = G + iffiT] is the
compl ex shear modul us. The compl ex shear modul us is anal ogousto the compl ex
stiffness described in Section B.6.3 of Appendix B. Using equation(5.92) to
el iminate frequency dependence, the compl ex shear modul us can al so beexpressed as
G* = G(l + 2iS). This equation of motion has the sol utionu (z, t) = Aei(rot-k*z) +
Bei(rot+k*z) (5.98)where A and B depend on the boundary conditions and k* = (0JP/
G* is the compl ex wavenumber. It can be shown (after Kol sky, 1963) that k* is given
by(5.99)178whereWave Propagation Chap. 5(5.100)2 kT = pO) (J1 +4~2+ 1)2G(1 + 4~2)
': >2k22= pO) (J1 + 4~2 1)2G(1 + 4~2) ': >and onl y the positive root of kl and the
negative root of k2 have physical significance. Notethat for the inviscid case (11
=~ =0), k2 =0 and k[ =k. For a wave propagating in the positivez-direction, the
sol ution can be written asut z, t) = Ae k2Zei(rot - k , z ) (5.101)which shows
(since k2 is negative) that material damping produces exponential attenuationof
wave ampl itude with distance.Al though the Kel vin-Voigt model is by far the most
commonl y used model for soil s,it represents onl y one of an infinite number of
rheol ogical model s. By rearranging and addingmore springs and dashpots, many
different types of behavior can be model ed, al thoughthe compl exity of the wave
equation sol ution increases dramatical l y as the number ofsprings and dashpots
increases.Exampl eS.SA harmonic pl ane wave with a period of 0.3 sec travel s through
a viscoel astic material (G =1.6x 106 psf, T] = 9000 Ib-sec/fr', Y= 140 pcf).
Determine the distance over which the displ acementampl itude of the pl ane wave woul d
be hal ved.Sol ution From equation (5.101), the displ acement ampl itude at Z = Zl is
u(Z\) = Aexp(kzz1)If Z=Zz represents the l ocation at which the displ acement
ampl itude is hal ved, thenwhich l eads toor_ In(l l 2)ZZ-Zl - -kz-For this case~ _ T]Ul
_ 21tT] _ 21t(9000 Ib-sec/ftZ) = 0.059- 2G - 2GT - 2(1.6 x 106 l b/ftZ)(0.3 sec)
Then, from equation (5.100)sok: =140 l b/ftZz(21t/0.3 sec)?32.2 ft/sec (,il +
4(0.059)Z - 1) = -0.0022(1.6 x 106 Ib/ftZ)(l + 4(0.059)Z)ZZ-Z\ = In(0.5) = 347ft-
0.002Sec. 5.5 Attenuation of Stress Waves 179(5.102)(5.103)5,,5.2 Radiation Damping
Since material damping absorbs some of the el astic energy of a stress wave, the
specificenergy (el astic energy per unit vol ume) decreases as the wave travel s
through a material .The reduction of specific energy causes the ampl itude of the
stress wave to decrease withdistance. The specific energy can al so decrease by
another common mechanism, which canbe il l ustrated by the propagation of stress
waves al ong an undamped conical rod.Consider the unconstrained conical rod of smal l
apex angl e shown in Figure 5.23 andassume that it is subjected to stress waves of
wavel ength considerabl y l arger than the diameterof the rod in the area of interest.
If the apex angl e is sufficientl y smal l , the normal stresswil l be uniform across
each of two spherical surfaces that bound an el ement of width dr, andwil l act in a
direction virtual l y paral l el to the axis of the rod. Letting u represent the
displ acementparal l el to the axis of the rod, the equation of motion in that
direction can be written,using exactl y the same approach used in Section 5.2.1.1,
aspr2a,dr~~ = ( o + ~~dr}r + dr)2a, - or'o.which simpl ifies tod2u dcr pr- = r-+2cr
dt2 drSubstituting the stress-strain and strain-displ acement rel ationships
(assuming now that theends of the el ement are pl anar) givesord2(ur) _ Ed2(ur)aT-Pa?
which is the now-famil iar wave equation. Its sol ution wil l be of the formu(r,t) =
-1[f(vt-r) +g(vt+r)]r(5.104)(5.105)(5.106)where v = JE/p. Equation (5.106)
indicates that the ampl itude of the wave wil l decreasewith distance (even though
the total el astic energy remains the same). The reduction is of8cro + 8r drFigure
5.23 Conical rod of apex angl e a.180 Wave Propagation Chap. 5purel y geometric
origin, resul ting from the decrease in specific energy that occurs as thearea of
the rod increases.Even though el astic energy is conserved (no conversion to other
forms of energy takespl ace), this reduction in ampl itude due to spreading of the
energy over a greater vol ume ofmaterial is often referred to as radiation damping
(al so as geometric damping and geometricattenuation). It shoul d be distinguished
from material damping in which el astic energy isactual l y dissipated by viscous,
hysteretic, or other mechanisms.When earthquake energy is rel eased from a faul t
bel ow the ground surface, bodywaves travel away from the source in al l directions.
If the rupture zone can be representedas a point source, the wavefronts wil l be
spherical and the preceding anal ysis can easil y beextended to show that geometric
attenuation causes the ampl itude to decrease at a rate ofl Ir. It can al so be shown
(Bul l en, 1953) that geometric attenuation of surface waves causestheir ampl itudes
to decrease at a rate of essential l y 1/(,Jr); in other words, surface waves
attenuate (geometrical l y) much more sl owl y than body waves. This expl ains the
greaterproportion of surface wave motion (rel ative to body wave motion) that is
commonl yobserved at l arge epicentral distances. This expl ains the advantages of the
surface wavemagnitude, rel ative to body wave magnitude, for characterization of
distant earthquakes.For probl ems in which energy is rel eased from a finite source,
ranging from the l argescal ecase of rupture al ong an earthquake faul t to the
smal l er-scal e case of a vibrating foundation,radiation damping can be extremel y
important. In such cases the effects of radiationdamping often dominate those of
material damping.5.6 SUMMARY1. Onl y body waves can travel through an unbounded,
homogeneous sol id. There aretwo types of body waves: p- and s-waves. P-waves are
irrotational , or dil atational ,waves-they induce vol umetric but not shearing
deformations in the material s theytravel through. The direction of particl e
movement caused by p-waves is paral l el tothe direction in which the wave is
travel ing. S-waves, al so known as shear waves,invol ve shearing but not vol umetric
deformations. The passage of an s-wave causesparticl e movement perpendicul ar to the
direction of wave travel .2. Body waves travel at vel ocities that depend on the
stiffness and density of the material they travel through. Because geol ogic
material s are stiffer in vol umetric compressionthan in shear, p-waves travel faster
through them than do s-waves.3. The interaction of incl ined body waves with the
stress-free surface of the earth producessurface waves. The motions produced by
surface waves are concentrated in ashal l ow zone near the surface.4. Rayl eigh waves
are the most important type of surface wave for earthquake engineeringappl ications.
In a homogeneous el astic hal f-space, Rayl eigh waves woul d travel sl ightl y more
sl owl y than s-waves and woul d produce both vertical and horizontal particl e motions
that fol l ow a retrograde el l iptical pattern.5. The depth to which Rayl eigh waves
induce significant motion is inversel y proportional to the frequency of the wave.
Low-frequency Rayl eigh waves can produce particl emotion at l arge depths, but the
motions produced by high-frequency Rayl eighwaves are confined to shal l ow depths.
Chap. 5 Homework Probl ems 1816. When body wave vel ocities increase with depth, as
they general l y do in the earth's crust,Rayl eigh wave vel ocities are frequency
dependent. Low-frequency Rayl eigh waves,which induce motion in deeper, stiffer
material s, travel faster than high-frequency Rayl eighwaves. Waves with frequency-
dependent vel ocities are said to be dispersive.7. Love waves are surface waves that
can devel op in the presence of a soft surficial l ayer. Love waves are dispersive-
their vel ocities vary with frequency between theshear wave vel ocity of the
surficial l ayer (at high frequencies) and the shear wavevel ocity of the underl ying
material (at l ow frequencies).8. When a body wave strikes a rigid boundary oriented
perpendicul ar to its direction oftravel , the wave is perfectl y refl ected as an
identical wave travel ing back in the oppositedirection. The zero-displ acement
boundary condition requires that the stress atthe boundary be twice that of the
wave away from the boundary. When a body wavestrikes a stress-free boundary
oriented perpendicul ar to its direction of travel , thewave is refl ected as an
identical wave of opposite pol arity travel l ing back in the samedirection. The zero-
stress boundary condition requires that the particl e motion at theboundary be twice
as l arge as the particl e motion away from the boundary.9. When a body wave strikes
a normal boundary between two different material s, part ofthe wave energy is
refl ected and part is transmitted across the boundary. The behaviorof
the wave at the boundary is governed by the ratio of the specific impedances of
thematerial s on either side of the boundary. This impedance ratio determines the
ampl itudesand pol arities of the refl ected and transmitted waves.10. When body waves
strike boundaries between different material s at angl es other than90°, part of the
wave energy is refl ected and part is refracted as it crosses the boundary.If the
direction of particl e motion is paral l el to the boundary, the refl ected and
refracted waves wil l be of the same form as the incident wave. If not, new types of
waves can be created; for exampl e, an incl ined p-wave that strikes a horizontal
boundarywil l produce refl ected p- and SV-waves, and al so refracted p- and SV-waves.
11. When an incl ined wave travel s upward through horizontal l ayers that become
successivel ysofter, the portion of the wave that crosses each l ayer boundary wil l
berefracted cl oser and cl oser to a vertical direction.12. The ampl itude ofa stress
wave decreases as the wave travel s through the earth's crust.There are two primary
mechanisms that cause this attenuation of wave ampl itude.The first, material
damping, is due to absorption of energy by the material s the waveis travel ing
through. The second, radiation damping, resul ts from the spreading ofwave energy
over a greater vol ume of material as it travel s away from its source.HOMEWORK
PROBLEMS5.,1 Determine the wave propagation vel ocity of a l ongitudinal wave in a
constrained rod of: (a)steel , (b) cast iron, (c) concrete with f~ =4,000 psi.5.2
Determine the wave propagation vel ocity of a torsional wave in a constrained rod
of: (a) steel ,(b) cast iron, (c) concrete with f~ = 4,000 psi.5.3 Derive an
expression for the wave propagation vel ocity of a l ongitudinal wave in an
unconstrainedel astic rod with Young's modul us, E, Poisson's ratio, Y, and density,
p. Negl ect the182 Wave Propagation Chap. 5effects of radial displ acements in your
derivation. How does this vel ocity compare with that ofa l ongitudinal wave in a
constrained rod?5.4 A constrained steel rod is subjected to a harmonic axial stress
at a frequency of 1 Hz.(a) Determine the wavel ength of the axial displ acements
al ong the rod.(b) Determine the phase angl e between displ acements measured at
points on the rod l ocated1 ft, 10 ft, 100 ft, 1000 ft, and 10000 ft apart.(c)
Repeat Part (a) for l oading frequencies of 10 Hz, 100 Hz, 1 kHz, and 1 MHz.5.5 A
constrained steel rod is subjected to a harmonic torsional stress at a frequency of
1 Hz.(a) Determine the wavel ength of the angul ar rotations al ong the rod.(b)
Determine the phase angl e between rotations measured at points on the rod l ocated 1
ft,10 ft, 100 ft, 1000 ft, and 10000 ft apart.(c) Repeat Part (a) for l oading
frequencies of 10 Hz. 100 Hz, 1 kHz, and 1 MHz.5.6 Determine the depth at which the
vertical displ acement ampl itude of a Rayl eigh wave is equal to one-hal f of the
vertical displ acement ampl itude at the ground surface. For Rayl eigh wavestravel ing
through crystal l ine bedrock with a constant shear modul us of 57 ksi and specific
gravity of 2.7, pl ot the variation of this depthwith frequency for f = 0.1 Hz tof=
20 Hz.Steel I+------'-XCast IronFigure P5.75.7 The infinite rod shown above is hal f
steel and hal f cast iron. If a stress pul se of ampl itude 100psi (compression
positive) and frequency 1000 Hz travel s through the steel in the +x-direction,
determine the displ acement ampl itude of the transmitted wave that travel s through
the cast iron.Steel Cast Iron~---- 10 m-----J.~II.........----10 m----~Figure P5.8
5.8 The finite rod shown above is subjected to an impact l oad that produces a
rectangul ar axial stress pul se of ampl itude 100 psi and duration 0.1 msec its l eft
end. The l eft end of the rod isfree and the right end is fixed. Assuming that the
impact began at t = 0, determine and pl ot theaxial stress at the mid-point of the
cast iron section from t = ato t= 15 msec.Chap. 5 Homework Probl ems 183v= 180 m/sec
~ 4 mv= 400 m/sec ~ 5 mv= 550 m/sec r7 m,.720 m1,,, )11 mv= 1000 m/sec /Incident p-
wave FigureP5.9S.9 Consider the l ayered soil deposit shown above. Determine and
pl ot the paths of the refractedand refl ected rays if an incident raypath strikes
the l owest boundary at a 4Y angl e. Show onl ythe first refl ection and refraction at
each l ayer boundary.S.10 Seismol ogists usual l y describe the effects of material
damping by the qual ity factor. Qual ity factorval ues on the order of 300 and 500 are
typical l y used for waves travel l ing through the earth'scrust in western North
America (WNA) and eastern North America (ENA), respectivel y. Why arehigher qual ity
factors used for ENA? What damping ratios do these qual ity factors correspond to?6
Dynamic Soil Properties6.1 INTRODUCTIONThe nature and distribution of earthquake
damage is strongl y infl uenced by the response ofsoil s to cycl ic l oading. This
response is control l ed in l arge part by the mechanical properties ofthe soil .
Geotechnical earthquake engineering encompasses a wide range of probl ems invol ving
many types of l oading and many potential mechanisms of fail ure, and different soil
propertiesinfl uence the behavior of the soil for different probl ems. For many
important probl ems,particul arl y those dominated by wave propagation effects, onl y
l ow l evel s of strain are inducedin the soil . For other important probl ems, such as
those invol ving the stabil ity of masses of soil ,l arge strains are induced in the
soil . The behavior of soil s subjected to dynamic l oading is governedby what have
come to be popul arl y known as dynamic soil properties. Whil e recognizingthat the
properties themsel ves are not dynamic (indeed, they appl y to a host of nondynamic
probl ems), that term wil l be used in this book because of its conciseness and
famil iarity.Detail ed treatment of every aspect of the behavior of cycl ical l y l oaded
soil s is beyondthe scope of a book such as this. This chapter addresses the most
important aspects of theirbehavior in the context of the various geotechnical
earthquake engineering probl emsaddressed in the fol l owing chapters. It presents a
variety of methods by which l ow- and highstrainsoil behavior can be measured in the
fiel d and in the l aboratory. The behavior of cycl ical l yl oaded soil s, and different
approaches to its characterization, are al so described.184Sec. 6.2 Representation
of Stress Conditions by the Mohr Circl e 1856.2 REPl l ESENTATION OF STRESS CONDITIONS
BY THE MOHR CIRCLEThe cycl ic properties of soil s depend on the state of stress in
the soil prior to l oading and onthe stresses imposed by the l oading. To discuss
soil properties and their rel ationship to thevarious types of cycl ic l oading
encountered in geotechnical earthquake engineering probl ems,the concepts and
terminol ogy used to describe stresses must be specified. Al thoughsuch concepts are
ordinaril y presented earl y in a first course on soil mechanics, their importancein
the understanding and sol ution of geotechnical earthquake engineering probl ems
(particul arl y the l iquefaction probl ems discussed in Chapter 9) is sufficient to
warrant theirrepetition here.The stress conditions at any point in a mass of soil
can be described by the normal andshear stresses acting on a particul ar pl ane
passing through that point. Because most normal stresses in soil s are compressive
(soil s cannot, in general , resist tensil e stresses), it is customaryin geotechnical
engineering to describe compressive stresses as positive. Consequentl y,positive
shear stresses are those that tend to cause countercl ockwise rotation of thebody
they act upon, and cl ockwise angl es are positive. Figure 6.1 il l ustrates the sign
conventionsfor normal and shear stresses; axand ay are the normal stresses acting
on pl anesnormal to the x- and y-axes, respectivel y, 'txy (and 'tyx)is the shear
stress in the y-direction (xdirection)on the pl ane normal to the x-axis (y-axis),
and aa and 't a are the normal and shearstresses on the pl ane incl ined at angl e a.
In structural mechanics, the opposite conventionsare general l y used.The notation
used to describe the foregoing stresses is different than that used todevel op the
equations of motion for three-dimensional wave propagation in Chapter 5. Forthat
probl em of sol id mechanics, the standard notation of sol id mechanics was used. For
thischapter and the remainder of the book, the notation above, which is most
commonl y used ingeotechnical engineering, is used. The equival ence of the two
notations was discussed inSection 5.2.2.1.It is often necessary to consider the
stresses on several different pl anes that passthrough a particul ar point.
Equil ibrium requirements can be used (e.g., Hol tz and Kovacs,1981) to express the
normal and shear stresses on a pl ane incl ined at an angl e, a, to the xaxisasax + ay
ay- ax .aa = --2- + --2-cos2a-'txysm2aax-ay .'t a = --2-sm2a-'txycos2a(6.1a)(6.1b)
Figure 6.1 Sign conventions for normal and shear stresses.186 Dynamic Soil
Properties Chap. 6Figure 6.2 Mohr circl e of stress for el ement subjected to major
principal stress, (Jy. andminor principal stress, (JX' Location of pol e denoted by
P.Equations 6.1 describe a circl e whose center is at [0" =(O"x+ O"y)/2, 1: =0] and
whose radiusis [( O"y - o JI2] 2 + 1: ~y . This circl e, shown in Figure 6.2, is the
wel l -known Mohr circl eof stress. The Mohr circl e simpl y il l ustrates the stress
conditions acting on an el ementgraphical l y and, as such, is very useful for
understanding states of stress and stressesinduced by external l oading. It wil l be
used often in the remainder of this book.Equations (6.1) al l ow the stresses on
pl anes of different incl ination to be determinedanal ytical l y, but they can al so be
determined graphical l y using the pol e of the Mohr circl e.The pol e has a useful
property: any l ine drawn through the pol e wil l intersect the Mohr circl eat
a point that describes the shear and normal stresses on a pl ane paral l el to that
l ine.Consider the el ement shown in Figure 6.2 subjected to a vertical normal
stress, o.; and ahorizontal normal stress, O"x- The shear stresses on the
boundaries are zero. The stress conditionson the horizontal pl ane are known: o =
O"y and 1: = O. Since the property of the pol estates that a horizontal l ine drawn
through it must intersect the Mohr circl e at a pointdescribing those stress
conditions, a horizontal l ine drawn through the point describingthose stress
conditions wil l intersect the Mohr circl e at the pol e. For the case of Figure 6.2,
the point of known stress conditions is point A and the pl ane for which the stress
conditionsare known is horizontal . Consequentl y, a horizontal l ine drawn through
point A must intersectthe Mohr circl e at the pol e, l abel ed as point P.Exampl e 6.1
Compute the normal and shear stresses on a pl ane passing through the el ement shown
in FigureE6.1a and incl ined at 45° cl ockwise from horizontal .Sol ution The stresses
on the horizontal pl ane are (J = 4 and 1: = +1. Drawing a horizontal l inethrough
this point reveal s the l ocation of the pol e at point P. Note that the known
stresses on the1r .-..--2~tD~"-2-J---1---(a)o-1Figure E6.1-": - ~_(4, 1)_ Stress
conditions(b)Sec. 6.2 Representation of Stress Conditions by the Mohr Circl e 187
vertical pl ane, (J = 2 and t = -I, coul d just as easil y have been used with a
vertical l ine to determinethe l ocation of the pol e. Once the pol e has been
identified, the stress conditions on anypl ane can be determined. Drawing a l ine
through the pol e paral l el to the pl ane of interest (FigureE6.1 b) shows that the
stresses on that pl ane are (J = 4 and t = -1.6 ..2.1 Principal StressesTwo points
on the Mohr circl e are of particul ar interest. The points where the circl e
intersectsthe normal stress axis describe the normal stresses on pl anes where no
shear stresses exist.Those pl anes are cal l ed principal stress pl anes and the normal
stresses that act on them arecal l ed principal stresses. The principal stress axes
are al igned in the directions of the principal stresses; therefore, they are
perpendicul ar to the principal stress pl anes. The l argest principal stress is the
majorprincipal stress, (Jj, and the smal l est is the minorprincipal stress, (h There
is al so an intermediate principal stress, (Jz, that can take on any val ue between
(Jt and (J3; acompl ete Mohr diagram woul d incl ude (J2' as shown in Figure 6.3.
Since the mechanical behavior of soil s is much more sensitive to the rel ationship
between (Jt and (J3 than to the val ueof (J2' and since (J2 and (J3 are often nearl y
equal , the val ue of (J2 is usual l y not shown.0"Figure 6.3 Mohr circl es of stress
incl uding intermediate principal stress, O"z.The pol e can be used to determine the
orientation of the principal stress pl anes. The factthat the angl e between two
l ines passing through any point on a semicircl e and the "corners"of the semicircl e
is 90° confirms that the major and minor principal stresses act perpendicul arto
each other (the intermediate principal stress acts mutual l y perpendicul ar to the
major andminor principal stresses). Figure 6.4a shows the orientation of the
principal stress axes on theel ement shown previousl y in Figure E6.l b. If the shear
stress, txy, is increased from I to 2-1 -1(2, -1)1: 2 1: 2o 00" 0"··2 -2(2, -2)(a)
(b)Figure 6.4 Orientation of principal stress axes for (a) 1: xy =1 and (b) 1: " =2.
Note therotation of principal stress axes that accompanies the change in 1: xy' .188
Dynamic Soil Properties Chap. 6with axand ayhel d constant, the Mohr circl e grows to
the size shown in Figure 6.4b. Note thatthe increase in shear stress is accompanied
by rotation of the principal stress axes.6.2.2 Stress PathsThe variation in stress
conditions acting on an el ement of soil can be tracked by pl otting theMohr circl e
at various stages in the l oading sequence, but such a pl ot can quickl y become
difficul t to decipher for many l oading sequences. It is much simpl er to observe the
stressconditions by pl otting the variation of the position of a singl e point on the
Mohr circl e. Thestress point usual l y sel ected is the very top of the Mohr circl e,
as shown in Figure 6.5. Thepath taken by the stress point during l oading is cal l ed
the stress path. Since many propertiesof soil are dependent on the stress path
induced by the appl ied l oading, the stress path is avery useful tool in
geotechnical engineering. [In a considerabl e body of geotechnical engineering
l iterature, particul arl y that rel ating to the constitutive model l ing of soil s, the
stresspoint is defined according to p =(al + a2 + (3)13 and q =al - a3. Al though
each form hasits own merits, they essential l y present the same information; the
form of Figure 6.5 is morecommonl y used in geotechnical earthquake engineering and
wil l be used hereafter.].(J = p = (J1 + (J321: = q = (J1 - (J321: ~ Stress pointq
---------------------(JFigure 6.5 Location and definition of stress point on which
stress path is based.Stress paths can be expressed in terms of total or effective
stresses. Since the effectivestress is equal to the difference between the total
stress and the pore pressure, the effectivestress path is described byp' =a' I +
a'3 = (al -u) + (a3-U) al +a3 2u (6.2a) ----- p-u2 2 2 2, a' 1 - a'3 (aj-u) - (a3-u)
a l -a3 (6.2b) q2 2 -2-=q(6.3)I-KoI +Koa'l (1- Ko)=a' j (1 + Ko) =(d a' I + a' 3 )2
dq = dp'Total and effective stress paths are often pl otted together; the horizontal
distance betweenthe two is equal to the pore pressure.Many soil deposits are formed
by sedimentation of soil particl es through water. Asmore and more soil is
deposited, consol idation causes the vol ume to decrease and the effectivestresses to
increase. If the process is one-dimensional (i.e., if the soil particl es moveonl y
in the vertical direction), the minor principal stress wil l be proportional to the
majorprincipal stress and the effective stress path of an el ement of soil bel ow the
ground surfacewil l move from A to B in Figure 6.6. The sl ope of the stress path in
this range is given by(d a' I - a'3 )2Sec. 6.2 Representation of Stress Conditions
by the Mohr Circl e 189p'LE ACFigure 6.6 Effective stress paths for K;consol idation
(A to B) and subsequentdrained l oading al ong axial compression(Ae), axial extension
(AE), l ateral compression (LC), and l ateral extension(LE) stress paths.where Ko=
03'/01' is the coefficient ofl ateral earth pressure at rest. If the ground surface
is l evel , the principal stress axes wil l be vertical and horizontal .After the soil
has consol idated, sl ow (drained) extemal l oading can cause the stress pathto move in
a variety of directions. If the vertical stress increases whil e the horizontal
stressremains constant (a condition often approximated by foundation l oading), the
stress path wil l move in the direction l abel ed AC in Figure 6.6. If the horizontal
stress decreases with constantvertical stress (as in the devel opment of active
earth pressure conditions), the stress pathmoves in the direction LE. If the
vertical stress is decreased with constant horizontal stress (asbeneath an
excavation), the stress path moves in the AE direction, and if the horizontal
stressincreases with constant vertical stress (passive earth pressure conditions),
the stress pathmoves in the direction l abel ed LC. For each of these ideal ized
conditions, no shear stresses areinduced on vertical or horizontal pl anes;
consequentl y, the principal stress axes remain vertical and horizontal (al though
they wil l instantaneousl y exchange positions if the p'-axis is crossed).Most
real istic l oading conditions invol ve simul taneous changes in horizontal and
vertical stresses and/or the devel opment of shear stresses on horizontal and
vertical pl anes. Consideran el ement of soil beneath a l evel ground surface (Figure
6.7a) subjected to vertical l ypropagating s-waves, At stage A, the el ement is under
at-rest conditions with the Mohr circl eas indicated in Figure 6.7b and the stress
path at point A in Figure 6.7c. Since the major principal stress is vertical , the
pol e of the Mohr circl e is at the point (a'h, 0) in Figure 6.7b. A vertical l y
propagating shear wave wil l produce shear stresses on horizontal and vertical
pl anesand distort the el ement as shown in stage B of Figure 6.7 a. Since the shear
stresses increasewhil e the vertical and horizontal stresses remain constant, the
radius of the Mohr circl eincreases but the center does not move (Figure 6.7b). The
stress path (Figure 6.7c) moves vertical l y,as does the position of the pol e (Figure
6.7b), which indicates that the principal stressaxes are rotated from their initial
vertical and horizontal positions. Since the horizontal shearstresses are cycl ic in
nature, their direction wil l reverse after going back through the 'thv ='tvh= 0
position in stage C. Note that the stress conditions at stage C are identical to
those of stageA, and that the principal stress axes have rotated back to the
vertical and horizontal positions.At stage D, the shear stresses act in the
opposite direction and the principal stress axes rotatein the opposite direction as
in stage B. Thus the l oading induced by vertical l y propagatingshear waves can be
described by the stress path of Figure 6.7c and a principal stress axis rotation.
Note that the stress path never indicates isotropic stress conditions (it never
reaches thep'-axis) and that the principal stress axes rotate continuousl y.The
nature of principal stress axis rotation is significant. Research (e.g., Wong and
Arthur, 1986; Symes et al ., 1988; Sayao and Vaid, 1989) has shown that principal
stressqA ..",~------------------..190 Dynamic Soil Properties Chap. 6cr~t 1: vh--~I
B "It~ crht(a)A,C(b)q IB,DA,C(c)P'Figure
6.7 (a) Stress and strain conditionsimposed on el ement of soil bel ow l evel ground
surface by vertical l y propagating shear waves at four different times; (b) Mohr
circl es, l ocations of pol es, and orientations of major principal stress axis; (c)
stress path.rotation can cause shear and vol umetric strain by itsel f (i.e., even if
the stress point does notmove). Hence some of the strain induced by vertical l y
propagating shear waves resul ts fromprincipal stress rotation; this effect is not
present in many fiel d and l aboratory tests.Exampl e 6.2A reconstituted triaxial
specimen of dry sand is consol idated isotropical l y to an effective confining
pressure of 200 kPa, and then l oaded in drained triaxial compression to a deviator
stress(o1 - (j3) of 200 kPa. At that point the specimen is subjected to a harmonic
deviator stress thatoscil l ates between 100 and 300 kPa. Pl ot the total and
estimated effective stress paths.Sol ution Because the sand is dry, no pore
pressures exist, so the total and effective stressesare equal . During preparation,
the stresses acting on the specimen are very smal l , so the stresspath is at point A
in Figure E6.2. Isotropic consol idation takes the stress path to point B, and
drained triaxial compression takes it to point C. Harmonic l oading then causes the
stress path tooscil l ate between points D and E.q (kPa)200 --100 -Ao100 200 300 400
Figure E6.2Sec. 6.3 Measurement of Dynamic Soil Properties 1916.3 ME}l ~SUREMENTOF
DYNAMIC SOIL PROPERTIESThe measurement of dynamic soil properties is a critical
task in the sol ution of geotechnical earthquake engineering probl ems. A wide variety
offiel d and l aboratory techniques are avail abl e,each with different advantages and
l imitations with respect to different probl ems. Manyare oriented toward measurement
of l ow-strain properties and many others toward propertiesmobil ized at l arger
strains. The sel ection of testing techniques for measurement of dynamicsoil
properties requires careful consideration and understanding of the specific probl em
athand. Efforts shoul d al ways be made to use tests or test procedures that
repl icate the initial stress conditions and the anticipated cycl ic l oading
conditions as cl osel y as possibl e.Soil properties that infl uence wave propagation
and other l ow-strain phenomenaincl ude stiffness, damping, Poisson's ratio, and
density. Of these, stiffness and damping arethe most important; the others have
l ess infl uence and tend to fal l within rel ativel y narrowranges. The stiffness and
damping characteristics of cycl ical l y l oaded soil s are critical to theeval uation of
many geotechnical earthquake engineering probl ems-not onl y at l ow strainsbut
because soil s are nonl inear material s, al so at intermediate and high strains. At
high l evel sof strain, the infl uence of the rate and number of cycl es of l oading on
shear strength mayal so be important. Vol ume change characteristics are al so
important at high strain l evel s.The measurement of these important soil properties
in fiel d and l aboratory tests is presentedin the fol l owing sections. Many of the
tests have been devel oped specifical l y to measuredynamic soil properties; others
are modified versions of tests commonl y used tomeasure soil behavior under
monotonic l oading conditions. The appl icabil ity of the varioustests to dynamic soil
properties is emphasized here-descriptions of their appl ications tostatic
properties may be found in standard geotechnical engineering texts (e.g., Lambe and
Whitman, 1969; Hol tz and Kovacs, 1981).Any investigation of dynamic soil properties
shoul d be performed with due recognitionof the inevitabl e uncertainty in measured
properties. Sources of uncertainty incl ude the inherentvariabil ity of soil s (a
resul t of the geol ogic environment in which they were deposited),inherent
anisotropy (a function of the soil structure or "fabric"), induced anisotropy
(causedby anisotropic stress conditions), dril l ing and sampl ing disturbance,
l imitations of fiel dand/or l aboratory testing equipment, testing errors, and
interpretation errors. Some of thesesources of uncertainty can be minimized by
careful attention to test detail s, but others cannot.61.3.1 Fiel d TestsFiel d tests
al l ow the properties of the soil to be measured in situ (i.e., in their existing
statewhere the compl ex effects of existing stress, chemical , thermal , and
structural conditions arerefl ected in the measured properties). The measurement of
dynamic soil properties by fiel dtests has a number of advantages. Fiel d tests do
not require sampl ing, which can al ter thestress, chemical , thermal , and structural
conditions in soil specimens. Many fiel d tests measurethe response of rel ativel y
l arge vol umes of soil , thereby minimizing the potential forbasing property
eval uation upon smal l , unrepresentative specimens. Many fiel d tests inducesoil
deformations that are simil ar to those of the probl em of interest, particul arl y for
wavepropagation and foundation design probl ems. On the other hand, fiel d tests do
not al l ow theeffects of conditions other than the in situ conditions to be
investigated easil y, nor do they192 Dynamic Soil Properties Chap. 6al l ow porewater
drainage to be control l ed. In many fiel d tests, the specific soil property of
interest is not measured but must be determined indirectl y, by theoretical anal ysis
or empirical correl ation.Some fiel d tests can be performed from the ground surface,
whil e others require thedril l ing of borehol es or the advancement of a probe into
the soil . Surface tests are often l essexpensive and can be performed rel ativel y
quickl y. They are particul arl y useful for material sin which dril l ing and sampl ing
or penetration is difficul t. Borehol e tests, on the otherhand, have the advantage
of the information gained directl y from the boring: visual and l aboratory-
determined soil characteristics, water tabl e l ocation, and so on. Al so, the
interpretationof borehol e tests is usual l y more direct than that of surface tests.
6.3.1.1 Low-Strain TestsLow-strain tests general l y operate at strain l evel s that
are not l arge enough to inducesignificant nonl inear stress-strain behavior in the
soil , typical l y at shear strains bel ow about0.001 %. As such, most are based on the
theory of wave propagation in l inear material s.Many invol ve the measurement of body
wave vel ocities which can easil y be rel ated to l owstrainsod modul i. Others invol ve
the devel opment of standing waves, whose measured frequenciesand/or wavel engths can
be used to compute l ow-strain modul i.Seismic geophysical tests represent an
important cl ass of fiel d tests for determinationof dynamic soil properties. Seismic
tests invol ve the creation of transient and/or steadystatestress waves and the
interpretation of their behavior from measurements made at oneor more different
l ocations. In many seismic tests, a source produces a "pul se" of waveswhose times
of arrival are measured at distant receivers. The source, which may range froma
sl edgehammer bl ow to the ground surface to a buried expl osive charge, wil l
general l y producep-waves, s-waves, and surface waves. The rel ative ampl itudes of
each depend on howthe impul se is generated. Expl osive sources and vertical impact
sources (Figure 6.8a and b)are rich in p-wave content. SH-waves are produced most
efficientl y by striking the end ofa beam pressed tightl y against the ground surface
(Figure 6.8c).Since p-waves travel fastest, their arrival s at distant receivers are
most easil y detectedand their arrival times most easil y measured. S-wave resol ution
can be improved markedl yby re\iersing the pol arity of the impul se, as is easil y
accompl ished for SH-waves by strikingthe other end of the beam of Figure 6.8c.
Since the pol arity of the train of p-waves is notreversed, subtracting the reversed
record from the original record wil l diminish the p-waveampl itudes whil e enhancing
the s-wave ampl itudes. Wave arrival s can al so be enhanced byadding, or "stacking,"
records from mul tipl e impul ses; the random noise portions of therecords tend to
cancel each other whil e the actual waves are reinforced.D==([;~); \\ ': .."./ /'- »>
(a) (b) (c)Figure 6.8 Different methods for creation of impul sive disturbances for
seismicgeophysical tests: (a) shal l ow expl osives; (b) vertical impact; (c)
horizontal impact.Sec. 6.3 Measurement of Dynamic Soil Properties 193Careful
consideration of groundwater conditions is essential for proper interpretation of
seismic geophysical test measurements. P-waves travel through groundwater at about
5000ft/sec, depending on temperature and sal inity. Soft, saturated soil s may
propagate p-waves atthese high vel ocities even though the vel ocity is not
indicative of the stiffness of the soil skel eton.Fail ure to consider groundwater
effects can resul t in significant overestimation of soil stiffness. The groundwater
probl em can l argel y be avoided by using s-waves which are propagatedby the soil
skel eton and not the groundwater.Seismic Refl ection Test. The seismic refl ection
test al l ows the wave propagationvel ocity and thickness of surficial l ayers to be
determined from the ground surface orin offshore environments. The test and its
interpretation are conceptual l y very simpl e. Therefl ection test is most useful for
investigation of l arge-scal e and/or very deep stratigraphy.It is rarel y used for
del ineation of shal l ow soil l ayers.For the simpl e profil e shown in Figure 6.9a, the
test is performed by producing animpul se (usual l y rich in p-waves) at the source,
S, and measuring the arrival time at thereceiver, R. The impul se produces stress
waves that radiate away from the source in al l directions with a hemispherical
wavefront. Some of the wave energy fol l ows a direct pathfrom S to R and arrives at
R attd= distance of travel = x (6.4)wave vel ocity Vpl By measuring x and td, the p-
wave vel ocity of the upper l ayer,
vpl , can easil y be determined.Another portion of the impul se energy travel s
downward and strikes the horizontal l ayerboundary at an angl e of incidencei = tan"
2~ (6.5)The part of that wave that is refl ected back toward the ground surface
arrives at the receiver atdistance of travel tr =wave vel ocity(6.6)s x RDirect wavex
~~1~ ___H(a) (b)Figure 6.9 (a) Ray path for incident and refl ected p-wave from
horizontal l ayerboundary; (b) variation of travel time for direct and refl ected
waves. Difficul ty inresol ution increases with increasing source-receiver
separation.194 Dynamic Soil Properties Chap. 6By measuring t, and knowing x and vp
I from the direct wave cal cul ation, the thickness of theupper l ayer can be
cal cul ated as(6.7)Figure 6.9b shows how the arrival times for the direct and
refl ected waves are rel ated; thedifference in arrival times decreases with
increasing distance. When, as is usual l y the case,actual conditions differ from the
simpl e assumptions of horizontal l ayering above, mul tipl emeasurements must be made.
In the case of the incl ined l ayer boundary shown in Figure6.10, for exampl e,
travel -time measurements at receivers A and B can be used to determinethe angl e of
incl ination asV~I (tRA + tRS)(tRS - tRA) XA + Xs sina = --- (6.8)4za ( xs -XA) 4za
where v,! =xAl tdA=xitds and tdA and tdsare the direct wave arrival times at
receivers A andB, resp~ctivel y. If receiver A is pl aced at the source (xA =O),then
tRAVpl Za = 2 (6.9)sina (6.10)V~I(tRA +tRS)(tRs-tRA)-X~2tRAVp l XSThe
characteristics of deeper l ayers may be eval uated using refl ections from deeper
interfaces(Griffiths and King, 1965; Ewing et al ., 1957; Kl eyn, 1983). The method
is l imited,however, by the difficul ty associated with determining the arrival time
of the refl ectedwaves, particul arl y for cases in which refl ected waves arrive whil e
the receivers are stil l responding to direct waves. Interpretation of resul ts for
profil es with l ow-vel ocity l ayersmay al so be difficul t.A Source BFigure 6.10 Source
and receiver l ayout for refl ection test with incl ined l ayer boundary.Exampl e 6.3A
l oose deposit of overconsol idated cl ay is underl ain by bedrock. Previous subsurface
investigationsin the area suggest that the bedrock surface is nearl y horizontal . A
seismic refl ectionSec. 6.3 Measurement of Dynamic Soil Properties 195survey shows
the arrival of distinct p-waves at a geophone 38 msec and 200 msec after an
impul sive l oad is appl ied at a point 20 m from the geophone. Determine the
thickness and thep-wave vel ocity of the cl ay deposit.Sol ution Assuming that the
first p-wave arrival is caused by the direct p-wave, the p-wavevel ocity is given by
x 20 mVpl = td 0.038 sec = 526 ml secIf the second p-wave is due to the refl ected
wave, the thickness of the deposit is given byH = !Jt~V~I-x2 = !,)
(0.200sec)2(526m1sec)2_(20m? = 51.6mSeismic Refraction Test. The seismic refraction
test el iminates the mostimportant l imitation of the seismic refl ection test by
using the arrival times of the firstwaves, regardl ess of path, to reach a given
receiver. The test invol ves measurement of thetravel times of p- and/or s-waves
from an impul se source to a l inear array of points al ong theground surface at
different distances from the source. Al though the seismic refraction test ismore
commonl y used than the seismic refl ection test, its greatest earthquake engineering
appl ication is al so for del ineation of major stratigraphic units.A typical test
setup is shown in Figure 6.11. An impul sive energy source, which canbe mechanical
or expl osive, is l ocated at or near the ground surface. A series of receivers,
usual l y geophones, are pl aced in a l inear array. One receiver is l ocated at the
source. Theoutput of al l of the receivers is recorded when the impul se l oad is
triggered. From theserecordings the arrival times of the first waves to reach each
receiver can be determined andTravel time ofCii first arrival s.1': 20~80
,----.,.-----c-------,.-----,---------,.------.200 300 400 500Distance (ft)100
O .....---;.--7"--;..----+--.......;---'---!---....;...-----IoRecording,equipment:
GeophonesExpl osive chargein shal l ow hol eFigure 6.11 Seismic refraction test setup.
(After Redpath, 1973.)196 Dynamic Soil Properties Chap. 6(6.11)pl otted as a
function of source-receiver distance, as shown in Figure 6.11. If the receiverswere
al l excited by a singl e stress wave travel ing al ong the ground surface at constant
vel ocity,v (the test can be performed and interpreted with p- or s-waves, so v
coul d be vp or vs ) 'the arrival time-distance pl ot woul d be a straight l ine of
sl ope l /v that passed through theorigin. Figure 6.11 shows that this is cl earl y not
the case-a different, more compl icatedmechanism is at work.Horizontal Layering.
Assume that the seismic refraction test is being conductedon the surface of a two-
l ayered el astic hal f-space as shown in Figure 6.13. Theimpul se produces stress
waves that travel away from the source in al l directions with ahemispherical
wavefront. Some of the energy travel s directl y from the source to the receiversin
the form of a direct wave, arriving at the nth receiver at a travel timeXntdn = VI
where vI is the wave propagation vel ocity of material 1. Other rays travel downward
towardthe boundary between material s 1 and 2. At that boundary, these rays are
refl ected andrefracted, with the directions ofthe refracted rays determined by
Snel l 's l aw (Section 5.4.2).At the critical angl e of incidence, i.; the refracted
ray wil l travel paral l el to the boundary.According to Huygens' principl e (which
says that any point on a wavefront acts as thesource of a new disturbance) and
Snel l 's l aw, this critical l y refracted wave wil l produce ahead wave in material 1
that wil l travel at vI in a direction incl ined at (900- iJ to the boundary.The
resul ting wavefront can have a portion control l ed by the direct wave and a portion
control l ed by the head wave, as il l ustrated in Figure 6.12. Note that the direct
wave producesthe first wave arrival at short source-receiver distances, but the
head wave arrivesbefore the direct wave at distances greater than the critical
distance, Xc-At distances greater than Xc, a ray that travel s downward at vel ocity
VI through material I, is critical l y refracted to travel in material 2 at vel ocity
Vl and is then critical l yrefracted back up through material 1 at vel ocity, VI wil l
reach a receiver faster than a ray thattravel s al ong the shorter direct path at
vel ocity VI' The travel time required for the head waveto reach the nth receiver
(Figure 6.13) can be written asxDirectwaves\<: ': ///\t~Figure 6.12 Wavefronts for
first-arriving waves in a seismic refraction survey. Note thatfirst arrival s near
the source are from direct waves but, at distances greater than the critical
distance, XC' the first arrival s are from head waves. (After Corps of Engineers,
1979.)Sec. 6.3 Measurement of Dynamic Soil Properties 197Figure 6.13 Travel path
for first arrival when x.o x.,Hsk-I~--r- H.i.._=': ~~,.: =~~~---': '----"-_H xn-2Htanie
Hthn = - .- -. + +--.-.VI COS Ie V2 VI COS Ie(6.12)Substituting the resul t of
Snel l 's l aw for critical incidence, sin ie =v/v2, and the trigonometricidentity
cos2ic = 1 - sin2iC' and rearranging yiel dsth = X n + 2H p: - _1 (6.13)n V2 vf v~If
a receiver was pl aced exactl y at the critical distance, XC' the direct wave and the
head wavewoul d reach it at exactl y the same time (i.e., tdn = thn)' Consequentl y,
from equations (6.i 1)and (6.13),(6.14)from which(6.15)Therefore, the travel time-
distance diagram al l ows three important characteristics of thesubsurface
conditions, namel y VI, V2, and H, to be obtained.For the case of mul tipl e
horizontal l ayers, the travel time-distance diagram wil l exhibit more than one break
in sl ope, as il l ustrated in Figure 6.14. The distances correspondingto these sl ope
breaks can be used, al ong with the sl opes themsel ves, to determinethe thicknesses
of deeper l ayers. The thickness of the kth l ayer, for exampl e, woul d be givenby
(Corps of Engineers, 1979)(k ~ 2) (6.16)k-I . ~. J 2 2H~ Xck Vk+I-Vk "Hj Vk + l
AjVk- V] - Vk Vk+l -Vjk - - +"", ~ --'------;: : : ¢======-----'-2 Vk+l +Vk Vj J~2 _v2
j=! k+l kIn the preceding paragraphs, it has been expl icitl y assumed that the
vel ocity of eachl ayer is smal l er than the l ayer immediatel y bel ow it. For marl y
geol ogic conditions this is agood assumption, but when it is not, the resul ts of a
seismic refraction test can be misl eading.A l ow-vel ocity l ayer underl ying a higher-
vel ocity l ayer (i.e., a vel ocity reversal ) wil l notappear as an individual segment
on the travel time-distance diagram. Instead, it wil l causethe computed depths of
the l ayer boundaries to be greater than the actual depths (Redpath,1973). Al so,
bl ind zones, where a subsurface l ayer exists but is not indicated by the travel
time-distance diagram, can be caused by insufficient l ayer thickness or
insufficient vel ocity198Xe2 xe3DistanceDynamic Soil Properties Chap. 6Figure 6.14
Travel time-distancediagram for mul tipl e horizontal l ayers.(After Redpath, 1973.)
contrast (Soske, 1959). In such cases the head wave from a deeper l ayer can
overtake thehead wave of an intermediate l ayer before it reaches the ground
surface. The undetectedexistence of a bl ind zone wil l cause the computed depth of
the deeper l ayer to be l ess than theactual depth (Redpath, 1973). In cases where
the vel ocity increases continuousl y with depth(e.g., sands, gravel s, normal l y
consol idated cl ays), ray paths wil l be curved rather thanstraight, as previousl y
assumed. For the case where vel ocity is proportional to depth (Figure6.15). the ray
paths wil l become circul ar arcs. The resul ting time-distance curves can be
transformed into curves of vel ocity versus depth (Redpath, 1973; Corps of
Engineers, 1979).Incl ined
or Irregul ar Layering. When the boundaries between l ayers arenot paral l el , the
travel time-distance diagram wil l not yiel d the true vel ocities of al l l ayers
directl y since the apparent vel ocity (the distance between adjacent receivers
divided by thedifference in their arrival times) is infl uenced by the sl ope of the
l ayer boundaries and thecritical angl es of incidence. Referring to Figure 6.16, the
apparent vel ocity from a seismicrefraction test in the down-dip direction, V2D, is
l ower than the apparent vel ocity, vw, froman identical test in the up-dip
direction. From Snel l 's l aw,VI sin u, + ex) =V2D(6.17a)Vj sin(ic - ex) = (6.17b)V2U
which can be rearranged to produce the apparent dip angl e (the dip angl e in the
vertical pl ane of the array of receivers)Sec. 6.3 Measurement of Dynamic Soil
Properties 199Q)E+=~Sl ope = ~1 : \SIOP8 = Vx/' i /'/ .,/' : : "" : : v.. ~x~~_I_s _
DistancezFigure 6.15 Ray paths and travel timedistancediagram for a singl e l ayer
withl inearl y increasing vel ocity. (After Redpath,1973.)'" = I( .-I (6.18) VI . -I
VI ) u, 2 sm --sm -V2D V2UThe apparent dip angl e is equal to the true dip angl e
onl y when the dip vector l ies within thepl ane of the receiver array. In other
cases, another survey util izing a nonparal l el array ofreceivers is required to
determine the true dip (Richart et al ., 1970). The thickness of the upperl ayer,
measured perpendicul ar to the l ayer boundary at each shot point, can be computed as
t, r,Q) E t.;+=~';: : -<TidFigure 6.16 Reverse profil ing used toidentity irregul ar
or sl oping interfaces.Concept of apparent vel ocity in underl yingl ayer is al so
il l ustrated. (After Redpath,1973.~200The true val ue of V2 is given byV2DdV,Tid2cos
aDuv.r.;2cos aDynamic Soil Properties Chap. 6(6.19a)(6.19b)(6.20)The preceding
discussion il l ustrates the importance of reverse profil ing for cases ofincl ined
l ayering. Since the nature of the l ayering is not often known in advance, as wel l
as toprovide additional data, reverse profil ing is routinel y performed.Exampl e 6.4A
seismic refraction survey with reverse profil ing between two shot points l ocated
120 m apartshows the p-wave arrival times l isted bel ow" Determine the thickness of
the surficial l ayer of soil .Distance from p-Wave Arrival Distance from p-Wave
Arrival Geophone Shot Point A Time (msec) Shot Point B Time (msec)A 0 0 120 88B 5 II
100 78C 10 26 80 67D 20 49 60 58E 40 65 40 47F 60 71 20 37G 80 76 10 26H 100 83 5
12I 120 88 0 0Sol ution The arrival time-distance diagrams for the forward and
reverse profil es are pl ottedin Figure E6.4.Distance from shot point B (m)120 100 80
60 40 20 0100 .. .. 100U 28 UQ) 96 Q)(fJ (fJ~ <. -- - -- -- - --- EQ)E 50··· 50 E
: ;: : : 55 : ;: : : ""iii 107 ""iii"e; 60 Tid> ";: : ~ ~24 3330··· ,- 00 20 40 60 80 100
120Distance from shot point A (m)FigureE6.4Sec. 6.3 Measurement of Dynamic Soil
Properties 2011945 m/sec3429 m/secFrom the initial sl opes of the travel -time
diagrams, the average p-wave vel ocity of the surficial l ayer is given byI( 24m 13m)
VI = - + = 397 m/sec2 0.060 sec 0.033 secThe apparent vel ocities in the down-dip
and up-dip directions are107mVw = 0.055 sec96 mVw = 0.028 secThe apparent dip angl e
can then be computed as_ 1( . -I VI . -I VI ) _ I ( . _I 397 . -I 397 ) = 2.560ex -
: 2 SIn V2D - SIn Vw -: 2 SIn 1945 - SIn 3429Then the depths of the l ayer boundary,
measured perpendicul ar to the boundary, at shot pointsA andB are5.8 mD_ vl TiuB -
--- 2 cos ex(397 m/sec)(0.029 sec)2 cos 2.56°(397 m/sec)(0.053 sec) = 10.5 m2 cos
2.56°A more general procedure can be devel oped to interpret the resul ts of tests on
soil profil eswith l ayers of variabl e thickness. Consider the site of Figure 6.17,
where waves fromthe two sources SP1 and SPz reach a receiver at point D at arrival
times TDI and Ttn- respectivel y.The total time [i.e., the total travel time from
the l ocation of SPI to that of SPz (andvice versa)], is given byr, = AB + BCEF +FG
VI Vz VI(6.21)<DE TD2F --1 ~V1o Arrival times from SP1n Arrival times from SP2I
Distance.. s------- _x_SP1 . ..... ,ZD. . .... .. SP2Z'~.~ Figure 6.17 Forward
andreverse profil esfor l ayerof irregul ar thickness. (After Corpsof Engineers,
1979.)202 Dynamic Soil Properties Chap. 6where ZI and Z2 are the vertical
thicknesses of the surficial l ayer at SPj and SP2 , respectivel y.The arrival times
at point D are given by(6.22)Tm = FG + EF + DE = Z2. + (S-XD) -Z2 tan ie-zDtan i e
+ ZD. (6.23)vI V2 vI VjCOSl e V2 VICOSl eThe vel ocity of the l ower l ayer, v2' is
inversel y proportional to the difference between thearrival times from the forward
and reverse profil es (i.e., a pl ot of arrival time difference versusdistance has a
sl ope of 2Iv2) . Combining equations (6.21) through (6.23) and substitutingv2 =
v/sin ic yiel dsTherefore2 1. 2.ZD - SIl l IeVI COS i e(6.24)(6.25)l (TD] + Tin -
Tt)vIZ = -"----'=--=-....: ..: .-....: D 2 cos i eBy moving the receiver (point D) to
different l ocations (or using mul tipl e geophones), thevariation of surficial l ayer
thickness between SPI and SP2 can be determined. Impl icit in thepreceding
derivation is the assumption that the first arrival s at D are from head waves from
the same l ayer rather than from direct waves. Long arrays may be required to ensure
theval idity of this assumption over distances of interest.Suspension Logging Test.
Suspension l ogging, al though commonl y used inpetrol eum expl oration, has onl y
recentl y been appl ied to geotechnical earthquake engineeringprobl ems. A probe 5 to
6 m (IS to 20 ft) l ong is l owered into an uncased borehol e fil l edwith water or
dril l ing fl uid. A horizontal reversibl e-pol arity sol enoid l ocated near the base of
the probe produces a sharp, impul sive pressure wave in the dril l ing fl uid. Upon
reaching theborehol e wal l , the pressure wave produces both P: and s-waves in the
surrounding soil . Thesewaves travel through the soil and eventual l y transmit energy
back through the dril l ing fl uid totwo biaxial geophones l ocated about 1 m (3 ft)
apart near the top of the probe. To enhanceidentification of p- and s-wave
arrival s, the procedure is repeated with an impul se of oppositepol arity.
Differences in arrival times are used to compute the average p- and s-wave
vel ocitiesof the soil between the geophones. The suspension test al l ows measurement
of wave propagationvel ocities in a singl e, uncased borehol e, but the frequencies of
the waves (500 to 2,000Hz for s-waves and 1000 to 3000 Hz for p-waves) are much
higher than those of interest ingeotechnical earthquake engineering.Because the
sol enoid travel s with the geophones down the borehol e, the ampl itude ofthe signal s
is rel ativel y constant at al l depths. As a resul t, the suspension l ogging test is
effective at great depths-up to 2 km (Nigbor and Imai, 1994). By overl apping
measurementpoints, resol utions of l ess than 1 m (3.1 ft) can be obtained. This
capabil ity is particul arl yuseful at sites that may have thin l ayers of soft or weak
soil . The presence of tubewaves may l imit the effectiveness of the suspension
l ogging test in cased borehol es.Sec. 6.3 Measurement of Dynamic Soil Properties 203
Steady-State Vibration (Rayl eigh Wave) Test. The probl em of detectingwave arrival s
and measuring arrival times is el iminated in tests that interpret propertiesfrom
the characteristics of steady-state vibrations. The displ acements al ong the ground
surfaceadjacent to a vertical l y vibrating circul ar footing are caused primaril y by
Rayl eighwaves (Mil l er and Pursey, 1955). Since Rayl eigh waves produce both vertical
and horizontal displ acements, the ground surface wil l , for a constant l oading
frequency, be distorted asshown in exaggerated form in Figure 6.18. By pl acing a
receiver at the center of the footingand moving another receiver to points at
different distances from the receiver, the l ocationsof points vibrating in-phase
can be determined. The horizontal distances between suchpoints are equal to the
wavel ength of the Rayl eigh wave. By measuring the Rayl eigh wavel ength,the Rayl eigh
wave phase vel ocity, VR, can be cal cul ated asVR = ffiARl 2rr = fAR (6.26)From the
phase vel ocity, Poisson's ratio, and Figure 5.10, the shear wave vel ocity can be
estimated. For many soil s, Vs "" l .09VR'For soil s whose stiffness varies with
depth, dispersion wil l cause the Rayl eigh wavephase vel ocity to vary with
frequency. The shape of the Rayl eigh wave displ acement profil e(Figure 5.10)
suggests that the measured phase vel ocity corresponds to the soil properties ata
depth of about AR/3 (Gazetas, 1991) to ARI2 (Heukel om and Foster, 1960; Richart et
aI.,1970). By varying the l oading frequency in the fiel d, the variation of shear
wave vel ocitywith depth can be estimated. The steady-state vibration test is useful
for determining thenear-surface shear wave vel ocity but cannot easil y provide
detail ed resol ution of highl yvariabl e vel ocity profil es. For geotechnical
earthquake engineering appl ications, the steadystatevibration test has l argel y been
suppl anted by the spectral anal ysis of surface waves test.---- AR ----....1...----
AR --------;~Figure 6.18 Rayl eigh wave-induced deformation of ground surface
adjacent tovertical l y vibrating footing. (After Richart et aI., 1970.)Spectral
Anal ysis of Surface Waves Test. The shape of a dispersioncurve [i.e., a pl ot of
Rayl eigh wave vel ocity versus frequency (or wavel ength)], at a particul arsite is
rel ated to the variation of body wave vel ocities with depth. The preceding
steadystatetest can be used to generate a dispersion curve by repeating the test at
different l oadingfrequencies. This process, however, tends to be quite time
consuming in the fiel d. With theuse of digital data acquisition and signal -
processing equipment, a dispersion curve can beobtained from an impul sive or random
noise l oad.
The measurement and interpretation ofdispersion curves obtained in this way, known
as spectral anal ysis of surface waves(SASW) (Heisey et aI., 1982; Nazarian and
Stokoe, 1983; Stokoe et aI., 1994), is one of themost significant recent advances
in shal l ow seismic expl oration.The SASW test is performed by pl acing two vertical
receivers on the ground surface inl ine with an impul sive or random noise source, as
il l ustrated in Figure 6.19. The output of both204Source + Receiver • 0Ie d l -Ie d2
Dynamic Soil PropertiesReceiverO~-IChap. 6Figure 6.19 Typical configuration of
source and receivers in a SASW test. Receiverspacing is changed in such a way that
d, + d2 remains constant.receivers is recorded and transformed to the frequency
domain using the fast Fourier transform.After transformation, the phase difference,
<\>(1), can be computed for each frequency.The corresponding travel time between
receivers can be cal cul ated for each frequency fromf1t(f) = <\>(!)2IT!(6.27)Since
the distance between receivers, Sd =d2- dj, is known, the Rayl eigh wave phase
vel ocity and wavel ength can be cal cul ated as functions of frequency: (6.28)(6.29)
With modern el ectronic instrumentation, these cal cul ations can be performed in the
fiel dvirtual l y in real time. The resul ts can be used to pl ot the experimental
dispersion curve (Figure6.20). Whil e the test shoul d, in theory, yiel d good resul ts
for a singl e receiver spacing,practical considerations dictate that several
different receiver spacings be used. At eachspacing, the midpoint between the two
receivers is kept at the same distance from the source.Identification of the
thickness and shear wave vel ocity of subsurface l ayers invol vesthe iterative
matching of a theoretical dispersion curve to the experimental dispersion curve.The
Haskel l -Thomson sol ution (Thomson, 1950; Haskel l , 1953) for a series of uniform
el asticl ayers of infinite horizontal extent is used to predict the theoretical
dispersion curve. Initial estimates of the thickness and shear wave vel ocity of each
l ayer are then adjusted until theval ues that produce the best fit to the
experimental dispersion curve are identified. This identificationprocedure is
usual l y referred to as inversion (Nazarian, 1984). For profil es in whichthe shear
wave vel ocity varies irregul arl y with depth, the dispersion curve may be infl uenced
by higher-mode Rayl eigh waves (Gucunski and Woods, 1991; Tokimatsu et aI., 1992).
SASW tests have a number of important advantages over other fiel d tests. They canbe
performed quickl y, they require no borehol e, they can detect l ow-vel ocity l ayers,
andthey can be used to considerabl e depth (> 100 m). Comparison of shear wave
vel ocity profil esobtained from SASW testing and cross-hol e testing (see the next
section) have showngood agreement (Hil tunen and Woods, 1988). SASW testing is
particul arl y useful at siteswhere dril l ing and sampl ing are difficul t; it has been
used successful l y in such material s asSec. 6.3 Measurement of Dynamic Soil
Properties 205oooro'ooPhase vel ocity (ft/sec)200 400 600 800Figure 6.20
Experimental dispersioncurve from SASW test. (After Gucunskiand Woods, 1991; used
by permission ofthe University of Missouri-Rol l a.)gravel s and debris fl ow deposits
(Stokoe et aI., 1988) and l andfil l s (Kavazanjian et aI.,1(94). The procedure does,
however require special ized equipment and experienced operators.Its appl icabil ity
is al so l imited to sites at which the assumptions of the Haskel l -Thomsonsol ution
(e.g., horiziontal l ayering) are at l east approximatel y satisfied.Seismic Cross-Hol e
Test. Seismic cross-hol e tests use two or more borehol esto measure wave propagation
vel ocities al ong horizontal paths. The simpl est cross-hol e testconfiguration
(Figure 6.21a) consists of two borehol es, one of which contains an impul seenergy
source and the other a receiver. By fixing both the source and the receiver at the
samedepth in each borehol e, the wave propagation vel ocity of the material between
the borehol esat that depth is measured. By testing at various depths, a vel ocity
profil e can be obtained.When possibl e, use of more than two borehol es is desirabl e
(Figure 6.21b) to minimize possibl einaccuracies resul ting from trigger time
measurement, casing and backfil l (material pl aced between the casing and the
borehol e wal l ) effects, and site anisotropy. Wave propagationvel ocities can then be
cal cul ated from differences in arrival times at adjacent pairsof borehol es. Arrival
times can be determined by eye using points of common phase (firstarrival , first
peak, first trough, etc.) or by the type of cross-correl ation techniques commonl y
used in petrol eum expl oration (Roesl er, 1977).Since the impul se sources must be
l ocated in the borehol e, variation of the p-wave/swavecontent is more difficul t
than for methods in which it is at the surface. When expl osivesources are used, the
wave content is shifted toward higher p-wave content when l argercharges are used,
particul arl y when detonated above the ground surface (Woods, 1978). Anumber of
mechanical impul se sources have been used, incl uding the driving of a standard
penetration test (Section 6.3.1.2) sampl er, vertical impact l oading of rods
connected to borehol epackers or jacks, torsional impact l oading of a torque foot at
the bottom of the borehol e(Stokoe and Hoar, 1978), and other techniques (Appl egate,
1974; Stokoe and Abdel -razzak,1975; Aul d, 1977). The best resul ts are general l y
obtained when the pol arity of the impul se206 Dynamic Soil Properties Chap. 6(a) (b)
Figure 6.21 Seismic cross-hol e test: (a) direct measurement using two-hol e
configuration; (b) interval measurement using three-hol e configuration.source is
reversibl e, hence the frequent preference for mechanical sources over expl osive
sources.The cross-hol e test often al l ows individual soil l ayers to be tested since
l ayer boundariesare frequentl y nearl y horizontal . It can al so detect hidden l ayers
that can be missed by seismicrefraction surveys. Cross-hol e tests can yiel d
rel iabl e vel ocity data to depths of 30 to 60 m(100 to 200 ft) using mechanical
impul se sources, and to greater depths with expl osivesources. On the other hand,
the sensitivity of the measured vel ocities to source-receiver distanceoften
requires borehol e deviation surveys, particul arl y for borehol es more than 15 to 20m
(50 to 65 ft) deep. The measured vel ocities may not be equal to the actual
vel ocities whenhigher-vel ocity l ayers exist nearby. In such cases, more advanced
methods of interpretationthat can account for refraction (e.g., Butl er et aI.,
1978) are required. Hryciw (1989) presentedmethods for correction of ray-path
curvature in material s of continuousl y varying vel ocity.Ampl itude attenuation
measurements from cross-hol e tests invol ving three or moreborehol es has been used
to compute the material damping ratio of soil s (Hoar and Stokoe,1984; Mok et aI.,
1988; EPRI, 1993). The procedure requires accuratel y cal ibrated and oriented
receivers that are wel l coupl ed to the borehol e wal l . By assuming a radiation
pattern, theeffects of geometric attenuation (radiation damping) can be separated
from the measuredattenuation to l eave the attenuation due to material damping. The
required assumptions rendersuch approaches best suited to sites of simpl e geometry
and homogeneous soil conditions.Exampl e 6.5Determine the SV-wave vel ocity from the
cross-hol e test trigger and geophone records shownin Figure E6.5. The trigger and
geophone are l ocated 5 m apart. The sol id l ine represents theresponse from a
downward impact on a mechanical source and the dotted l ine represents theresponse
from an upward impact.Sol ution There is an obvious wave arrival at the geophone at
about 2 msec after impact at thesource. However, cl ose examination of the geophone
record shows that the pol arity of this earl yarrival was not infl uenced by the
pol arity of the impact; consequentl y, the earl y arrival can beidentified as a p-
wave. At a l ater point, the arrival of waves whose pol arity is reversed by the
reversal of impact pol arity is observed. These represent SV-waves-the arrival time
of23 msecafter impact (determined graphical l y) indicates an SV-wave vel ocity ofx 5m
Vs = J'it = 0.023 sec = 217 ml secSec. 6.3 Measurement of Dynamic Soil Properties
20750 60----- Upward hit-- Downward hit20 30 40Time (msec)FigureE6.5o 10Trigger
-----,~ • • 1\ Vertical II"Geophone -----"\IRecordsSeismic Down-Hol e (Up-Hol e) Test.
Seismic down-hol e (or up-hol e) testscan be performed in a singl e borehol e. In the
down-hol e test, an impul se source is l ocatedon the ground surface adjacent to the
borehol e. A singl e receiver that can be moved to differentdepths, or a string of
mul tipl e receivers at predetermined depths, is fixed against thewal l s of the
borehol e, and a singl e triggering receiver is l ocated at the energy source (Figure
6.22). Al l receivers are connected to a high-speed recording system so that their
output canbe measured as a function of time. In the up-hol e test, a movabl e energy
source is l ocated inthe borehol e with a singl e receiver on the ground surface
adjacent to the borehol e.Receiver Source(a) (b)Figure 6.22 (a) Seismic up-hol e
test, and(b) seismic down-hol e test.The objective of the down-hol e (or up-hol e)
test is to measure the travel times of pand/or s-waves from the energy source to
the receiver(s). By properl y l ocating the receiverpositions, a pl ot of travel time
versus depth can be generated (Figure 6.23). The sl ope of thetravel -time curve at
any depth represents the wave propagation vel ocity at that depth.S-waves can be
generated much more easil y in the down-hol e test than the up-hol e test;
consequentl y, the down-hol e test is more commonl y used. With an SH-wave source, the
down-hol e test measures the vel ocity of waves simil ar to those that carry most
seismic
energyto the ground surface. Because the waves must travel through al l material s
between theimpul se source and the receivers, the down-hol e test al l ows detection of
l ayers that can be hiddenin seismic refraction surveys. Potential difficul ties with
down-hol e (and up-hol e) tests andtheir interpretation can resul t from disturbance
of the soil during dril l ing of the borehol e, casingand borehol e fl uid effects,
insufficient or excessivel y l arge impul se sources, background208 Dynamic Soil
Properties Chap. 6Time (msec)o 20 40 60 80 100 120 140 160 180 200 220Vp = 1950 v«
= 350vp =2500 Vs =800Geol ogic l og0Fil l ----------------------- 21Bay mud
----------------------- 42Uppersands----------------------- 64.5Stiff cl ayand sands
----------------------- 91 5Firm cl ay J: : --------._------------- 15. 110.5 Q)Stiff
cl ay 0---------------.------- 130.5Sand----._-.-._------------ 144.5Rock200Vs = 750
Vp =5500 Vs =1100Vp = 5000Vp =9000 v« =37003060180 -1505 90J: : 15.Q) o120210 p -wave
arrival ss - wavearrival sFigure 6.23 Travel -time curve from down-hol e test in San
Francisco Bay area. (AfterSchwarz and Musser, 1972.)noise effects, and groundwater
tabl e effects. The effects of material and radiation damping onwaveforms can make
identification of s-wave arrival s difficul t at depths greater than 30 to 60m (100
to 200 ft). Efforts toward measurement of damping ratios in the down-hol e test have
al so been made (Redpath et al ., 1982; Redpath and Lee, 1986; EPRI, 1993).Seismic
Cone Test. The seismic cone test (Robertson et aI., 1985) is very simil arto the
down-hol e test, except that no borehol e is required. A seismic cone penetrometer
consistsof a conventional cone penetrometer (Section 6.3.1.2) outfitted with a
geophone or accel erometermounted just above the friction sl eeve. At different
stages in the cone penetrationsounding, penetration is stopped l ong enough to
generate impul ses at the ground surface, oftenby striking each end of a beam
pressed against the ground by the outriggers of the cone rig,with an instrumented
hammer. Travel time-depth curves can be generated and interpreted inthe same way as
for down-hol e tests. Al though down-hol e (and up-hol e) tests have usual l ybeen
performed to compl ement other tests or to provide redundancy, the efficiency of the
seismiccone test may l ead to its more common use. Cross-hol e seismic tests using
two seismiccones have al so been performed (e.g., Bal di et aI., 1988).6.3.1.2 High-
Strain TestsThe fiel d of in situ testing of soil s has expanded tremendousl y in the
past 25 years asa number of new devices and procedures have been devel oped and
impl emented. Whil eSec. 6.3 Measurement of Dynamic Soil Properties 209these tests
are most commonl y used to measure high-strain characteristics such as soil strength,
their resul ts have al so been correl ated to l ow-strain soil properties. For
geotechnical earthquake engineering probl ems, the standard penetration test, cone
penetration test,dil atometer test, and pressuremeter test are of particul ar
interest.Standard Penetration Test. The standard penetration test (SPT) is by far
the ol dest and most commonl y used in situ test in geotechnical engineering. It is
al so commonl yused in a number of geotechnical earthquake engineering appl ications.
In the SPT, astandard spl it-barrel sampl er (Figure 6.24) is driven into the soil at
the bottom of a borehol eby repeated bl ows (30 to 40 bl ows per minute) of a 140-l b
(63.6 kg) hammer rel eased froma height of 30 in (76 em). The standard SPT sampl er
shoul d have a constant inside diameter;the use of sampl ers designed to accommodate
internal sampl e l iners can underestimate penetrationresistance by 10 to 20% when
the l iners are not in pl ace. The sampl er is usual l ydriven 18 in. (46 ern); the
number ofbl ows required to achieve the l ast 12 in. (30 em) of penetrationis taken
as the standard penetration resistance, N. The N val ue is a function of thesoil
type, confining pressure, and soil density, but is al so infl uenced by the test
equipmentand procedures. In fact, studies have shown that different equipment and
procedures arequite common both within the United States and other countries, and
that they strongl yinfl uence the energy del ivered to the soil by each bl ow of the
hammer (Kovacs et aI., 1977;Schmertmann et aI., 1978; Kovacs and Sal omone, 1982).
Seed et aI. (1985) recommendedthat the test be performed in 4- to 5-in.-diameter
(10 to 13 ern) rotary borehol es with upwarddefl ection of bentonite dril l ing mud
using a tricone or baffl ed drag bit. The recommendedsampl er shoul d have a constant
inside diameter and be connected to A or AW [for depthsl ess than 15 m (50 ft)] or N
or NW (for greater depths) dril l rods. Driving at a rate of 30 to40 bl ows per
minute with 60% of the theoretical free-fal l energy del ivered to the sampl erwas
al so recommended. It has become common to normal ize the N val ue to an overburden
pressure of 1 ton/tr' (100 kPa) and to correct it to an energy ratio of 60% (the
average ratio ofthe actual energy del ivered by safety hammers to the theoretical
free-fal l energy) according to(6.30)where Nm is the measured penetration
resistance, eN an overburden correction factor (Figure6.25), Em the actual hammer
energy, and Ef f the theoretical free-fal l hammer energy.The corrected standard
penetration resistance, (Nj )60, has been correl ated to many importantproperties of
coarse-grained soil s. Correl ations for the properties of fine-grained soil sare much
l ess rel iabl e.Exampl e 6.6A site in Japan had the measured 8PT resistances indicated
in tabl e E6.6. Ishihara (1993) indicatedthat the 8PT procedures used in Japan
del iver about 72% of the theoretical free-fal l energy to the sampl er. Assuming that
the sands have an average void ratio of 0.44 and that thewater tabl e is at a depth
of 1.5 m, compute the corresponding (N j )60 val ues.Sol ution Given the void ratio
of 0.44, and assuming that Gs == 2.7, the average dry and submergeddensities of the
sand are 1.874 Mg/rrr' and 1.180 Mg/m', respectivel y. These densitiescan be used to
compute the vertical effective stress val ues that correspond to each depth atwhich
the 8PT resistance was measured (col umn 3 bel ow). For exampl e, the vertical
effectivestress at a depth of 6.2 m is given by210 Dynamic Soil Properties Chap. 6
Tabl e E6-6Depth (m) Nm Depth (m) Nm1.2 7 11.2 232.2 4 12.2 133.2 3 13.2 II4.2 3
14.2 115.2 5 15.2 246.2 9 16.2 277.2 12 17.2 58.2 12 18.2 69.2 14 19.2 410.2 9 20.2
38SPT correction factor, c,0 0.5 1.0 1.5 2.0 2.5cr 0EcBazaraa~ (1967) ~Q.~ oeenn
Liao & Whitman~ 2 (1986) Ul c Q) CN=~"E3 avo: : J.0iii>0>Q) 4 : gQ)iIi 5 I. .1 2 in.
Figure 6.24SPT sampl er.Figure 6.25 SPT overburden correction factor. (AfterLiao and
Whitman, 1986.)Sec. 6.3 Measurement of Dynamic Soil Properties 211a'vo = [(l .s m)
( 1.874~n+ (4.7 m) (1.180~~) J(9.81 s: 2) = 82.0 kPaBy converting the vertical
effective stresses from units of kPa to units of tons/ft.", the val ue ofthe
correction factor at each depth can be computed by the rel ationship of Liao and
Whitman(1986) (col umn 4 bel ow). For exampl e, the correction factor at a depth of
6.2 m is given by1 = 1.08(82.0 kPa)[(0.01044 tons/ft 2)/kPa]Then the corrected SPT
resistances can be computed using equation (6.30) (col umn 6 bel ow).For a depth of
6.2 m,Em O.72Eff(N 1) 60 = NmCNo.60Eff = (9)(1.08)0.60Eff= 11.7The appropriate
val ues for al l depths are tabul ated bel ow.(I) (2) (3) (4) (5) (6)Depth (m) Nm (J~o
(kPa) eN E,JEff (N j )601.2 7 22.\ 2.08 0.72 17.32.2 4 35.7 1.64 0.72 7.93.2 3 47.3
1.42 0.72 5.\4.2 3 58.8 1.28 0.72 4.65.2 5 70.4 1.16 0.72 7.06.2 9 82.0 1.08 0.72 \
1.77.2 12 93.6 1.01 0.72 14.68.2 12 105.3 0.95 0.72 13.79.2 14 116.8 0.91 0.72 15.2
10.2 9 128.4 0.86 0.72 9.311.2 23 140.0 0.83 0.72 22.812.2 13 151.5 0.79 0.72 12.4
13.2 11 163.1 0.77 0.72 10.114.2 11 174.7 0.74 0.72 9.8\5.2 24 186.3 0.72 0.72 20.7
16.2 27 197.9 0.70 0.72 22.517.2 5 209.4 0.68 0.72 4.118.2 6 221.1 0.66 0.72 4.7
19.2 4 232.7 0.64 0.72 3.120.2 38 244.3 0.63 0.72 28.5For gravel l y soil s, the
Becker hammer penetration test (BPT) can be used in the sameway as the SPT is for
sands. In a recommended BPT procedure (Harder and Seed, 1986), acl osed 6.6.in.-OD
dril l bit at the end of a 6.6.in. (16.8 em) OD steel casing is driven into thesoil
by an ICE 180 diesel pil e-driving hammer (8100 ft-l b/bl ow (110 N-ml bl ow) rated
energy). The BPT resistance is taken as the number of bl ows per foot of
penetration, correctedfor variations in the diesel hammer bounce chamber pressure
(which refl ect theeffects of soil resistance and combustion conditions on hammer
energy). By comparing theresul ts from the BPT and SPT at the same sandy sites,
Harder and Seed (1986) found that theBPT and SPT resistances were rel ated as shown
in Tabl e 6-1.Rather than rel ying on bounce chamber pressures, Sy and Campanel l a
(1994) used apil e driving anal yzer to measure the maximum transferred energy,
ENTHRU. With energy212 Dynamic Soil Properties Chap. 6Tabl e 6.1 Equival ence of
CorrectedBPT and SPT ResistancesCorrected BPTBl owcount, NBCCorrected SPTBl owcount,
N60o20406080100o-20-34-46-58-70measurements, the BPT resistance can be corrected to
a reference ENTHRU l evel of 30%of the rated energy of an ICE 180 diesel hammerNb30
= NbEN~~RUwhere ENTHRU is expressed as a percentage. Considering the effects of
skin resistanceal ong the sides of the BPT casing, the graphical correl ation with
SPT resistance shown inFigure 6.26 was devel oped.
100 ..,------,--------;-----,-------,------,Figure 6.26 BPT-SPT correl ations for
different BPT shaft resistances. (After Sy100 and Campanel l a, 1984; used with
permission.)20 40 60 80BPT Nb30 (bl ows/0.3m)o --~~~§5: : ===-+-___+--Jo8020 ...EC')~
60~e.}40 ..Ii: : enCone Penetration Test. In recent years,
use of the cone penetration test(CPT) in geotechnical engineering practice has
increased sharpl y. The CPT invol ves thesteady penetration of a standard cone
penetrometer (Figure 6.27a) into the ground. The standardcone penetrometer has a
conical tip of 10 cnr' (1.55 irr') area and 60° apex angl e immediatel ybel ow a
cyl indrical friction sl eeve of 150 cm 2 (23.3 in ') surface area. Thepenetrometer
is pushed into the ground at a constant rate of 2 ern/sec (0.8 in/sec). The tipand
friction sl eeve are each connected to l oad cel l s that measure the tip resistance,
qC' andsl eeve resistance.j., during penetration (Figure 6.27b). The friction ratio,
FR = fs/qc' is al soSec. 6.3 Measurement of Dynamic Soil Properties 213Frictionratio
(%)o 2 4 6 8(b)Bearingresistance(tons/tt)100 200 300 400 500~ : ~,...(--r ---
---,_.---F=: " »---- ---,i..J~Frictionresistance(tons/tt)642 0o40605010g 20.<: : g-
30oBiaxial incl inometerPushing rodsBiaxial geophoneSignal ampl ifierFrictionsl eeve
(150 cm2),"y=,;: J.--~- Frictionsl eevel oad cel l Tip l oad cel l (a)Figure 6.27 (a)
Typical cone penetrometer with seismic cone capabil ities (after Bal diet aI., 1988);
(b) resul ts of cone penetration sounding (after Ward, 1986).a useful parameter-it
is high in cohesive soil s and l ow in cohesionl ess soil s. The absol uteand rel ative
magnitudes of the penetration resistances can be correl ated to many ofthe same
properties as the SPT, and al so to soil type.The CPT can be performed rapidl y
(usual l y about four times faster than dril l ing andsampl ing) and rel ativel y
inexpensivel y. It provides a continuous profil e of penetration resistancethat can
detect the presence of thin l ayers or seams that are easil y missed in SPT testing.
The capabil ities of cone penetrometers can be enhanced by adding additional
transducers to measure additional variabl es such as porewater pressure (in the
piezocone) orwave propagation vel ocity [in the seismic CPT(Section 6.3.1.1)].
However, the CPT cannotbe used at sites with very stiff and/or very dense soil s
without damaging the probe or rods.The presence of gravel -size particl es can al so
l imit the use of the CPT.Dil atometer Test. The dil atometer test (DMT) uses a fl at
dil atometer (Figure6.28), a stainl ess steel bl ade with a thin fl at circul ar
expandabl e steel membrane on one side(Marchetti, 1980). The dil atometer is jacked
into the ground with the membrane surfacefl ush with the surrounding bl ade surface.
At interval s of 10 to 20 em (4 to 8 in.), penetrationis stopped and the membrane
infl ated by pressurized gas. The pressure at which the membranemoves by 0.05 mm
(0.002 in.) (the l ift-off pressure, Po) and the pressure at which itscenter moves
1.1 mm (0.043 in.) (PI) are recorded, corrected, and used with the hydrostatic
pressure, u.; and the effective overburden pressure, ()~o' to compute various
indices towhich soil properties can be correl ated. The most commonl y used of these
indices are thematerial index, [D, the horizontal stress index, KD, and the
dil atometer modul us, ED' given by(6.31)214 Dynamic Soil PropertiesWireChap. 6
Fl exibl emembrane60mmPneumatictubingI~--I 1--- 14 mm1.1 mmFl exibl emembraneFigure
6.28 Front and side views of the Marchetti fl at dil atometer. (After Bal di et al .,
1986.)Po- Uoa'vO(6.32)ED = « i», - Po) (6.33)where a is equal to 34.7 for a 60-mm
(2.4 in.) membrane diameter and a membrane defl ectionof 1.1 mm (0.043 in.).
Dil atometer parameters have been correl ated to l ow-strain soil stiffness (Section
6.4.2.1) and l iquefaction resistance (Section 9.5.3.1).Pressuremeter Test. The
pressuremeter test (PMT) is the onl y in situ testcapabl e of measuring stress-
strain, as wel l as strength, behavior (Mair and Wood, 1987).The pressuremeter
(Figure 6.29a) is a cyl indrical device that uses a fl exibl e membrane toControl unit
Borehol e(a)Cyl indrical fl exibl emembranepressurizedby fl uid- - -- - - - - - - - -- --
--"" -------- -~- - _--IPl asticdeformationEl asticdeformationSeating
......-=------------'----Corrected vol ume of cavity, V(b)Figure 6.29 Pressuremeter
test: (a) test setup; (b) typical pressuremeter curve. (AfterMair and Wood, 1987.)
Sec. 6.3 Measurement of Dynamic Soil Properties 215appl y a uniform pressure to the
wal l s of a borehol e. Deformation of the soil can be measuredby the vol ume of fl uid
injected into the fl exibl e membrane or by feel er arms for pressuremetersthat use
compressed gas. After correcting the measured pressures and vol ume changesfor
system compl iance, el evation differences, and membrane effects, a pressure-vol ume
curve (Figure 6.29b) can be devel oped. Using cavity expansion theory, the pressure-
vol umecurve can be used to compute the stress-strain behavior. Sel f-boring
pressuremeters,which minimize soil disturbance, and push-in pressuremeters, which
can penetrate soft soil svery quickl y, have al so been devel oped.Other Fiel d Tests. A
number of variations of the preceding tests have beenused to measure dynamic soil
properties, and new tests are al so being devel oped. The conceptof geotomography
fol l ows from advances in the medical imaging fiel d (Johnson et aI.,1978; Lytl e,
1978). Using mul tipl e sources and receivers, a l arge matrix of source-receiver
travel times can be measured and compared with predictions of a ray-tracing model
that canaccount for refraction and refl ection at possibl e material boundaries
between the sourcesand receivers. The number, position, and incl ination of material
boundaries are adjusteduntil the computed travel -time matrix matches the observed
matrix. Al though geotomographyis stil l in its infancy, it offers the potential for
rapid, detail ed resol ution of two- andthree-dimensional subsurface structures. An
apparatus that measures el ectrical conductivityof soil -porewater systems (Arul mol i
et aI., 1985) has bee~ used to estimate in situ densityand l ow-strain stiffness and
to eval uate l iquefaction potential . An in situ borehol e torsional test system (Henke
and Henke, 1991) advances two thin, concentric tubes into the soil at thebottom of
a borehol e. Records of rotation versus cycl ic torque appl ied to the inner tube have
shown behavior simil ar to that of stress-strain curves in control l ed l aboratory
tests.~'.3.2 Laboratory TestsLaboratory tests are usual l y performed on rel ativel y
smal l specimens that are assumed to berepresentative of a l arger body of soil . The
specimens are tested as el ements (i.e., they aresubjected to unifortn initial
stresses and uniform changes in stress or strain conditions). Inother l aboratory
tests, specimens are tested as model s and the resul ts must be interpreted interms
of the nonuniform boundary conditions acting on the model .The abil ity of l aboratory
tests to provide accurate measurements of soil propertiesdepends on their abil ity
to repl icate the initial conditions and l oading conditions of the probl emof
interest. No l aboratory test can represent al l possibl e stress and strain paths
with general rotation of principal stress axes; consequentl y, different tests wil l
be most suitabl e fordifferent probl ems.6.3.2.1 Sampl ing .El ement tests are
performed on soil specimens, For probl ems invol ving the responseof soil s to be
pl aced as fil l s, specimens can be constructed from bul k or disturbed sampl esby
simul ating the compaction process as cl osel y as possibl e in the l aboratory. When
theproperties of an existing soil are needed, however, the probl em becomes more
difficul t.Tests on existing soil s can be performed on undisturbed or reconstituted
specimens. However,in many instances the resul ts wil l be different between these
tests because of differencesin soil fabric between natural and reconstituted soil
specimens, even when densitiesand appl ied stresses are simil ar.216 Dynamic Soil
Properties Chap. 6Dynamic soil properties are infl uenced by many factors, incl uding
density and stressconditions, and other factors, such as soil fabric or structure,
age, stress and strain history,and cementation. Whil e the void ratio and stress
conditions can be recreated in a reconstitutedspecimen, the effects of the other
factors cannot. Since the effects of these other factorsare manifested primaril y at
l ow strain l evel s, they are easil y destroyed by sampl edisturbance. For the resul ts
ofl aboratory tests to refl ect the actual behavior ofthe in situ soil as cl osel y as
possibl e, high-qual ity undisturbed sampl es must be obtained.For cohesive soil s,
procedures for the preparation of test specimens by careful l y trimmingthin-wal l ed
tube or bl ock sampl es are fairl y wel l establ ished. Undisturbed sampl ing of
cohesionl ess soil s such as cl ean sands and gravel s is more difficul t. Even thin-
wal l ed sampl ingtubes can cause significant disturbance of cl ean sands, causing
densification ofl oose sands anddil ation of dense sands (Marcuson et al ., 1977). The
use of bl ock sampl ing (Hom, 1978) hasproven effective, but the process is l aborious
and may not be practical bel ow certain depths.The use of freezing and coring was
first described by Hvorsl ev (1949). This approach has beenshown to be effective
when the confining pressure is maintained and when free drainage ismaintained
continuousl y at the freezing front (Singh et al ., 1979). Konno et al . (1993), for
exampl e, described the successful retrieval of undisturbed sampl es by coring
through a 140ern-diameter col umn of sandy gravel frozen by l iquid nitrogen
circul ating through a central 73mmsteel tube for a period of about 160 hrs.6.3.2.2
Low-Strain El ement TestsOnl y a l imited number of l aboratory tests are abl e to
determine the properties of soil sat l ow strain l evel s. These incl ude the resonant
col umn test, the ul trasonic pul se test, and thepiezoel ectric bender el ement test.
Resonant Col umn Test. The resonant col umn test is the most commonl yused l aboratory
test for measuring the l ow-strain properties of soil s.
It subjects sol id or hol l owcyl indrical specimens to harmonic torsional or axial
l oading by an el ectromagneticl oading system (Figure 6.30a). The l oading systems
usual l y appl y harmonic l oads for whichthe frequency and ampl itude can be
control l ed, but random noise l oading (Al -Sanad andAggour, 1984) and impul se l oading
(Tawfiq et al ., 1988) have al so been used.Drive coil MagnetLocking ringInner
cyl inderProximitor targetAccel erometer4<---1+-- Fl uid bath\ Base pedestal LVDT(b)
SpecimenDrive coil hol derProximitorhol der -==JL~!~".dJ=[@ Drive coil Top
cap\Ifi;m: ;« l f--Magnet Level ing& securingscrewAccel erometerAccel erometer
counterweightProximitor probe(a)Figure 6.30 Typical resonant col umn test apparatus:
(a) top view of l oading system,and (b) profil e view ofl oading system and soil
specimen. (After EPRI, 1993.)Sec. 6.3 Measurement of Dynamic Soil Properties 217
After the resonant col umn specimen has been prepared and consol idated, cycl ic
l oadingis begun. The l oading frequency is initial l y set at a l ow val ue and is then
gradual l yincreased until the response (strain ampl itude) reaches a maximum. The
l owest frequencyat which the response is l ocal l y maximized is the fundamental
frequency of the specimen.The fundamental frequency is a function of the l ow-strain
stiffness of the soil , the geometryof the specimen, and certain characteristics of
the resonant col umn apparatus.The shear modul us can be rel ated to the fundamental
frequency by the fol l owing procedure.Consider a resonant col umn specimen of height,
h, fixed against rotation of its base,with pol ar moment of inertia, J, subjected to
harmonic torsional l oading (Figure 6.30b).From equation (5.8), the el astic
resistance of the specimen produces a torque at its top,T = GJ de = G! dedZ p dZ
(6.34)(6.35)where 1is the mass pol ar moment of inertia of the specimen. This torque
must be equal to theinertial torque of the l oading system. If the el ements of the
torsional l oading system connectedto the top of the specimen have a mass pol ar
moment of inertia, l a, the inertial torque isT = -1 hdzea dtZAssuming that the
rotations of the specimen are al so harmonic, they can be described bye(Z, t) = e(z)
(C j cos ffit+ Cz sin mr) (6.36)where e(z) =C3 cos kz + C4sin kz: The zero rotation
boundary condition at the base (z=0)requires C3 = 0, and the equal ity of equations
(6.34) and (6.35) requires, at the fundamental frequency, ffin =knvs' so that
GiC4kncos knh(Cjcos ffint+ Czsin ffint) p= -l oh(-ffi~C4sin knh) (Cj cos ffint+
Czsin ffint)which can be expressed as1 ffinh ffinh- = -tan - (6.37)10 Vs VsFor a
given specimen, 1,l a, and h are general l y known at the time that cycl ic l oading
begins.The fundamental frequency is then obtained experimental l y, and equation
(6.37) is used tocal cul ate vs' The shear modul us is then obtained from G = Pv~ .
Damping can be determinedfrom the frequency response curve using the hal f-power
bandwidth method (Section B.6.1 ofAppendix B) or from the l ogarithmic decrement by
pl acing the specimen in free vibration.For l ongitudinal l oading, the anal ogous
equation isW _ ffinh ffinh-tan - (6.38)Wo - VI VIwhere W is the weight of the
specimen, Wo the weight of the l oading system, andVI = JEiP the l ongitudinal wave
propagation vel ocity.If the l oading system was massl ess (/0 = 0), equation (6.37)
woul d degrade to218 Dynamic Soil Properties Chap. 6V s == 2: nh == 4inh (6.39)where
in is the fundamental frequency in hertz. In this case the rotations woul d fol l ow a
quarter-sine-wave pattern over the height of the specimen at the fundamental
frequency.Adding the mass of the l oading system resul ts in a more l inear variation
of rotation and, consequentl y,more uniform strain conditions over the height of the
specimen.The shear strain in a sol id cyl indrical resonant col umn specimen l oaded in
torsion variesfrom zero at the centerl ine of the specimen to a maximum val ue at its
outer edge. In situationsin which the shear modul us varies with shear strain
ampl itude, the effects of nonuniform straincan be significant (Drnevich, 1967,
1972). The use of hol l ow specimens minimizes the variationof shear strain ampl itude
across the specimen.Large-diameter resonant col umn devices have been used for
gravel l y soil s (Woods,1991) and rock (Prange, 1981). Konno et al . (1993) performed
what coul d be described as insitu resonant col umn tests of gravel l y soil s at a
potential nucl ear power pl ant site in Japan. Inthese tests the material in two l O-
m-ID (32.8 ft) circul ar trenches was excavated and repl acedwith water-fil l ed bags
to depths of 5 m (16 ft) and 9 m (30 ft). Concrete cap bl ocks 3 (10 ft)and 5 m (16
ft) thick were cast on top of each of the resul ting soil col umns. Vibratory shakers
were pl aced on top of the bl ocks near the edges; cycl ic torsional l oading was
appl ied to thesoil col umn by operating the jacks 1800 out of phase. By performing
frequency sweeps, theresponse of the soil col umns coul d be measured at shear strain
ampl itudes up to 0.01 %.Exampl e 6.7A 6-in.-high specimen of soft sil ty cl ay with a
unit weight of 105 Ib/ft3 is tested in a resonantcol umn device with I1Io= 0.4. From
the frequency response curve shown in Figure E6.7, determinethe shear modul us of
the specimen.40 45 50 55 60
OL---'----'-------'--'---'-------'-----'---'-----'--------'35Frequency (Hz)
FigureE6.7Sol ution The maximum ampl itude of the accel erometer output occurs at the
fundamental frequencyof the specimen, determined graphical l y to bein = 41 HzThen
equation (6.37) can be written asII owhich is satisfied when wnh/vs = 0.593. Then
Sec. 6.3 Measurement of Dynamic Soil Propertiesffinh = 2nfnh = 2n(41)(0.5) = 212
ft/sec0.593 0.593 0.593G = pvt = 105 Ib/ft3/ 212 ft/sec) 2 = 146,5571b/ft232.2
ft/sec219The resonant col umn test al l ows stiffness and damping characteristics to
be measuredunder control l ed conditions. The effects of effective confining
pressure, strain ampl itude,and time can readil y be investigated. However,
measurement of porewater pressure is difficul t,and the material properties are
usual l y measured at frequencies above those of mostearthquake motions.Ul trasonic
Pul se Test. Wave propagation vel ocities can be measured in thel aboratory by means
of the ul trasonic pul se test (Lawrence, 1963; Nacci and Tayl or, 1967).Ul trasonic
transmitters and receivers are attached to pl atens that can be pl aced at each endof
a specimen with the distance separating them careful l y measured. The transmitters
andreceivers are made of piezoel ectric material s which exhibit changes in
dimensions whensubjected to a vol tage across their faces, and which produce a
vol tage across their faceswhen distorted. A high-frequency el ectrical pul se appl ied
to the transmitter causes it todeform rapidl y and produce a stress wave that
travel s through the specimen toward thereceiver. When the stress wave reaches the
receiver, it generates a vol tage pul se that is measured.The distance between the
transmitter and receiver is divided by the time differencebetween the vol tage
pul ses to obtain the wave propagation vel ocity. The ul trasonic pul setest is
particul arl y useful for very soft material s, such as seafl oor sediments, since it
can beperformed whil e the soil is stil l in the sampl ing tube (Woods, 1978).
Piezoel ectric Bender El ement Test. Another type of test that al l ows measurementof
shear wave vel ocity on l aboratory specimens makes use of piezoel ectric bender
el ements (Shirl ey and Anderson, 1975; De Al ba et al ., 1984; Dyvik and Madshus,
1985).Bender el ements are constructed by bonding two piezoel ectric material s
together in such away that a vol tage appl ied to their faces causes one to expand
whil e the other contracts, causingthe entire el ement to bend as shown in Figure
6.31. Simil arl y, a l ateral disturbance of thebender el ement wil l produce a vol tage,
so the bender el ements can be used as both s-wavetransmitters and receivers.In most
setups, the bender el ements protrude into opposite ends of a soil specimen. A
vol tage pul se is appl ied to the transmitter el ement, which causes it to produce an
s-wave.Direction I of shear wavepropagationDirection of el ements tip andsoil
particl e movementtil ( • Zero vol tageFigure 6.31 Piezoel ectric bender el ement.
Positive vol tage causes el ement to bend oneway, negative vol tage causes it to bend
theother.220 Dynamic Soil Properties Chap. 6When the s-wave reaches the other end
of the specimen, distortion of the receiver el ementproduces another vol tage pul se.
The time difference between the two vol tage pul ses is measuredwith an oscil l oscope
and divided into the distance between the tips of the bender el ementsto give the s-
wave vel ocity of the specimen.Piezoel ectric bender el ements have been incorporated
into conventional and cubical triaxial devices, direct simpl e shear devices,
oedometers, and model tests. Since the specimenis not disturbed during the bender
el ement test, it can be subsequentl y tested for othersoil characteristics.6.3.2.3
High-Strain-El ement TestsAt high shear strain ampl itudes, soil s general l y exhibit
vol ume change tendencies.Under drained l oading conditions, these tendencies are
al l owed to manifest themsel ves inthe form of vol umetric strain, but under undrained
conditions they resul t in changes in porepressure (and effective stress). Since
soil behavior is governed by effective stresses, al l methods of testing soil s at
high strain l evel s must be capabl e of control l ing porewater drainagefrom the
specimen and measuring vol ume changes and/or pore pressures accuratel y.The probl em
of system compl iance (vol ume changes due to the testing apparatus rather thanthe
soil ), which can l ead to errors in vol ume change/pore pressure measurement, is
importantin the interpretation of high-strain test resul ts. Membrane penetration in
coarse-grainedsoil is an important contributor
to system compl iance.Cycl ic Triaxial Test. Just as the triaxial compression test
is the most commonl yused l aboratory test for measurement of soil properties under
static l oading conditions,the cycl ic triaxial test has been the most commonl y used
test for measurement ofdynamic soil properties at high strain l evel s. In the
triaxial test, a cyl indrical specimen ispl aced between top and bottom l oading
pl atens and surrounded by a thin rubber membrane(Figure 6.32). The specimen is
subjected to a radial stress, usual l y appl ied pneumatical l y,and an axial stress. By
virtue of these boundary conditions, the principal stresses in the specimenare
al ways vertical and horizontal .Cel l pressure ----+Axial • l oadLoad cel l LVDTO-ring
seal Rubber membraneSoil specimen____________ Cel l wal l Pore pressuretransducerFigure
6.32 Typical triaxial apparatus.Sec. 6.3 Measurement of Dynamic Soil Properties 221
The difference between the axial stress and the radial stress is cal l ed the
deviatorstress. In the cycl ic triaxial test, the deviator stress is appl ied
cycl ical l y, either under stresscontrol l edconditions (typical l y by pneumatic or
hydraul ic l oaders), or under strain-control l edconditions (by servohydraul ic or
mechanical l oaders). Cycl ic triaxial tests are mostcommonl y performed with the
radial stress hel d consant and the axial stress cycl ed at a frequencyof about 1 Hz.
As with the static triaxial test, the cycl ic triaxial test can be performed under
isotropical l yconsol idated or anisotropical l y consol idated conditions, thereby
producing thestress paths shown in Figure 6.33. Figure 6.33a shows the cycl ic
deviator stress and total stress path for an isotropical l y consol idated specimen.
Isotropical l y consol idated tests arecommonl y used to represent l evel -ground sites
where no initial shear stresses exist on horizontal pl anes. The test begins with
zero shear stress (point A) and the deviator stress is initial l yincreased. Since
the axial stress is then greater than the radial stress, the major andminor
principal stress axes are vertical and horizontal , respectivel y. After the deviator
stressreaches its maximum val ue (point B), it decreases and approaches a val ue of
zero (point C).Just before it reaches point C, the major principal stress axis is
stil l vertical , but it rotatesinstantaneousl y to horizontal as point C is passed
and the deviator stress becomes negative.At point C, no shear stress exists on the
specimen. This process of stress reversal repeatsitsel f throughout the test, with
instantaneous 90° rotations of the principal stress axes occurringevery time the
deviator stress passes through zero,To model conditions in and beneath sl opes where
initial static shear stresses exist, anisotropical l yconsol idated triaxial tests are
performed. Figure 6.33b refers to an anisotropical l yconsol idated specimen for which
the cycl ic deviator stress ampl itude is greater than the deviatorstress during
consol idation. Stress reversal s al so exist in this situation, even though thecycl ic
deviator stress is no l onger symmetric about the p-axis. If the cycl ic deviator
stress(Jd B 2~C r-; r-. r-. q1 A,C~ B A\J\..../\J\j / p• t(a)(Jd 2~QL\J\r ql 7,/ • P
(b)"2'~qJ / • p(c)Figure 6.33 Time histories of deviator stress and stress paths for
(a) isotropical l yconsol idated conditions, (b) anisotropical l y consol idated
conditions with cycl ic deviatorstress ampl itude greater than deviator stress during
consol idation (producing stressreversal s), and (c) anisotropical l y consol idated
conditions with cycl ic deviator stressampl itude l ess than deviator stress during
consol idation (no stress reversal s).222 Dynamic Soil Properties Chap. 6ampl itude is
smal l er than the deviator stress during consol idation (Figure 6.33c), no stress
reversal s wil l occur. For this case, the principal stress axes wil l not rotate and
the specimen wil l never reach the zero shear stress condition. The stress paths in
Figure 6.33 are obviousl y differentwith respect to initial stress conditions,
stress path, and principal stress axis rotation thanthose imposed on the el ement of
soil subjected to vertical l y propagating s-waves shown in Figure6.7. These
differences il l ustrate the fundamental difficul ty in the direct appl ication of
propertiesobtained from the cycl ic triaxial test to actual wave propagation
probl ems.In some cases, the cel l pressure is al so appl ied cycl ical l y. By decreasing
(or increasing)the cel l pressure by the same amount that the deviator stress is
increased (or decreased)by, the Mohr circl e can be made to expand and contract
about a constant center point. Theresul ting stress path wil l then oscil l ate
vertical l y, much l ike that shown for the case of vertical l ypropagating s-waves
(Figure 6.7). Al hough the stress path of such a triaxial test canbe made to match
that induced by a vertical l y propagating s-wave, the principal stresses inthe
triaxial test remain constrained to the vertical and horizontal direction rather
than rotatingcontinuousl y as caused by the s-wave.The stresses and strains measured
in the cycl ic triaxial test can be used to compute theshear modul us and damping
ratio (Section 6.4.2). The cycl ic triaxial test al l ows stresses tobe appl ied
uniforml y, al though stress concentrations can exist at the cap and base, andal l ows
drainage conditions to be accuratel y control l ed (when the effects of membrane
penetrationare mitigated). It requires onl y minor modification of standard triaxial
testingequipment. On the other hand, the cycl ic triaxial test cannot model stress
conditions thatexist in most actual seismic wave propagation probl ems. Bedding
errors and system compl ianceeffects general l y l imit measurements to shear strains
greater than about 0.0 I%,al though l ocal strain measurement (e.g., Burl and and
Symes, 1982; Ladd and Dutko, 1985;Goto et al ., 1991) can produce accurate
measurements at strain l evel s as smal l as 0.0001 %.Membrane penetration effects can
be important in cycl ic triaxial tests of coarse sandsand gravel s. After
consol idation, the thin triaxial membrane wil l penetrate the perimetervoids of
coarse sand and gravel specimens. As excess pore pressures devel op during cycl ic
l oading, the net pressure on the membrane decreases and its penetration decreases.
When thishappens, the effective vol ume ofthe voids increases and the excess pore
pressure drops bel owthe l evel it woul d have had if true constant-vol ume conditions
had been maintained. Becausethey al l ow the effective stresses to be higher than
they woul d be under constant-vol ume conditions,membrane penetration effects Canl ead
to inaccurate stiffness and damping measurementsand unconservative estimation of
l iquefaction resistance (Chapter 9). Procedures havebeen devel oped for measurement
(Vaid and Negussey, 1984; Kramer and Sivaneswaran,1989a), minimization (Lade and
Hernandez, 1977; Raju and Venkataramana, 1980), compensation(Seed and Anwar, 1986;
Tokimatsu and Nakamura, 1986), and posttest correction(Martin et aI., 1978; Kramer
and Sivaneswaran, 1989b) of membrane penetration effects.Exampl e 6.8A cycl ic
triaxial test on a saturated cl ay specimen produces the stress-strain l oop shown in
FigureE6.8. Determine the secant shear modul us and damping ratio.Sol ution
Graphical l y, the sl ope of a straight l ine between the ends of the stress-strain
curveshows thatEsec236 kPa0.01416,857 kPaSec. 6.3 Measurement of Dynamic Soil
Properties 223(Jdc(kPa)200---2 fa ('Yo)Figure E6.8Then, assuming that the saturated
cl ay is l oaded under undrained conditions, v = 0.5, soG = __E_ = 16,857 kPa = 5619
kPasec 2(1 + v) 2(1 + 0.5)The area of the hysteresis l oop is 4.52 kPa and the area
of the triangl e denoting the maximumstrain energy is 1.65 kPa. Then~ = -.l area of
hysteresis l oop = 4.52 kPa = 02184n: area of triangl e 1.65 kPa .Cycl ic Direct
Simpl e Shear Test. The cycl ic direct simpl e shear test is capabl eof reproducing
earthquake stress conditions much more accuratel y than is the cycl ic triaxial test.
It is most commonl y used for l iquefaction testing. In the cycl ic direct simpl e
sheartest, a short, cyl indrical specimen is restrained against l ateral expansion by
rigid boundarypl atens (Cambridge-type device), a wire-reinforced membrane (NGI-type
device), or aseries of stacked rings (SGI-type device). By appl ying cycl ic
horizontal shear stresses to thetop or bottom of the specimen, the test specimen is
deformed (Figure 6.34) in much the sameway as an el ement of soil subjected to
vertical l y propagating s-waves (Figure 6.7).~Horizontal l oad cel l ~/LVDTs for
vertical ~ , / displ acementVertical l oad cel l Soil specimenLVDT for horizontal
displ acementc=: ==== To vol ume change device/porepressure transducerFigure 6.34 NGI
cycl ic simpl e shear apparatus. Soil specimen is contained within wirereinforced
rubber membrane. (After Airey and Wood, 1987.)224 Dynamic Soil Properties Chap. 6
The simpl e shear apparatus, however, appl ies shear stresses onl y on the top and
bottomsurfaces of the specimen. Since no compl ementary shear stresses are imposed
on thevertical sides, the moment caused by the horizontal shear stresses must be
bal anced by nonuniforml ydistributed shear and normal stresses. The effects of
nonuniformity of stressescan be reduced by increasing the diameter/height ratio of
the specimen; such effects aresmal l at diameter/height ratios greater than about
8: 1 (Kovacs and Leo, 1981). Conventional simpl e shear apparatuses are l imited by
their inabil ity to impose initial stresses otherthan those corresponding to Ko
conditions. In recent years, simpl e shear devices that al l owindependent control of
vertical and horizontal stresses have been devel oped. To better simul ateactual
earthquake conditions, Pyke (1973) used a l arge shaking tabl e to produce a
twodirectional l arge-scal e simpl e shear apparatus with a diameter/height ratio of
about
9 in onedirection and 20 in the other. Smal l -scal e, bidirectional , cycl ic simpl e
shear apparatuseshave al so been devel oped (Boul anger et al ., 1993).Cycl ic Torsional
Shear Test. Many of the difficul ties associated with thecycl ic triaxial and cycl ic
direct simpl e shear tests can be avoided by l oading cyl indrical soil specimens in
torsion. Cycl ic torsional shear tests al l ow isotropic or anisotropic initial stress
conditions and can impose cycl ic shear stresses on horizontal pl anes with
continuous rotationof principal stress axes. They are most commonl y used to measure
stiffness and dampingcharacteristics over a wide range of strain l evel s.Ishihara
and Li (1972) devel oped a torsional triaxial test that used sol id specimens.Dobry
et al . (1985) used strain-control l ed cycl ic torsional l oading al ong with stress-
control l edaxial l oading of sol id specimens to devel op a CyT-CAD test that has
proven effectivefor measurement of l iquefaction behavior. Torsional testing of
sol id specimens,however, produces shear strains that range from zero al ong the axis
of the specimen to amaximum val ue at the outer edge. To increase the radial
uniformity of shear strains, others(e.g., Drnevich, 1967, 1972) devel oped hol l ow
cyl inder cycl ic torsional shear apparatuses(Figure 6.35). Whil e hol l ow cyl inder
tests offer perhaps the best uniformity and control Axial stress Torque.--_.--Figure
6.35 Hol l ow cyl inder apparatus. The specimen is encl osed within internal and
external membranes on which internal and external pressures can be appl ied
independentl y.Appl ication of cycl ic torque induces cycl ic shear stresses on
horizontal pl anes.Sec. 6.3 Measurement of Dynamic Soil Properties 225over stresses
and drainage, specimen preparation can be difficul t and the equipment is notwidel y
avail abl e.6.3.2.4 Model TestsIn contrast to el ement tests, model tests usual l y
attempt to reproduce the boundaryconditions of a particul ar probl em by subjecting a
smal l -scal e physical model of a ful l -scal eprototype structure to cycl ic l oading.
Model tests may be used to eval uate the performanceof a particul ar prototype or to
study the effects of different parameters on a general probl em.Whil e model testing
is very useful for identification of important phenomena and verificationof
predictive theories, it has not yet devel oped to the point of being used directl y
fordesign of significant structures or facil ities.The behavior of soil s is
sensitive to stress l evel ; soil s that exhibit contractive behaviorunder high normal
stresses may exhibit dil ative behavior at l ower stress l evel s. One of themost
significant chal l enges in model testing, therefore, is the probl em of testing
model swhose stress dependency matches that of the ful l -scal e prototype. Because
this is very difficul tunder the gravitational fiel d of the earth, one common
approach invol ves testingunder increased gravitational fiel ds. Model tests can
therefore be divided into those performedunder the gravitational fiel d of the earth
(l g model tests) and those performed underhigher gravitational accel erations. The
19 tests are most commonl y performed with the useof shaking tabl es; tests under
increased gravitational fiel ds are usual l y performed in a geotechnical centrifuge.
Both shaking tabl e and centrifuge model tests share certain drawbacks, among the
most important of which are simil itude and boundary effects. Because different
aspects ofthe response of a l IN-scal e model are governed by different scal e
factors, simil itude cannotbe assured for al l parameters simul taneousl y. Boundary
effects are usual l y associated withthe metal l ic bins or boxes in which shaking
tabl e and centrifuge model s are usual l y constructed.The sidewal l s can restrain soil
movement and refl ect energy that woul d radiateaway in the prototype probl em. The
industrial fil l er material Duxseal has been used as anabsorbent wal l l ining with
some success (Steedman, 1991).Shaking Tabl e Tests. In the earl y years of
geotechnical earthquake engineering,virtual l y al l physical model testing was
performed on shaking tabl es. Shaking tabl eresearch has provided val uabl e insight
into l iquefaction, postearthquake settl ement, foundationresponse, and l ateral earth
pressure probl ems. Most shaking tabl es util ize a singl ehorizontal transl ation
degree of freedom, but shaking tabl es with mul tipl e degrees of freedomhave al so
been devel oped. Shaking tabl es are usual l y driven by servohydraul ic actuators
(Figure 6.36); their dynamic l oading capacities are control l ed by the capacity of
thehydraul ic pumps that serve the actuators. Large pumps and l arge actuators are
required toproduce l arge displ acements of heavy model s at moderate or high
frequencies.Shaking tabl es of many sizes have been used for geotechnical earthquake
engineeringresearch. Some are quite l arge, al l owing model s with dimensions of
several meters to betested. Thus shaking tabl es can often util ize actual , prototype
soil s rather than resorting tothe smal l er particl e sizes often required for smal l er
scal e model tests. For these l arge model s,soil s can be pl aced, compacted rel ativel y
easil y, and instrumented rel ativel y easil y.Shaking tabl e model s can be easil y
viewed from different perspectives during testing.On the other hand, high
gravitational stresses cannot be produced in a shaking tabl etest. Though the
contractive behavior associated with high normal stresses at significant226
Instrumented wal l Dynamic Soil Properties Chap. 6Al uminumpl atesActuator8 ItShaking
tabl e base(supported on l inear bearings)Side wal l s,rigid pl astic boardsFigure 6.36
Shaking tabl e with soil bin used for dynamic earth pressure research.(After Sherif
et aI., t 982.)depths can be simul ated by pl acing soil very l oosel y during model
preparation, the processof preparing such model s is quite difficul t. Because of the
l ow normal stress l evel s, the contributionof factors that produce a cohesive
component of strength wil l be greater in themodel than in the prototype. Correction
procedures (e.g., Hettl er and Gudehus, 1985; Iai,1989) have been devel oped to aid
in the interpretation of shaking tabl e test resul ts.Centrifuge Tests. In a
centrifuge test, a l iN-scal e model l ocated at a distance,r, from the axis of a
centrifuge (Figure 6.37) is rotated at a rotational speed, Q = IN/ r ,which is
sufficient to raise the accel eration fiel d at the l ocation of the model to N times
theaccel eration of gravity. In principl e, the stress conditions at any point in the
model shoul dthen be identical to those at the corresponding point in the ful l -scal e
prototype. The overal l behavior (e.g., displ acements, fail ure mechanisms, etc.)
shoul d al so be identical .Centrifuge tests are restricted to much smal l er model s
than even moderate-sizedshaking tabl es. Since the gravitational fiel d increases
with radial distance, the gravitational CounterweightMotorBearingBearingSl ip rings
TestbucketFigure 6.37 Cross section through a geotechnical centrifuge. (After
O'Reil l y, 1991.)Sec. 6.3 Measurement of Dynamic Soil Properties 227accel eration at
the top of the model is l ower than that at the bottom of the model . Since the
gravitational fiel d acts in the radial direction, the horizontal pl ane is curved
(O'Reil l y,1991) by an amount that decreases with increasing centrifuge radius.
Simil itude considerations are very important in the pl anning and interpretation of
centrifuge tests. Scal ing factors for a number of parameters are shown in Tabl e 6-
2. Thescal ing factors show how dynamic events are speeded up in the centrifuge. For
exampl e,the stresses and strains in a 30-m (100-ft)-high prototype earth dam coul d
be model ed witha 30-cm (1-ft)-high centrifuge model accel erated to 100g (of course,
the particl e size ofthe model soil wil l correspond to a prototype particl e size
that is 100 times l arger, too). Aharmonic I-Hz base motion l asting 10 sec at the
prototype scal e woul d be model ed by a100-Hz motion l asting 0.1 sec in the model .
The dissipation of any generated pore pressures,however, woul d occur 10,000 times
as fast in the model as in the fiel d. For this reason,viscous fl uids such as
gl ycerin or sil icon oil are often used as pore fl uids in centrifugemodel s. Improved
estimates of prototype behavior can be obtained using the model ing ofmodel s
technique (Schofiel d, 1980), which invol ves comparing the response of model s of
different size at the same prototype scal e.Obviousl y, high-speed transducers and
data acquisition systems are required to obtainuseful resul ts in dynamic centrifuge
tests. Because the scal ing l aws appl y to al l parts of themodel , miniaturized
transducers and cabl es are required to minimize their infl uence on theresponse of
the model .Tabl e 6-2 Scal ing Factors for Centrifuge Model inq" .Type of EventAl l
eventsDynamic eventsDiffusion eventsQuantityStressStrainLengthMassDensityForce
GravityTimeFrequencyAccel erationStrain rateTimeStrain rateModel DimensionPrototype
DimensionIIl iNl IN3Il IN2Nl iNNNNl IN2N2Source: After Kutter and James (1989)."Val ues
are based on the assumption that the same soil s and fl uid areused in the model and
the prototype and that the soil properties are notrate dependent.Other Model Tests.
In recent years the hydraul ic gradient simil itude test,original l y devel oped by
Zel ikson (1969), has been used to measure dynamic soil properties(e.g., Yan and
Byrne, 1990, 1991). The hydraul ic gradient simil itude test achieves high bodyforces
by subjecting a model to a control l ed, downward-acting hydraul ic gradient. Since
thebody forces are al igned with fl ow l ines, the presence of fl ow obstructions or
nonuniform soil 228 Dynamic Soil Properties Chap. 6(6.40)conditions can cause
"gravity" to act in different directions at different l ocations. Sucheffects
compl icate interpretation of the test resul ts and restrict the use of the test to a
rel ativel ynarrow
range of probl ems.6.3.3 Interpretation of Observed Ground ResponseInterpretation
of the response of instrumented, ful l -scal e structures subjected to dynamicl oading
or earthquakes can provide inval uabl e information on dynamic soil properties and
other geotechnical earthquake engineering parameters. This approach requires wel l -
instrumentedsites, either with vertical arrays or cl osel y speced instruments at
soil and rock sites.At such sites the actual motions at the base and surface of a
soil deposit can be used with asuitabl e ground response model (Chapter 7) to
identify the dynamic soil properties (specifical l y,shear modul us and damping) that
produce the best agreement between predicted andactual motions (e.g., Abdel -Ghaffar
and Scott, 1979; Tokimatsu and Midorikawa, 1981;Chang et al ., 1991; Gl aser, 1995).
Obviousl y, this approach requires an earthquake to produceground motions, and the
strain l evel s to which the dynamic soil properties can be measuredare control l ed by
the strains produced by the earthquake. Exampl es of observedground response from
which dynamic soil properties can be identified are presented in Sections8.2.2 and
8.2.3. As instrumentation and remote data acquisition systems continue toimprove,
more ful l -scal e response data are l ikel y to become avail abl e.6.4 STRESS-STRAIN
BEHAVIOR OF CYCLICALLY LOADED SOILSThe mechanical behavior of soil s can be quite
compl ex under static, l et al one seismic l oadingconditions. Geotechnical engineers
are constantl y chal l enged by the need to characterizethe most important aspects of
cycl ic soil behavior as accuratel y as possibl e with simpl e,rational model s. The
point at which the confl icting requirements of simpl icity and accuracyare bal anced
depends on many factors, and many combinations have been proposed.For the purposes
of this book, three broad cl asses of soil model s wil l be discussed: equival ent
l inear model s, cycl ic nonl inear model s, and advanced constitutive model s. Ofthese,
equival ent l inear model s are the simpl est and most commonl y used but have l imited
abil ity to represent many aspects of soil behavior under cycl ic l oading conditions.
At the otherend of the spectrum, advanced constitutive model s can represent many
detail s of dynamic soil behavior, but their compl exity and difficul ty of cal ibration
currentl y renders them impractical for many common geotechnical earthquake
engineering probl ems. Neverthel ess, each cl assof soil model reveal s important
information about the cycl ic behavior of soil s.6.4.1 Some Basic Aspects of
Particul ate Matter BehaviorBefore considering specific stress-strain model s, it is
useful to examine some basic apsectsof the mechanical behavior of particul ate
media. Several important aspects of l ow-strainsoil behavior can be il l ustrated by
considering the soil as an assembl age of discrete el asticparticl es. Hertz (1881)
studied the behavior of identical spheres of radius, R, compressedagainst each
other by a normal force, N (Timoshenko and Goodier, 1951), and showed thatN =
2J2CR3/283/23(l -v) NSec. 6.4 Stress-Strain Behavior of Cycl ical l y Loaded Soil s 229
(6.41)(6.42)where G and V are the el astic constants of the spheres and ON is the
change in distancebetween the centers of the spheres. For a cubical l y packed array
of spheres l oaded al ong oneof the packing axes (Figure 6.38), the average normal
stress is obtained by dividing the normal force by its tributary area, that is,a =-
N- = N(2R)2 4R2Then the tangent modul us for uniaxial l oading is given byde dN/4R2 1
dN 3 [ 2G J2/3113El an = dE = doNI2R = 2RdoN = 2: 3(l -V) awhich suggests that the
stiffness shoul d, theoretical l y, vary with the cube root of the axial stress. When a
tangential force, T, is appl ied, el astic distortion causes the centers of the
spheres to be displ aced perpendicul ar to their original axis (Mindl in and
Deresiewicz, 1953;Dobry et al ., 1982) by an amount0T = [1 - ( 1 - ;,Y3J {31: (2 - v)
(1 + v) [3(1-4~) NRr1/3} T ~ fN (6.43)wherefis the coefficient offriction between
the spheres (note that OTis a nonl inear functionof T). When T becomes equal to [N,
gross sl iding of the particl e contacts occurs (thoughsl ippage of part of the
contact can occur before this point). This gross sl iding is required forpermanent
particl e reorientation; consequentl y, vol ume changes (drained conditions) cannot
occur and excess pore pressures (undrained conditions) cannot be generated when
grosssl iding does not occur. The shear strain corresponding to the initiation of
gross sl idingrYtv =0T(T = fN)2R(6.44)'t .....f---- LFigure 6.38 Cubical l y packed
assembl ageof spheres subjected to normal stress, o, andshear stress, 't, that
produce interparticl econtact forces Nand T. (After Dobry et al .,1982.)230 Dynamic
Soil Properties Chap. 6is cal l ed the vol umetric threshol d shear strain. When the
properties of quartz (E =11 x 106psi (7.6 x 107 kPa), v = 0.31,/= 0.50) are
substituted into equation (6.44), the threshol dshear strain is given byYtv(%) =
0.00017502/3 (6.45)where o is in psf. For confining pressures of practical interest
(500 to 4000 psf (25 to 200kPa)), equation (6.44) woul d predict a threshol d shear
strain between 0.01 and 0.04%. Real soil s, of course, do not consist of regul ar
arrays of spherical particl es, but the existence ofa threshol d shear strain very
cl ose to that predicted by equation (6.44) has been observedexperimental l y for
sands under both drained (Drnevich and Richart, 1970; Youd, 1972;Pyke, 1973) and
undrained (Park and Sil ver, 1975; Dobry and Ladd, 1980; Dobry et aI.,1982) l oading
conditions. The simpl e ideal ized anal ysis of a regul ar array of spheres hel ps
il l ustrate the reason for its existence. Experimental evidence suggests that the
vol umetricthreshol d shear strain increases with pl asticity index (PI); the
vol umetric threshol d shearstrain of a cl ay with PI =50 is approximatel y one order
of magnitude greater than that of asand, with PI = 0 (Vucetic, 1994). Experimental
evidence al so indicates (Vucetic, 1994) thatsoil s exhibit l inear el astic behavior
bel ow a l inear cycl ic threshol d shear strain, Ytl ' that isapproximatel y 30 times
smal l er than Ytv' .Such anal yses of the interaction of individual soil particl es
can provide insight intothe mechanical behavior of soil s. In recent years,
tremendous advances have been made inmicromechanical model ing of soil s.
Micromechanical model s account for the kinematicsand contact interactions of
individual soil particl es, thereby el iminating the need for a gl obal constitutive
model . A number of computational model s have been devel oped with various
restrictions on dimensional ity, particl e shape and size distribution, particl e
kinematics,and contact behavior (e.g., Cundal l and Strack, 1979; Ting et aI., 1989;
Ng and Dobry,1994). Al though the computational effort invol ved in tracking the
motion of each individual particl e is currentl y too l arge to al l ow their use for
practical probl ems with compl icatedboundary conditions, they have provided useful
insight into several aspects of soil behavior.6.4.2 Equival ent Linear Model A
typical soil subjected to symmetric cycl ic l oading as woul d be expected beneath a
l evel ground surface far from adjacent structures, might exhibit a hysteresis l oop
of the typeshown in Figure 6.39. This hysteresis l oop can be described in two ways:
first, by the actual path of the l oop itsel f, and second, by parameters that
describe its general shape. In general terms, two important characteristics of the
shape of a hysteresis l oop are its incl ination andits breadth. The incl ination of
the l oop depends on the stiffness of the soil , which can bedescribed at any point
during the l oading process by the tangent shear modul us, Gtan- Obviousl y,Gta n
varies throughout a cycl e of l oading, but its average val ue over the entire l oopcan
be approximated by the secant shear modul usG =: SO sec Yc(6.46)where "c. and Yc are
the shear stress and shear strain ampl itudes, respectivel y. Thus Gsecdescribes the
general incl ination of the hysteresis l oop. The breadth of the hysteresis l oop is
Sec. 6.4 Stress-Strain Behavior of Cycl ical l y Loaded Soil s 231tFigure 6.39 Secant
shear modul us, Gseeandtangent shear modul us, Gtawrel ated to the area, which as a
measure of energy dissipation, can convenientl y be described(Section B.6.1) by the
damping ratio(6.47)where WD is the dissipated energy, Ws the maximum strain energy,
and A100p the area of thehysteresis l oop. The parameters Gsec and ~ are often
referred to as equival ent l inear material parameters. For certain types of ground
response anal yses, they are used directl y to describethe soil behavior; other types
of anal yses require the actual path of the hysteresis l oop asdescribed by a cycl ic
nonl inear or advanced constitutive model . Both types of groundresponse anal ysis are
described in Chapter 7.Because some of the most commonl y used methods of ground
response anal ysis arebased on the use of equival ent l inear properties, considerabl e
attention has been given to thecharacterization of Gsec and ~ for different soil s,
It is important to recognize, however, thatthe equival ent l inear model is onl y an
approximation of the actual nonl inear behavior of thesoil . The assumption of
l inearity embedded in its use has important impl ications when it isused for ground
response anal ysis, as discussed in Chapter 7. It al so means that it cannot beused
directl y for probl ems invol ving permanent deformation or fail ure; equival ent l inear
model s impl y that the strain wil l al ways return to zero after cycl ic l oading, and
since a l inearmaterial has no l imiting strength, fail ure cannot occur.
Neverthel ess, the assumption ofl inearityal l ows a very efficient cl ass of
computational model s to be used for ground responseanal yses, and it is commonl y
empl oyed for that reason.6.4.2.1 Shear Modul usLaboratory tests have shown
that soil stiffness is infl uenced by cycl ic strain ampl itude,void ratio, mean
principal effective stress, pl asticity index, overconsol idation ratio,and number of
l oading cycl es. The secant shear modul us of an el ement of soil varies withcycl ic
shear strain ampl itude. At l ow strain ampl itudes, the secant shear modul us is high,
but it decreases as the strain ampl itude increases. The l ocus of points
corresponding to thetips of hysteresis l oops of various cycl ic strain ampl itudes is
cal l ed a backbone (or skel eton)232 Dynamic Soil Properties Chap. 6Modul us reduction
curve(a)GGmax1.0 -r---.Yo(b)l ogyFigure 6.40 Backbone curve showing typical
variation of Gsec with shear strain.curve (Figure 6040a); its sl ope atthe origin
(zero cycl ic strain ampl itude) represents the l argestval ue of the shear modul us,
Gmax • At greater cycl ic strain ampl itudes, the modul us ratioGseciGmax (the "sec"
subscript wil l hereafter be dropped for convenience and consistencywith the
notation commonl y used in the geotechnical earthquake engineering l iterature)drops
to val ues of l ess than 1. Characterization of the stiffness of an el ement of soil
thereforerequires consideration of both Gmaxand the manner in which the modul us
ratio G/Gmaxvarieswith cycl ic strain ampl itude and other parameters. The variation
of the modul us ratiowith shear strain is described graphical l y by a modul us
reduction curve (Figure 6040b). Themodul us reduction curve presents the same
information as the backbone curve; either onecan be determined from the other.
Maximum Shear Modul us, Gmax • Since most seismic geophysical testsinduce shear
strains l ower than about 3 x 10-4%, the measured shear wave vel ocities can beused
to compute Gmax as(6.48)The use of measured shear wave vel ocities is general l y the
most rel iabl e means of eval uatingthe in situ val ue of Gmax for a particul ar soil
deposit, and the seismic geophysical testsdescribed in Section 6.3.1.1 are commonl y
used for that purpose. Care must be taken in theinterpretation of shear wave
vel ocity, particul arl y at sites with anisotropic stress conditions,which can cause
measured shear wave vel ocities to vary with the direction of wave propagationand
particl e movement (Roesl er, 1979; Stokoe et al ., 1985; Yan and Byrne, 1991).When
shear wave vel ocity measurements are not avail abl e, Gmaxcan be estimated inseveral
different ways. Laboratory test data suggest that the maximum shear modul us canbe
expressed as(6.49)where F(e) is a function of the void ratio, OCR the
overconsol idation ratio, k an overconsol idationratio exponent (Tabl e 6-3), G: n the
mean principal effective stress [G;n = ( Gi + G;+ G3 )/3], n a stress exponent, and
Pa is atmospheric pressure in the same units as G: n andGmax. Hardin (1978) proposed
that F(e) = 1/(0.3 + 0.7e2) , whil e Jamiol kowski et al . (1991)suggested that F(e)
=1/eu .The stress exponent is often taken as n =0.5 but can be computedSec. 6.4
Stress-Strain Behavior of Cycl ical l y Loaded Soil sTabl e 6-3 Overconsol idation Ratio
Exponent, K233Pl asticity Indexo204060802100k0.000.180.300.410.480.50Source: After
Hardin and Dmevich (1972b).for individual soil s from the resul ts of l aboratory
tests at different effective confining pressures.It shoul d be apparent that Gmax,
Pa' and cr;" must be expressed in the .same units. Equation(6.49) can al so be used
to adjust measured Gmax val ues to represent conditions that aredifferent (e.g.,
increased effective stresses) from those at which the measurements were made.Other
empirical rel ationships have been proposed for specific soil types. The maximum
shear modul us of sand, for exampl e, is often estimated asGmax = 1000 Kz,max (cr;")
0.5 (6.50)where Kz max is determined from the void ratio or rel ative density (Tabl e
6-4) and cr;" is inl b/ftz(Seed and Idriss, 1970). Fiel d tests have consistentl y
shown that shear wave vel ocities ofgravel s are significantl y higher than those of
sands, indicating that Gmax of gravel is higher thanthat of sand. Kz,max val ues for
gravel s are typical l y in the range 80 to 180 (Seed et al ., 1984). Forfine-grained
soil s, prel iminary estimates of the maximum shear modul us can be obtained from
pl asticity index, overconsol idation ratio, and undrained strength (Tabl e 6-5).
Because undrainedstrengths are highl y variabl e and because shear modul i an
undrained strengths vary differentl ywith effective confining pressure, these
resul ts must be used careful l y.Tabl e 6-4 Estimation of K2,maxe K2• max Dr (%) K2,max
0.4 70 30 340.5 60 40 400.6 51 45 430.7 44 60 520.8 39 75 590.9 34 90 70Source:
Adapted from Seed and Idriss (1970).The maximum shear modul us can al so be estimated
from in situ test parameters. Anumber of empirical rel ationships between Gmax and
various in situ test parameters havebeen devel oped; some of the more widel y
publ ished are presented in Tabl e 6-6. The inherentdifficul ty of correl ating a smal l
strain parameter such as Gmax with penetration parametersthat rel ate to much l arger
strains is evident from the scatter in the data on which theyare based and from the
variabil ity of the resul ts obtained by different investigators. As such,the
useful ness of such correl ations is currentl y l imited to prel iminary estimates of
Gmax.234 Dynamic Soil PropertiesTabl e 6-5 Val ues of Gmax/suaOverconsol idation
Ratio, OCRChap. 6Pl asticity Index15-2020-2535-45l l OO70045029006003805600500300
Source: After Weil er (1988).a Undrained strength measured in CU triaxial
compression.However, the appl ication of in situ testing to geotechnical earthquake
engineering probl emsis onl y in its earl y stages, and significant advances can be
expected as additional databecome avail abl e.Eval uation of shear modul us can be
compl icated by rate and time effects (Andersonand Woods, 1975, 1976; Anderson and
Stokoe, 1978; Isenhower and Stokoe, 1981). Rateeffects can cause G max to increase
with increasing strain rate. The infl uence of strain rate onG max increases with
increasing soil pl asticity; for San Francisco Bay mud (PI", 40), G maxincreases
about 4% per tenfol d increase in strain rate. Rate effects can be significant when
comparing G max val ues obtained from fiel d shear wave vel ocity measurements
(usual l ymade with the use of impul sive disturbances which produce rel ativel y high
frequencies)with val ues obtained from l aboratory tests. The shear wave vel ocity,
and hence G max'increases approximatel y l inearl y with the l ogarithm of time past
the end of primary consol idationto an extent that cannot be attributed sol el y to
the effects of secondary compression.The change of stiffness with time can be
described byAGmax = NG(Gmax)l OOO (6.51)where AGmax is the increase in G max over
one l og cycl e of time and (Gmax)l OOO is the val ueof G max at a time of 1000 min
past the end of primary consol idation. NG increases withincreasing pl asticity
index, PI, and decreases with increasing OCR (Kokushu et aI., 1982).For normal l y
consol idated cl ays, NG can be estimated from the rel ationshipNG ", 0.027 JPI
Anderson and Woods (1975) showed that some of the discrepancy between G max val ues
fromfiel d and l aboratory tests coul d be expl ained by time effects, and thatNG coul d
be used to correctthe G max val ues from l aboratory tests to better represent actual
in situ conditions.A brief summary of the effects of environmental and l oading
conditions on the maximumshear modul us of normal l y and moderatel y overconsol idated
soil s is presented inTabl e 6-7.Modul us Reduction, GJ'Gmax ' In the earl y years of
geotechnical earthquakeengineering, the modul us reduction behaviors of coarse- and
fine-grained soil s were treatedseparatel y (e.g., Seed and Idriss, 1970). Recent
research, however, has reveal ed a gradual transition between the modul us reduction
behavior of nonpl astic coarse-grained soil andpl astic fine-grained soil .Zen et al .
(1978) and Kokushu et al . (1982) first noted the infl uence of soil pl asticityon the
shape of the modul us reduction curve; the shear modul us of highl y pl astic soil s was
Tabl e 6-6 Empirical Rel ationships between Gmax and In Situ Test Parameters
Rel ationship Soil Type ReferencesGmax = 20,000(Nl )~~33(cr;n)0.5 Sand Ohta and Goto
(1976),Seed et al . (1986)Gmax = 325N2068 Sand Imai and Tonouchi(1982)Gmax =
1634(qc)0.250 (cr~ )0.375 Quartz sand Rix and Stokoe (1991)Gmax and cr;"in Ib/ft2
Gmax in kips/ft2Gmax' qc' and cr~in kPa;Based on fiel d tests inItal y and on
cal ibrationchamber testsGmax' qc' and cr~ in kPa;Based on fiel d tests inItal yGmax'
qc' and cr~in kPa;Based on fiel d tests atworl dwide sitesBased on cal ibrationchamber
testsBased on fiel d testsCommentsGmax• Pa' (j~in sameunits; Yo is dil atometerbased
unit weight of soil ;based on fiel d testsGur• c is correctedunl oading-rel oading
modul us from cycl ic PMTGur is secant modul us ofunl oading-rel oadingportion ofPMG;
(XI' isfactor that depends onunl oading-rel oadingstress conditions; basedon theory
and fiel d test dataHryciw (1990)Bal di et al . (1986)Mayne and Rix (1993)Bel l otti et
al . (1986)Bal di et al . (1986)Bel l otti et al . (1986)Byrne et al . (1991)Cl aySil ica
sandSandSandSand, sil t, cl aySandSandGmax = 406(qc )0.695 e-1.130Gmax / Ed = 2.72 ±
0.59Gmax / Ed = 2.2 ± 0.7Gmax= 530 Yo/Yw -I 025 05 (cr~ / Pa)0.25 2.7 - Yo/Yw Ko'
(Pacr~)'3.6 $ gmax $ 4.8ur,cGmax = 1.68 G(X urp(Figure 6.41)SPTCPTDMTPMTIn SituTest
N'-IIIII236 Dynamic Soil Properties Chap. 6Figure 6.41 Estimation of Gmax from CPT
tip resistance for uncemented sil ica sands.(After Bal di et al ., 1989.)500 1000
2000300026 r---;----: -----: -----,------,-----,,-----,----,.--.,----,24 ..22 ..20··
18 ..16 ".14 ".12 ""10 ..8 ..6 ""4 ""2 ."
0'-----'--'--'--'--'-'---'--''-----'------'200Tabl e 6-7 Effect of Environmental and
Loading Conditions onMaximum Shear Modul us of Normal l y Consol idated and Moderatel y
Overconsol idated
Soil sIncreasing FactorEffective confining pressure, cr;"Void ratio, eGeol ogic age,
'« Cementation, cOverconsol idation ratio, OCRPl asticity index, PIStrain rate, y
Number of l oading cycl es, NIncreases with cr;"Decreases with eIncreases with tg
Increases with cIncreases with OCRIncreases with PI if OCR> 1;stays about constant
if OCR = 1No effect for non-pl astic soil s;increases with y for pl astic soil s(up to
-10% increase per l og cycl eincrease in y)Decreases after N cycl es of l arge Ye'but
recovers l ater with time in cl ays;increases with N for sandSource: Modified from
Dobry and Vucetic (1987).observed to degrade more sl owl y with shear strain than did
l ow-pl asticity soil s. Afterreviewing experimental resul ts from a broad range of
material s, Dobry and Vucetic (1987)and Sun et al . (1988) concl uded that the shape
of the modul us reduction curve is infl uencedmore by the pl asticity index than by
the void ratio and presented curves of the type shownin Figure 6.42. These curves
show that the l inear cycl ic threshol d shear strain, Ytt, is greater forhighl y
pl astic soil s than for soil s of l ow pl asticity. This characteristic is extremel y
important;it can strongl y infl uence the manner in which a soil deposit wil l ampl ify
or attenuate earthquakemotions. The PI = 0 modul us reduction curve from Figure 6.42
is very simil ar to theSec. 6.4 Stress-Strain Behavior of Cycl ical l y Loaded Soil s1.0
0.8G0.6 ....PI = 200Gmax o~ 0.40.2IOCR=1-1SI0.00.0001 0.001 0.01 0.1 10Cycl ic shear
strain, Yc (%)Figure 6.42 Modul us reduction curves for fine-grained soil s of
different pl asticity.(After Vucetic and Dobry (1991). Effect of soil pl asticity on
cycl ic response. Journal ofGeotechnical Engineering, Vol . 117, No. I. Reprinted by
permission of ASCE.)237average modul us reduction curve that was commonl y used for
sands (Seed and Idriss, 1970)when coarse- and fine-grained soil s were treated
separatel y. This simil arity suggests that themodul us reduction curves of Figure
6.42 may be appl icabl e to both fine- and coarse-grainedsoil s (this concl usion
shoul d be confirmed for individual coarse-grained soil s, particul arl ythose that
coul d exhibit aging or cementation effects). The difficul ty of testing very l arge
specimens has precl uded the widespread testing of gravel l y soil s in the l aboratory,
but avail abl etest data indicate that the average modul us reduction curve for gravel
is simil ar to,though sl ightl y fl atter than, that of sand (Seed et al ., 1986; Yasuda
and Matsumoto, 1993).Modul us reduction behavior is al so infl uenced by effective
confining pressure, particul arl yfor soil s of l ow pl asticity (Iwasaki et al ., 1978;
Kokoshu, 1980). The l inear cycl icthreshol d shear strain, Ytl , is greater at high
effective confining pressures than at l ow effectiveconfining pressures. The effects
of effective confining pressure and pl asticity index onmodul us reduction behavior
were combined by Ishibashi and Zhang (1993) in the formwhereK(y, PI)m(y, PI) -maG =
K(y, PI)()~)m(y,PI)-moGmax0.5 { 1 + tanh [In( 0.0001O~+ n (PI) Y.492J}{ [ (0 000556
)O.4J } 0.272 1 - tanh In . y exp(-0.0145PI1.3)(6.52)n(PI) 10.03.37 X 10-6 PI J.404
7.0 X 10-7 PI1.9762.7 X 10-5 PIl . l l 5for PI = 0for 0 < PI ~ 15for 15 < PI s 70for
PI> 70238 Dynamic Soil Properties Chap. 60.201.000.0010-6 10'5 10-4 10'3 10-2 10-1
Cycl ic shear strain ampl itude, y0.80G 0.60Gmax0.40(J~ ~ 400 k~m(J~ ~ 200k~m(J~ ~
50k~m....-----'l .~A_(J~ = 1k~m0.0010-6 10-5 10'4 10'3 10'2 10-1Cycl ic shear strain
ampl itude, y1.000.800.20G 0.60Gmax0.40(a) (b)Figure 6.43 Infl uence of mean
effective confining pressure on modul us reductioncurves for (a) nonpl astic (PI =0)
soil , and (b) pl astic (PI =50) soil . (After Ishibashi(1992). Discussion, Journal
ofGeotechnical Engineering, Vol . 118, NO.5. Reprinted bypermission of ASCE.)The
effect of confining pressure on modul us reduction behavior of l ow- and high-
pl asticitysoil s is il l ustrated in Figure 6.43.Under stress-control l ed harmonic
l oading conditions, pore pressure generation andstructural changes can cause the
shear strain ampl itude of a soil specimen to increase withincreasing number of
cycl es. If cl ay or saturated sand specimens are l oaded harmonical l yunder strain-
control l ed undrained conditions, the shear stress ampl itude woul d be observedto
decrease with increasing number of cycl es. Both conditions il l ustrate the tendency
forrepeated cycl ic l oads to degrade the stiffness of the specimen. For cohesive
soil s, the val ueof the shear modul us after N cycl es, GN, can be rel ated to its
val ue in the first cycl e, GJ, by(6.53)where the degradation index, 0, is given by 0
= N-t and t is the degradation parameter(Idriss et al ., 1978). The degradation
parameter has been shown to decrease with increasingPI and increasing
overconsol idation ratio, and to increase with increasing cycl ic strainampl itude
(Idriss et al ., 1980; Vucetic and Dobry, 1989; Tan and Vucetic, 1989). Theeffects
of stiffness degradation on modul us reduction behavior is shown in Figure 6.44.The
infl uence of various environmental and l oading conditions on the modul us ratioof
normal l y consol idated and moderatel y overconsol idated cl ays is described in Tabl e
6-8.6.4.2.2 Damping RatioTheoretical l y, no hysteretic dissipation of energy takes
pl ace at strains bel ow the l inearcycl ic threshol d shear strain. Experimental
evidence, however, shows that some energyis dissipated even at very l ow strain
l evel s (the mechanism is not wel l understood), so thedamping ratio is never zero.
Above the threshol d strain, the breadth of the hysteresis l oopsexhibited by a
cycl ical l y l oaded soil increase with increasing cycl ic strain ampl itude, which
indicates that the damping ratio increases with increasing strain ampl itude.Just as
modul us reduction behavior is infl uenced by pl asticity characteristics, so is
damping behavior (Kokushu et al ., 1982; Dobry and Vucetic, 1987; Sun et aI., 1988).
Damping ratios of highl y pl astic soil s are l ower than those of l ow pl asticity soil s
at the sameSec. 6.4 Stress-Strain Behavior of Cycl ical l y Loaded Soil s 23910N=110100
10000.01 0.1N=11010010000.0010.0 '-- -'---- --'- ~ __'_ ___J0.00010.2 ....0.8 ....
0.40.6Cycl ic shear strain, Yc (%)Figure 6.44 Effect of cycl ic degradation on shear
modul us. (After Vucetic and Dobry(1991). Effect of soil pl asticity on cycl ic
response, Journal ofGeotechnical Engineering,Vol 117, No.1. Reprinted by permission
of ASCE.)Tabl e 6-8. Effect of Environmental and Loading Conditions onModul us Ratio
(at a Given Strain Le\(~1) of Normal l y Consol idatedand Moderatel y Overconsol idated
Soil sIncreasing Factor G/GmaxConfining pressure, a;"Void ratio, eGeol ogic age, tg
Cementation, cOverconsol idation ratio, OCRPl asticity index, PICycl ic strain, Yc
Strain rate, yNumber of l oading cycl es, NIncreases with a;n; effect decreaseswith
increasing PIIncreases witb eMay increase with tgMay increase with cNot affected
Increases with PIDecreases with YcG increases with Yo but G/Gmaxprobabl y not
affected if G and Gmaxare measured at same yDecreases after N cycl es of l arge Yc
(Gmax measured before N cycl es) forcl ays; for sands, can increase (underdrained
conditions) or decrease(under undrained conditions)Source: Modified from Dobry and
Vucetic (1987).cycl ic strain ampl itude (Figure 6.45). The PI = 0 damping curve from
Figure 6.45 is nearl yidentical to the average damping curve that was used for
coarse-grained soil s when they240 Dynamic Soil Properties Chap. 625 ....20~~o~15Ol
c"0.~ o 105 ....o0.0001 0.001 0.01 0.1PI =015305010020010(6.54)Cycl ic shear strain,
Yc(%)Figure 6.45 Variation of damping ratio of fine-grained soil with cycl ic shear
strainampl itude and pl asticity index. (After Vucetic and Dobry (1991). Effect of
soil pl asticityon cycl ic response, Journal a/Geotechnical Engineering, Vol . 117,
No. 1. Reprinted bypermission of ASCE.)were treated separatel y from fine-grained
soil s. This simil arity suggests that the dampingcurves of Figure 6.45 can be
appl ied to both fine- and coarse-grained soil s. The dampingbehavior of gravel is
very simil ar to that of sand (Seed et aI., 1984).Damping behavior is al so
infl uenced by effective confining pressure, particul arl y forsoil s of l ow
pl asticity. Ishibashi and Zhang (1993) devel oped an empirical expression forthe
damping ratio of pl astic and nonpl astic soil s. Using equation (6.52) to compute the
modul usreduction factor, GIGmax' the damping ratio is given by~ = 0.333 1+ expC-
0.0145PI1.3) [0.586(...£)2 -1.547...£ + IJ2 Gmax GmaxThe infl uence of various
environmental and l oading conditions on the damping ratio of normal l yconsol idated
and moderatel y overconsol idated soil s is described in Tabl e 6-9.6.4.3 Cycl ic
Nonl inear Model sThe nonl inear stress-strain behavior of soil s can be represented
more accuratel y by cycl ic nonl inearmodel s that fol l ow the actual stress-strain path
during cycl ic l oading. Such model s areabl e to represent the shear strength of the
soil , and with an appropriate pore pressure generationmodel , changes in effective
stress during undrained cycl ic l oading. A variety of cycl ic nonl inearmodel s have
been devel oped; al l are characterized by (1) a backbone curve and (2) a seriesof
"rul es" that govern unl oading-rel oading behavior, stiffness degradation, and other
effects.The simpl est of these model s have rel ativel y simpl e backbone curves and
onl y a few basicrul es. More compl ex model s may incorporate many additional rul es
that al l ow the model tobetter represent the effects of irregul ar l oading,
densification, pore pressure generation, orSec. 6.4 Stress-Strain Behavior of
Cycl ical l y Loaded Soil sTabl e 6-9. Effect of Environmental and Loading Conditions on
Damping Ratio of Normal l y Consol idated and Moderatel yOverconsol idated Soil s241
Increasing FactorConfining pressure,
cr;"Void ratio, eGeol ogic age, tgCementation, cOverconsol idation ratio, OCR
Pl asticity index, PICycl ic strain, YcStrain rate, yNumber of l oading cycl es, N
Damping ratio, ~Decreases with cr;" ; effect decreaseswith increasing PIDecreases
with eDecreases with tgMay decrease with cNot affectedDecreases with PIIncreases
with YcStays constant or may increase with yNot significant for moderate YcandN
(6.55)Source: Modified from Dobry and Vucetic (1987).other effects. The
appl icabil ity of cycl ic nonl inear model s, however, is general l y restricted toa
fairl y narrow, al beit important range of initial conditions and stress paths.The
performance of cycl ic nonl inear model s can be il l ustrated by a very simpl e exampIe
in which the shape of the backbone curve is described by 't =Fbb(Y)' The shape of
anybackbone curve is tied to two parameters, the initial (l ow-strain) stiffness and
the (highstrain)shear strength of the soil . For the simpl e exampl e, the backbone
function, Fbb(y), canbe described by a hyperbol aGmaxY1 + (Gmax/'tmax) l yl The shape
of the hyperbol ic backbone curve is il l ustrated in Figure 6.46. Other expressions
[e.g., the Ramberg-Osgood model (Ramberg and Osgood, 1943)] can al so be used to
describe the backbone curve. Al ternativel y, backbone curves can be constructed from
modul usreduction curves.The quantities G max and 'tmax may be measured directl y,
computed, or obtained byempirical correl ation. For the exampl e model , the response
of the soil to cycl ic l oading isgoverned by the fol l owing four rul es: 1. For initial
l oading, the stress-strain curve fol l ows the backbone curve.yFigure 6.46 Hyperbol ic
backbone curveasymptotic to 1: = Gm• x yand to 1: = 1: m• x(and 1: = -1: m• x) '242
Dynamic Soil Properties Chap. 62. If a stress reversal occurs at a point defined by
(y" r.), the stress-strain curve fol l owsa path given by't"-'t"r _F (Y-Yr) -2- - bb
-2-In other words, the unl oading and rel oading curves have the same shape as the
backbonecurve (with the origin shifted to the l oading reversal point) but is
enl arged by afactor of 2. These first two rul es, which describe Masing behavior
(Masing, 1926), arenot sufficient to describe soil response under general cycl ic
l oading. As a resul t, additional rul es are needed.3. If the unl oading or rel oading
curve exceeds the maximum past strain and intersectsthe backbone curve, it fol l ows
the backbone curve until the next stress reversal .4. If an unl oading or rel oading
curve crosses an unl oading or rel oading curve from theprevious cycl e, the stress-
strain curve fol l ows that of the previous cycl e.Model s that fol l ow these four rul es
are often cal l ed extended Masing model s. Anexampl e of the extended Masing model is
shown in Figure 6.47. Cycl ic l oading begins atpointA, and the stress-strain curve
during initial l oading (from A to B) fol l ows the backbonecurve as required by rul e
1. At point B, the l oading is reversed and the unl oading portion ofthe stress-
strain curve moves away from B al ong the path required by rul e 2. Note that the
initial unl oading modul us is equal to Gmax. The unl oading path intersects the
backbonecurve at point C and, according to rul e 3, continues al ong the backbone
curve until the nextl oading reversal at point D. The rel oading curve then moves
away from D as required byrul e 2, and the process is repeated for the remainder of
the appl ied l oading. Al though thismodel is very simpl e and is expressed onl y in
terms of effective stresses, it inherentl y incorporatesthe hysteretic nature of
damping and the strain-dependence of the shear modul usand damping ratio. Other
unl oading-rel oading model s are avail abl e (e.g., Iwan, 1967; Finnet al ., 1977;
Vucetic, 1990); the Cundal l -Pyke model (Pyke, 1979) is particul arl y straightforward
and easil y impl emented into ground response anal yses. To avoid spurious responseat
very l ow strain l evel s, some cycl ic nonl inear model s require the addition of a
smal l yF~Backbonecurve"t "t(a) (b)Figure 6.47 Extended Masing rul es: (a) variation
of shear stress with time; (b)resul ting stress-strain behavior (backbone curve
indicated by dashed l ine).Sec. 6.4 Stress-Strain Behavior of Cycl ical l y Loaded
Soil s 243amount of l ow strain damping. Note that the cycl ic nonl inear model does
not require theshear strain to be zero when the shear stress is zero. The abil ity
to represent the devel opmentof permanent strains is one of the most important
advantages of cycl ic nonl inear model sover equival ent l inear model s.This simpl e
exampl e model does not, however, al l ow for the determination of shearinduced
vol umetric strains that can l ead to hardening under drained conditions or to pore
pressure devel opment with attendant stiffness degradation under undrained
conditions.Such factors are accounted for in the majority ofthe cycl ic nonl inear
model s commonl y usedin geotechnical earthquake engineering practice (e.g., Finn et
al ., 1977; Pyke, 1979, 1985).The abil ity to compute changes in pore pressure, hence
al so changes in effectivestress, represents another significant advantage of cycl ic
nonl inear model s over equival entl inear model s. As pore pressures increase,
effective stresses decrease, and consequentl y theval ues of Gmax and 'tmax decrease.
Since the shape and position of the backbone curvedepends on Gmax and 'tmax, the
backbone curve "degrades" with increasing pore pressure. Aswith actual soil s, the
stiffness in a stress-strain model depends not onl y on the cycl ic strainampl itude,
as impl ied by the equival ent l inear model , but al so on the stress history of the
soil . When incorporated into computational model s for ground response anal ysis,
cycl icnonl inear model s al l ow prediction of the generation, redistribution, and
eventual dissipationof pore pressures during and after earthquake shaking. These
capabil ities are very useful for eval uation of l iquefaction hazards (Chapter 9).
6.4.4 Advanced Constitutive Model sThe most accurate and general methods for
representation of soil behavior are based onadvanced constitutive model s that use
basic principl es of mechanics to describe observedsoil behavior for (a) general
initial stress conditions, (b) a wide variety of stress paths, (c)rotating
principal stress axes, (d) cycl ic or monotonic l oading, (e) high or l ow strain
rates,and (f) drained or undrained conditions.Such model s general l y require a yiel d
surface that describes the l imiting stress conditionsfor which el astic behavior is
observed, a hardening l aw that describes changes in thesize and shape of the yiel d
surface as pl astic deformation occurs, and afl ow rul e that rel atesincrements of
pl astic strain to increments of stress. The Cam-Cl ay (Roscoe and Schofiel d,1963)
and modified Cam-Cl ay (Roscoe and Burl and, 1968) model s were among the first ofthis
type. Improvements in the prediction of shear strains have resul ted from the use of
mul tipl enested yiel d l oci within the yiel d surface (Mroz, 1967; Prevost, 1977) and
the devel opmentof bounding surface model s (Dafal ias and Popov, 1979) which
incorporate asmooth transition from el astic to pl astic behavior. Detail ed treatment
of such advanced constitutivemodel s is beyond the scope of this book. The
interested reader can refer to a numberof sources, incl uding Desai and Siriwardane
(1984), Dafal ias and Herrmann (1982),Wroth and Houl sl ey (1985), Lade (1988), and
Wood (1991).Al though advanced constitutive model s al l ow considerabl e fl exibil ity
and general ityin model ing the response of soil s to cycl ic l oading, their
description usual l y requires manymore parameters than equival ent l inear model s or
cycl ic nonl inear model s. Eval uation ofthese parameters can be difficul t, and the
parameters obtained from one type of test can bedifferent from those obtained from
another. Al though the use of advanced constitutive model l swil l undoubtedl y
increase, these practical probl ems have, to date, l imited their use ingeotechnical
earthquake engineering practice.2446.4.5 DiscussionDynamic Soil Properties Chap. 6A
hierarchy of model s are avail abl e for characterization of the stress-strain
behavior ofcycl ical l y l oaded soil s. The model s range considerabl y in compl exity and
accuracy; amodel that is appropriate for one type of probl em may not be appropriate
for another. Nosingl e stress-strain model is appropriate for al l probl ems.
Sel ection of a stress-strain model requires careful consideration of the probl em to
which it is to be appl ied, recognition of theassumptions and l imitations of the
avail abl e model s, and a good understanding of how themodel is used in al l required
anal yses.6.5 STRENGTH OF CYCLICALLY LOADED SOILSThe effect of cycl ic l oading on the
l imiting strength of soil s is of considerabl e importancein geotechnical earthquake
engineering. Probl ems of sl ope stabil ity, foundation performance,and retaining wal l
behavior, among others, are strongl y infl uenced by the strengththat the soil can
mobil ize at l arge strains.Soil strength behavior is most convenientl y discussed in
terms of coarse-grained cohesionl esssoil s and fine-grained cohesive soil s under
drained and undrained conditions. Earthquakel oading is general l y appl ied so rapidl y
that al l but the most permeabl e of soil s arel oaded under undrained conditions. The
strength of cohesionl ess soil s is inextricabl y tied tothe phenomenon of
l iquefaction, a probl em so important in geotechnical earthquake engineeringthat it
is treated in a separate chapter (Chapter 9). The fol l owing discussion isdirected
toward the effect of cycl ic l oading on the undrained shear strength of cohesive
soil s.6.5.1 Definitions of Fail ureThe shear strength of an el ement of soil is
typical l y defined as the shear stress mobil ized atthe point of "fail ure," but
fail ure can be defined in many different ways. In the fiel d, fail ureis usual l y
associated with deformations that exceed some serviceabil ity l imit. Since
deformationresul ts from the integration of strains over some vol ume of soil , the
point of fail ureof an el ement of soil is often defined in terms of a l imiting
strain.Consider an el ement of soil in drained equil ibrium under anisotropic stress
conditionsin a cycl ic direct simpl e shear test (point A in Figure 6.48). The
appl ication of a cycl ic shearstress, 'teye' produces (under stress-control l ed
conditions) a cycl ic shear strain, Yeye, but al soan increase in the average strain,
Yave' The average shear strain increases with increasingnumbers of l oading cycl es.
Cl earl y, the strength of the soil during cycl ic l oading coul d bedefined in terms of
l imiting val ues of Ycyc or Yave or of some combination of the two. Theavail abl e
strength of the soil under monotonic l oading (after the cycl ic l oading has ended)
may al so be of interest.The fol l owing sections discuss two measures of the strength
of a cycl ical l y l oadedsoil . The "cycl ic" strength is based on a l imiting val ue of
cycl ic and/or average strain duringcycl ic l oading (al though the soil may not
actual l y be in a state offail ure as defined by effectivestress conditions). The
"monotonic" strength is the ul timate static strength that can bemobil ized after
cycl ic l oading has ended.Sec. 6.5 Strength of Cycl ical l y Loaded Soil s 245Shear
stressShearstrainoy! Excitation..TimeResponseOil ~~IConsol idation Creep'yShear6.5.2
Cycl ic StrengthTime Figure 6.48 Definitions of average andcycl ic shear stress and
shear strain. (AfterGoul ois et al ., 1985. Used by permission ofASTM.)The l evel s of
both cycl ic and permanent deformations are of interest in a number of geotechnical
earthquake engineering probl ems. They are al so important in the design of
foundationsfor marine structures subjected to wave l oading, and much of the current
state ofknowl edge of cycl ic strength has come from research in that area. The
cycl ic strength of anel l ement of soil depends on the rel ationship between the
average shear stress, "Cave, and thecycl ic shear stress, "Ccyc' When the average
shear stress is l ow, unidirectional strains wil l accumul ate sl owl y, so the average
shear strain wil l al so be l ow. The ampl itude of the cycl icstrain, however, may
become l arge if the cycl ic shear stress is l arge. If, on the other hand,the average
shear stress is high (rel ative to the static shear strength, su), substantial
unidirectional strains can devel op even when the cycl ic shear stress is smal l .For
the case of "Cave = 0, no unidirectional strain wil l devel op, so fail ure must be
defined in terms of the cycl ic shear strain, Ycyc' When fail ure is defined in terms
of a specificl evel of cycl ic shear strain (often 3 percent), the cycl ic strength
ratio, defined as s.;»;decreases with increasing numbers of cycl es, as shown in
Figure 6.49. At cycl ic stress ratiosbel ow some l imiting val ue, however, the fail ure
strain wil l never be reached (i.e., stabl eresponse wil l be achieved). This l imiting
cycl ic stress ratio, referred to as the critical l evel of repeated l oading (CLRL) by
Sangrey et al . (1969), increases with increasing soil pl asticity.Hermann and
Houston (1980) found CLRL val ues ranging from 0.05 for a nonpl asticsil t to 0.55 for
San Francisco Bay mud.For cases in which "Cave is greater than zero, both Ycyc and
Yave wil l depend on "Ccyc and"Cave (Seed and Chan, 1966). Investigations of the
cycl ic response of marine cl ays (e.g., Meimonand Hicher, 1980; Goul ois et al .,
1985; Andersen et al ., 1988) have shown that Ycycdepends predominantl y on "Ccyc and
the number of cycl es, and Yave depends predominantl yon "Cave and the number of
cycl es (Figure 6.50).In the devel opment of a procedure for estimating earthquake-
induced permanentdeformations in dams and embankments, Makdisi and Seed (1978)
defined the dynamicyiel d strength of soil s that exhibit smal l changes in pore
pressure under undrained l oadingas 80% of the undrained strength of the soil .
Substantial permanent deformations candevel op when the total (static pl us cycl ic)
shear stress exceeds the dynamic yiel d strength.Dynamic Soil Properties----£f>5%..
- - - - - - - £f ",,3%..............Chap. 61000 10,000---- "''' -.. .3 10 10 100
100246120~~100~~l fJJ~0 80~s: 0,c 60~Ci5g 40(3>,020aCycl es to fail ure, NFigure 6.49
Variation of cycl ic strength ratio with number of cycl es for different soil s.(After
Lee and Focht, 1976.)1.1.. 1.00.9.. 0.8.. 0.7~_--,0.6~~~"'--l " 0.5/[ .. 0.40.3..
0.26 4 32.5-Yave(%) .. 0.11.1 1.00.90.80.70.60.50.4 0.3 0.2 0.1 0~'t strength~'t
strengthFigure 6.50 Variation of average shearstrain with average shear stress,
cycl ic shearstress, and number of cycl es in cycl ic directsimpl e shear tests on
pl astic Drammen cl ay.'tstrength is a reference strength measuredunder sl ow l oading
conditions with 't ave = O.Cycl ic l oading appl ied at a period of 10 sec.(After
Goul ois et al ., 1985. Used bypermission of ASTM.)6.5.3 Monotonic StrengthEval uation
of the static stabil ity of sl opes and retaining wal l s and the capacity of
foundationsafter earthquake shaking has ended is another important probl em in
geotechnical earthquake engineering. Such probl ems require eval uation of the
avail abl e shear strength ofthe soil after the earthquake has ended. This
postearthquake strength must refl ect anyeffects of cycl ic l oading imposed by the
earthquake.As pointed out by Castro and Christian (1976), the ul timate (residual ,
high-strain)undrained shear strength of a saturated soil is control l ed by its void
ratio and structure. Barringany change in soil structure, a saturated soil at a
particul ar void ratio wil l mobil ize aSec. 6.5 Strength of Cycl ical l y Loaded Soil s
2472 3 4p' (ksc)5 052 3 4Axial strain (%)22specific undrained strength, with l ittl e
infl uence of the history of stresses and strains bywhich that strength is arrived
at. For such soil conditions, the undrained strength after cycl icl oading wil l be
equal to the undrained strength before undrained l oading (at the same strainrate).
Since cycl ic l oading induces positive excess pore pressures, the effective stress
in anel ement of soil sheared monotonical l y after being subjected to cycl icl oading
wil l be l owerthan that in an identical el ement that is sheared monotonical l y
without prior cycl ic l oading.Consequentl y, the el ement that had been cycl ed woul d
be expected to exhibit more dil ativebehavior but to have a l ower stiffness in the
earl y stages of monotonic undrained l oadingthan the el ement that had not been
cycl ed.Changes in monotonic strength cal l be caused by disturbance ofthe soil
structure duringcycl ic l oading. The extent to which the structure of the soil is
disturbed is infl uenced bythe rel ationship between the cycl ic strain ampl itude and
the strain at which fail ure occursunder monotonic l oading conditions (Thiers and
Seed, 1969). Substantial structural disturbancecan modify the stress-strain
behavior and reduce the monotonic shear strength. Thesix triaxial specimens shown
in Figure 6.51 had simil ar void ratios (except specimen 6,which had a somewhat
higher void ratio than the rest) at the end of consol idation. Specimen1 was sheared
monotonical l y immediatel y after consol idation, but specimens 2 to 6 werefirst
subjected to varying l evel s of cycl ic l oading. Since the void ratios were nearl y
thesame, the specimens woul d therefore be expected to have simil ar monotonic
strengths. Asshown by the stress-strain curves and stress paths, they behaved
l argel y as woul d beexpected. After being subjected to different l evel s of cycl ic
strain, their ul timate (l argestrain) strengths were simil ar (except specimen 6,
which was l ower than the others). Differencesin the ul timate strength can be
expl ained by smal l differences in the void ratios andal so by differences in the
extent of structural disturbance induced by the cycl ic l oading.Thiers and Seed
(1969) found that the ul timate strength of three cl ays decreased byl ess than 10%
when the cycl ic strain ampl itude was l ess than one-hal f of the fail ure strainfrom
monotonic tests. At higher cycl ic strain ampl itudes, the reduction in strength was
more3r----------------,Figure 6.51 Effect of cycl ic l oading on subsequent monotonic
undrained l oadingbehavior of triaxial specimens of a sl ightl y pl astic sil t: (a)
stress-strain behavior; (b)effective stress path behavior. Specimen I was tested in
conventional CD test with noprior cycl ic l oading. Specimens 2 to 6 were subjected
to different l evel s of cycl ic l oadingprior to monotonic l oading. Note the dil ative
nature of the stress paths of specimens 2 to6 compared to specimen I. (After Castro
and Christian, 1976.)248 Dynamic Soil Properties Chap. 6dramatic, as il l ustrated in
Figure 6.52. Simil ar resul ts have been obtained by others (e.g.,Koutsoftas, 1978;
Byrne et aI., 1984).0.8 , , ,\0 ; __ . .;0.6 c - -., '\- -: .: .: .; - .0.4 , _ ,
-",,.: ,0.2 , _ , _ -j- - ·······co San Bay mud'" Anchorage sil ty cl ay, zoneU
Ancnoraqe suty cl ay, zone 36.6 SUMMARY1. Earthquake damage is strongl y infl uenced
by the dynamic response of soil deposits.The cycl ic nonl inear and strength
characteristics of soil s govern their dynamic responseduring earthquakes.2. The
measurement of dynamic soil properties is an important aspect of geotechnical
earthquake engineering. A variety of fiel d and l aboratory techniques are avail abl e;
some are oriented toward measurement of l ow-strain behavior and others toward
measurement of soil behavior at high strain l evel s.3. Fiel d tests al l ow measurement
of soil properties in situ; the compl ex effects of existingstress, chemical ,
thermal , and structural conditions are therefore refl ected in themeasured soil
properties. Many fiel d tests measure the response of l arge vol umes ofsoil and
induce soil deformation simil ar to those induced by earthquakes. Because insitu
conditions cannot be easil y control l ed
or varied, fiel d tests do not al l ow measurementof the behavior of the soil under
other stress states or soil conditions.4. A number of fiel d tests measure l ow-
strain soil properties, particul arl y wave propagationvel ocities. These tests
incl ude seismic refl ection, seismic refraction, steady-statevibration, spectral
anal ysis of surface waves, seismic crosshol e, seismic downhol e(and uphol e), and
seismic cone tests. Other fiel d tests, such as the standard penetration,cone
penetration, dil atometer, and pressuremeter tests, measure the properties of the
soil at higher strain l evel s.Sec. 6.5 Strength of Cycl ical l y Loaded Soil s 2495.
Laboratory tests al l ow the control and measurement of stresses, strains, and
porewaterpressures. As a resul t, they can often simul ate anticipated initial and
dynamic stressconditions better than fiel d tests. The resul ts of l aboratory tests,
however, may beinfl uenced by sampl e disturbance. Sampl e disturbance has a
particul arl y strong effecton l ow-strain properties.6. Several l aboratory tests
measure l ow-strain soil properties. Among these are the resonantcol umn test, the
ul trasonic pul se test, and the piezoel ectric bender el ement test.The cycl ic
triaxial test, cycl ic direct shear test, and cycl ic torsional shear test can beused
to measure dynamic soil properties at higher strain l evel s.7. Dynamic soil
properties may al so be inferred from the resul ts of model tests. Shakingtabl e tests
can accommodate rel ativel y l arge model s, but their inabil ity to producehigh
gravitational stresses can make extrapol ation to prototype conditions difficul t.
Centrifuge tests can satisfy simil itude requirements much better than shaking tabl e
tests but must be performed on rel ativel y smal l model s.8. Soil s exhibit nonl inear,
inel astic, stress-strain behavior under cycl ic l oading conditions.At l ow strain
l evel s, the stiffness of a soil is greatest and the damping is smal l est.At higher
strain l evel s, the effects of nonl inearity and inel asticity increase, producing
l ower stiffness and greater damping. Compl ete characterization of such behavior is
very compl icated. For the great majority of geotechnical earthquake engineering
anal yses,however, approximate characterization is sufficient. Three broad cl asses
ofstress-strain model s are used for geotechnical earthquake engineering anal yses:
equival ent l inear model s, cycl ic nonl inear model s, and advanced constitutive
model s.9. Equival ent l inear model s treat soil s as a l inear viscoel astic material s.
Nonl inear behavioris accounted for by the use of strain-dependent stiffness and
damping parameters.The stiffness of the soil is usual l y characterized by the
maximum shear modul us,which is mobil ized at l ow strains, and a modul us reduction
curve, which shows howthe shear modul us decreases at l arger strains. Damping
behavior is characterized bythe damping ratio, which increases with increasing
strain ampl itude. The shapes of themodul us reduction and damping curves are
infl uenced by soil pl asticity and, for soil sof very l ow pl asticity, by effective
confining pressure.10. Cycl ic nonl inear model s represent the nonl inear, inel astic
behavior of soil s using anonl inear backbone curve and a series of rul es that govern
unl oading-rel oadingbehavior. Backbone curves are usual l y described by simpl e
functions that refl ect thetransition from the initial stiffness (at l ow strain
l evel s) to the ul timate strength (athigh strain l evel s). The unl oading-rel oading
rul es control the behavior of the model during stress reversal s and ensure that it
behaves in a manner simil ar to that exhibitedby actual soil s subjected to irregul ar
cycl ic l oading. In contrast to equival ent l inearmodel s, cycl ic nonl inear model s
al l ow permanent strains to devel op. Cycl ic nonl inearmodel s can al so be coupl ed with
pore pressure generation model s to predict changesin effective stress during cycl ic
l oading. Model ing such behavior requires that theoriginal backbone curve be
degraded (softened) as pore pressures increase.250 Dynamic Soil Properties Chap. 6
11. Advanced constitutive model s use basic principl es of mechanics to describe soil
behavior for general initial stress conditions, a wide variety of stress paths with
rotatingprincipal stresses, cycl ic or monotonic l oading, high or l ow strain rates,
anddrained or undrained conditions. As such, they are much more general than
equival entl inear or cycl ic nonl inear model s. The penal ty for this increased
general ity comes inthe form of increased compl exity, an increased number of model
parameters (some ofwhich can be difficul t to determine), and increased
computational effort when incorporatedinto ground response or soil -structure
interaction anal yses.12. The shear strength of soil can be infl uenced by cycl ic
l oading. The l evel of permanentstrain that devel ops in a cycl ical l y l oaded el ement
of soil depends on the rel ationshipbetween the average (static) shear stress and
the cycl ic shear stress. If the averageshear stress is zero, onl y cycl ic strains
wil l devel op and fail ure is defined in terms ofa l imiting cycl ic strain l evel . If
the average shear stress is greater than zero, cycl icstresses can produce
unidirectional as wel l as cycl ic shear strains. The rate at whichunidirectional
strain devel ops is infl uenced by the rel ative magnitudes of the averageand cycl ic
shear stresses. The ul timate strength of a soil l oaded monotonical l y after an
episode of cycl ic l oading is al so important. Because the undrained of a saturated
soil is control l ed by its density and structure, the postearthquake undrained
strength wil l (barring structural changes) be essential l y the same as the static
undrained strength,even if excess pore pressures have devel oped during cycl ic
l oading.HOMEWORK PROBLEMS6.1 At a l evel site, bedrock is overl ain by a l ayer of
overconsol idated cl ay of variabl e thickness. Aseismic refraction survey is
conducted with 13 receivers pl aced on a straight l ine between twoshot points
l ocated 1,000 ft apart. From the p-wave arrival times l isted bel ow, determine and
pl ot the subsurface profil e in the central 800 ft portion of the survey. Determine
the p-wavevel ocities of the overconsol idated cl ay and the underl ying bedrock.
Distance Arrival time Distance Arrival timeReceiver fromSP 1 fromSP I (msec) fromSP
2 fromSP 2 (msec)A 0 0 1000 124B 50 15 950 119C 100 30 900 114D 200 48 800 105E 300
58 700 96F 400 67 600 85G 500 78 500 75H 600 91 400 73I 700 108 300 68J 800 113 200
47K 900 115 100 30L 950 119 50 15M 1000 124 0 0Ol ap.6 Homework Probl ems 2511;.2 The
figure bel ow shows the vertical response of geophones l ocated at the same depths in
vertical borehol es spaced 5 m apart. The waves recorded at the geophones resul ted
from downward(sol id l ines) and upward (dashed l ines) impacts on a mechanical source
at the same depth in athird borehol e that was col inear with the other two
borehol es. Determine the average SV-wavevel ocity of the soil between the geophones.
ReceiverR1ReceiverR2o 10 20 30 40Time (msec)FigureP6.250Upward hitDownward hit606.3
Estimate the damping ratio of the resonant col umn specimen from which the frequency
response curve of Figure E6.7 was obtained.6.4 Portions of the time histories of
deviator stress and axial strain from a stress-control l ed cycl ictriaxial test are
shown bel ow. Compute the secant shear modul us and damping ratio of the test
specimen. Assuming that the soil is saturated and its response is consistent with
the Masingcriteria, estimate the maximum shear modul us of the soil .-100 -0.5100 0.5
ctl l l .-" »xeenn ![~0 0 ~1ii iil ~ S·.~ ~00Figure P6.4252 Dynamic Soil Properties
Chap. 66.5 A site is underl ain by about 14 m of hydraul ic fil l sand with the
fol l owing properties.Grain size characteristics: Dl O < 0.074 mmD30=0.14mmDso= 0.17
mmD60 = 0.20 mmSaturated unit weight =18.8 kN/m3Average uncorrected SPT resistance
= 6 bl ows/ftAverage cone tip resistance =3 MPaThe groundwater tabl e is l ocated at a
depth of 2m.(a) Estimate the shear wave vel ocity of the hydraul ic fil l at a depth
of 5 m by as many proceduresas the avail abl e data wil l support.(b) Shear wave
vel ocity measurements at this site have indicated hydraul ic fil l shear wave
vel ocities ranging from about 120 m/sec - 200 m/sec at 5 m depth with an average
val ue ofabout 170 m/sec. Comment on the l evel of agreement between the estimated
shear wavevel ocities from Part (a) and the measured shear wave vel ocities.6.6 At
the site described in Probl em 6.5, the hydraul ic fil l is underl ain by a 15 m thick
deposit ofmedium stiff, normal l y consol idated sil ty cl ay with the fol l owing
properties: Water content = 40%Liquid l imit = 46Pl astic l imit = 23Saturated unit
weight =15.9 kN/m3Cone tip resistance =8 to 14 kg/crrr'Undrained shear strength =
27 to 61 kPa(a) Estimate the shear wave vel ocity of the sil ty cl ay at a depth of 20
m by as many proceduresas the avail abl e data wil l support.(b) Shear wave vel ocity
measurements at this site have indicated sil ty cl ay shear wave vel ocitiesranging
from about 120 m/sec - 180 m/sec at 20 m depth with an average val ue ofabout 155
m/sec. Comment on the l evel of agreement between the estimated shear wavevel ocities
from Part (a) and the measured shear wave vel ocities.6.7 Determine and pl ot the
backbone curve that woul d correspond to the Vucetic-Dobry modul usreduction curve
for a cl ay with PI =15.6.8 Compute and pl ot the modul us reduction curve for a sil ty
cl ay (PI = 30) at a mean effective confiningpressure of 40 kPa using the
rel ationship ofIshibashi and Zhang (1993). How does thiscurve compare with that
shown in Figure 6.42?Chap. 6 Homework Probl ems 2536,,9 Compute and pl ot the damping
ratio curves that correspond to the modul us reduction curvesshown in Figure 6.43.
Comment on the infl uence of effective confining pressure on dampingcharacteristics
of l ow and high
pl asticity soil s.6.10 An el ement of soil with a maximum shear modul us of700,000
psf and an undrained strength of1000 psf is subjected to the time history of shear
stress shown bel ow. Assuming that the backbonecurve is hyperbol ic, and that the
soil fol l ows the extended Masing criteria, pl ot the resul tingstress-strain
behavior. Mark al l points at which the unl oading-rel oading rul es governing the
stress-strain behavior change (i.e. mark as 1-2 at the point where control over the
stress-strainresponse shifts from Rul e 1 to Rul e 2).1: - psf1000o-1000TimeFigure
P6.107Ground Response Anal ysis7.1 INTRODUCTIONOne of the most important and most
commonl y encountered probl ems in geotechnical earthquakeengineering is the
eval uation of ground response. Ground response anal yses are usedto predict ground
surface motions for devel opment of design response spectra, to eval uatedynamic
stresses and strains for eval uation of l iquefaction hazards, and to determine the
earthquake-induced forces that can l ead to instabil ity of earth and earth-retaining
structures.Under ideal conditions, a compl ete ground response anal ysis woul d model
the rupturemechanism at the source of an earthquake, the propagation of stress
waves through the earthto the top of bedrock beneath a particul ar site, and woul d
then determine how the groundsurface motion is infl uenced by the soil s that l ie
above the bedrock. In real ity, the mechanismof faul t rupture is so compl icated and
the nature of energy transmission between thesource and the site so uncertain that
this approach is not practical for common engineeringappl ications. In practice,
empirical methods based on the characteristics of recorded earthquakesare used to
devel op predictive rel ationships of the types discussed in Chapter 3.These
predictive rel ationships are often used in conjunction with a seismic hazard
anal ysisto predict bedrock motion characteristics at the site. The probl em of
ground response anal ysisthen becomes one of determining the response of the soil
deposit to the motion of thebedrock immediatel y beneath it. Despite the fact that
seismic waves may travel through tens254Sec. 7.2 One-Dimensional Ground
ResponseAnal ysis 255of kil ometers of rock and often l ess than 100 m of soil , the
soil pl ays a very important rol ein determining the characteristics of the ground
surface motion.The infl uence of l ocal soil conditions on the nature of earthquake
damage has beenrecognized for many years. Since the 1920s, seismol ogists and, more
recentl y, geotechnical earthquake engineers have worked toward the devel opment of
quantitative methods for predictingthe infl uence of l ocal soil conditions on strong
ground motion. Over the years, anumber of techniques have been devel oped for ground
response anal ysis. The techniquesare often grouped according to the dimensional ity
of the probl ems they can address,al though many of the two- and three-dimensional
techniques are rel ativel y straightforwardextensions of corresponding one-
dimensional techniques. This chapter describes the mostcommonl y used methods for
one-, two-, and three-dimensional ground response, and introducesthe probl em of
soil -structure interaction.7.2 ONI: -DIMENSIONAL GROUND RESPONSE ANALYSISWhen a
faul t ruptures bel ow the earth's surface, body waves travel away from the source in
al l directions. As they reach boundaries between different geol ogic material s, they
arerefl ected and refracted. Since the wave propagation vel ocities of shal l ower
material s aregeneral l y l ower than the material s beneath them, incl ined rays that
strike horizontal l ayerboundaries are usual l y refl ected to a more vertical
direction. By the time the rays reach theground surface, mul tipl e refractions have
often bent them to a nearl y vertical direction (Figure7.1). One-dimensional ground
response anal yses are based on the assumption that al l boundaries are horizontal and
that the response of a soil deposit is predominantl y caused bySH-waves propagating
vertical l y from the underl ying bedrock. For one-dimensional ground response
anal ysis, the soil and bedrock surface are assumed to extend infinitel y inthe
horizontal direction. Procedures based on this assumption have been shown to
predictground response that is in reasonabl e agreement with measured response in
many cases.l ,fl SiteFigure 7.1 Refraction process that produces uearl y vertical
wave propagation near theground surface.Before describing any of the ground
response model s, it is necessary to define several terms that are commonl y used to
describe ground motions. With reference to Figure 7.231,the motion at the surface
of a soil deposit is the free surface motion. The motion at the baseof the soil
deposit (al so the top of bedrock) is cal l ed a bedrock motion. The motion at a
l ocationwhere bedrock is exposed at the ground surface is cal l ed a rock outcropping
motion. Ifthe soil deposit was not present (Figure 7.2b), the motion at the top of
bedrock woul d be thebedrock outcropping motion.256 Ground Response Anal ysis Chap. 7
Free surface motion~=l f~ -\-~Rockoutcroppingmotion-~c- -\--Rockoutcroppingmotion
Bedrockmotion(a)Bedrockoutcroppingmotion(b)Figure 7.2 Ground response nomencl ature:
(a) soil overl ying bedrock; (b) no soil overl ying bedrock. Vertical scal e is
exaggerated.7.2.1 Linear ApproachThe manner in which transfer functions can be used
to compute the response of singl edegree-of-freedom systems is il l ustrated in
Appendix B (Section B.5.4.2). An importantcl ass of techniques for ground response
anal ysis is al so based on the use of transfer functions.For the ground response
probl em, transfer functions can be used to express variousresponse parameters, such
as displ acement, vel ocity, accel eration, shear stress, and shearstrain, to an input
motion parameter such as bedrock accel eration. Because it rel ies on theprincipl e of
superposition, this approach is l imited to the anal ysis of l inear systems.
Nonl inearbehavior can be approximated, however, using an iterative procedure with
equival entl inear soil properties.The mathematical aspects of the transfer function
approach were described in SectionB.5.4.2 of Appendix B. Al though the cal cul ation
invol ve manipul ation of compl ex numbers,the approach itsel f is quite simpl e. A
known time history of bedrock (input) motion isrepresented as a Fourier series,
usual l y using the FFT (Section A.3.4). Each term in the Fourierseries of the
bedrock (input) motion is then mul tipl ied by the transfer function to producethe
Fourier series of the ground surface (output) motion. The ground surface (output)
motion can then be expressed in the time domain using the inverse FFT. Thus the
transferfunction determines how each frequency in the bedrock (input) motion is
ampl ified, ordeampl ified, by the soil deposit.7.2.1.1 Eval uation of Transfer
FunctionsThe key to the l inear approach is the eval uation of transfer functions. In
the fol l owingsections, transfer functions are derived for a series of successivel y
more compl icated geotechnical conditions. Al though the simpl est of these may onl y
rarel y be appl icabl e to actual probl ems, they il l ustrate some of the important
effects of soil deposits on ground motioncharacteristics without undue mathematical
compl exity. The more compl ex are capabl e ofdescribing the most important aspects of
ground response and are very commonl y used ingeotechnical earthquake engineering
practice.Sec. 7.2 One-Dimensional Ground Response Anal ysis 257(7.2)Uniform Undamped
Soil on Rigid Rock. First, consider a uniform l ayerof isotropic, l inear el astic
soil overl ying rigid bedrock as shown in Figure 7.3. Harmonichorizontal motion of
the bedrock wil l produce vertical l y propagating shear waves in theoverl ying soil .
The resul ting horizontal displ acement can be expressed, using the resul ts ofSection
5.2.1.3, asU(z, t) = Aei(ffi/+kz) + Bei(ffi/-kz) (7.1)where 00 is the circul ar
frequency of ground shaking, k the wave number (= oo/vs) and A andB the ampl itudes
of waves travel ing in the -z (upward) and +z (downward) directions,respectivel y. At
the free surface (z = 0), the shear stress, and consequentl y the shear strain,must
vanish; that is,'teo, t) = Gy(O, t) = Gau~~, t) = °Substituting (7.1) into (7.2)
and differentiating yiel dsGik(Aeik(O) - Be-ik(O))eiffi/ = Gik(A - B)eiffi/ = °
(7.3)which is satisfied (nontrivial l y) when A = B. The displ acement can then be
expressed aseikz+ e- ikz .u(z, t) = 2A e' ffi / = 2A cos kz eiffi/2(7.4)which
describes a standing wave of ampl itude 2Acos kz. The standing wave is produced by
the constructive interference of the upward and downward travel ing waves and has a
fixedshape with respect to depth. Equation (7.4) can be used to define a transfer
function thatdescribes the ratio of displ acement ampl itudes at any two points in
the soil l ayer. Choosingthese two points to be the top and bottom of the soil l ayer
gives the transfer functionF ( ) _ Umax(0, t) _ 2Aeiffi/ 1 1I 00 - umax(H, t) - 2A
cos kHeiffi/ = cos kH = cos (ooH/vs ) (7.5)The modul us of the transfer function is
the ampl ification functionIF I (00)[ = J{Re [FI (oo)]}2 + {1m [FI (oo)]}2 = I
(1 / ) I (7.6)cos ooH Vswhich indicates that the surface displ acement is al ways at
l east as l arge as the bedrock displ acement(since the denominator can never be
greater than 1) and, at certain frequencies, is1u rz tAei(wt+kZ)Figure 7.3 Linear
el astic soil deposit of thickness H underl ain by rigid bedrock.258 Ground Response
Anal ysis Chap. 7much l arger. Thus IFl (w)1 is the ratio of the free surface motion
ampl itude to the bedrockmotion ampl itude (or, since the bedrock is rigid in this
case, the bedrock outcropping motion).As wHl vs approaches nl 2 + nn; the denominator
of equation (7.6) approaches zero, whichimpl ies that infinite ampl ification, or
resonance, wil l occur (Figure 7.4). Even this
very simpl emodel il l ustrates that the response of a soil deposit is highl y
dependent upon the frequencyof the base motion, and that the frequencies at which
strong ampl ification occurs depend onthe geometry (thickness) and material
properties (s-wave vel ocity) of the soil l ayer.0n 3n 5n 7n 9n 2 2 2 2 2 kH0nv s
3nvs 5nvs 7nvs 9nvs (j) 2H 2H 2H 2H 2HFigure 7.4 Infl uence of frequency on steady-
state response of undamped l inearel asticl ayer.Exampl e 7.1Compute the time history
of accel eration at the surface of the l inear el astic soil deposit shown inFigure
E7.l a in response to the E-W component of the Gil roy No. I (rock) motion (Figure
3.1).I Vs = 1050 Ib/ft 210 It Y = 110 Ib/ft3';=0Figure E7.1af '. ~~ i ~~! ~~
Sol ution Computation of the ground surface motion from the bedrock motion can be
accompl ishedin the fol l owing series of steps: 1. Obtain the time history of
accel eration of the input motion. In this case the input motionis the E-W component
of the Gil roy No. I (rock) motion shown in FigureE7.l b. The Gil royNo. I record
consists of 2000 accel eration val ues at 0.02-sec interval s.2. Compute the Fourier
series of the bedrock (input) motion. The Fourier series is compl exval ued; its one-
sided Fourier ampl itude spectrum (Section A.3 of Appendix A) is shownin Figure
E7.l c. The Fourier ampl itude spectrum is defined for frequencies up to 1/2~t =25
Hz, but most of the energy in the bedrock motion is at frequencies l ess than 5 to
10Hz.3. Compute the transfer function that rel ates the ground surface (output)
motion to the bedrock(input) motion. From equation (7.5), the transfer function
(Figure E7.1d) for the caseof undamped soil is real .val ucdThcjransfcr function has
val ues of I bel ow frequencies.'.: ! - "Sec. 7.2 One-Dimensional Ground Response
Anal ysis 259of about 10Hz. However, at frequencies that approach the fundamental
frequency of thesoil deposit (fo =Vs /4H=26.25 Hz), the transfer function begins to
take on l arge val ues.4. Compute the Fourier series of the ground surface (output)
motion as the product of thetransfer function and the Fourier series of the bedrock
(input) motion. At frequencies l essthan 5 to 10Hz, the Fourier spectrum of the
ground surface motion is virtual l y the sameas that of the bedrock motion. Al though
the transfer function indicates that frequenciesabove 20 Hz or so wil l be ampl ified
strongl y, the input motion is weak in that frequencyrange. The one-sided ampl itude
spectrum is shown in Figure E7.1e. Examination of thisFourier ampl itude spectrum
indicates that the ground surface motion has somewhat morehigh frequency motion,
but is general l y simil ar to that of the bedrock motion.5. Obtain the time history
of the ground surface motion by inverting its Fourier series. Asil l ustrated in
Figure E7.1f, the time history of ground surface motion has a somewhatgreater
content of high-frequency components, but is general l y simil ar to the time history
of bedrock motion.: ] o 5 10 15 20 25 30 35 40Time (sec)20 25Frequency (Hz)5 10 15
(d)20 25Frequency (Hz)(.)jl !~~o 5 10 15 20 25Frequency (Hz): ' : (I)o 5 10 15 20 25
30 35 40Time (sec)F'260 Ground Response Anal ysis Chap. 7(7.9)Because the soil l ayer
had a high natural frequency (intentional l y sel ected to avoid theextremel y l arge
transfer function val ues that exist near the natural frequency of an undamped
l ayer), its effect on the ground surface motion was rel ativel y smal l .Uniform,
Damped Soil on Rigid Rock. Obviousl y, the type of unboundedampl ification predicted
by the previous anal ysis cannot physical l y occur. The previous anal ysisassumed no
dissipation of energy, or damping, in the soil . Since damping is present in al l
material s, more real istic resul ts can be obtained by repeating the anal ysis with
damping.Assuming the soil to have the shearing characteristics of a Ke1vin-Voigt
sol id, the waveequation can be written [equation (5.94)] as(J2 u (J2u (J3 up (Jt2 =
G(JZ2 + 11 (JZ2(Jt (7.7)As shown in equation (5.94), the sol ution to this wave
equation is of the formu(z, t) = Aei(rot+k*z) + Bei(rot-k*z)where k* is a compl ex
wave number with real part kr and imaginary part k2. Repeating theprevious
al gebraic manipul ations with the compl ex wave number, the transfer function forthe
case of damped soil over rigid rock can be expressed as1 1F2 (w) = cos k*H cos (wH/
v~*) (7.8)Since the frequency-independent compl ex shear modul us (Section 5.5.1) is
given by G* =G(l + i2~), the compl ex shear wave vel ocity can be expressed as* - ~*
- JG(l + i2~) - ~(I .~) - (1 .~) v - - - - - +1 - V +1s P P P sfor smal l ~. Then
the compl ex wave number can be written, again for smal l ~, as(7.10)(7.12)and
final l y, the transfer function, asF ( ) 1 1 (7.11)2 ro = cosk(I- i~)H = cos
[wH/vs(l + i~)]Using the identity Icos(x + iy)1 = ,Jcos2x + sinh 2y, the
ampl ification function can beexpressed as!F2(w)1 = 1,Jcos 2kH + sinh? ~kHSince
sinh2y "" l for smal l y, the ampl ification function can be simpl ified to!F2(w)1 ""
1 = 1 (7.13),Jcos2 kH + (~kH)2 Jcos 2 (wH/vs) + [~(wH/vs)]2For smal l damping
ratios, equation (7.13) indicates that ampl ification by a damped soil l ayer al so
varies with frequency. The ampl ification wil l reach a l ocal maximum wheneverkl l »
TtI2 + nt: but wil l never reach a val ue of infinity since (for ~ > 0) the
denominator wil l Sec. 7.2 One-Dimensional Ground Response Anal ysis 261al ways be
greater than zero. The frequencies that correspond to the l ocal maxirna are the
natural frequencies of the soil deposit. The variation of ampl ification factor with
frequencyis shown for different l evel s ofdamping in Figure 7.5. This ampl ification
factor is al so equal to the ratio of the free surface motion ampl itude to the
bedrock (or bedrock outcropping)motion ampl itude. Comparing Figures 7.4 and 7.5
shows that damping affects the responseat high frequencies more than at l ower
frequencies.12<5 10t5~ 8 S=5% c0~6o~Q. 4E<t: 220kHFigure 7.5 Infl uence of frequency
on steady-state response of damped, l inear el astic l ayer.n = 0, 1,2, ... ,00The nth
natural frequency of the soil deposit is given byVs(1t ) con"" H 2+ nt: (7.14)Since
the peak ampl ification factor decreases with increasing natural frequency, the
greatestampl ification factor wil l occur approximatel y at the l owest natural
frequency, al so knownas. the fundamental frequency(7.15)The period of vibration
corresponding to the fundamental frequency is cal l ed the characteristicsite period,
(7.16)The characteristic site period, which depends onl y on the thickness and shear
wave vel ocityof the soil , provides a very useful indication of the period of
vibration at which the most significantampl ification can be expected.At each
natural frequency, a standing wave devel ops in the soil . Normal izeddeformed shapes,
or mode shapes, for the first three natural frequencies are shown in Figure7.. 6.
Note that the soil displ acements are in phase at al l depths in the fundamental
mode, butnot in the higher modes. At frequencies above the fundamental frequency,
part of the soil deposit may be moving in one direction whil e another part is moving
in the opposite direction.This phenomenon must be considered in the eval uation of
inertial forces in soil massesrequired for seismic stabil ity anal yses (Chapter 10).
262 Ground Response Anal ysis Chap. 71.0 "-- -----"'--- ---.J1.0Normal ized
displ acement-1.0 -0.5 0.0 0.50.0 r--------,------,----,-----,---,-----,z/H 0.5
GroundTT SUrface ....: ".",.,': ',"'"': .: : ""'-"."• ..• .• • ...• ....• • • ..• • • • • ...,Figure
7.6 Displ acement patterns for standing wavesat fundamental (n = 0), second(n =I)
and third (n = 2) natural frequencies for a soil l ayer with~ = 5%. Displ acements are
normal ized by maximum displ acement at the fundamental frequency.Exampl e 7.2The site
at which the Gil roy No.2 (soil ) earthquake motion shown in Figure 3.1 was recorded
is underl ain by some 540 ft of soil underl ain by shal e and serpentinite bedrock.
The shear wavevel ocity of the soil varies from about 1000 ft/sec at depths l ess
than about 130 ft to about 2000ft/sec at greater depths. Assuming an average shear
wave vel ocity of 1500 It/sec", an averageunit weight of 125Ib/ft3, a damping ratio
of 5%, and rigid bedrock (Figure E7.2a), compute theground surface motion that
woul d occur if the bedrock was subjected to the E-W component ofthe Gil roy No. 1
(rock) motion.540 ftvs = 1500 ft/secy = 125 Ib/ft3~ = 5%Rigid bedrock Figure E7.2a
Sol ution Computation of the ground surface motion from the bedrock motion can be
accompl ishedin the same five steps as described in Exampl e 7.1: 1. Obtain the time
history of accel eration of the input motion. This step is identical to step1 in
Exampl e 7.1; the motion is shown in Figure E7.2b.2. Compute the Fourier series of
the bedrock (input) motion. This step is al so identical tostep 2 in Exampl e 7.1;
the resul t is shown in Figure E7.2c.3. Compute the transfer function that rel ates
the ground surface (output) motion to the bedrock(input) motion. From equation
(7.11), the transfer function for the case of dampedsoil is compl ex val ued. The
modul us of the transfer function is shown in Figure E7.2d.The shape of the transfer
function indicates that significant ampl ification wil l occur atseveral natural
frequencies, and that higher frequencies (greater than about 10 Hz) wil l be
suppressed.Sec 7.2 One-Dimensional Ground Response Anal ysis 2634. Compute the
Fourier series of the ground surface (output) motion as the product of thetransfer
function and the Fourier series of the bedrock (input) motion. The Fourier spectrum
of the ground surface motion (Figure E7.2e) shows ampl ification at the natural
frequenciesof the soil deposit and l ittl e high-frequency motion.5. Obtain the time
history of the ground surface motion by inverting the Fourier series. Thetime
history of ground surface motion (Figure
E7.2f) indicates that peak accel erations atthe ground surface and bedrock l evel s
are simil ar, but the frequency contents are different.Because the ground surface
motion is weighted toward l ower frequencies, the peakvel ocity and displ acement at
the ground surface are l ikel y to be considerabl y greater thanat bedrock.The rigid
bedrock anal ysis predicts a peak ground surface accel eration of 0.452 g, whichis
considerabl y greater than the peak accel eration of 0.322g actual l y recorded at the
Gil roy No.2 (soil ) station.Uniform, Damped Soil on El astic Rock. The preceding two
sectionsdevel oped expressions for ampl ification factors for soil s overl ying rigid
bedrock. If the bedrockis rigid, its motion wil l be unaffected by motions in, or
even the presence of, the overl ying35 40Time (sec)20 25Frequency (Hz)20 25 30(b)
I§O: ~Il ""~"""""'_"""'''''': _''_'~ _ <t: -O.5~ -o 5 10 15(')Ii!: : ~<t: 0 . <'V",~o 5
10 1520 25Frequency (Hz)10 15(~ ~ : : C====='==~_.l ....- ~ 1o 55 10 15 20 25Frequency
(Hz)30 35 40Time (sec)(~E: ~---3o 5 10 15 20 25Figure E7.2b-f264 Ground Response
Anal ysis Chap. 7soil . It acts as a fixed end (Section 5.4.1) boundary. Any
downward-travel ing waves in the soil wil l be compl etel y refl ected back toward the
ground surface by the rigid l ayer, thereby trappingal l of the el astic wave energy
within the soil l ayer.If the rock is el astic, however, downward-travel ing stress
waves that reach the soil rockboundary wil l be refl ected onl y partial l y; part of
their energy wil l be transmittedthrough the boundary to continue travel ing downward
through the rock. If the rock extendsto great depth (l arge enough that waves
refl ected from any deeper material boundaries donot return to the soil -rock
boundary soon enough, or with sufficient ampl itude, to infl uencethe response of the
soil deposit), the el astic energy of these waves wil l effectivel y beremoved from
the soil l ayer. This is a form of radiation damping, and it causes the free surface
motion ampl itudes to be smal l er than those for the case of rigid bedrock.Consider
the case of a soil l ayer overl ying a hal fspace of el astic rock (Figure 7.7). Ifthe
subscripts sand r refer to soil and rock, respectivel y, the displ acements due to
vertical l ypropagating s-waves in each material can be written as(7.17a)Ur(z" t) =
Arei(WI+k;z,) + Brei (wl -k;z,) (7.17b)The free surface effect, as before, requires
that As=Bs,and compatibil ity of displ acementsand continuity of stresses at the
soil -rock boundary require thatUs(zs = H)'tsCzs = H) = 'tr(zr = 0)Substituting
equations (7.17) into equation (7,18) yiel dsA (eik;H + e-ik;H) = A + B s . r rFrom
equation (7.19) and the definition of shear stress (r = G dU/dz)AsiGsk;(eik;H - e-
ik;H) = iGrk;(A r- Br)(7.18)(7.19)(7.20)(7.21)Soil Rock GrPrHFigure 7.7 Nomencl ature
for the case of a soil l ayer overl ying a hal f-space of el astic rock.Sec. 7.2 One-
Dimensional Ground Response Anal ysis 265or(7.22)The ratio(7.23)where v;s and v;r
are the compl ex shear wave vel ocities ofthe soil and rock, respectivel y,and a; is
the compl ex impedance ratio (see Section 5.4.1). Sol ving equations (7.20) and(7.22)
simul taneousl y gives(7.24a)s, = ~As[(1- a;)eik;H + (1 + a;)e-ik;H] (7.24b)Suppose
that a vertical l y propagating shear wave of ampl itude, A, travel ed upwardthrough
the rock. If the soil was not present, the free surface effect at the rock outcrop
woul dproduce a bedrock outcropping motion of ampl itude 2A. With the soil present,
the free surfacemotion ampl itude woul d be4A2A s =(1 + a;)eik;H + (1 - a;)e-ik;H
Defining the transfer function, F3, as the ratio of the soil surface ampl itude to
the rock outcropampl itude,F3 (m) = 2(1 + a;)eik;H + (1 - a;)e-ik;Hwhich, using
Eul er's l aw, can be rewritten as(7.25)(7.26) 1cos (mH/v;s) + i a; sin(mH/v;s)1F3
( co) = * *. * = ----,...---.,....-----cos ksH + i a z sm ksHThe modul us of F3(m)
cannot be expressed in a very compact form when soil dampingexists. To il l ustrate
the important effect of bedrock el asticity, however, the ampl ificationfactor for
undamped soil can be expressed as1F3(m,~ =0)1 = 1 . (7.27)Jcos2ksH + a~ sin2ksHNote
that resonance cannot occur (the denominator is al ways greater than zero, even when
the soil is undamped). The effect of the bedrock stiffness, as refl ected by the
impedance ratio,on ampl ification behavior is il l ustrated in Figure 7.8. Note the
simil arity between the effectsof soil damping and bedrock el asticity by comparing
the shapes of the ampl ification factorcurves in Figure 7.8 and those in Figure 7.5.
The el asticity of the rock affects ampl ification266 Ground Response Anal ysis Chap.
78,---------r--r---------------,,---,----------,62Impedance ratio = 0.02 3 4 5 6 7
O'--------'------"------'---------'-----..!-------.L--------'okHFigure 7.8 Effectof
impedance ratioon ampl ification factorfor case of undamped soil .simil arl y to the
damping ratio of the soil -both prevent the denominator from reaching zero.This
radiation damping effect has significant practical importance, particul arl y in the
easternUnited States, where bedrock is substantial l y harder than that typical l y
found in the westernstates. The stiffer bedrock means that greater ampl ification
may occur in the east and thatdesign criteria establ ished on the basis of empirical
evidence from western earthquakes maybe somewhat unconservative in the east.Exampl e
7.3Repeat Exampl e 7.2 assuming that the bedrock is not rigid. Assume a shear wave
vel ocity of5000 ft/sec, a unit weight of 160 l b/ft ', and 2% damping for bedrock at
the Gil roy No.2 site(Figure E7.3a).540 ftSoil vs = 1500 It/secy =125 Ib/tt3~ = 5%
Rockvs = 5000 tt/secy = 160 Ib/tt3~ = 2% Figure E7.3aSol ution Computation ofthe
ground surface motion from the bedrock motion can be accompl ishedin the same five
steps described in Exampl e 7.2. The onl y difference is that the transferfunction in
this exampl e wil l incl ude the effects of bedrock compl iance.1. Obtain the time
history of accel eration of the input motion. This step is identical to stepI in
Exampl es 7.1 and 7.2; the motion is shown in Figure E7.3b.2. Compute the Fourier
series of the bedrock (input) motion. Again, this step is identical tostep 2 in
Exampl es 7.1 and 7.2; the resul t is shown in Figure E7.3c.3. Compute the transfer
function that rel ates the ground surface (output) motion to the bedrock(input)
motion. From Equation 7.25, the transfer function for the case of damped267: 130 35
40Time (sec)j20 25Frequency (Hz)j20 25Frequency (Hz)j20 25Frequency (Hz)2515152010
15.: .o 5One-Dimensional Ground Response Anal ysisg 0.5BE: : : ~ § 0 ....,......--~--
~--------------« -0.5 ~--'--------'-----'-------'----~--o 5 10(d) ~ ': = o(b)1Sec.
7.235 40Time (sec)20 25 30~0'5~CD Oi 0 ~----------------] ~-0.5 __--'--- .L-
__-----''--__----'- _o 5 10 15(f)l Figure E7.3b-fsoil is compl ex val ued. The modul us
of the transfer function, shown in Figure E7.3d,indicates that l ess ampl ification
wil l occur than in the case of rigid bedrock.4. Compute the Fourier series of the
ground surface (output) motion as the product of thetransfer function and the
Fourier series of the bedrock (input) motion. The Fourier seriesofthe ground
surface motion (Figure E7.3e) shows l ess ampl ification than in the case ofrigid
bedrock.5. Obtain the time history of the ground surface motion by inverting the
Fourier series. Thetime history of ground surface motion (Figure E7.3f) indicates
that the peak accel erationsat the ground surface is l ower than the peak
accel eration at the bedrock l evel ; thefrequency contents are al so different.The
compl iant bedrock anal ysis predicts a peak ground surface accel eration of O.339g,
which agrees wel l with the peak accel eration ofO.322g recorded at the Gil roy No.2
(soil ) station.The good agreement between peak accel erations, however, does not
mean that this simpl eanal ysis has predicted al l aspects of the Gil roy No.2 (soil )
motion. Comparison of the Fourierampl itude spectrum of the predicted motion (Figure
E7.3e) with that of the recorded motion(Figure 3.13b) shows significant differences
in frequency content.268 Ground Response Anal ysis Chap. 7(7.29)Layered, Damped Soil
on El astic Rock. Whil e the uniform el astic l ayermodel s are useful for il l ustration
of the infl uence of soil conditions on several groundmotion characteristics, they
are sel dom suitabl e for anal ysis of practical ground responseprobl ems. Real ground
response probl ems usual l y invol ve soil deposits with l ayers of differentstiffness
and damping characteristics with boundaries at which el astic wave energywil l be
refl ected and/or transmitted. Such conditions require the devel opment of transfer
functions for l ayered soil deposits.Consider a soil deposit consisting of N
horizontal l ayers where the Nth l ayer is bedrock(Figure 7.9). Assuming that each
l ayer of soil behaves as a Kel vin-Voigt sol id, thewave equation is of the form
given in equation (5.94). The sol ution to the wave equation canbe expressed in the
formu(z, t) = Aei(Ol t+k"Z)+ Bei(Ol t-k'"Z) (7.28)where A and B represent the
ampl itudes of waves travel ing in the -z (upward) and +z (downward)directions,
respectivel y. The shear stress is then given by the product of the compl exshear
modul us, G*, and the shear strain, so*au au au't(z, t) = G az = (G + ian!) az = G(l
+ 2i~) azIntroducing a l ocal coordinate system, Z, for each l ayer, the displ acement
at the topand bottom of l ayer m wil l be(7.30a)um(Zm =n.; t) = (Ameik': 'h m + Bme-
ik;~hm) eioot (7.30b)Displ acements at l ayer boundaries must be compatibl e (i.e.,
the displ acement at the top ofa particul ar l ayer must be equal to the displ acement
at the bottom of the overl ying l ayer).Appl ying the compatibil ity requirement to the
boundary between l ayer m and l ayer m + 1,that is,U1Layer 1 • • G1 ~1 Z, U2 P1 !h
1Z2 t'-m f~ Gm~m Pm !Zm hmm+ 1 Zm+1' ~m+' Gm+1 ~m+1 Pm+1 ~ hm+1 .-Zm+2. Um+ 2N f~N
GN~NPN tzN hN = =Figure 7.9 Nomencl ature for l ayered soil deposit on el astic
bedrock.Sec. 7.2yiel dsOne-Dimensional Ground Response Anal ysis 269(7.31)The shear
stresses at the top and bottom of l ayer m are= ik * G* (A eik~,hm - B e-ik~,hm) e
iffi{m m m m(7.32a)(7.32b)Since stresses must be continuous at l ayer boundaries,
'tm(Zm=hm,t) = 'tm+l (Zm+1 =O,t)so(7.33)Adding (7.31) and (7.33) and subtracting
(7.33) from (7.31) gives the recursion formul asBm+.1=!A (l _c/)eik~,hm+!B (l +(J.: )e-
ik,~hm2 m m 2m m(7.34a)(7.34b)where ex: is the compl ex impedance ratio at the
boundary between l ayers m and m + 1: (7.35)At the ground surface, the shear stress
must be equal to zero, which requires [fromequation (7.32a)] that Al =B I . If the
recursion formul as of equation (7.34) are appl iedrepeatedl y for al l l ayers from 1
to m, functions rel ating the ampl itudes inl ayer m to those inl ayer 1 can be
expressed byAm = am(m)A,s; = bm(m)B I(7.36a)(7.36): »The transfer function rel ating
the displ acement ampl itude at l ayer i to that at l ayer j is given byp..(m) = ~ =
ai(m) + biCm) (7.37)I) IUjl a/m) + bj(m)Because l ui = ml zil = m21uI for harmonic
motion, equation (7.37) al so describes theampl ification of accel erations and
vel ocities from l ayer i to l ayer j. Equation (7.37) indicatesthat the motion in any
l ayer can be determined from the motion in any other l ayer.Hence if the motion at
anyone point in the soil profil e is known, the motion at any otherpoint can be
contributed. This resul t al l ows a very useful operation cal l ed deconvol ution
(Section 7.2.1.4) to be performed.Exampl e 7.4As part of a comprehensive
investigation of ground motion estimation techniques, the El ectricPower Research
Institute performed a detail ed subsurface investigation at the site of the Gil roy
270 Ground Response Anal ysis Chap. 7NO.2 (soil ) recording station (EPRI, 1993). A
rough approximation to the measured shear wavevel ocity profil e is l isted bel ow.
Depth Range (ft)0-2020-454S-7070-13013O-S40>S40Average ShearWave Vel ocity (ft/sec)
SOO700ISOO10002000SOOOAssuming, as in Exampl es 7.2 and 7.3, an average soil unit
weight of 125 l b/ft 3 and 5% soil damping, compute the expected ground surface
response when the bedrock is subjected to theGil roy No.1 (rock) motion.Sol ution As
in the previous exampl es of this chapter, this probl em requires eval uation of the
transfer function that rel ates the ground surface motion to the bedrock motion.
Because of mul tipl erefl ections within the l ayered system, the transferfunction
[equation (7.36)] for this exampl eis considerabl y more compl icated than for the
singl e-l ayered cases of the previousexampl es. Whil e the transfer function can be
eval uated by hand, it has al so been coded in thecomputer program SHAKE (Schnabel et
al ., 1972). SHAKE was used, with constant soil stiffnessand damping ratio, to
obtain the transfer function shown in Figure E7Ac. As in the previousexampl es, the
Fourier series of the ground surface motion (Figure E7Ad) was computed asthe
product of the transfer function and the Fourier series of the bedrock motion.
Inversion of thisFourier series produces the time history of ground surface
accel eration (Figure E7Ae).Examination of Figure E7 Ac shows that the transfer
function for the l ayered system isindeed more compl icated than the transfer
functions for the singl e-l ayered cases of Exampl es 7.1,7.2, and 7.3. The spikes in
the transfer function at frequencies of about 3.5 and 5.5 Hz hel p producea peak
accel eration ofOA99g that is considerabl y l arger than the peak accel eration
ofO.322gthat was actual l y recorded at the Gil roy No.2 (soil ) station. Differences
in the frequency contentsof the predicted and recorded motions can be seen by
comparing Figures E7Ad and 3.13b.7.2.1.2 Equival ent Linear Approximationof
Nonl inear ResponseSince the nonl inearity of soil behavior is wel l known, the l inear
approach must bemodified to provide reasonabl e estimates of ground response for
practical probl ems ofinterest. As discussed in Chapter 6, the actual nonl inear
hysteretic stress-strain behavior ofcycl ical l y l oaded soil s can be approximated by
equival ent l inear soil properties. The equival entl inear shear modul us, G, is
general l y taken as a secant shear modul us and the equival entl inear damping ratio,
~, as the damping ratio that produces the same energy l oss in asingl e cycl e as the
actual hysteresis l oop. The strain-dependent nature of these equival entl inear
properties was described in Section 6.4.2.Since the l inear approach requires that G
and ~ be constant for each soil l ayer, theprobl em becomes one of determining the
val ues that are consistent with the l evel of straininduced in each l ayer. To
address this probl em, an objective definition of strain l evel isneeded. The
l aboratory tests from which modul us reduction and damping ratio curves (e.g.,those
shown in Figures 6.47 and 6.50) have been devel oped used simpl e harmonic l oading
Sec. 7.2 One-Dimensional Ground Response Anal ysis 27180 90Time (sec)20 25Frequency
(Hz)20 30 40 50 60 70.1l ~5OC: ==== ·~2 s(c: ) ~l ~ 0o 5 10 15(bl j§O: ~"« -0.5
DC__.L...----'-----'---~--~----'---~---~--o 10J! II10 15 20 25Frequency (Hz)10 15
20 25Frequency (Hz)(n E: ~': : • ' : jo 10 20 30 40 50 60 70 80 90Time (sec)
FigureE7.4and characterized the strain l evel by the peak shear strain ampl itude.
The time history ofshear strain for a typical earthquake motion, however, is highl y
irregul ar with a peak ampl itudethat may onl y be approached by a few spikes in the
record. Figure 7.10 shows both harmonic(as in a typical l aboratory test) and
transient (as in a typical earthquake) shear straintime histories that have the
same peak cycl ic shear strain. Cl earl y, the harmonic record representsa more severe
l oading condition than the transient record, al though their peak val uesare
identical . As a resul t, it is common to characterize the strain l evel of the
transient recordin terms of an effective shear strain which has been empirical l y
found to vary between about50 and 70% of the maximum shear strain. The computed
response is not particul arl y sensitiveto this percentage, however, and the
effective shear strain is often taken as 65% of thepeak strain.Since the computed
strain l evel depends on the val ues of the equival ent l inear properties,an iterative
procedure is required to ensure that the properties used in the anal ysis are
compatibl e with the computed strain l evel s in al l l ayers. Referring to Figure 7.11,
the iterativeprocedure operates as fol l ows: 272 Ground Response Anal ysis Chap. 7c: :
.~1i5co(I)s: enFigure 7.10 Two shear strain time histories with identical peak
shear strains. For thetransient motion of an actual earthquake, the effective shear
strain is usual l y taken as65% of the peak strain.Shear strain (l og scal e)~(1)
L-- .I.-.O.--'-- _s(: 1)Shear strain (l og scal e)~ G(1) r---.-..-=- - - - - - .. (1)
(JJ- I -S (3) I "8 d 3) • IE Ico d 2)(I)s: enFigure 7.11 Iteration toward strain-
compatibl e shear modul us and damping ratio inequival ent l inear anal ysis. Using
initial estimates, G(l )and ~(l ), the equival ent l inearanal ysis predicts an
effective shear strain, Ydil .Because this strain is greater than those
corresponding to G(l )and ~(l ). an iteration is required. The next iteration uses
parameters,G(2) and ~(2), that are compatibl e with y~il .The equival ent l inear
anal ysis is repeatedand the parameters checked until strain-compatibl e val ue of G
and ~ are obtained.1. Initial estimates of G and ~ are made for each l ayer. The
initial l y estimated val uesusual l y correspond to the same strain l evel ; the l ow-
strain val ues are often used forthe initial estimate.2. The estimated G and ~
val ues are used to compute the ground response, incl udingtime histories of shear
strain for each l ayer.3. The effective shear strain in each l ayer is determined
from the maximum shear strainin the computed shear strain time history. For l ayer j
'Y~~j = Ry'Y~~Xjwhere the superscript refers to the iteration number and Ry is the
ratio of the effectiveshear strain to maximum shear strain. Ry depends on
earthquake magnitude (Idrissand Sun, 1992) and can be estimated fromSec. 7.2 One-
Dimensional Ground Response Anal ysis 273M-l Ry = -----u>4. From this effective shear
strain, new equival ent l inear val ues, G(i+1) and ~(i+l ) are chosenfor the next
iteration.5. Steps 2 to 4 are repeated until differences between the computed shear
modul us anddamping ratio val ues in two successive iterations fal l bel ow some
predeterminedval ue in al l l ayers. Al though convergence is not absol utel y
guaranteed, differences ofl ess than 5 to 10% are usual l y achieved in three to five
iterations (Schnabel et al .,1972).Even though the process of iteration toward
strain-compatibl e soil properties al l owsnonl inear soil behavior to be approximated,
it is important to remember that the compl exresponse method is stil l a l inear
method of anal ysis. The strain-compatibl e soil propertiesare constant throughout
the duration of the earthquake, regardl ess of whether the strains ata particul ar
time are smal l or l arge. The method is incapabl e of representing the changes insoil
stiffness that actual l y occur during the earthquake. The equival ent l inear approach
toone-dimensional ground response anal ysis of l ayered sites has been coded into a
widel yused computer program cal l ed SHAKE (Schnabel et al ., 1972).Exampl e 7.5An
extensive l aboratory testing program conducted by EPRI (1993) produced detail ed
informationon the modul us reduction and damping characteristics of the soil s
beneath the Gil royNo.2 (soil ) recording station. Al though the soil conditions
varied with depth, a rough approximationto the average modul us reduction and
damping characteristics is
given bel ow.Strain(%)10-4 10-3.5 10-3 10-2.5 10-2 10-1.5 10- 1 10-0.5 10° 10°·5
101G/Gmax 1.00 1.00 1.00 0.99 0.90 0.71 0.47 0.24 0.10 0.05 0.04S(%) 3.00 3.00 3.00
3.53 4.83 7.68 12.3 18.5 24.4 27.0 30.0Repeat the anal ysis of Exampl e 7.4 with the
data l isted above using the iterative equival ent l inearapproach.Sol ution As in the
case of Exampl e 7.4, the transfer function was eval uated using the computerprogram
SHAKE (Schnabel et a!., 1972). In this exampl e the first iteration used the same
stiffness and damping val ues used in Exampl e 7.4. Subsequent iterations used
stiffness anddamping val ues that were consistent with the modul us reduction and
damping behavior l istedabove. After a total of eight iterations, the shear modul i
and damping ratios had converged towithin 1% of strain-compatibl e val ues. Because
the strain-compatibl e shear modul i weresmal l er than the l ow-strain shear modul i on
which the anal ysis of Exampl e 7.4 was based (theiterations converged to strains at
which G/Gmax val ues were l ess than 1.0), the transfer function(Figure E7.5c) was
shifted toward l owerfrequencies. As in the previous exampl es, the Fourierseries of
the ground surface motion (Figure E7.5d) was computed as the product of thetransfer
function and the Fourier series of the bedrock motion. Inversion of this Fourier
seriesproduced the time history of ground surface accel eration shown in Figure
E7.5e.274 Ground Response Anal ysis Chap. 780 90Time (sec)20 25Frequency (Hz)30 40
50 60 70: : : : : : 20(C)iiI: : ~c 0 ~~o 5 10 15.~ : : J ': ~(d) 0-(Jl -;LL o A-0 5 10 15
20 25Frequency (Hz)5 10 15 20 25Frequency (Hz)E: ~: : : : : 'j (f)(e) il l : : ~'--''---
J."""",'~_.J.- .J.- --,-- --,-- joo 10 20 30 40 50 60 70 80 90Time (sec)FigureE7.5
The softer soil behavior indicated by the iterative equival ent l inear anal ysis is
cl earl yrefl ected in the transfer function (Figure E7.5c), the Fourier series of the
ground surface motion(Figure E7.5d), and the time history of ground surface
accel eration (Figure E7.5e). The predictedpeak ground surface accel eration of
O.304gcompares wel l with the peak accel eration ofO.322grecorded at the Gil roy No.2
(soil ) station, but comparison of the overal l motions in thetime domain (Figure
E7.5e versus Figure 3.1) and frequency domain (Figure E7.5d versus Figure3.13b)
shows significant differences.The anal yis described in this exampl e was intended to
il l ustrate the effects of the equival entl inear approach to approximation of
nonl inear soil behavior on computed ground motions.Like Exampl es 7.2 to 7.4, it
characterized the actual soil conditions at the Gil roy No.2 (soil ) sitein a
simpl ified manner. More detail ed characterization (EPRI, 1993) produced
significantl ybetter agreement between predicted and recorded motions.7.2.1.3
Deconvol utionBecause the equival ent l inear approach util izes a l inear anal ysis, the
response at anypoint can be rel ated to the response at any other point. Al though
the transfer functionsdevel oped in Section 7.1.1.1 rel ated to the computation of
free surface motion from bedrockSec. 7.2 One-Dimensional Ground Response Anal ysis
275motion, transfer functions rel ating motions at other depths can al so be derived
without difficul ty.An important probl em of practical interest invol ves the
computation of bedrockmotion from a known free surface motion. This process, known
as deconvol ution, is particul arl yuseful in the interpretation of actual ground
motions recorded on the surfaces of soil deposits.Al though deconvol ution of a l inear
el astic system shoul d theoretical l y produce aunique sol ution, practical
difficul ties often arise. Some of these may be of a numerical nature, particul arl y
when iteration toward strain-compatibl e soil properties are required(Roesset, 1977)
and strain l evel s are l arge. Others are associated with l imitations in theaccuracy
of the assumption that al l motion resul ts from vertical l y propagating shear waves.
Sil va (1988) found that about 75% of the power (87% of the ampl itude) in a free
surfacemotion coul d be attributed to vertical l y propagating shear waves at
frequencies up to 15 Hz;the remainder was attributed to scattered waves and surface
waves. Sil va suggested adeconvol ution procedure based on the use of a prefil tered
(15 Hz l ow-pass to el iminate thetendency to devel op unreal istical l y l arge
accel erations at depth) free surface motion anditeration toward strain-compatibl e
properties using 87% of the input motion ampl itudesbefore deconvol ving using the
fil tered free surface motion at ful l (100%) ampl itude. Deconvol utionshoul d be
performed with great care and the reasonabl eness of any deconvol vedmotion eval uated
careful l y.']'.2.2 Nonl inear ApproachAl though the equival ent l inear approach is
computational l y convenient and provides reasonabl eresul ts for many practical
probl ems, it remains an approximation to the actual nonl inearprocess of seismic
ground response. An al ternative approach is to anal yze the actual nonl inear response
of a soil deposit using direct numerical integration in the time domain.By
integrating the equation of motion in smal l time steps, any l inear or nonl inear
stressstrainmodel (Section 6.4.3) or advanced constitutive model (Section 6.4.4)
can be used. Atthe beginning of each time step, the stress-strain rel ationship is
referred to to obtain theappropriate soil properties to be used in that time step.
By this method, a nonl inear inel asticstress-strain rel ationship can be fol l owed in
a set of smal l incremental l y l inear steps.Most currentl y avail abl e nonl inear one-
dimensional ground response anal ysis computerprograms characterize the stress-
strain behavior of the soil by cycl ic stress-strainmodel s (Section 6.4.3) such as
the hyperbol ic model , modified hyperbol ic model , Ramberg-Osgood model , Hardin-
Drnevich-Cundal l -Pyke (HDCP) model , Martin-Davidenkovmodel , and Iwan-type model .
Others have been based on advanced constitutive model s(Section 6.4.4), such as the
nested yiel d surface model . Some of the most commonl y usedcomputer programs for
nonl inear one-dimensional ground response anal ysis are l isted inTabl e 7-1. A number
of techniques can be used to integrate the equations of motion. Ofthese, the
expl icit finite-difference technique is most easil y expl ained.Consider the soil
deposit of infinite l ateral extent shown in Figure 7.12a. If the soil l ayer is
subjected to horizontal motion at the bedrock l evel , the response wil l be governed
by the equation of motionciT cPuaz = P at2auPat (7.38)276 Ground Response Anal ysis
Tabl e 7-1 Computer Programs for Nonl inear OneDimensional Ground Response Anal ysis
Chap. 7ProgramCHARSOILDESRA-2DYNAIDMASHNONLI3TESS]xSoil Model Ramberg-Osgood
Hyperbol icNested yiel d surfaceMartin-DavidenkovIwan-typeHDCPReferenceStreeter et
al . (1973)Lee and Finn (1978)Prevost (1989)Martin and Seed (1978)Joyner (1977)Pyke
(1985)12i-------------- i+ 1(a) (b)Figure 7.12 (a) Nomencl ature for uniform soil
deposit of infinite l ateral extentoverl ying bedrock; (b) discretization of soil
deposit into N subl ayers.To introduce the expl icit finite-difference method,
consider the functionf(x) shown in Figure7.13. The first derivative off(x) at x =x
is given bydf(x) _ l im~ - Lix~Of(x + Lix) - f(x)Lix(7.39)A reasonabl e approximation
to the first derivative can be made by removing the restrictionof the l imit and
using a smal l but finite val ue of Lix. In this way the expression of equation(7.39)
is referred to as aforward-difference approximation to df (x) / dx. Figure 7.13
il l ustratesthat the forward difference actual l y provides a better approximation to
the derivativeat x = x+ Lix/2 than at x = x.Dividing the soil l ayer into N
subl ayers of thickness, Liz(Figure 7.12b), and proceedingthrough time in smal l time
increments of l ength, Lit, the notation Ui,1 =U (z =i Liz, t) canbe used to write
finite-difference approximations to the derivativesc/'t 'ti+ 1,1-'ti,1 (7.40a) -"=
az Lizau Ui, I+M - « ;1 at (7.40b) "= LitSubstituting equations (7.40) into the
equation of motion al l ows that differential equation tobe approximated by the
expl icit finite-difference equationPu,,t+M- u,,1Lit(7.41)Sec. 7.2F(x)One-
Dimensional Ground Response Anal ysis--TI t(x + Ll X) - t(x)______________ ~ 11I1II11
I1Ll x .. I1X X+ Ll x x277(7.42)(7.44)Figure 7.13 Forward-difference approximation
ofl '(x) is given by sl ope of l inepassing through function at x = x and x = x+ Ax .
Approximation becomes exact asfu -7 O.Sol ving for »: 1+!11 gives. ~t U ( ) i, I +!11
= U i, I + -P-L--l ;Z- 'ti + I ' I - 't i, IEquation (7.42) simpl y shows how the
conditions at time, t, can be used to determinethe conditions at time, t + M. Using
equation (7.42) for al l i, the vel ocity profil e can be determinedat time t + ~t.
Using the computed vel ocities at the end of each time step as the initial vel ocities
for the next time step, the repeated appl ication of equation (7.42) al l ows the
equationof motion to be integrated in a series of smal l time steps.As with any
integration probl em, however, the boundary conditions must be satisfied.Since the
ground surface is a free surface, 'tl =0, soU1,1+111 = UI, 1+ P~Z't2, I (7.43)The
boundary condition at the bottom of the soil deposit depends on the nature of the
underl yingbedrock. If the bedrock is rigid, its particl e vel ocity, ub(t) = UN+ I,
I' can be specifieddirectl y as the input motion. If the bedrock is el astic,
continuity of stresses requires that theshear stress at the bottom of the soil
l ayer , 'tN+ 1,1' be equal to the shear stress at the top of therock l ayer, 'tr,t.
Thus. . M ( ) UN+ 1,1+111 =uN+ 1,1+P-~Z-'tr.-t 'tN,tIf an incident wave travel ing
upward through the rock has a particl e vel ocity u, (t) at thesoil -rock boundary,
the shear stress at the boundary is approximated (Joyner and Chen,1975) by(7.45)278
Ground
Response Anal ysis Chap. 7Substituting equation (7.45) into equation (7.44) and
sol ving for UN + 1, 1+ M givesUN+l 'I+~[2PrVsrUr(t+M) -'tN,I]1 + (M/p~z)Prvsr(7.46)
Once the boundary conditions have been establ ished, the integration cal cul ations
proceedfrom the bottom (i =N + 1) to the top (i =1) of the soil deposit in each
time step, and stepby step in time. Computation of the vel ocity at the end of each
time step, however, requiresknowl edge of the shear stress in that time step.If the
soil deposit is initial l y at rest, then Ui, I = 0 = 0 and 'ti,1 = 0 for al l i. When
theinput motion, in the form of Ub(t) (rigid bedrock) or ur(t) (el astic bedrock),
imparts somevel ocity to the base of the soil deposit, UN +! wil l take on a nonzero
val ue. In subsequenttime steps, UN' UN _ j, UN _2, ... wil l al l take on nonzero
val ues as the soil deposit moves inresponse to the input motion. The incremental
displ acement in each time step is given by(7.47)Summing the incremental
displ acements al l ows the total displ acement, Ui,I' to be determinedat the beginning
of each time step. The shear strain in each subl ayer is given bydUi I Ui+! I-ui IY·
=-' '" ' ,1,1 dz ~z (7.48)If the soil is assumed to be l inear el astic, the shear
stress depends onl y on the current shearstrain, (i.e., 'tu =G,Yu). If the soil is
nonl inear and inel astic, however, the shear stress wil l depend on the current shear
strain and the stress-strain history. In such cases the computedshear strain, Yi,I'
and the cycl ic stress-strain rel ationship (or advanced constitutive model )are used
to determine the corresponding shear stress, 'ti,I' The integration process can
thenbe summarized as fol l ows: 1. At the beginning of each time step, the particl e
vel ocity, Ui, I' and total displ acement,Ui,I' are known at each l ayer boundary.2.
The particl e displ acement profil e is used to determine the shear strain, Yi,!'
within eachl ayer.3. The stress-strain rel ationship is used to determine the shear
stress, 'ti,!' in each l ayer.The stress-strain curve may be l inear or nonl inear. If
nonl inear inel astic soil behavioris assumed, stress reversal s are checked and
accounted for (e.g., by appl ication of theMasing criteria) in each l ayer.4. The
input motion is used to determine the motion of the base of the soil l ayer at timet
+ ~t.5. The motion of each l ayer boundary at time t + ~t is cal cul ated, working
from bottomto top. The process is then repeated from step 1 to compute the response
in the nexttime step.Because the particl e vel ocities are computed at times that
differ by one-hal f time stepfrom those at which the shear stresses are best
approximated, the expl icit method canbecome numerical l y unstabl e if the time step
is too l arge (i.e., if ~t > &/vss) (Davis, 1986).Sec. 7.2 One-Dimensional Ground
Response Anal ysis 279Using different difference expressions, an impl icit finite-
difference formul ation can el iminatethe stabil ity probl em, thereby al l owing the use
of l onger time steps. Impl icit formul ations,however, invol ve the sol ution of a set
of N + 1 simul taneous equations (which can becomputational l y time consuming for
l arge N) at each time step. Whether it is more efficientto perform the rapid
cal cul ations of the expl icit method at a l arge number of time steps, orthe more
time-consuming cal cul ations of the impl icit method at fewer time steps, is often
difficul t to predict. Most existing computer programs for nonl inear ground response
anal ysisuse the expl icit formul ation.7 ..2.3 Comparison of One-Dimensional Ground
Response Anal ysesAl though equival ent l inear and nonl inear methods are both used to
sol ve one-dimensional ground response anal ysis probl ems, their formul ations and
underl ying assumptions are quitedifferent. Consequentl y, it is reasonabl e.to expect
to find some differences in their resul ts.The resul ts of equival ent l inear and
nonl inear ground response anal yses have beencompared on a number of occasions
(e.g., Joyner and Chen, 1975; Martin and Seed, 1978;Dikmen and Ghaboussi, 1984)
with the fol l owing general concl usions: 1. The inherent l inearity of equival ent
l inear anal yses can l ead to spurious resonances(i.e., high l evel s of ampl ification
that resul t from coincidence of a strong componentof the input motion with one of
the natural frequencies of the equival ent l inear soil deposit). Since the stiffness
of an actual nonl inear soil changes over the duration of al arge earthquake, such
high ampl ification l evel s wil l not devel op in the fiel d.2. The use of an effective
shear strain in an equival ent l inear anal ysis can l ead to an oversoftenedand
overdamped system when the peak shear strain is much l arger than theremainder of
the shear strains, or to an undersoftened, underdamped system when theshear strain
ampl itude is nearl y uniform.3. Equival ent l inear anal yses can be much more
efficient than nonl inear anal yses, particul arl ywhen the input motion can be
characterized with acceptabl e accuracy by asmal l number of terms in a Fourier
series. For exampl e, most earthquakes contain rel ativel yl ittl e el astic wave energy
at frequencies above 15 to 20 Hz. Consequentl y, theresponse can usual l y be computed
with reasonabl e accuracy by considering onl y thefrequencies bel ow 15 to 20 Hz (or
l ower, in some cases). As the power, speed, andaccessibil ity of computers have
increased in recent years, the practical significance ofdifferences In the
efficiency of one-dimensional ground response anal yses hasdecreased substantial l y.
4. Nonl inear methods can be formul ated in terms of effective stresses to al l ow
model ingof the generation, redistribution, and eventual dissipation of excess pore
pressure duringand after earthquake shaking. Equival ent l inear methods do not have
this capabil ity.5. Nonl inear methods require a rel iabl e stress-strain or
constitutive model . The parametersthat describe such model s are not as wel l
establ ished as those of the equival entl inear model . A substantial fiel d and
l aboratory testing program may be required toeval uate nonl inear model parameters.
280 Ground Response Anal ysis Chap. 76. Differences between the resul ts of
equival ent l inear and nonl inear anal yses depend onthe degree of nonl inearity in the
actual soil response. For probl ems where strain l evel sremain l ow (stiff soil
profil es and/or rel ativel y weak input motions), both anal ysescan produce reasonabl e
estimates of ground response. For probl ems invol ving highstrain l evel s,
particul arl y probl ems in which the induced shear stresses approach theavail abl e
shear strength of the soil , nonl inear anal yses are l ikel y to provide reasonabl e
resul ts.In summary, both equival ent l inear and nonl inear techniques can and have
been usedsuccessful l y for one-dimensional ground response anal ysis. The use and
interpretation ofeach requires knowl edge of their underl ying assumptions,
understanding of their operation,and recognition of their l imitations. Neither can
be considered mathematical l y rigorous orprecise, yet their accuracy is not
inconsistent with the variabil ity in soil conditions, uncertaintyin soil
properties, and scatter in the experimental data upon which many of their input
parameters are based.7.3 TWO-DIMENSIONAL DYNAMIC RESPONSE ANALYSISThe methods of
one-dimensional ground response anal ysis described in previous sectionsare useful
for l evel or gentl y sl oping sites with paral l el material boundaries. Such
conditionsare not uncommon and one-dimensional anal yses are widel y used in
geotechnical earthquakeengineering practice. For many other probl ems of interest,
however, the assumptionsof one-dimensional wave propagation are not acceptabl e.
Sl oping or irregul ar ground surfaces,the presence of heavy structures or stiff,
embedded structures, or wal l s and tunnel s al l require two-dimensional or possibl y
even three-dimensional anal ysis. Probl ems in whichone dimension is considerabl y
greater than others can often be treated as two-dimensional pl ane strain probl ems. A
number of common cases are shown in Figure 7.14.The fol l owing sections discuss a
number of methods that can be appl ied to twodimensional dynamic response probl ems.
Techniques for the sol ution of such probl emshave been devel oped using both
frequency-domain (compl ex response) methods and timedomain(direct integration)
methods. These techniques have been appl ied to many practical probl ems such as those
shown in Figure 7.14. Two- and three-dimensional dynamicresponse and soil -structure
interaction probl ems are most commonl y sol ved using dynamicfinite-el ement anal yses.
Al though a detail ed treatment of the finite-el ement method is(a) (b) (c)Figure 7.14
Exampl es of common probl ems typical l y anal yzed by two-dimensional pl ane strain
dynamic response anal yses: (a) cantil ever retaining wal l ; (b) earth dam; (c)tunnel .
Sec. 7.3 Two-Dimensional Dynamic Response Anal ysis 281beyond the scope of this
book, a brief introduction precedes the description of two- andthree-dimensional
dynamic response anal yses.7.3.1 Dynamic Finite-El ement Anal ysisThe finite-el ement
method treats a continuum as an assembl age of discrete el ements whoseboundaries are
defined by nodal points, and assumes that the response of the continuum canbe
described by the response of the nodal points. The fol l owing section presents a
very briefsummary of the finite-el ement method that is intended onl y to provide a
basic description ofits principl es to the unfamil iar reader. Compl ete descriptions
of the finite-el ement methodmay be found in a number of books devoted to that topic
(e.g., Desai and Abel , 1972; Bathe,1982; Zienkiewicz and Tayl or, 1989).7.3.1.1
El emental Equations of MotionIn the finite-el ement method, the probl em of interest
is first discretized by dividing itinto el ements as shown in Figure 7.15. The
displ acement of the soil at any point within anel ement, {v}T ={u v}, is expressed
in terms of the nodal point displ acements,
{q} T ={UI U2 U3 U4 VI V2 V3 V4}, by{v} = [N] {q} (7.49)where [N] is a matrix of
shape functions. The strain-displ acement matrix, [B], al l ows thestrains to be
determined from the nodal point displ acements{E} = [B] {q}and the stress-strain
matrix, [D], rel ates stresses to strains: {O'} = [D] {e}(7.50)(7.51)(7.52)Defining a
l ocal coordinate system, (s, t), that maps the quadril ateral el ements into squares
as shown in Figure 7.16, and using the strain-displ acement and stress-strain
rel ationships,an el ement stiffness matrix can be written (assuming unit thickness
in the z-direction) asI I[ke] = f f [B]T[D] [B]IJl dsdt-1 -IxFigure 7.15 Finite-
el ement discretization of retaining structure il l ustrating the degreesof freedom of
a typical four-noded el ement.282 Ground Response Anal ysis Chap. 7->t1, 1 1, 14 3s1
21, -1 1, -1(7.53)(7.54)xFigure 7.16 Mapping of quadril ateral el ement from
irregul ar shape in x-y coordinatesystem to square shape in s-t coordinate system.
where the Jacobian4 4IJI = " " x. (dNidNj _ dNidNi ) .L..J L..J I ds dt dt ds Y;i ~
Ij ~ 1A consistent el ement mass matrix can be written, assuming constant density
within the el ement,as1 1[me] = Pf f [N]T [N] [Jl ds dt-1 -IAs an al ternative, a
l umped-el ement mass matrix can be devel oped by assuming that themass of the el ement
is concentrated at the nodal points. Experience has shown that use of theconsistent
el ement matrix tends to overestimate the natural frequencies of a system and that
the l umped mass matrix tends to underestimate them by about the same amount. Lysmer
etal . (1974) suggested the use of a mixed el ement mass matrix, which is simpl y the
average ofthe consistent and l umped mass matrices.Damping matrices can be
troubl esome because of the impl ications of various formul ationson the frequency
dependence of damping. For nonl inear ground response anal yses,however, damping
resul ts primaril y from the hysteretic behavior of the soil and is therefore
accounted for by variations in the stiffness matrix under cycl ic l oading
conditions. Somesmal l amount of viscous damping may be incl uded in a two-
dimensional ground responseanal ysis to account for damping at very smal l strains
and to minimize numerical probl emsthat can arise in the compl ete absence of
damping. A consistent damping matrix can beobtained from[eel = Pf f [B]T ['l J] [B]
IJl ds dt-1 -Iwhere ['l J] is a matrix of damping terms. The equations of motion for
the el ement can thenbe written as(7.55)Sec. 7.3 Two-Dimensional Dynamic Response
Anal ysis 283where the el ement force vector is given by{Q (t)} = rr[N]T {W} IJl ds dt
+f[N] T{T} dS-1 -1 sand {W} is the vector of prescribed body forces and {T} is a
vector of external tractions thatmay be appl ied to some surface, S.7.3.1.2 Gl obal
Equations of MotionOnce the equations of motion for each el ement are obtained, they
are combined in away that satisfies compatibil ity of displ acements to obtain the
gl obal equations ofmotion[M] {u} + [C] {u} + [K] {u} = {R (t)} (7.56a)where [M] is
the gl obal mass matrix, [C] the gl obal damping matrix, [K] the gl obal stiffness
matrix, {u} the gl obal nodal point displ acement vector, and {R(t)} the gl obal nodal
pointforce vector. For the case of l oading induced by base motion, the gl obal
equation of motion is[M] {u} + [C] {u} + [K] {u} = - [M] [1] Ub (t) (7.56b)7.3.1.3
Discretization ConsiderationsThe response of both equival ent l inear and nonl inear
finite-el ement model s can beinfl uenced by discretization. In particul ar, the use of
coarse finite-el ement meshes can resul tin the fil tering of high-frequency
components whose short wavel engths cannot be model edby widel y spaced nodal points.
The maximum dimension of any el ement shoul d be l imitedto one-eighth (Kuhl emeyer and
Lysmer, 1973) to one-fifth (Lysmer et al ., 1975) of theshortest wavel ength
considered in the anal ysis.7.3.1.4 Boundary ConditionsFor computational efficiency
it is desirabl e to minimize the number of el ements in afinite-el ement anal ysis.
Since the maximum dimensions of the el ements are general l y control l edby the wave
propagation vel ocity and frequency range of interest, minimizing thenumber of
el ements usual l y becomes a matter of minimizing the size of the discretizedregion.
As the size of the discretized region decreases, the infl uence of boundary
conditionsbecomes more significant.For many dynamic response and soil -structure
interaction probl ems, rigid or nearrigidboundaries such as bedrock are l ocated at
considerabl e distances, particul arl y in thehorizontal direction, from the region of
interest. As a resul t, wave energy that travel s awayfrom the region of interest may
effectivel y be permanentl y removed from that region. In adl ynamicfinite-el ement
anal ysis, it is important to simul ate this type of radiation dampingbehavior. The
most commonl y used boundaries for finite-el ement anal yses can be dividedinto three
groups (Christian et al ., 1977; Wol f, 1985).El ementary Boundaries. Conditions of
zero displ acement or zero stress arespecified at el ementary boundaries (Figure
7.l 7a). El ementary boundaries can be used tomodel the ground surface accuratel y as
a free (zero stress) boundary. For l ateral or l ower284 Ground Response Anal ysis
Chap. 7(a)--+---+-ffl --+---+-ffl --+-----+-ffl ~(b) (c)Figure 7.17 Three types of
finite-el ement mesh boundaries: (a) el ementary boundaryin which zero displ acements
are specified; (b) l ocal boundary consisting of viscousdashpots; (c) l umped-
parameter consistent boundary (actual l umped parameter woul dconsist of more masses,
springs, and dashpots than shown).boundaries, however, the perfect refl ection
characteristics of el ementary boundaries cantrap energy in the mesh that in real ity
woul d radiate past the boundaries and away from theregion of interest. The
resul ting "box effect" can produce serious errors in a groundresponse or soil -
structure interaction anal ysis. If el ementary boundaries are pl aced farenough from
the region of interest, refl ected waves may be damped sufficientl y to negatetheir
infl uence.Local Boundaries. Section 5.4.1 showed how a viscous dashpot coul d be
usedto simul ate a semi-infinite region for the case of normal l y incident body
waves. The use ofviscous dashpots (Figure 7.17b) represents a common type of l ocal
boundary. It can beshown (e.g., Wol f, 1985) that the val ue of the dashpot
coefficient necessary for perfectenergy absorption depends on the angl e of
incidence of the impinging wave. Since wavesare l ikel y to strike the boundary at
different angl es of incidence, a l ocal boundary with specificdashpot coefficients
wil l al ways refl ect some of the incident wave energy. Additional difficul ties arise
when dispersive surface waves reach a l ocal boundary; since their phasevel ocity
depends on frequency, a frequency-dependent dashpot woul d be required to absorbal l
their energy. The effects of refl ections from l ocal boundaries can be reduced by
increasingthe distance between the boundary and the region of interest.Consistent
Boundaries. Boundaries that can absorb al l types of body wavesand surface waves at
al l angl es of incidence and al l frequencies are cal l ed consistent boundaries.
Consistent boundaries can be represented by frequency-dependent boundary stiffness
matrices obtained from boundary integral equations or the boundary el ement method.
Wol f(1991), for exampl e, devel oped a l umped-parameter model consisting of an
assembl age ofdiscrete springs, masses, and dashpots which can approximate the
behavior of a consistentboundary. A greatl y simpl ified exampl e of such an
assembl age is shown in Figure 7.17c.7.3.2 Equival ent Linear ApproachThe two-
dimensional equival ent l inear approach is very simil ar to the one-dimensional
approach. A soil -structure system is represented by a two-dimensional finite-
el ementmodel . The input motion is represented by a Fourier series and the equations
of motion aresol ved for each frequency of the series (or at sel ected frequencies
with interpretation inbetween) with the resul ts summed to obtain the total
response.Sec. 7.3 Two-Dimensional Dynamic Response Anal ysis 285(a) (b)Figure 7.18
Practical situation where two-dimensional ground response anal yses areused: (a)
pl ane strain conditions can be assumed at center of l ong dam, al l owing (b)center
section of dam to be model ed in two dimensions.Consider the probl em of the earth
dam shown in Figure 7.18a. Assuming that the axisof the dam is l ong rel ative to its
height, the response of the center portion of the dam can beassumed to be two-
dimensional . Dynamic equil ibrium of the model shown in Figure 7.l 8brequires that
the fol l owing equations of motion be satisfied: [M] {u} + [K*] {u} = - [M] [1] ~b(t)
(7.57)(7.59)where [M] is the mass matrix, [K*] the compl ex stiffness matrix, [K*]
=[K] + im[C], {u}the vector of unknown nodal point displ acements (rel ative to the
base), and iib(t) is the timehistory of base accel eration. The mass and stiffness
matrices are assembl ed from the correspondingel ement stiffness matrices using
standard finite-el ement procedures and dampingis introduced into the anal ysis
through the use of compl ex shear modul i when forming thecompl ex el ement stiffness
matrices.If the base motion is assumed to be harmonic, the rel ative displ acement
vector can beexpressed as{u} = {H (oij } l ib (ro) e'?' (7.58)where {H( m)} is a
vector of transfer functions and l ib ( co) is the Fourier transform of iib(t).
Substituting equation (7.58) into the equation of motion gives- m2[M] {H (m)} ub
(ro) ei ffi( + [K*] {H (m)} ub (co) ei ffi( = - [M] [1] ub(m) ei ffi(Rearranging
al l ows the transfer function vector to be expressed as{ H (m) } - [M]- m2 [M] -
[K*]Once the transfer function vector has been obtained, computation of the
response fol l owsthe same procedures used for one-dimensional compl ex response
anal ysis
(Section7.1.1.2). In this approach the primary computational effort is associated
with eval uation ofthe transfer functions. For l arge probl ems, the mass and
stiffness matrices are l arge, andeval uation of the transfer functions can be quite
time consuming. For computational efficiency,the transfer functions are often
expl icitl y eval uated at onl y a l imited number of frequencies,with val ues at
intermediate frequencies obtained by interpol ation (Lysmer et al .,286 Ground
Response Anal ysis Chap. 71975; Tajirian, 1981). Iteration toward strain-compatibl e
material properties (Section7.1.1.3) can be incorporated on an el ement-by-el ement
basis. Anal yses of this type are performedby the widel y used computer program FLUSH
(Lysmer et al ., 1975). The use of general izedhyperel ements in the computer program
GROUND2D greatl y reduces the numberof degrees of freedom required for two-
dimensional dynamic response anal yses (Deng et al .,1995).7.3.3 Nonl inear Approach
Two-dimensional nonl inear anal yses can be used to estimate permanent displ acements
ofsl opes (Chapter 10), retaining structures (Chapter 11), and other constructed
facil ities. Twodimensional nonl inear dynamic response anal yses are performed by
writing the gl obal equations of motion [equation (7.57)] from a finite-el ement
ideal ization in incremental form and then integrating them in the time domain. Such
anal yses can be divided into twomain groups according to the manner in which the
soil behavior is represented. One groupuses cycl ic nonl inear stress-strain model s
(Section 6.4.3) and the other uses advanced constitutivemodel s (Section 6.4.4).Finn
et al . (1986) extended the hyperbol ic model with extended Masing criteria anda
residual pore pressure model from one to two dimensions in the computer program
TARA-3. The program has been used to back-anal yze ful l -scal e case histories and the
resul ts of various centrifuge model tests. Even with its rel ativel y simpl e cycl ic
stress-strainmodel , the program has been abl e to capture the most important aspects
of ground responsewith good accuracy. According to Finn (1988), the simpl icity of
the stress-strain model canproduce substantial computational efficiency rel ative to
model s based on more compl icatedsoil model s.Methods based on advanced constitutive
model s have been devel oped. The programDYNAFLOW (Prevost, 1981) uses a mul tipl e
yiel d surface model to predict deformationsand pore pressures. It has been appl ied
to probl ems of earth dams with good resul ts. A consortiumof Japanese construction
firms have devel oped the program DIANA (Kawai, 1985)that can perform static and
dynamic anal yses with different advanced constitutive model s.A number of other
programs incorporating advanced constitutive model s have been devel opedin recent
years.7.3.4 Other Approaches to Two-Dimensional DynamicResponse Probl emsIn an
attempt to capture the most significant aspects of various two-dimensional dynamic
response probl ems without the computational cost and compl exity of dynamicfinite-
el ement anal yses, a number of investigators have devel oped al ternative approaches
tospecific probl ems. These approaches typical l y invol ve simpl ifying assumptions
that al l owtwo-dimensional probl ems to be sol ved by one-dimensional anal yses.7.3.4.1
Shear Beam ApproachOne of the earl iest approaches to the dynamic anal ysis of two-
dimensional geotechnical systems was the shear beam anal ysis appl ied to earth dams
by Mononobe et al . (1936). Theapproach has since been verified and extended to
cover a variety of conditions; a comprehensivereviewwas prepared by Gazetas (1987).
The shear beam approach is based on the assumptionthat a dam deforms in simpl e
shear, thereby producing onl y horizontal displ acements.Sec. 7.3 Two-Dimensional
Dynamic Response Anal ysis 287Hatanaka (1952) and others have verified the accuracy
of this assumption, at l east for rigidfoundation conditions. The shear beam
approach al so assumes that either shear stresses orshear strains are uniform across
horizontal pl anes. These assumptions have al so been verified(Chopra, 1966;
Dakoul as, 1985); stresses and strain are nearl y constant across the dam exceptin
smal l zones near the upstream and downstream faces where they decrease to zero
(Gazetas,1987).Consider the homogeneous, infinitel y l ong dam shown in Figure 7.19.
Assuming horizontal displ acements to be constant at a given depth, the horizontal
displ acement rel ative tothe base, u(z, t), is independent of x. The resul tant
shearing force on the upper surface of asl ice of thickness, dz: isXdSz(t) = Lu't
(x, z, t) dxThe corresponding resul tant shearing force on the bottom of the sl ice
isXdSZ+dz(t) = Lu['t(X,z,t)+d't(~/,t) dzJdxThe resul tant inertial force acting on
the sl ice depends on the total accel eration, that is,For equil ibrium in the x-
direction,Sz+ dz(t) - Sz(t)or(7.60)-: z[r: : ,(x, 4 z)d+Z = P[ii(Z' t) + ",(tl ]2~ZdZ
Substituting 't(x, z, t) =G(x, z)Y(z, t) and y(z, t) =dU(z, t)l dz, the shear beam
equation canbe written asBFigure 7.19 Earth dam, showing stresses acting on an
el ement of thickness, dz: 288 Ground Response Anal ysis Chap. 7p(ii+ii,,) =
l ~[o(z)ZduJz dz dzwhere the average shear modul us, G, is given by(7.61)(7.62)O(z) =
_I_l x dC(x, z) dxxl l +xd-XtfEquation (7.62) is simpl y a one-dimensional wave
equation (i.e., the shear beam approachal l ows the two-dimensional dam section to be
represented as a one-dimensional system).Gazetas (1982) devel oped sol utions to the
shear beam wave equation for the casewhere the shear modul us increases as a power
function of depth according to C(z) =C,,(zIH)II1, where C" is the average shear
modul us at the base of the dam. For such conditions,the nth natural circul ar
frequency (assuming hl H =I) is given by(Oil = ~. ~'(4+m)(2-m)where vss is the
average shear wave vel ocity of the soil in the dam and ~11 is the nth root ofa
period rel ation (Dakoul as and Gazetas, 1985) tabul ated in Tabl e 7-2 for the first
fivemodes of vibration.Tabl e 7-2 Val ues of ~11 for First Five Modesof Vibration of
an Earth DamIJ150 Itm 20 2.404 5.520I : 2 2.903 6.0334 2.999 6.133 7~2 3.142 6.283 3
3.382 7.1063 4 58.654 11.792 14.9319.171 12.310 15.4519.273 12.413 15.5449.525
12.566 15.70810.174 13.324 16.471Figure E7.6Equation (7.62) produces a fundamental
period ofT _ 161t HI - (4+m)(2-m)~1 V's(7.63)Exampl e 7.6The earth dam shown in
Figure E 7.6 is constructed ofcompacted cl ay with a shear wave vel ocityof 1200
ftl sec. Compute the first three natural frequencies of the dam.Sol ution Because the
crest of the dam is so narrow, H ~ h. Then, from equation (7.62), thefirst three
natural frequencies can be cal cul ated as(01 = v"~(4 + m)(2 _ m) = 1200 2.404 (4)
(2) = 19.2 rad/secH 8 150 811 = 3.1 Hz(02 = %'~(4+m)(2-m)(0, = v"~(4+m)(2-m)H8
12005.520(4) (2) = 44.2 rad/sec150 812008.654(4) (2) = 69.2 rad/sec150 812 = 7.0 Hz
I, = 11.0 HzSec. 7.3 Two-Dimensional Dynamic Response Anal ysis 289The mode shape at
the nth natural frequency is given byUn(z) = ( IZi )-mI2Jq[ ~n(ZIi )1-mI2J (7.64)
where Jq is aBessel function of the first kind of order, q = m/(2 - m), which can
be eval uatedfrom_ co (_l )k (xr: Jq(x) - It k!r(q + k + 1) 2 (7.65)k~Owhere ro is
the gamma function (Tabl e 7-3).Tabl e 7-3 Recursion Rel ationship and Val ues of Gamma
Functionr (x) = [ e-xxn-1dx; r (x + 1) =xl " (x)x l (x) x r(x) x r(x) x r(x)1.00
1.00000 1.25 0.90640 1.50 0.88623 1.75 0.919061.01 0.99433 1.26 0.90440 1.51
0.88659 1.76 0.921371.02 0.98884 1.27 0.90250 1.52 0.88704 1.77 0.923761.03 0.98335
1.28 0.90072 1.53 0.88757 1.78 0.926231.04 0.97884 1.29 0.89904 1.54 0.88818 1.79
0.928771.05 0.97350 1.30 0.89747 1.55 0.88887 1.80 0.931381.06 0.96874 1.31 0.89600
1.56 0.88964 1.81 0.934081.07 0.96415 1.32 0.89464 1.57 0.89049 1.82 0.936851.08
0.95973 1.33 0.89338 1.58 0.89142 1.83 0.939691.09 0.95546 1.34 0.89222 1.59
0.89243 1.84 0.942611.10 0.95135 1.35 0.89115 1.60 0.89352 1.85 0.945611.11 0.94739
1.36 0.89018 1.61 0.89468 1.86 0.948691.12 0.94359 1.37 0.88931 1.62 0.89592 1.87
0.951841.13 0.93993 1.38 0.88854 1.63 0.89724 1.88 0.955071.14 0.93642 1.39 0.88785
1.64 0.89864 1.89 0.958381.15 0.93304 1.40 0.88726 1.65 0.90012 1.90 0.961771.16
0.92980 1.41 0.88676 1.66 0.90167 1.91 0.965231.17 0.92670 1.42 0.88636 1.67
0.90330 1.92 0.968781.18 0.92373 1.43 0.88604 1.68 0.90500 1.93 0.972401.19 0.92088
1.44 0.88580 1.69 0.90678 1.94 0.97610120 0.91817 1.45 0.88565 1.70 0.90864 1.95
0.979881.21 0.91558 1.46 0.88560 1.71 0.91057 1.96 0.98374122 0.91311 1.47 0.88563
1.72 0.91258 1.97 0.987681.23 0.91075 1.48 0.88575 1.73 0.91466 1.98 0.991711.24
0.90852 1.49 0.88595 1.74 0.91683 1.99 0.995812.00 1.00000The first and second mode
shapes are shown in Figure 7.20 for various val ues of thestiffness parameter, m. As
m increases, the shear beam anal ysis produces a "whipl asheffect" characterized by
l arge shear strain and high accel eration near the crest of the dam inthe second and
higher modes. Under strong shaking, nonl inear behavior of the material s inan actual
earth dam may prevent the devel opment of the high accel erations predicted by the
shear beam anal ysis.290 Ground Response Anal ysis Chap. 7
0.0 ,-----,------,--,---------,------", 0.0 ,---------,-----,----: : : ;;;;110.2 0.2
0.4 0.4z zR R0.6 0.60.8 0.81.0 1.00.0 0.2 0.4 0.6 0.8 1.0 -0.5 0.0 0.5 1.0U1 U2(a)
(b)Figure 7.20 Mode shapes for (a) first mode and (b) second mode of earth dam
response. Note rapid change of U with depth near crest of dam for second mode at
111 = 1(whipl ash effect). (After Dakoul as and Gazetas, 1985.)The midcrest and base
accel erations are rel ated by the transfer functionU (z =0) + uiJH(w) = ..UiJ(aon)"
Ircq + I )J,,(ao) iaJ" + I (ao)(7.66)where ao = wHIP,s' and a~ = P.,·v,./Prv,.,- is
the impedance ratio at the soil -rock interface.In the preceding derivation, the
soil
was assumed to be l inear and undamped. Theeffects of soil damping can easil y be
incl uded by repeating the derivation with the soil characterizedby a compl ex
stiffness. The resul ts, as expressed in equations (7.62) through(7.66), woul d be
identical except for the repl acement of v" by the compl ex shear wavevel ocity, v;",.
=v".(1 + i ~).The shear beam approach is an excel l ent exampl e of a procedure that
through the judicioususe of appropriate assumptions, greatl y simpl ifies an
important cl ass of groundresponse probl ems. It al l ows rapid estimation of many
important response parameters, andcan be used to check the reasonabl eness of the
resul ts of more sophisticated anal yses. Theshear beam transfer function can be used
in an equival ent l inear anal ysis, or nonl inearinel astic stress-strain behavior can
be assumed in an incremental nonl inear anal ysis.7.3.4.2 Other ApproachesThe l ayered
inel astic shear beam (Stara-Gazetas, 1986) combines the shear beamapproach with a
one-dimensional nonl inear ground response anal ysis. In this approach, aseries of
static nonl inear finite-el ement anal yses are performed with incremental l y
increasinghorizontal pseudoinertial forces to determine the nonl inear backbone
curve for an entirehorizontal l ayer, or "superel ernent." The backbone curves for
each l ayer are then used withthe extended Masing criteria (Section 6.4.3) in a
nonl inear shear beam anal ysis of the dam.Other approaches incl ude the simpl ified
nonl inear method of Dakoul as (1985) and the nonl inearhysteretic method of El gamel
et al . (1985).Sec. 7.4 Three-Dimensional Dynamic Response Anal ysis 2917.3.5
Comparison of Two-Dimensional DynamicResponse Anal ysesDifferences in the underl ying
assumptions and formul ations of two-dimensional dynamicresponse anal yses l ead to
differences in their resul ts. The proper use of these anal ysesrequires
understanding of these differences. The two-dimensional equival ent l inear methodcan
suffer from the spurious resonances and difficul ties associated with effective
straindetermination described for the one-dimensional equival ent l inear method in
Section 7.2.3.In addition, the different modes of vibration associated with the
extra degrees of freedom(vertical transl ation and rotation) in the two-dimensional
case compl icate the computationof the maximum shear strain, require the use of
another material parameter (such as Poisson'sratio) in addition to the shear
modul us, and produce much more compl icated stresspaths. Again, the equival ent
l inear approach is restricted to total stress anal yses.Two-dimensional nonl inear
methods have the enormousl y beneficial capabil ity ofcomputing pore pressures (hence
effective stresses) and permanent deformations. The accuracywith which they can be
computed, however, depends on the accuracy of the constitutivemodel s on which they
are based. Whil e great progress in the constitutive model ing ofsoil s has been made
in the past 20 years, additional refinement is required before precise apriori
predictions of permanent displ acement are possibl e.The shear beam model s are
fundamental l y different from the equival ent l inear andnonl inear finite-el ement
model s in that they restrict particl e movement to the horizontal pl ane. Finite-
el ement anal yses are capabl e of model ing an actual dam's tendency to respond
vertical l y as wel l as horizontal l y, but the shear beam model forces al l of the
el astic waveenergy to produce horizontal deformations. As a resul t, shear beam
model s general l y overestimatethe fundamental frequency of most dams by about 5%,
and higher natural frequenciesby increasingl y greater amounts. Shear beam
displ acements compare wel l (withinabout 10%) with those computed by finite-el ement
anal yses, but shear beam crest accel erationsfor fl exibl e dams can be up to 50%
greater than those from finite-el ement anal yses.This discrepancy is rel ated to the
whipl ash effect produced by the higher shear beam modes.For stiff dams where these
higher modes are associated with frequencies greater than thoseassociated with
earthquake motions, the computed accel erations match much more cl osel y.7.4 THn~EE-
DIMENSIONALDYNAMIC RESPONSE ANALYSISIn some instances the two-dimensional
ideal izations of the preceding sections may not beappropriate and three-dimensional
dynamic response anal yses are necessary. Such conditionscan arise, as il l ustrated
in Figure 7.21, when soil conditions vary three-dimensional l y,when probl em
boundaries vary three-dimensional l y, and when the response of three-dimensional
structures is of interest.Three-dimensional dynamic response probl ems are treated
in much the same way astwo-dimensional probl ems. Dynamic finite-el ement anal yses
are avail abl e, using bothequival ent l inear and nonl inear approaches. A number of
three-dimensional anal yses havebeen devel oped with an emphasis on soil -structure
interaction probl ems (Section 7.5.2).For the important probl em of earth dams, shear
beam anal yses have been devel oped for theapproximate response of dams in narrow
canyons.292 Ground Response Anal ysis Chap. 7(a) (b) (c)Figure 7.21 Three situations
requiring three-dimensional dynamic response or soil structural interaction anal ysis:
(a) site where soil conditions vary significantl y in threedimensions; (b) earth dam
in narrow canyon; (c) site where response of soil is infl uencedby response of
structures (and vice versa) and where response of one structure mayinfl uence
response of another.7.4.1 Equival ent Linear Finite-El ement ApproachComputational l y,
three-dimensional equival ent l inear dynamic response anal yses are virtual l y
identical to the two-dimensional anal yses described in Section 7.2.2. Three-
dimensional finite el ements (usual l y in the shapes of "bricks" or tetrahedra) have
more nodal points with more degrees of freedom than corresponding two-dimensional
el ements, but thebasic process of el ement mass, damping, and stiffness formul ation,
and assembl y into gl obal equations of motion, is identical .To anal yze the seismic
response of earth dams in narrow canyons, Kagawa (1977)impl emented three-
dimensional finite el ements into a predecessor of the two-dimensional equival ent
l inear program FLUSH (Lysmer et al ., 1975). Mejia and Seed (1981) el iminatedseveral
restrictions that resul ted from Kagawa's original assumptions whil e devel oping the
three-dimensional equival ent l inear program TLUSH (Kagawa et al ., 1981).Three-
dimensional equival ent l inear anal yses have al so been devel oped for soil structure
interaction (Section 7.4) probl ems. Computer programs such as SASSI (Lysmer etal .,
1981), CLASSI (Luco and Wong, 1982), and HASSI (Katayama et al ., 1991) use bothsoil
and structural el ements to model soil -structure systems.7.4.2 Nonl inear Finite-
El ement ApproachThree-dimensional nonl inear ground response anal yses are al so
rel ativel y straightforwardextensions of their two-dimensional counterparts. The
computer program DYNAFLOW (Prevost,(981), based on a mul tipl e yiel d surface
pl asticity model , has three-dimensional as wel l as two-dimensional capabil ities.
When appl ied to the back-anal ysis of earth dams (e.g., Prevostet al ., 1985), it
real istical l y represented dam response whil e accounting for al l three componentsof
ground shaking and al l modes of vibration. Three-dimensional nonl inear
soil structureinteraction anal yses can al so be performed with programs such as TRANL
(Bayl oret a!', 1974)and FLEX (Vaughan, 1983). Many of thel arge mul tipurpose finite-
el ement codes(e.g., ANSYS, ADINA, etc.) are al so capabl e of performing three-
dimensional nonl inearSec. 7.4 Three-Dimensional Dynamic Response Anal ysis 293
anal yses, al though they are not specifical l y oriented toward soil -structure
probl ems or earthquakel oading.7,,4.3 Shear Beam ApproachMany earth dams are
constructed in rel ativel y narrow canyons where abutment boundaryconditions
infl uence the response of the dam. Hatanaka (1952) and Ambraseys (1960)devel oped
shear beam-type sol utions for earth dams in rectangul ar canyons of different
width/height ratios. Dakoul as and Gazetas (1986) devel oped a cl osed-form sol ution
for ahomogeneous (vss=constant) earth dam in a semicyl indrical val l ey of width, L,
and depth,H, showingn1tvssH2H (7.67)Un(Y = 0, z)sin(n1tzl H)n1tzl Hand( (7.68)ii(y =
0, z = 0) + iib H 0)) = ..ubwhere al l variabl es are as defined for equations (7.62)
to (7.66) and the origin of axes is atthe center of the crest of the dam. Note that
the naturai frequencies of a three-dimensional dam are al ways greater than the
corresponding natural frequencies of its two-dimensional maximum section, thereby
il l ustrating the "stiffening" effects of the abutments. This stiffeningeffect is
al so observed for dams in canyons of different shapes, al though the effectdecreases
as the crest l ength/dam height ratio increases (Figure 7.22a).2 3 4 5Mode number, n
B ~ ~ .~ /~ 0, . '\ Rectangul ar~". Semicyl indrical ~",Wide trapezoidal '",Narrow
trapezoidal Triangul ar canyonOL-l .-_---'--__l .-_---'--_---'.L.J(Tn )3_0(Tn)2_00.52.0
4.0 6.0UHOL-_...l -__L-_--'-__...l -__L-_-'0.0(T1)3_0(71)2-01.5Figure 7.22 Effect of
three-dimensional boundary effects on (a) fundamental frequencies for different
crest l ength/dam height ratios, and (b) natural periods for firstfive modes of
earth dams (UH =2) in canyons of different shape. (After Gazetas, 1987.Used with
kind permission from El sevier Science, Ltd, The Boul evard, Langford Lane,Kidl ington
OX5 1GB UK.)294 Ground Response Anal ysis Chap. 77.4.4 Comparison of Three-
Dimensional DynamicResponse Anal ysesThe rel ative capabil ities and l imitations of
the various approaches to three-dimensional dynamic response anal ysis are very
simil ar to those of their two-dimensional counterparts.Of special interest for
three-dimensional probl ems, however, are the rel ative computational efficiencies of
the various approaches.The shear beam approach
is simpl e and consequentl y very efficient for earth damprobl ems, but it makes
quite restrictive assumptions about the mode of dam response. Thenonl inear finite-
el ement approach offers the greatest potential , in terms of its abil ity to compute
permanent displ acements and variations in effective stress, but it requires
sol ution ofthe incremental equations of motion at every time step. For l arge three-
dimensional probl ems,this can invol ve very l arge systems of simul taneous equations
and many time steps.The equival ent l inear approach, through the use of the
cutofffrequency and interpol ation ofthe transfer functions, requires l arge matrix
computations at onl y a rel ativel y smal l numberof frequencies. The equival ent l inear
approach can therefore be more computational l y efficientthan the nonl inear
approach.7.5 SOIL-STRUCTURE INTERACTIONAl l of the ground response anal yses
described thus far have considered cases in which nostructures are present. Ground
motions that are not infl uenced by the presence of structuresare referred to es
free-fiel d motions. When a structure founded on sol id rock is subjected toan
earthquake, the extremel y high stiffness of the rock constrains the rock motion to
be verycl ose to the free-fiel d motion. Structures founded on rock are considered to
be fixed-basestructures. Computation of their response is rel ativel y simpl e using
the methods describedin Appendix B.On the other hand, the same structure woul d
respond differentl y if supported on a softsoil deposit. First, the inabil ity of the
foundation to conform to the deformations of the freefiel dmotion woul d cause the
motion of the base of the structure to deviate from the freefiel dmotion. Second,
the dynamic response of the structure itsel f woul d induce deformationof the
supporting soil . This process, in which the response of the soil infl uences the
motionof the structure and the response of the structure infl uences the motion of
the soil , is referredto as soil -structure interaction.Soil -structure interaction
has l ittl e effect on the dynamic response of many structuresand foundation systems.
In other cases, however, its effects can be significant. Whether thenegl ect of its
effects is conservative or unconservative depends on the detail s of the probl emat
hand and must be eval uated on a case-by-case basis. The anal ysis of soil -structure
interactioncan be quite compl icated; a detail ed treatment is beyond the scope of
this book. Formore detail s, the work of Wol f (1985) and the Committee on Nucl ear
Structures and Material s(1979) are hel pful . The purpose of the fol l owing sections
is simpl y to provide a basicdescription of soil -structure interaction phenomena, to
provide insight into the conditionsunder which it can be significant, and to
introduce some of the methods that are commonl yused to eval uate its effects.Sec.
7.5 Soil -Structure Interaction 2957,,5.1 Il l ustration of Soil -Structure Interaction
EffectsA simpl e anal ysis is sufficient to il l ustrate the most important effects of
soil -structure interaction.Fol l owing the approach of Wol f (1985), consider the case
of the simpl e SDOF systemmounted on a rigid, massl ess, L-shaped foundation (Figure
7.23a) supported on anel astic soil deposit. The structure is characterized by its
mass, m, stiffness, k, and dampingcoefficient, c. If the material supporting the
foundation is rigid, the natural frequency of theresul ting fixed-base system woul d
depend onl y on the mass and stiffness of the structure,that is,Wo = ~and the
hysteretic damping ratio woul d be~ = cwo2k(7.69)(7.70)If the supporting material is
compl iant, however, the foundation can transl ate and rotate.The stiffness and
damping characteristics of the compl iant soil -foundation system can berepresented
by the transl ational and rotational springs and dashpots shown in Figure 7.23b.The
foundation dashpots represent two sources of damping: material damping caused by
inel astic behavior of the soil supporting the foundation, and radiation damping
that occursas dynamic forces in the structure cause the foundation to deform the
soil , producing stresswaves that travel away from the foundation. The amount of
material damping wil l dependon the l evel of strain induced in the soil ; if the
strains are high, material damping can be substantial ,but if they are l ow, the
material damping may be negl igibl e. In contrast, radiationdamping is a purel y
geometric effect that exists at l ow as wel l as high strain ampl itudes. Fortypical
foundations, radiation damping is often much greater than material damping.The
total displ acements of the mass and the base of the structure can be spl it into
theirindividual components (Figure 7.23c): k.. ..Ub(t)(a)G,v ... ..Ub(t)(b)h(c)
Figure 7.23 Compl iant base model with one dynamic degree of freedom: (a) SDOFsystem
on an el astic soil deposit; (b) ideal ized discrete system in which compl iance of
base is represented by transl ational and rotational springs and dashpots; (c)
componentsof motion of base and mass.296 Ground Response Anal ysis Chap_ 7(7.71a)
(7.71b)where the symbol s are defined as shown in Figure 7.23c. Negl ecting material
damping inthe soil (~g = 0), the horizontal force imposed on the soil by the
foundation woul d be(7.72)where the subscript x refers to the horizontal direction
for ~g =0 conditions. For harmonicexcitation at frequency CD, material damping can
be introduced by the use of a compl ex stiffness,so that .Since PhPh = kx(l + i2~x +
i2~g)uo (7.73)khUo+ chuO', the horizontal stiffness and damping coefficients arekh
= k; (7.74a)(7.74b)The first term on the right side of equation (7.74b) corresponds
to radiation damping, andthe second to material damping. If the structure was rigid
(k = 00)and the foundation unabl eto rotate (k; = 00), the natural frequency for
transl ational vibration woul d beCDh = !!i; (7.75)Repeating the same process for the
rocking mode of vibration produces(7.76a)(7.76b)(7.77)where the subscript 8 denotes
the absence of material damping. If the structure was rigid (k =00)and the
foundation unabl e to transl ate (kh =00),the natural frequency for rocking woul d be
CDr = J: ~2Insight into the soil -structure interaction probl em can now be gained by
devel oping anequival ent SDOF system (i.e., a SDOF system that responds in
essential l y the same way asthe system of Figure 7.23). Using the subscript e to
describe the properties of this equival entsystem, the equation of motion (for
harmonic motion) can be written as(- mCD2 + iCDCe + ke)u = mCD2Ubwhere Ub is the
equival ent seismic input motion. Note that the mass is the same for theequival ent
and actual model s. For the equival ent systemSec. 7.5 Soil -Structure Interaction 297
~e(7.79)Ce(O2keThe natural frequency of the equival ent model , (Oe, is the frequency
at which the response ofthe equival ent system goes to infinity for ~e = O. This
occurs when1 111 = -+-+-(02 (02 (02 (02e o h ror at(0 -(00 e - (7.78)Jl + k/kh + kh
2/krEquation (7.78) indicates that the natural frequency of the equival ent system
is al waysl ower than that of the fixed-base structure. Inother words, an important
effect of soil -structureinteraction is to reduce the natural frequency of the soil -
structure system to a val uel ower than that of the structure under fixed-base
conditions.By negl ecting second-order damping ratio terms, the equival ent
hysteretic dampingratio can be expressed as(02 l (02J (02 (02 ~e = --1~ + 1- --1 ~g
+ --1~x + --1~e(00 (00 (Oh (OrEquation (7.79) indicates that the damping ratio of
the equival ent system wil l , for typical soil s and foundations, be l arger than the
damping ratio of the structure itsel f. Consequentl y,another important effect of
soil -structure interaction is to increase the effective dampingratio to a val ue
greater than that of the structure itsel f.For the fixed-base case, no transl ation
or rotation of the base is possibl e, but the basetransl ation of the equival ent
system can be shown to be(7.80)and the base rotation(7.81)Then the motion of the
mass rel ative to the free-fiel d motion is given by the sum of the basedispl acement,
Uo, the displ acement of the top of the rod due to rotation of the base, he, andthe
displ acement due to distortion of the structure, u,2 [ 1 ~ ~ .( 1 1) 2~xi 2~ei]
u+uo+he = (00 -+2(..,-..,)1 --- ------ U(02 g (02 (02 (02 (02e e 0 h r(7.82)298
Ground Response Anal ysis Chap. 7The effects of soil -structure interaction are
easil y il l ustrated in terms of the fol l owingdimensionl ess parameters: Stiffness
ratio: Swahwhere Vs is the soil .shear wave vel ocityVsSl enderness ratio: h h where a
is characteristic l ength of the foundationa(radius for circul ar shape)Mass ratio:
ffim where p is the mass density of the soil pa 34.6 22 -- pv sa-V0.4 4Ce = -l -pvsa-
VLarge val ues of the stiffness ratio correspond to situations where a rel ativel y
stiff structurerests on a rel ativel y soft soil . The fixed-base condition is
real ized at zero stiffness ratio.Actual foundation stiffness and damping
coefficients are frequency dependent. Toil l ustrate the effects of soil -structure
interaction, however, the fol l owing simpl ified, frequency-independent expressions
can be used to estimate the stiffness and damping coefficientsof a rigid circul ar
footing of radius ak _ 8Gax - 2-v8Ga3ke = 3(l -v)The graphs in Figures 7.24 and 7.25
show the infl uence of soil -structure interaction on thenatural frequency, damping
ratio, and displ acement characteristics of the equival ent SDOFsystem. Comparing the
response characteristics of the equival ent system with the fixedbasesystem
il l ustrates the effects of soil -structure interaction.5 10I0.5 1(b),./IiI,~/I;/.1//
~~/ / ..7: ;: 0;;'0 _0... - No radiationdamping~eo0.05 0.10.25·0.20·0.10 .0.15 .0.05·
0.5 1 5 10(a)---m =0.5-----m =1··············..·m = 3----m =5----m=10"~'~.'\."\''""
.• • . \ \\..• ..• . " " "\ \ .\\ ...... '\"0 \'. \ .....• ' \: \ "" -'0 ,,,~-, '"~~o .
'. '-,-,,-,.,---,---,--,--,~c-: -----c---c--,..,.-,-,-,.,.,-----0.05 0.10.40.8·0.2·
0.6·Figure 7.24 Effect of stiffness ratio and mass ratio on (a) natural frequency,
and (b)damping ratio of soil -structure systems (h = I, v = 0.33, ~ = 0.025, ~g =
0.05). (AfterWol f, 1985.)Sec. 7.5 Soil -Structure Interaction 299Umax[mm]20000 / 2n
= 3 Hz- - - 000 /2n = 4 Hz................ 000 l Zt: = 5 Hz1 5 10(a)0.5so.L-~
===: ~~~~ s0.110 .(u + Uo + he)max[mm]Figure 7.25 Response of equival ent
soil structuresystem to artificial time history(Fz = I, iii = 3, v = 0.33, ~ =
0.025, ~g =0.05): (a) maximum structural distortion;(b) maximum displ acement of
mass rel ativeto free fiel d. (After Wol f, 1985.)1 5 10(b)0.5III/ .' ...(.......
/ ...• • ..o .=...=..==..===..='..-=-..=r..=-.~.~~..~~.: -~: : : : ~....~...~~~~__ S0.12~)
.50·l ei·100Figure 7.24a shows how the natural frequency of the equival ent SDOF
system dropsbel ow that of the fixed-base system as the stiffness ratio increases.
The effects of soil -structureinteraction on the natural frequency is smal l at l ow
stiffness ratios, i.e., when the stiffnessof the soil (as refl ected in the shear
wave vel ocity) is l arge rel ative to the stiffness of thestructure. For the fixed-
base condition (S =0) , the natural frequency of the equival entsystem is equal to
the fixed-base natural frequency. Figure 7.24b il l ustrates the infl uence ofsoil -
structure interaction on the damping ratio of the equival ent SDOF system. For the
fixed-base condition, the damping ratio of the equival ent system is equal to the
structural damping ratio, but as the stiffness ratio increases, the effects of
radiation and soil dampingbecome more apparent. At high stiffness ratios,
structural damping represents onl y a smal l part of the total damping of the system.
The effects of soil -structure interaction on displ acements is il l ustrated in Figure
7.25.The indicated maximum responses are for an artificial input motion that
produced an NRCresponse spectrum scal ed to amax=1.0g. In this case, the effects of
soil -structure interactionwere to reduce the maximum structural distortion, Umax,
by an amount that increased withincreasing stiffness ratio and to increase the
overal l displ acement (rel ative to the free fiel d)by an amount that increased with
increasing stiffness ratio. Thus soil -structure interactiontends to reduce the
demands on the structure, but because the foundation can transl ate androtate,
increase the overal l displ acement. These effects can be important for tal l , sl ender
300 Ground Response Anal ysis Chap. 7structures or for cl osel y spaced structures
that may be subject to pounding when rel ative displ acementsbecome l arge.7.5.2
Methods of Anal ysisWhil e the preceding anal ysis il l ustrated several important
effects of soil -structure interaction,it is not suitabl e for the detail ed anal ysis
of practical soil -structure interaction probl ems.Methods for the anal ysis of soil -
structure interaction can be divided into two maincategories: direct methods and
mul tistep methods.7.5.2.1 Direct MethodIn the direct method, the entire soil -
foundation-structure system is model ed and anal yzedin a singl e step. As il l ustrated
in Figure 7.26, free-fiel d input motions are specifiedal ong the base and sides of
the model and the resul ting response of the interacting system iscomputed (for a
finite-el ement model ) from the equations of motion[M]{n} + [K*]{u} = -[M] { urr(t)}
(7.83)where {tiff (t)} are the specified free-fiel d accel erations at the boundary
nodal points. Theuse of the direct method requires a computer program that can
treat the behavior of both thesoil and the structure with equal rigor...... .....
UffUff .....Figure 7.26 Direct method of soil -structure interaction anal ysis.
Entire probl em ismodel ed and response to free-fiel d motion appl ied at boundaries is
determined in asingl e step.7.5.2.2 MUl tistep MethodMul tistep methods use the
principl e of superposition to isol ate the two primary causesof soil -structure
interaction: the inabil ity of the foundation to match the free-fiel d deformationand
the effect of the dynamic response of the structure-foundation system on the
movementof the supporting soil . Because they rel y on superposition, they are
l imited to theanal ysis of l inear (or equival ent l inear) systems.I{inematic
Interaction. In the free fiel d, an earthquake wil l cause soil displ acementsin both
the horizontal and vertical directions. If a foundation on the surface of,or
embedded in, a soil deposit is so stiff that it cannot fol l ow the free-fiel d
deformation pattern,its motion wil l be infl uenced by kinematic interaction, even if
it has no mass. ForSec. 7.5 Soil -Structure Interaction 301
....-.........,s ...............(a) (b)-(--y\}((c)Figure 7.27 Kinematic interaction
with free-fiel d motions indicated by dashed l ines: (a) fl exural stiffness of surface
foundation prevents it from fol l owing vertical componentof free-fiel d displ acement;
(b) rigidity of bl ock foundation prevents it from fol l owinghorizontal component of
free-fiel d displ acement: (c) axial stiffness of surface foundationprevents
immediatel y underl ying soil from deforming incoherentl y.exampl e, the fl exural
stiffness of the massl ess mat foundation in Figure 7.27a prevents itfrom fol l owing
the horizontal l y varying vertical component of the free-fiel d motion. Therigidity
of the massl ess embedded foundation in Figure 7.27b keeps it from fol l owing the
vertical l y varying horizontal free-fiel d motion. The axial stiffness of the sl ab in
Figure7.27c prevents devel opment of the incoherent free-fiel d motion. In each of
these cases, themotion of the foundation is infl uenced by kinematic interaction.
Kinematic interaction,however, is not present in al l probl ems. If the surface
foundation of Figure 7.27a was subjectedto vertical l y propagating s-waves
(horizontal particl e motion onl y), it woul d notrestrict the movement of the soil ;
consequentl y, no kinematic interaction woul d exist. Kinematicinteraction wil l occur
whenever the stiffness of the foundation system impedes devel opmentof the free-
fiel d motions.Kinematic interaction can al so induce different modes of vibration in
a structure.Consider the embedded foundation of Figure 7.28a. When subjected to
vertical l y propagatings-waves with a wavel ength equal to the depth of embedment, a
net overturning momentcan be appl ied to the foundation, thereby causing the
foundation to rock as wel l as transl ate,even though the free-fiel d motion is purel y
transl ational . At a different frequency (Figure7.28b), the wavel ength may be such
that rotation is inhibited. Horizontal l y propagatingwaves can, in a simil ar manner,
induce torsional vibration of the foundation.The deformations due to kinematic
interaction al one can be computed by assumingthat the foundation has stiffness but
no mass. The equations of motion for this case are(7.84)(a)~ff .... ....(b)Figure
7.28 Excitation of rocking vibrations in an embedded foundation by vertical l y
propagating s-waves: (a) at certain frequencies, the wavel ength is such that
unbal ancedoverturning moments cause rocking; (b) at other frequencies (and
wavel engths), rockingmay be suppressed.302 Ground Response Anal ysis Chap. 7dn, , -
Massl ess Inertial forces appl ier r -.- - <r -.- - structure .- to
structure/foundatio -,. -e- - - - ....... - -...tL-: _-L-. -- 0 "'.''-r- -~ ~Fixed
boundary(a) (b)Figure 7.29 (a) Kinematic interaction anal ysis and (b) inertial
interaction anal ysis.Mass of structure in inertial interaction anal ysis shown as
being l umped at the center ofthe structure.where [Msoil ] is the mass matrix
assuming that the structure and foundation are massl ess.Equation (7.84) describes
the probl em il l ustrated in Figure 7.29a. Equation (7.84) is sol vedfor {UKI}, which
is referred to as the foundation input motion.Inertial Interaction. The structure
and foundation do have mass, however,and this mass causes them to respond
dynamical l y. If the supporting soil is compl iant, theforces transmitted to it by
the foundation wil l produce foundation movement that woul d notoccur in a fixed-base
structure. The effects of soil compl iance on the resul ting response aredue to
inertial interaction.The deformations due to inertial interaction can be computed
from the equations ofmotion(7.85)where [Mstructure] is the mass matrix assuming
that the soil is massl ess. Note that the rightside of equation (7.85) represents
the inertial l oading on the structure-foundation system.This inertial l oading
depends on the base motion and the foundation input motion, whichrefl ects the
effects of kinematic interaction. In the inertial interaction anal ysis, the
inertial l oading is appl ied onl y to the structure; the base of the soil deposit is
stationary. Equation(7.85) corresponds to the probl em il l ustrated in Figure 7.29b.
Combination of I<inematic and Inertial Interaction. The kinematicinteraction
anal ysis produces the motion of the massl ess foundation-structure system (rel ative
to the base) due to kinematic interaction. This motion is combined with the base
motion to produce the total kinematic motion of the foundation-structure system.
When theinertial l oading that resul ts from this kinematic motion is appl ied to the
foundation-structuresystem resting on massl ess soil , equation (7.85) al l ows
computation of the rel ative (tothe total kinematic) motion. Adding equations (7.84)
and (7.85) gives[Msoil ]{UKI} + [M]{urr} + [K*] ({uKI } + {urr}) (7.86)= - ([Msoil ]
+ [Mstructure]) u b } - [Mstructure]{u KI }Since {uKd + {urr} ={u} and [Msoil ] +
[Mstructure] =[M], equation (7.86) is equival ent tothe original equations of motion
[M]{u} + [K*]{u} = -[M]{iibCt)}Sec. 7.6 Summary 303which proves that the sol ution
to the entire soil -structure
interaction probl em is equal to thesum of the sol utions of the kinematic and
inertial interaction anal yses. Therefore, the mul tistepprocedure can be summarized
as fol l ows: 1. A kinematic interaction anal ysis, in which the foundation-structure
system isassumed to have stiffness but no mass, is performed. This anal ysis
produces the foundationinput motion.2. The foundation input motion is used to appl y
inertial l oads to the structure in an inertial interaction anal ysis, in which the
soil , foundation, and structure are al l assumed tohave stiffness and mass.If the
foundation itsel f is rigid, t~e soil can be repl aced by a set of equival ent springs
and dashpots in the inertial interaction anal ysis. The inertial interaction
anal ysis can then beperformed by appl ying inertial forces to the masses of the
structure (Figure 7.30a) or byappl ying the input motion to the supports (Figure
7.30b); the two methods are mathematical l yequival ent. Considerabl e research,
invol ving anal ytical , numerical , and experimental model ing, has produced a variety
of techniques for the eval uation of spring and dashpotconstants for foundation
systems. Gazetas (1991) provided a very useful , practical series ofcharts and
tabl es for estimation of spring and dashpot coefficients for a variety of
foundationtypes and soil conditions. For more compl icated soil and foundation
conditions, computerprograms such as DYNA4 (Novak et al ., 1993) al l ow computation
of compl etefoundation stiffness and damping matrices.DbFigure 7.30 Equival ent
formul ations of inertial interaction anal ysis for structures withrigid foundation:
(a) inertia forces appl ied to each el ement; (b) foundation motionappl ied through
frequency-dependent springs and dashpots (not shown).7.6 SUMMARY1. Anal ysis of the
response of soil deposits and earthen structures to earthquake motionsis one of the
most important aspects of geotechnical earthquake engineering practice.Ground
response probl ems are often grouped into categories for which one-, two-,and three-
dimensional ground response anal yses are appropriate.304 Ground Response Anal ysis
Chap. 72. Ground response anal yses are general l y performed by one of two methods:
equival entl inear anal ysis or nonl inear anal ysis. Equival ent l inear anal yses are
l inear anal yses inwhich the soil stiffness and damping characteristics are adjusted
until they are compatibl ewith the l evel of strain induced in the soil . Nonl inear
anal yses actual l y consider thenonl inear inel astic stress-strain behavior of soil s
by integrating the equations of motionin smal l time steps. Equival ent l inear
anal yses are general l y more computational l y efficientthan nonl inear anal yses.3.
One-dimensional ground response anal yses are based on the assumption that theground
surface and al l material boundaries bel ow the ground surface are horizontal and
extend infinitel y in al l l ateral directions. Al though these assumptions can never
bestrictl y satisfied, they are satisfied sufficientl y for engineering purposes at
many sites.4. The equival ent l inear approach makes use of transfer functions to
rel ate parameters ofinterest (such as ground surface accel eration or vel ocity) to
known parameters (suchas bedrock accel eration). The nature of a transfer function
is infl uenced by the thickness,stiffness, and damping characteristics of each soil
l ayer. The transfer function isal so infl uenced by bedrock properties.S. During an
earthquake, stress waves that reach the ground surface are perfectl yrefl ected
(i.e., with no l oss of energy) to travel downward toward bedrock. When adownward
travel ing wave reaches bedrock, part of the wave wil l be refl ected, andunl ess the
rock is perfectl y rigid, part wil l be transmitted to continue travel ing downward
through the rock. If the el astic bedrock is rel ativel y homogeneous, the transmitted
wave effectivel y removes energy from the soil deposit and causes its response tobe
l ower than it woul d have been for the case of rigid bedrock. This phenomenon
representsa form of radiation damping.6. Because the equival ent l inear approach is
fundamental l y l inear, the response at anytwo points in a soil deposit are uniquel y
rel ated to each other. As a resul t, an objectmotion can be specified at any point
in the soil deposit and the corresponding motioncomputed at any other point. This
feature al l ows use of a motion recorded on the surfaceof a soil deposit to compute
the bedrock motion that woul d have caused it. Thisuseful operation, known as
deconvol ution, does not provide a unique sol ution whenperformed with nonl inear
anal yses.7. Nonl inear anal yses integrate the equations of motion in smal l time
steps. At the beginningof each time step, the stress-strain rel ationship is
referred to to obtain the appropriatesoil properties to be used in that time step.
By this method, a nonl inear inel asticstress-strain rel ationship can be fol l owed in
a set of smal l incremental l y l inear steps.8. Both equival ent l inear and nonl inear
approaches have been used successful l y forground response anal ysis. Their use and
interpretation, however, requires knowl edgeof their underl ying assumptions,
understanding of their operation, and recognition oftheir l imitations. Neither can
be considered mathematical l y rigorous or precise, yettheir accuracy is not
inconsistent with the variabil ity in soil conditions, uncertainty insoil
properties, and scatter in the experimental data upon which many of their input
parameters are based.9. Two- and three-dimensional ground response anal yses are
usual l y performed usingdynamic finite-el ement anal yses. These anal yses can be
performed using equival entChap. 7 Homework Probl ems 305l inear or nonl inear
approaches. Both equival ent l inear and nonl inear dynamic finiteel ement anal yses are
anal ogous to their one-dimensional counterparts, and many ofthe same advantages and
l imitations appl y to each.10. Anal ysis of the response of certain two- and three-
dimensional earth structures, suchas earth dams and embankments, can be greatl y
simpl ified by the shear beamapproach. The shear beam approach al l ows two- and
three-dimensional probl ems tobe ideal ized as equival ent one-dimensional probl ems.
11. Ground motions that are not infl uenced by the presence of structures are cal l ed
freefiel dmotions. When structures are present, they interact with the soil through
a processreferred to as soil -structure interaction. Soil -structure interaction has
l ittl e effecton the response of some systems and a l arge effect on the response of
others. Itseffects are most significant for stiff and/or heavy structures supported
on rel ativel ysoft soil s. For soft and/or l ight structures founded on stiff soil s,
soil -structure interactioneffects are general l y smal l .][2. In general , soil -
structure interaction wil l cause the natural frequency of a soil -structuresystem to
be l ower than the natural frequency of the structure itsel f. Al so, radiationdamping
wil l general l y cause the total damping of a soil -structure system to begreater than
that of the structure itsel f. Because of these effects, soil -structure interaction
tends to reduce the demands on the structure, but because the foundation can
transl ate and rotate, can increase the overal l displ acement.13. Soil -structure
interaction is caused by two phenomena: the inabil ity of the foundationto match the
free-fiel d deformation (kinematic interaction) and the effect of thedynamic
response of the structure-foundation system on the movement of the supportingsoil
(inertial interaction). Kinematic interaction can induce modes of deformation(e.g.,
rocking and torsion) that are not present in a free-fiel d motion. Inertial
interaction occurs when the forces transmitted to the soil by the dynamic response
ofthe structure produce foundation movements that woul d not occur in a fixed-base
structure. The effects of inertial interaction are usual l y more pronounced than the
effects of kinematic interaction.14. For l inear or equival ent l inear anal yses, the
effects of kinematic interaction and inertial interaction can be separated. The
effects of kinematic interaction are first determinedand then used as input to an
inertial interaction anal ysis. Combining the resul tsof both anal yses al l ows the
overal l response to be determined.HOMEWORK PROBLEMS7.1 For the case of a uniform
l ayer of undamped soil overl ying rigid bedrock, devel op a transferfunction that
rel ates shear stress, 't (z = HI2, t), to the bedrock accel eration, ii b (t), Pl ot
themodul us of the transfer function from kH =0 to kH =2n: .7.2 An accel eration
reduction factor can be defined as the ratio of the peak accel eration at depth,z,
to the peak ground surface accel eration, i.e.Umax(Z)rd = -: -: -------'--Umax(Z = 0)
306 Ground Response Anal ysis Chap. 7For the case of a uniform l ayer of undamped
soil overl ying rigid bedrock, devel op an expressionfor the reduction factor as a
function of the thickness and shear wave vel ocity of the soil l ayer, and the
frequency of the input motion.7.3 Pl ot the reduction factor determined in Probl em
7.2 with depth for a 15 m thick soil depositwith a shear wave vel ocity of 300
rnfsec at input motion frequencies of 0.5 Hz, 1.0 Hz, 2.0 Hz,3.0 Hz, and 5.0 Hz.7.4
Derive a transfer function that rel ates the displ acement of the ground surface to
the displ acementof bedrock for the site il l ustrated bel ow.(a) Define the compl ex
wave numbers, kj and ki .(b) Define the compl ex impedance ratio, u" .(c) Express
the transfer function in terms of hI, hz, kj, ki, and o".FigureP7.47.s The shear
modul us of the soil within the earth dam shown bel ow increases l inearl y with depth.
The shear wave vel ocity of the soil just above the base of the dam is 800rnfsec.
Compute thefirst two natural frequencies of the dam.7.6 l fthe dam section shown in
Probl em 7.5 represented the maximum section of a dam in a semicyl indrical channel ,
estimate the first two natural frequencies of the dam.Chap. 7 Homework Probl ems
3077.7 The SDOF system shown bel ow is supported by a 10 ft diameter circul ar
footing. The mass ofthe footing is smal l compared to the mass of the structure.
Compute the damped natural periodof the soil -structure system: (a) assuming fixed
base conditions,(b) assuming that the footing can transl ate horizontal l y (but not
rotate), and(c) assuming that the footing can both transl ate and rotate.Figure P7.7
8Local Site Effectsand Design Ground Motions8.1 INTRODUCTIONThe infl uence of l ocal
geol ogic and soil conditions on the intensity of ground shaking andearthquake
damage has been known for many years. MacMurdo (1824) noted that "buil dingssituated
on rock were not by any means so much affected ... as those whose foundationsdid
not reach to the bottom of the soil " in the 1819 earthquake in Cutch, India. In his
report on the 1857 Neapol itan earthquake, Mal l et (1862) noted the effect of l ocal
geol ogicconditions on damage. Wood (1908) and Reid (1910) showed that the intensity
of groundshaking in the 1906 San Francisco earthquake was rel ated to l ocal soil and
geol ogic conditions.Gutenberg (1927) devel oped site-dependent ampl ification factors
from recordings ofmicroseisms at sites with different subsurface conditions. Since
these earl y observations,the effects of l ocal site conditions on ground motions
have been il l ustrated in earthquakesaround the worl d. More recentl y, the
avail abil ity of strong-motion instruments has al l owedl ocal site effects to be
measured quantitativel y in recent years.Local site effects pl ay an important rol e
in earthquake-resistant design and must beaccounted for on a case-by-case basis.
This is usual l y accompl ished by the devel opment ofone or more design ground motions
(i.e., motions that refl ect the l evel s of strong motionampl itude, frequency
content, and duration that a structure or facil ity at a particul ar siteshoul d be
designed for). The devel opment of site-specific design ground motions invol ves308
Sec. 8.2 Effects of Local Site Conditions on Ground Motion 309concepts presented in
al l of the preceding chapters of this book; it is one of the most importantaspects
of geotechnical earthquake engineering.Despite considerabl e evidence, the existence
of l ocal site effects was a matter of somedebate in past years. Indeed, provisions
specifical l y accounting for l ocal site effects did notappear in buil ding codes
until the 1970s. This chapter presents theoretical , instrumental ,and historical
evidence for the existence of l ocal site effects. It discusses procedures that are
commonl y used for the devel opment of site-specific design ground motions and
reviews themanner in which l ocal site effects are treated in the specification of
design ground motionsby contemporary buil ding codes and standards.8.2 EFFECTS OF
LOCAL SITE CONDITIONS ON GROUND MOTIONLocal site conditions can profoundl y
infl uence al l of the important characteristics-ampl itude,frequency content, and
duration-of strong ground motion. The extent of their infl uencedepends on the
geometry and material properties of the subsurface material s, on sitetopography,
and on the characteristics of the input motion. The nature of l ocal site effectscan
be il l ustrated in several ways: by simpl e, theoretical ground response anal yses, by
measurementsof actual surface and subsurface motions at the same site, and by
measurementsof ground surface motions from sites with different subsurface
conditions.8,,2.1 Evidence from Theoretical Ground ResponseAnal ysesThere are
important theoretical reasons why ground surface motions shoul d be infl uencedby
l ocal site conditions. At most sites the density and s-wave vel ocity of material s
near thesurface are smal l er than at greater depths. If the effects of scattering
and material dampingare negl ected, the conservation of el astic wave energy requires
that the fl ow of energy(energy fl ux, pvsu2 ) from depth to the ground surface be
constant. Therefore, since p andVs decrease as waves approach the ground surface,
the particl e vel ocity, U, must increase.The characteristics of l ocal soil deposits
can al so infl uence the extent to which groundmotion ampl ification wil l occur when
the specific impedance is constant. The basis for suchampl ification can be
il l ustrated anal ytical l y using simpl e, theoretical ground response anal yses.
Consider, for exampl e, the two soil deposits shown in Figure 8.1; their geometries
areidentical , but one is considerabl y stiffer than the other. If each soil is
assumed to be l inearl y40 ftY= 110 Ib/ft3Vs = 400 It/secS=0.10 40 ItY= 110 Ib/ft3Vs
=1600 ft/secS=0.10(a) (b)Figure 8.1 Two hypothetical soil deposits overl ying rigid
bedrock: (a) site A; (b) siteB. Soil s are identical , except the s-wave vel ocity of
the soil at site B is four times greaterthan that at site A.310 Local Site Effects
and Design Ground Motions Chap. 8el astic and bedrock to be rigid, the ampl ification
functions of each site [from equation (7.13)]are as il l ustrated in Figure 8.2.
Cl earl y, the softer site (site A) wil l ampl ify l ow-frequency(l ong-period) bedrock
motions more than the stiff site (site B); the reverse woul d be observedfor high-
frequency (short-period) motions. Since earthquakes produce bedrock motion overa
range offrequencies, some components of an actual bedrock motion wil l be ampl ified
morethan others, as il l ustrated in Exampl es 7.1 to 7.5. For the more real istic
condition of el asticbedrock, the nature of the l ocal site ampl ification wil l be
infl uenced by the specific impedanceof the bedrock. Consequentl y, any description
ofl ocal site conditions shoul d incl ude thedensity and stiffness of the bedrock. For
exampl e, the harder crystal l ine bedrock found inmuch of the eastern United States
woul d be expected to produce ampl ification factors about50% higher than those
associated with the softer rock conditions typical l y found in Cal ifornia(Jacob,
1991) for equival ent soil conditions.760ts 5.l Qc 4 0~~ 3D.. 2 E-c00 5 10 15 20
Frequency (Hz)Figure 8.2 Ampl ification functions for sites A and B. Note that the
softer soil at site Awil l ampl ify l ow-frequency input motions much more strongl y
than wil l the stiffer soil sof site B. At higher frequencies, the opposite behavior
woul d be expected.8.2.2 Evidence from Measured Ampl ification FunctionsThe ideal ized
assumptions of simpl e one-dimensional ground response anal yses (uniformmaterial s,
horizontal l ayering, vertical l y propagating s-waves, etc.) produce smooth
ampl ificationfunctions such as those shown in Figure 8.2. Since these conditions
rarel y exist inthe fiel d, actual ampl ification functions are not so smooth.
Interpretation of strong motion data from sites where both surface and subsurface
instruments had been instal l ed al l ows actual ampl ification functions to be computed
(e.g.,Joyner et al ., 1976; Johnson and Sil va, 1981; Chang et al ., 1986). The strong
ampl ificationat the natural frequencies of the soil deposit shown in Figure 8.3
cl earl y il l ustrates theimportance of l ocal soil conditions on ground response. The
site consists of various soil s ofrel ativel y uniform shear wave vel ocity overl ying
bedrock; consequentl y, the frequencydependence of the actual ampl ification function
is qual itativel y simil ar to that predicted bysimpl e ground response anal yses. For
sites with more compl icated subsurface conditions, orSec. 8.2 Effects of Local Site
Conditions on Ground Motion 311(a)5 10 15 20Frequency (Hz)(b)10 o~~~Q.(f)Sil ty cl ay
sandShal eSil ty cl ay sandSil ty cl ay sand gravel Cl aySil ty cl ay gravel Cl aySil ty cl ay
gravel Sandy cl ayCl ayey sand gravel Cl ay sandWave vel ocity (m/sec)1000 2000~ 0
Location of accel erometers510Vp15][ 20 Vs.l =1~ 25()30354045Figure 8.3 (a)
Subsurface profil e at Richmond fiel d station (after Johnson and Sil va,1981). (b)
Surface-bedrock ampl ification functions from response measured during theBriones
Hil l s (ML =4.3) earthquake (after Sil va, 1988). Fourier spectra were smoothedover a
I-Hz window prior to spectral ratio computation.for stronger earthquakes in which
soil nonl inearity may become significant, the abil ity ofsimpl e ground response
anal yses to predict the irregul ar peaks and val l eys of actual ampl ification
functions decreases. The effects of soil nonl inearity al so cause ampl ification
functionsfrom strong motions to differ from those of weak motions (e.g., Aki,
1993).Exampl e 8.1Compute the spectral ratio of the Gil roy No.1 (rock) and Gil roy
No.2 (soil ) motions.Sol ution The actual transfer function rel ating the Gil roy No. I
(rock) and Gil roy No.2 (soil )motions can be computed from the Fourier series ofthe
individual motions. The process can beperformed in three steps: 1. Compute the
Fourier series for the Gil roy No.1 (rock) motion. The Fourier ampl itudespectrum of
this motion was original l y shown in Figure 3.13a, and is repeated in FigureE8.1a.
Al so shown is a smoothed version of the spectrum (the smoothed version wassmoothed
numerical l y, but not as extensivel y as required for eval uation of the predominant
period in Exampl e 3.3).2. Compute the Fourier series for the Gil roy No.2 (soil )
motion. Raw and smoothed Fourierampl itude spectra for this motion are shown in
Figure E8.l b.3. Compute the transfer function as the ratio of the Fourier
ampl itudes of the Gil roy No.2(soil ) motion to the Fourier ampl itudes of the Gil roy
No.1 (rock) motion. The resul t,shown in Figure E8.l c, shows that l ow-frequency
components ofthe Gil roy No.1 (rock)motion were ampl ified by the soil s underl ying
the Gil roy No.2 (soil ) station; highfrequencycomponents were attenuated.312 Local
Site Effects and Design Ground Motions Chap. 810 15Frequency (Hz)Gil roy No 30~: 'I
FAS 20··(a)(g-sec) ~O ~~'V, 0.: \Aio 5 20 253°1~: ': ~'Gil rOYNO'21 FAS 20 -- .-(b)
(g-sec) 1: ~\I\l q;;VQ=0
~--o 5 10 15 20 25Frequency (Hz)5 10 15Frequency (Hz)Figure E8.120 258.2.3
Evidence from Measured Surface MotionsFurther evidence of the importance of l ocal
site conditions can be gained by comparingground surface motions measured at
different sites. For exampl e, recordings of ground motionat several l ocations in
San Francisco were made during a nearby M ""5.3 earthquake in 1957.Variations in
ground motion, expressed in terms of peak horizontal accel eration and response
spectra, are shown al ong with variation in soil conditions al ong a 4-mil e section
through thecity in Figure 8.4. Ground surface motions at the rock outcrops (Market
and Guerrero, Masonand Pine, Harrison and Main) were quite simil ar, but the
ampl itude and frequency content ofthe motions at sites underl ain by thick soil
deposits were markedl y different.Simil ar effects have been observed in many other
earthquakes. From the standpoint ofl ocal site effects, two of the most significant
recent earthquakes were the 1985 Michoacan(Mexico) earthquake (Stone et al ., 1987)
and the 1989 Lorna Prieta (Cal ifornia) earthquake(Seed et al ., 1990). These wel l -
documented earthquakes produced strong motion records atsites underl ain by a
variety of different subsurface conditions in Mexico City and the SanFrancisco Bay
area. A brief examination of these case histories il l ustrates the importance of
l ocal site effects.Sec. 8.2 Effects of Local Site Conditions on Ground Motion 313
~OO'8~," 0.6 ----~---.!----~---- i ~: : ---i---~ 0 : : : 90.5§ 0.4[- --J'~"--~,---J-
~,----15 ,',': ~ 0.3 +--+.-->---~ : : : ~ 0.2 - -: ----~----~.---~ 01---L+-L00 ' 1 ' 290
.5§ 0.4M-'-"-,J-',- j I ~: ~l T~_=ro.,: : « 0.1 ----~-. -,--,---00 ' ; , 2il ~~~1T~§§
0.O4 -,-K-' 1--W"---, l -J-"J-r--u~-J-,J-",-- J 0.3: --;C"-+--;f---: ---!~-- - -'i---f
1l 0.2 -;-+-,- -,+---c- -,-,----,- ,-+-c--- o : : : : : : : : : : : : « 0.1 ~--'---~~ m
'..;__ , " ", ,.' ,,'00 ' 1 ' 2 ' 1 - 2 ' ; , 2 0 ' 1 ' 2Period (sec) Period (sec)
Period (sec) Period (sec)0.10gPeriod (sec) Period (sec)Market andGuerreroStreets
o;"--"",~--_,...d,,.-r-400g£; 200 fj}0300100 -co'@ 0.15r----,""'=-----~*-------
+-----+_---+_--_____;: _: _: : -_____,~ 0.10-~0.05 - 5 O'-_+ + +- +- +- -+_---.JE
'"OJ : ;;Figure 8.4 Variation of spectral vel ocity, spectral accel eration, and peak
horizontal accel eration al ong a 4-mil e section through Sau Francisco in the 1957 San
Franciscoearthquake. (After Idriss and Seed. 1968. Used by permission of the
Seismol ogical Society of America.)8.2.3.1 Mexico City, 1985The September 19, 1985
Michoacan (M, =8.1) earthquake caused onl y moderatedamage in the vicinity of its
epicenter (near the Pacific coast of Mexico) but caused extensivedamage some 350 km
away in Mexico City. Studies of ground motions recorded at differentsites in Mexico
City il l ustrated the significant rel ationship between l ocal soil conditions and
damaging ground motions and l ed to important advances in understandingthe cycl ic
response of pl astic cl ays (e.g., Dobry and Vucetic, 1987).For seismic zonation
purposes, Mexico City is often divided into three zones with differentsubsurface
conditions (Figure 8.5a). Shal l ow, compact deposits of mostl y granul arsoil , basal t,
or vol canic tuff are found in the Foothil l Zone, l ocated west of downtown. In the
Lake Zone, thick deposits ofvery soft soil s formed from the pl uviation of airborne
sil t, cl ay,and ash from nearby vol canoes through the waters of ancient Lake Texcoco
extend to considerabl edepths, as shown by the contours of Figure 8.5b. These soft
soil s general l y consistof two soft cl ay (Mexico City Cl ay) l ayers separated by a 0-
to 6-m-thick (0 to 20 ft) compactsandy l ayer cal l ed the capa dura. Groundwater is
general l y found at a depth of about2 m over most of the Lake Zone. Between the
Foothil l and Lake Zones l ies the TransitionZone, where the soft soil deposits are
thin and interspersed erratical l y with al l uvial deposits.314 Local Site Effects and
Design Ground Motions Chap. 81 58~(5650. SCT', 40.. ...~'2'-0L'·3/0.'.·0..·· '''\ \
• UNA \ '): ..--(a) (b)Figure 8.5 Strong-motion instruments and geotechnical
conditions in Mexico City: (a)l ocations of strong motion instruments rel ative to
Foothil l , Transition, and Lake Zones;(b) contours of soft soil thickness. (After
Stone et al ., 1987.)Prior to 1985, a number of strong-motion instruments had been
depl oyed in MexicoCity. Shown in Figure 8.5 are the l ocations of those at the
Universidad Nacional Autonomade Mexico (UNAM) and the Secretary of Communications
and Transportation (SCT) site.The UNAM site was l ocated in the Foothil l Zone on 3
to 5 m (l O to 16 ft) of basal tic rockunderl ain by softer strata of unknown
thickness. The SCT site was l ocated on the soft soil sof the Lake Zone.Al though the
Michoacan earthquake was quite l arge, its great distance from MexicoCity produced
accel erations at the UNAM (rock) site of onl y O.03gto O.04g (Figure 8.6). InEast-
West accel erationUNAMSCTooo 10 20 30 40Time (sec)50 60Figure 8.6 Time histories of
accel erationrecorded by strong motion instruments atUNAM and SCT sites. (After
Stone et al .,1987.)Sec. 8.2 Effects of Local Site Conditions on Ground Motion 315
the Transition Zone, peak accel erations at the VIV site were sl ightl y greater than
those atUNAM but stil l quite l ow. In the Lake Zone, however, peak accel erations at
the CDA andSCT sites were up to five times greater than those at UNAM. The
frequency contents of theSCT and CDA motions were al so much different than that of
the UNAM motion; the predominantperiod was about 2 sec at SCT and sl ightl y l onger
at CDA. Strong l evel s of shakingpersisted over a very l ong duration at the SCT and
CDA sites. The response spectrashown in Figure 8.7 il l ustrate the pronounced
effects of the Lake Zone soil s: at periods ofapproximatel y 2 sec, spectral
accel erations at the SCT site were about 10 times greater thanthose at the UNAM
site. The SCT site was underl ain by 35 to 40 m (115 to 131 ft) of softcl ay with an
average s-wave vel ocity of about 75 m/sec (250 ft/sec). As a resul t, its
characteristicsite period [equation (7.16)] was T, =4H/vs "" 4 (37.5)/75 =2 sec, a
val ue consistentwith the peak in the SCT response spectrum of Figure 8.7.Structural
damage in Mexico City was highl y sel ective; l arge parts of the city experiencedno
damage whil e other areas suffered pronounced damage. Damage was negl igibl ein the
Foothil l Zone and minimal in the Transition Zone. The greatest damage occurred in
those portions of the Lake Zone underl ain by 38 to 50 m (125 to 164 ft) of soft
soil (Stoneet al ., 1987), where the characteristic site periods were estimated at
1.9 to 2.8 sec. Evenwithin this area, damage to buil dings of l ess than five stories
and modem buil dings greaterthan 30 stories was sl ight. Most buil dings in the five-
to 20-story range, however, either col l apsedor were badl y damaged. Using the crude
rul e of thumb that the fundamental periodof an N-story buil ding is approximatel y
N/l O sec, most of the damaged buil dings had fundamental periods equal to or somewhat
l ess than the characteristic site period. Accountingfor the period-l engthening
effect ofsoil -structure interaction (Section 7.4) and the tendencyfor the
fundamental period of a structure to increase during a strong earthquake (due to
thereduction in stiffness caused by cumul ative architectural and structural
damage), it seemsl ikel y that the damaged structures were subjected to many cycl es
ofl arge dynamic forces atperiods near their fundamental periods. This "doubl e-
resonance" condition (ampl ificationof bedrock motion by the soil deposit and
ampl ification of the soil motion by the structure)combined with structural design
and construction deficiencies to cause l ocal l y devastatingdamage.Buil ding damping
=5%0.80.6 .Sa (g)0.4 ..0.2 .oo 2 3Period (sec)4 5Figure 8.7 Response spectra
computedfrom recorded motions at UNAM and SeTsites. (After Ramo and Seed, 1986.)316
Local Site Effects and Design Ground Motions Chap. 88.2.3.2 San Francisco Bay Area,
1989On October 19, 1989, a M, =7.1 earthquake occurred near Mt. Lorna Prieta
l ocatedabout 100 km south of San Francisco and Oakl and, Cal ifornia. The Lorna
Prieta earthquakeproduced MMI VIII shaking in the epicentral region, but
intensities were actual l y higher,MMI IX, in portions of San Francisco and Oakl and.
The fact that the earthquake causedextensive damage in certain areas, and
rel ativel y l ittl e damage in others, suggested that l ocal site effects were
important.The San Francisco Bay basin is l argel y fil l ed with al l uvial deposits of
cl ays and sil ty tosandy cl ays with some l ayers of sandy and gravel l y soil s. The
deeper deposits were overconsol idatedby historical gl acial sea-l evel drawdown, but
the upper unit was deposited after thel ast drawdown episode. This material , known
as San Francisco Bay Mud, is a normal l y consol idatedsil ty cl ay. It is highl y
compressibl e and its strength grades from soft near the groundsurface to medium
stiff at depth. For purposes of seismic zonation, the Bay area can bedivided into
the three zones shown in Figure 8.8. The San Francisco Bay Mud is general l yfound at
the margins of the bay, where its thickness varies from zero up to several tens of
feet.Both the epicentral region and the San Francisco Bay area were wel l
instrumented withseismographs and accel erometers. Peak horizontal accel erations
were recorded at the l ocationsshown in Figure 8.8. These accel erations were high
near the epicenter but attenuatedwith distance from the source. The attenuation,
however, occurs much more rapidl y for sitesin the Rock/Shal l ow Residual Soil zone
than in the Al l uvium or Bay Mud zones (Figure 8.8).The response of two instruments,
those l ocated at Yerba Buena Isl and and TreasureIsl and in the middl e of San
Francisco Bay, are particul arl y instructive.
Yerba Buena Isl andBay mudAl l uviumD Rock and shal l ow residual soil oScal e (mil es)10
20Figure 8.8 Measured peak horizontal accel erations (in gs) in the San Francisco
Bayarea during the 1989 Lorna Prieta earthquake. Note the variation of peak
accel eration fordifferent site conditions. (After Seed et aI., 1990.)Sec. 8.2
Effects of Local Site Conditions on Ground Motion 31745% dampingTreasureIsl and
A._,,1 \Verba Buena.,J " Isl ~nd---' '-----------": '-------023Period (sec)(b)o0.4§c
0.6 ""o@Q) wooCl l ~ts 2i 0.2(f)15Verba Buena Isl and (E-W)200 I"(~020:
-: -.,...----.~.~ 0 5 10 20§ Time (sec)I_: : ~.o 5 10 15 20Time (sec)(a)is a rock
outcrop and Treasure Isl and is a 400-acre man-made hydraul ic fil l pl aced partial l y
on the Yerba Buena shoal s, a sandbar l ocated immediatel y northwest of Yerba Buena
Isl and. Treasure Isl and is underl ain by a variabl e thickness of San Francisco Bay
Mud; theTreasure Isl and seismograph was underl ain by 45 ft (13.7 m) of l oose sandy
soil (hydraul icfil l and natural soil s) over 55 ft (16.8 m) of San Francisco Bay
Mud. The Yerba BuenaIsl and seismograph was l ocated directl y on rock. Though the
Yerba Buena Isl and and TreasureIsl and instruments were virtual l y the same distance
from the source, they recorded dramatical l ydifferent ground surface motions (Figure
8.9a). Peak accel erations at YerbaBuena Isl and were 0.06g in the E-W direction and
O.03g in the N-S direction; the correspondingval ues at Treasure Isl and were 0.16g
and O.l l g. Response spectra for the two sitesare shown in Figure 8.9b. Cl earl y, the
presence of the soft soil s at the Treasure Isl and sitecaused significant
ampl ification of the underl ying bedrock motion.Ampl ification of ground motion by
soft soil deposits in other areas contributed significantl yto damage in other parts
of the San Francisco Bay area. The northern portion ofthe 1-880 Cypress Viaduct
that col l apsed in the earthquake was underl ain by San FranciscoBay Mud; the
southern part that remained standing was not.o.8,--------------,Figure 8.9 Ground
surface motions at Yerba Buena Isl and and Treasure Isl and in the 1989Lorna Prieta
earthquake: (a) time histories; (b) response spectra. (After Seed et al ., 1990.)
~S.2.4 Compil ations of Data on Local Site EffectsThe preceding section il l ustrates
the important infl uence ofl oca1 site conditions on the characteristicsof ground
surface motions. Case histories of ground response in Mexico City, theSan Francisco
Bay area, and many other l ocations have cl earl y shown that l ocal site conditions
strongl y infl uence peak accel eration ampl itudes and the ampl itudes and shapes of
response spectra.Comparisons of peak accel eration attenuation rel ationsips for
sites underl ain by differenttypes of soil profil es show distinct trends in
ampl ification behavior (Seed et al ., 1976).Al though attenuation data are scattered,
overal l trends suggest that peak accel erations at thesurfaces of soil deposits are
sl ightl y greater than on rock when peak accel erations are smal l and somewhat smal l er
at higher accel eration l evel s (Figure 8.10). Based on data from Mexico318 Local
Site Effects and Design Ground Motions Chap. 8Figure 8.10 Approximate rel ationships
between peak accel erations on rock andother l ocal site conditions. (After Seed et
aI., 1976.),,ROCk>/' »",/'/,/,/'Stiff soil s0.1 0.2 0.3 0.4 0.5 0.6 0.7Maximum
accel eration on rock (g)Deepcohesion l esssoil s
0.6 ,--,----,----.,------,-------,----,--.,..-,0.40.30.20.5§15~C])~el l E: : JE.~2 0.1
City and the San Francisco Bay area, and on additional ground response anal yses,
Idriss(1990) rel ated peak accel erations on soft soil sites to those on rock sites
(Figure 8.11). At l owto moderate accel eration l evel s (l ess than about O.4g), peak
accel erations at soft sites arel ikel y to be greater than on rock sites. In some
cases, such as Mexico City in 1985 and the SanFrancisco Bay area in 1989,
rel ativel y smal l rock accel erations may cause high accel erationsat the surfaces of
soft soil deposits. At higher accel eration l evel s, however, the l ow stiffnessand
nonl inearity of soft soil s often prevent them from devel oping peak accel erations as
l argeas those observed on rock.Local site conditions al so infl uence the frequency
content of surface motions andhence the response spectra they produce. Seed et al .
(1976) computed response spectra fromground motions recorded at sites underl ain by
four categories of site conditions: rock sites,stiff soil sites (l ess than 200 ft
(61 m) deep), deep cohesionl ess soil sites (greater than 250 ft(76 m) deep), and
sites underl ain by soft to medium-stiff cl ay deposits. Normal izing the computed
spectra (by dividing spectral accel erations by the peak ground accel eration)
il l ustrates0.6§(j) 0.52'(jj'0 0.4tnc 0.3 0c0~ 0.2C])Qi 0.1 -- oo-c 0.00.0Basedon
cal cul ations0.1 0.2 0.3 0.4 0.5Accel eration on rock sites (g)Figure 8.11
Approximate rel ationshipbetween peak accel erations on rock and softsoil sites.
(After Idriss, 1990, H. Bol tonSeed Memorial Symposium Proceedings,0.6 Vol . 2, p.
285. Used by permission ofBiTech Publ ishers, Ltd.)Sec. 8.2 Effects of Local Site
Conditions on Ground Motion 3194,---------------------,1.5 2.0 2.5 3.0Period (sec)
Soft to medium cl ay~and sand-15 recordsDeep cohesionl ess soil s(>250 ft)-30 records
Stiff soil s« 200 ft)-31 recordsRock-~8 records.0.5 1.0OL-----'----~--~--~---~-----'a
Figure 8.12 Average normal ized response spectra (5% damping) for different l ocal
siteconditions. (After Seed et aI., 1976.)the effects of l ocal soil conditions on
the shapes of the spectra (Figure 8.12). The effects areapparent: at periods above
about 0.5 sec, spectral ampl ifications are much higher for soil sitesthan for rock
sites. At l onger periods, the spectral ampl ification increases with decreasing
subsurface profil e stiffness. Figure 8.12 cl earl y shows that deep and soft soil
deposits producegreater proportions ofl ong-period (l ow-frequency) motion. This
effect can be very significant,particul arl y when l ong-period structures such as
bridges and tal l buil dings arefounded on such deposits. These resul ts al so show
that the use of a singl e response spectrumshape for al l site conditions is not
appropriate, a finding that has strongl y infl uenced thedevel opment of buil ding
codes and standards (Section 8.4.2).131.2.5 Effects of Surface Topography and Basin
GeometryThe effects of topographic irregul arities and al l uvial basin geometry on
ground motions canbe significant. Perhaps the best known exampl e of apparent
topographic effects was producedby an accel erograph on the abutment of Pacoima Dam
in southern Cal ifornia. ThePacoima Dam accel erograph recorded peak horizontal
accel erations of about 1.25g in eachof two perpendicul ar directions in the 1971 San
Fernando (ML = 6.4) earthquake, val ues thatwere considerabl y l arger than expected
for an earthquake of this magnitude. The acce1erograph,however, was l ocated at the
crest of a narrow, rocky ridge adjacent to the dam(Trifunac and Hudson, 1971).
Subsequent investigations have attributed a good part of theunusual l y high peak
accel erations to dynamic response of the ridge itse1f-a topographiceffect. In the
cases where al l uvial val l eys have been fil l ed with soft soil s, one-dimensional
ground response anal yses are often abl e to capture most essential aspects of
response. Theymay not, however, be abl e to describe the compl ex wave fiel ds and
l ong durations producedby mul tipl e refl ections in some of these basins. The effects
of surface topography and basingeometry are il l ustrated in the fol l owing sections.
320 Local Site Effects and Design Ground Motions Chap. 88.2.5.1 TopographyThe
topographic effects caused by simpl e irregul arities can be estimated from exact
sol utions to ideal ized probl ems (Aki, 1988). For a triangul ar infinite wedge
subjected to vertical l ypropagating SH-waves (with particl e motion paral l el to its
axis), apex displ acementsare ampl ified by a factor 21t/<j>, where <j> is the vertex
angl e of the wedge (Figure 8.13a). Thisapproach can be used to approximate
topographic effects for certain cases of ridge-val l eyterrain (Figure 8. I 3b).
Different geometries and different wave fiel ds have al so been considered(e.g., Gel i
et aI., 1988; Sanchez-Sesma, 1990; Facciol i, 1991).xz(a) (b)Figure 8.13
Characterization of simpl e topographic irregul arities: (a) notation for atriangul ar
wedge; (b) approximation of actual ground surface (sol id l ine) at trough andcrest
by wedges. (After Facciol i, 1991.)Increased ampl ification near the crest of a ridge
was measured in five earthquakes inMatsuzaki, Japan (Jibson, 1987). Figure 8.14
shows how the normal ized peak accel erationvaried at different points al ong the
ridge. The average peak crest accel eration was about 2.5times the average base
accel eration. Simil ar patterns of ampl ification on ridges are suggested1.2 r-
c---.,----------,------,------,1.0"" CIl c~gO.8.9 @"*"..Q2~ 0.6 _Q.(C)Il 0_ o (fJf,j
~ 0.4a: 0.251 --~crest3. 50m 15 4 50ml BaseRidge profil e4oo 50 100El evation (m)150
Figure 8.14 Normal ized peakaccel erations (means and error bars)recorded on mountain
ridge at Matsuzaki,Japan. (After Jibson, 1987.)Sec. 8.2 Effects of Local Site
Conditions on Ground Motion 321by damage patterns in earthquakes in Ital y and Chil e
(Finn, 1991). Anal ysis of topographicirregul arities is a compl icated probl em; the
interaction of waves can produce compl ex patternsof ampl ification and
deampl ification, depending on the geometry of the irregul arity andon the types,
frequencies, and angl es of incidence of the incoming waves (Sanchez-Sesmaand
Campil l o, 1993).8.2.5.2 BasinsSince many l arge cities are l ocated on or near
al l uvial val l eys, the effects of basingeometry on ground motion is of great
interest in geotechnical earthquake engineering. Thecurvature
of a basin in which softer al l uvial soil s have been deposited can trap body waves
and cause some incident body waves to propagate through the al l uvium as surface
waves(Vidal e and Hel mberger, 1988). These waves can produce stronger shaking and
l ongerdurations than woul d be predicted by one-dimensional anal yses that consider
onl y vertical l ypropagating s-waves.King and Tucker (1984) measured ground motions
al ong transverse and l ongitudinal profil es across the Chusal Val l ey near the
Afghanistan border of the former Soviet Union.Interpretation ofthe response in a
series of smal l (ML ~ 4.0) earthquakes suggested that onedimensional ground response
anal yses coul d predict the average response of sediments nearthe center of the
val l ey but not at the edges. Significant differences between the ampl ification
functions at the center and edges of the val l ey were observed, expl aining why the
motions at those l ocations were considerabl y different. Simil ar effects have been
observedfor other val l eys (e.g., Caracas in 1967, San Fernando in 1971, and
Leninakan, Armenia in1988) in different earthquakes.Bard and Gariel (1986) used an
anal ytical approach to study the two-dimensional response of shal l ow and deep
al l uvial val l eys. By comparing computed ampl ification functionsfor the two-
dimensional case with those based on the assumption of one-dimensional wave
propagation, the accuracy of the one-dimensional assumption coul d be demonstrated.
As shown in Figure 8.15a, the one- and two-dimensional ampl ification functions at
the centerof a shal l ow, fl at val l ey (Station 8) were quite simil ar, which indicates
that one-dimensional anal yses woul d be appropriate in that area. Cl oser to the edge
of the val l ey (Station4), however, the ampl ification functions were considerabl y
different. For the deep val l eyshown in Figure 8.l 5b, agreement between the one- and
two-dimensional ampl ificationfuctions was much better at the center of the val l ey
than near the edges, but was not as goodas for the shal l ow val l ey. For al l uvial
val l eys of irregul ar shape, such combined concave/convex regions, theoretical
studies (e.g., Rial et al ., 1992) indicate that very compl ex,even chaotic, motions
can resul t.The potential for significant differential motion across such al l uvial
val l eys hasimportant impl ications for the design of l ong-span structures, such as
bridges and pipel ines,that often cross val l eys. Differential movements can induce
l arge l oads and cause heavydamage to these types of structures.8.2.5.3 Eval uation
of EffectsEval uation of the effects of topographic and subsurface irregul arities
requires twoandin some cases, three-dimensional anal yses. Such anal yses are often
compl icated andtime consuming and may require more detail ed site characterization
than may be feasibl e.322 Local Site Effects and Design Ground Motions Chap. 810
Station4 Station810~-,. .. .....1 '1 2 : 3 4Vs (km/sec)0.3 0.7 3.5i~TStation424
113[ 5 6 7 84111!!: . • • • ...........2D, Vs increasing2D, VS constant1D, VS
increasing(a)Station810Vs (km/sec)0.3 0.7 3.5 insT3.52D, Vs increasing2D, VS
constant1D, VS increasing(b)Figure 8.15 Comparison of ampl ification factors for
one- and two-dimensional anal yses of (a) shal l ow, fl at basin, and (b) deep basin.
(After Bard and Gariel , 1986.Used by permission of the Seismol ogical Society of
America.)Al though they may be difficul t to predict, there is l ittl e doubt that such
effects exist. Sil va(1988) summarized the effects of topographic and subsurface
irregul arities, with commentson their quantitative predictabil ity, as shown in
Tabl e 8- 1.Al though provisions for considering topographic effects have been
incorporated intoa French buil ding code (French Association for Earthquake
Engineering, 1990), participantsin a recent site effects workshop (Whitman, 1992)
considered the introduction of suchprovisions into U.S. codes to be premature.Sec.
8.3 Design Parameters 323Tabl e 8-1 Effects of Topographic and Subsurface
Irregul aritiesQuantitativeStructure Conditions Type Size Predictabil itySurface
Sensitive to shape Ampl ification at top Ranges up to a factor Poor: general l y
topography ratio, l argest for of structure, of 30 but general l y underpredict size;
ratio between 0.2 and ampl ication and from about 2 to 10 maybe due to0.6; most
pronounced deampl ication at ridge-ridgewhen Ie ~ mountain base, rapid changes
interaction and threewidthin ampl itude phase dimensional effectsal ong sl opes
Sediment- Local changes in Increased duration Duration of Fairfil l ed val l eys
shal l ow sediment significant motionsthickness can be doubl edGeneration of l ong-
Increased Duration and Good at periodsperiod surface waves ampl ification
ampl ification of exceeding I secfrom body waves at and duration due to significant
motionsshal l ow incidence trapped surface may be increasedangl es waves over
onedimensional projectionsShal l ow and wide Effects most pronounced Broadband One-
dimensional Good: away from(depth/width near edges; l argel y ampl ification model s
may under- edges one< 0.25) sediment- vertical l y propagating near edges due to
predict at higher dimension worksfil l ed val l eys shear waves away generation of
frequencies by about wel l , near edgesfrom edges surface waves 2 near edges extend
onedimension tohigher frequenciesDeep and narrow Effects throughout Broadband One-
dimensional Fair: given detail ed(depth/width val l ey width ampl ification model s may
under- description of> 0.25) sediment- across val l ey due predict for a wide
vertical and l ateral fil l ed val l eys to whol e val l ey bandwidth by about changes in
material modes 2 to 4; resonant propertiesfrequencies shiftedfrom one dimension
Source: After Sil va (1988).8.3 DE!il GN PARAMETERSEarthquake-resistant design of new
structures and eval uation of the safety of existing structuresrequires anal ysis
oftheir response to earthquake shaking. Eval uation of geotechnical hazards,such as
l iquefaction and sl ope fail ure, al so requires anal ysis with respect to some l evel
ofshaking. The l evel of shaking for which satisfactory performance is expected is
often referredto as a design l evel of shaking and is described by a design ground
motion. Design groundmotions can, depending on how they are to be used, be
specified in many different ways. Manyanal yses of soil and structural response
require an entire time history of motion; others requireonl y one or more of the
ground motion parameters described in Chapter 3. The parametersmost commonl y used
to specify design ground motions are peak horizontal accel eration, peakhorizontal
vel ocity, predominant period, response spectrum ordinates, and duration.324 Local
Site Effects and Design Ground Motions Chap. 88.3.1 Design Earthqua'ces
Historical l y, design parameters were most commonl y determined from a specified
designearthquake, and some regul atory agencies stil l require that earthquake-
resistant design beperformed with respect to the motion produced by a design
earthquake. The specification ofa design earthquake impl ies a l evel of determinism
in the seismic hazard anal ysis; that is,after the design earthquake is
characterized (which can be done deterministical l y or probabil istical l y),its
effects at the site of interest are computed deterministical l y. Historical l y,design
earthquakes have been associated with two-l evel design, in which a structure or
facil ityis required both to remain operational at one l evel of motion, and to avoid
catastrophicfail ure at another, more severe l evel .Many different terms have been
used to describe the l evel s of severity associated withdesign earthquakes; some of
them have been defined differentl y by different organizations.The Maximum Credibl e
Earthquake (MCE) is usual l y defined as the l argest earthquake thatcan reasonabl y be
expected (al though the meaning of the word reasonabl y may be open tointerpretation)
from a particul ar source. The Safe Shutdown Earthquake (SSE), used in thedesign of
nucl ear power pl ants, is specifical l y defined (Christian, 1988) as the earthquake
thatproduces the maximum peak horizontal accel eration for the fol l owing cases: (1)
moving theepicenter of the l argest anticipated event in the surrounding
seismotectonic province (regionof uniform seismicity) to the site, (2) moving the
epicenters of the l argest events in the adjacentseismotectonic provinces to the
nearest points on their boundaries and attenuating theirmotions to the site, and
(3) moving the foci of the l argest events on any capabl e faul ts to thecl osest
points on the faul ts to the site and then attenuating their motions to the site. In
manygeographic regions, the MCE and SSE have simil ar characteristics. Other terms
that havebeen used to describe simil ar worst-case l evel s of shaking incl ude Maximum
Capabl e Earthquake,Maximum Design Earthquake, Contingency Level Earthquake, Safety
Level Earthquake,Credibl e Design Earthquake, and Contingency Design Earthquake. The
two-l evel design approach general l y requires that structures of facil ities be
designed to avoid catastrophicfail ure at the l evel s of shaking produced by these
upper-l evel design earthquakes.A l ower but more l ikel y l evel of shaking woul d be
produced by an Operating BasisEarthquake (aBE); it is an earthquake that shoul d be
expected during the l ife of a structure(Krinitzsky et aI., 1993). The aBE has been
taken as an earthquake with hal f the peak accel erationof the SSE (Christian, 1988),
as an earthquake that produces motion with a 50% probabil ityof exceedance in 50
years (USCOLD, 1985), and as an earthquake with a returnperiod of about 110 years
(Christian et al ., I978). Other terms that have been used to describedesign
earthquakes of simil ar size are Operating Level Earthquake, Maximum Probabl e
Earthquake, Probabl e Design Earthquake, and Strength Level Earthquake. Two-l evel
designrequires that structures and facil ities
be designed to remain operational after being subjectedto the l evel s of shaking
associated with these l ower-l evel design earthquakes.Design earthquakes are often
specified without regard to their l ikel ihood of occurrence,even though a MCE may
have a return period of 200 years in one l ocation and 10,000years in another. Their
use in the devel opment of design ground motions has decreased asprobabil istic
seismic hazard anal ysis techniques have devel oped. Stil l , some bel ieve (e.g.,
Krinitszky et al ., 1993) that the assumptions of the probabil istic approach are
insufficientl ysupported by observational data to al l ow its rel iabl e use for
critical structures and facil ities.This l ack of data is particul arl y significant
for l arge earthquakes.Sec. 8.3 Design Parameters 3258 ..3.2 Design SpectraResponse
spectra are often used to represent seismic l oading for the dynamic anal ysis of
structures.As a resul t, design ground motions are often expressed in terms of
design spectra.Design spectra and the response spectra of actual earthquakes are
not the same. Response spectrafrom earthquakes are highl y irregul ar (e.g., Figure
3.16); their shapes refl ect the detail s oftheir specific frequency contents and
phasing. Design spectra, on the other hand, are general l yquite smooth; they are
usual l y determined by smoothing, averaging, or envel oping theresponse spectra of
mul tipl e motions. The use of smooth design respose spectra impl icitl y recognizesthe
uncertainty with which soil and structural properties are known by avoiding sharp
fl uctuations in spectral accel erations with smal l changes in structural period.
Newmark and Hal l (1973), for exampl e, recommended that design response spectrabe
devel oped from a series of straight l ines on a tripartite pl ot (Section A.2.2 of
AppendixA) corresponding to the accel eration-, vel ocity-, and displ acement-
control l ed portions(Section 3.3.2.1) of the spectrum. A Newmark-Hal l design
spectrum is obtained by mul tipl yingthe peak ground accel eration, vel ocity, and
displ acement val ues by the factors shownin Tabl e 8-2. At periods bel ow about 0.17
sec (frequencies above about 6 Hz), the spectral accel erations are tapered down to
the peak ground accel eration. A peak ground vel ocity of48 in./sec (122 ern/sec) and
peak ground displ acement of 36 in. (91 ern) are assumed to beconsistent with a peak
ground accel eration of 1.0g; each of these parameters can be scal edby the ratio of
the design peak ground accel eration to l .Og to produce a design spectrum.Tabl e 8-2
Ampl ification factors for Newmark-Hal l DesignSpectraAmpl ication Factors for:
Structural Damping Ratioo251020Displ acement2.51.81.41.11.0Vel ocity4.02.81.91.31.1
Accel eration6.44.32.61.51.2Source: After Newmark and Hal l (1973).Exampl e 8.2Devel op
a Newmark-Hal l design spectrum for 5% damping and a peak ground acceieration of
O.25g.Sol ution For 5% damping, the accel eration, vel ocity, and displ acement
ampl ification factorsare 2.6, 1.9, and 1.4, respectivel y. Consequentl y, the peak
spectral val ues are: Spectral accel eration: Sa = 2.6(O.25g) = O.65gSpectral
vel ocity: S; = 1.9~: ~~~(48 in/sec) = 22.8 in/secSpectral displ acement: S« =
1.4~: ~~~ (36 in) = 12.6 inA tripartite pl ot of the resul ting design spectrum is
shown in Figure E8.2.326 Local Site Effects and Design Ground Motions Chap. 8
···7'····,··,······;'""·······,··y····,,'.···+, ,..• ....; ,-,, ,.", +/..,.,./
: c.: i/..: ": .. ,· +· .. '·'(: 10 ..'···/ "/''': ''7': ·1100
f,/..·,·,···: ......: h·.: '<j',,·.· .. i /·..'V'·..-i;,L..·
..: ··: .·: ....i..: '··,..·..·h·: ·: : i i
: ··,: : : ··/..·: ..: .···: ..·..·: ···c·: : ..·,,··,·.: ·..·..,,·/: : ··: ,....,··..: : ;.·".·,·.
./ ....(·: /··180.04 .06 .08 .1 .2 .4 .6 .8 1Period (sec)Figure E8.22 4 6 8 10 20For
sites that coul d be subjected to shaking from more than one seismic source,
devel opmentof a design spectrum can be compl icated. A site in Cal ifornia, for
exampl e, may besubjected to strong short-period (high-frequency) motion from smal l
earthquakes on nearbyfaul ts and strong l ong-period (l ow-frequency) motion from
l arge earthquakes on more distantfaul ts. The devel opment of predictive
rel ationships for the ordinates of response spectraSec. 8.4 Devel opment of Design
Parameters 327(Section 3.4.4.4) have al l owed uniform risk spectra (e.g., Trifunac
et al ., 1987), in whichspectral ordinates are obtained by individual PSHAs, to be
generated with proper considerationof al l seismic sources. With this approach, the
design spectrum has an equal probabil ityof exceedance at al l periods of vibration.
Because of the averaging proceduresinherent in the attenuation rel ationships on
which they are based, uniform risk spectra havesmooth shapes that are unl ike the
response spectra from individual ground motions.8.4 DEVELOPMENT OF DESIGN
PARAMETERSThe characteristics of the design ground motion at a particul ar site are
infl uenced by thel ocation of the site rel ative to potential seismic sources, the
seismicity of those sources, thenature of rupture at the source, travel path
effects between the source and the site, l ocal siteeffects, and the importance of
the structure or facil ity for which the ground motion is to beused. The manner in
which the motion is to be used shoul d al so be considered.Design ground motions are
usual l y devel oped in one of two ways: from site-specificanal yses or from the
provisions of buil ding codes and standards.8.4.1 Site-Specific Devel opmentSite-
specific design ground motions refl ect the detail ed effects of the particul ar
subsurfaceconditions at the sites of interest. The usual process for devel oping
site-specific groundmotions invol ves a seismic hazard anal ysis and a ground
response anal ysis. The seismichazard anal ysis can be performed deterministical l y or
probabil istical l y using the techniquesdescribed in Chapter 4. The use of
probabil istic seismic hazard anal yses requires that thedesign motion be associated
with some l evel of risk, or return period. Sel ection of such aquantity can be quite
compl ex; various social , economic, and pol itical considerations areoften invol ved.
For many structures and facil ities in the United States, design groundmotions have
been based on parameters with a 10% probabil ity of exceedance in 50 years(or a 475-
year return period). The requirements of most contemporary buil ding codes, for
exampl e, are based on that l evel of risk though some experts bel ieve that higher
return periodswoul d be more appropriate (Matthiesen et al ., 1982; Whitman, 1989).
Different regul atoryagencies may require different return periods for different
types of structures andfacil ities.The seismic hazard anal ysis wil l produce a set of
ground motion parameters that may,or may not, correspond to the subsurface
conditions at the site of interest. Both deterministicand probabil istic seismic
hazard anal yses util ize predictive rel ationships (Section 3.4) thatusual l y
correspond to a fairl y narrow range of subsurface conditions; if the site of
interest isl ocated on a simil ar profil e, these parameters may be taken directl y as
the design groundmotion parameters. If it is not, however, the parameters from the
seismic hazard anal ysismust be modified to account for the effects of l ocal site
conditions.This parameter modification process may be performed empirical l y, using
priorobservations such as those shown in Figures 8.12 and 8.13, or anal ytical l y. In
the anal ytical approach, both deconvol ution (Section 7.1.1.4) and conventional
ground response anal ysesmay be required, as il l ustrated in Figure 8.16. The seismic
hazard anal ysis wil l produceparameters that describe a ground motion at the surface
(point A) of a site with subsurface328 Local Site Effects and Design Ground Motions
Chap. 8Soil profil ecorrespondingto predictiverel ationshipRockoutcroppingSurface
motion from motionse: m/zard anal ysis CAc: ;;BBedrock motion atsite correspondingto
predictiverel ationshipSurface motionat site of interest \EConventional ground
responseanal ysis(convol ution)Bedrock motion atsite of interestFigure 8.16 Procedure
for modifying ground motion parameters from a seismic hazardanal ysis to account for
the effects of l ocal site conditions.conditions that corresponds to those of the
sites in the database from which the predictiverel ationship was devel oped. To
determine the corresponding parameters at the surface ofthe site of interest, a
time history of ground surface motion that is consistent with the predicted
parameters is generated; procedures for generating such time histories are
describedin Section 8.5. This motion is then deconvol ved through a soil profil e
corresponding to thepredictive rel ationship to determine the time history of
bedrock motion (at point B) thatwoul d produce the time history of motion at point
A. The corresponding rock at croppingmotion produces the bedrock motion appl ied at
the base (point D) of the soil profil e at thesite of interest. A conventional
ground response anal ysis is then performed to predict themotion at the surface of
the soil profil e of interest (point E). This motion, which is consistentwith the
resul ts of the seismic hazard anal ysis and al so with the l ocal site conditions, can
betaken as the design ground motion. It can be used to compute site-specific design
parameterssuch as peak accel eration and vel ocity, response spectral ordinates, and
duration.8.4.2 Code-Based Devel opmentAl ternativel y, design ground motions can be
devel oped on the basis of buil ding code provisions.Consideration of earthquake and
other effects in the design of new structures is mandatedby modem buil ding codes
which may be adopted as l aw by various city, county, andstate governments. The
purpose of codes such as the Uniform Buil ding Code (UBC) is "to provideminimum
standards to safeguard l ife or l imb, heal th, property, and publ ic wel fare by
regul atingand control l ing
the design, construction, qual ity of material s, use and occupancy,l ocation and
maintenance of buil dings" (ICBO, 1991). Codes for highway bridges, nucl earpower
pl ants, and other types of structures have been produced by various regul atory
agencies.The provisions of buil ding codes are devel oped by consensus of a broad
group ofexperienced practitioners and researchers. Al though contemporary codes do
consider siteeffects, they usual l y do SQ by l umping groups of simil ar soil profil es
together so that theirprovisions appl y to broad ranges of soil conditions within
which the l ocal conditions of aparticul ar site are expected to fal l . Because of
this, design ground motions devel oped fromSec. 8.4 Devel opment of Design Parameters
329code provisions are usual l y more conservative (i.e., correspond to stronger
l evel s of shaking)than those devel oped from site-specific anal yses. This
rel ationship between site-specificand code-based design ground motions provides an
economic incentive that impl icitl yencourages the devel opment of site-specific
design ground motions.8.4.2.1 BackgroundThe first buil ding code, motivated
primaril y by insurance l osses due to fire, was publ ishedin 1905. In 1927, the first
edition of the VBC was publ ished by what is now the International Conference of
Buil ding Official s. The VBChas become widel y adopted, particul arl yin the western
United States, where seismic hazards are high. As a resul t, the VBC has beennear
the forefront of activity on issues rel ating to earthquake-resistant design. In the
UnitedStates, the Buil ding Official s and Code Administrators International (BOCA)
Code, StandardBuil ding Code (SBC), and National Buil ding Code (NBC) are frequentl y
used in the midwest,south, and northeast, respectivel y.Historical l y, buil ding code
provisions rel ating to earthquake safety have devel opedincremental l y, with each
increment occurring shortl y after the occurrence of a damagingearthquake in the
United States. Al though buil ding codes go back as far as about 2100 B.C.(Berg,
1983), expl icit provisions for earthquake resistance did not appear in U.S.
buil dingcodes until after the 1925 Santa Barbara earthquake (ML =6.3). Those
provisions, however,were l isted onl y in an appendix and were not mandatory. After
the 1933 Long Beach, Cal iforniaearthquake (ML =6.3), in which the potential l y
tragic consequences of the col l apseof many school buil dings were averted by the
earthquake's occurrence after school hours,the state of Cal ifornia passed a bil l
known as the Fiel d Act, which required that al l newschool construction be designed
for higher l evel s of earthquake resistance and that school construction be cl osel y
supervised in the fiel d. Shortl y thereafter, Cal ifornia al so passed theRil ey Act,
which establ ished mandatory design requirements for nearl y al l occupied buil dingsin
the state.In 1959, the Structural Engineers Association of Cal ifornia (SEAOC)
publ ished a documenttitl ed Recommended Lateral Force Requirements and Tentative
Commentary, morecommonl y known as the SEAOC Bl ue Book, whose provisions were
incorporated into the1961 VBe. Since then the Bl ue Book has been modified about
every three years with itsrevised provisions adopted first by the ICBO in the UBC
and then by the other major regional codes. These codes devel oped in an incremental ,
evol utionary manner as new research andexperience was gained. The provisions of the
VBC, for exampl e, have historical l y been basedon those of the SEAOC Bl ue Book
(e.g., SEAOC, 1990). In 1975 the Appl ied Technol ogyCouncil began, with support from
the National Science Foundation and the National Bureauof Standards (now the
National Institute for Standards and Technol ogy), to take a fresh l ookat al l
aspects of earthquake-resistant design, with the aim of devel oping code provisions
thatcoul d be adopted nationwide. In 1978, ATC publ ished a report titl ed Tentative
Provisions forthe Devel opment ofSeismic Regul ationsfor Buil dings (ATC, 1978),
popul arl y known asATC3-06. This document incl uded many innovations, incl uding
several rel ating to the treatmentof l ocal site effects and devel opment of design
ground motions. Years l ater, many of the provisionsof ATC 3-06 were incorporated
into the NEHRP (National Earthquake HazardsReduction Program) Recommended
Provisions for the Devel opment ofSeismic Regul ationsfor New Buil dings produced by
the Buil ding Seismic Safety Council (BSSC, 1991a, b).330 Local Site Effects and
Design Ground Motions Chap. 8The VBe and NEHRP Provisions are the most infl uential
contemporary documentsthat describe minimum standards for earthquake-resistant
design of buil dings in the UnitedStates. Al though codes and provisions change with
time, both the VBe and NEHRP Provisionsaddress the issue of l ocal site conditions
and present commonl y used approaches tothe devel opment of design ground motions.
The fol l owing sections describe the manner inwhich l ocal site effects and design
ground motions are specified in the most recent (1994)versions of the VBe and NEHRP
Provisions. These descriptions are intended onl y to il l ustratethe phil osophies and
basic approaches of the documents; they are not compl ete.8.4.2.2 Uniform Buil ding
CodeBuil ding codes are not intended to el iminate earthquake damage compl etel y.
Indeed,the commentary to the 1990SEAOe Bl ue Book, upon which the 1991 VBeis based,
says that"structures designed in conformance [with the Bl ue Book recommendations]
shoul d, in general ,be abl e to: (1) resist a minor l evel of earthquake ground motion
without damage; (2)resist a moderate l evel of earthquake ground motions without
structural damage, but possibl yexperience some nonstructural damage; (3) resist a
major l evel of earthquake ground motionhaving an intensity equal to the strongest
either experienced or forecast for the buil ding site,without col l apse, but possibl y
with some structural as wel l as nonstructural damage"(SEAOC, 1990). These
expectations il l ustrate the basic phil osophy of earthquake-resistantstructural
design, as the commentary goes on to state that it "woul d in most cases be
economical l yprohibitive to design buil dings to remain el astic for al l l evel s of
earthquake groundmotions. A fundamental tenet of seismic design is that inel astic
yiel ding is al l owed to accomodateseismic l oadings as l ong as such yiel ding does not
impair the vertical l oad capacityof the structure. In other words, damage is
al l owed in the maximum expected earthquakel oading case onl y if it does not pose a
significant probabil ity of the structure's col l apse."The UBC al l ows two basic
approaches to the earthquake-resistant design of buil dings: a static approach in
which the effects of ground motions are represented by static l ateral forces, and a
dynamic approach in which ground motion is characterized by a designresponse
spectrum. The simpl er static approach is al l owed onl y for certain conditions of
geometric regul arity, occupancy, and height.Static Approach. The static approach is
based on determination of a designbase shear force, which is then distributed in a
specified pattern over the height of the structurefor structural anal ysis of
l ateral l oad resistance. The total design base shear in a givendirection (the
structure must be abl e to resist it in any direction) is given by(8.1)where Z is a
seismic zone/actor (Figure 8.17), I is an importance/actor (Tabl e 8-3), Rw is a
numerical coefficient that refl ects the ductil ity of the structure (Tabl e 8-4), W
is the seismicdead l oad (which incl udes permanent equipment and portions of l ive
l oads, partition l oads,and snow l oads), ande = 1.25S (8.2)T2I3In equation (8.2), S
is a soil coefficient (Tabl e 8-5) and T is the fundamental period of thestructure
in seconds. The val ue of e need not exceed 2.75, and the minimum val ue of C/Rw'
Al aska~v~ .e: « : • 3 P;;eutian'-': ",. D_ Isl ands"/)G1/~A26 ?6 Hawaiib!,4Puerto Ricoo
Zoneo12A2834z0.00.0750.150.200.300040l .Il l ...I.l Figure 8.17 Map and tabl e for
eval uation of VBC seismic zone factor, Z (Reproduced from the 1994 edition of the
UniformBuil ding Code'", copyright © 1994, with permission of the publ isher, the
International Conference of Buil ding Official s.)332 Local Site Effects and Design
Ground Motions Chap. 8Tabl e 8-3 USC Importance Factor, I.Occupancy Category
Essential Facil ities"Hazardous Facil itiesbSpecial Occupancy Structures"Standard
Occupancy Structures dMiscel l aneous Structures"I1.251.251.001.00a Incl udes
hospital s and other medical facil ities having surgery and emergency treatment
areas; fire andpol ice stations; tanks or other structures containing, housing or
supporting water or other fire-suppressionmaterial s or equipment required for the
protection of essential or hazardous facil ities, or special occupancystructures;
structures and shel ters in emergency-preparedness centers; standby power-generating
equipmentfor essential facil ities; structures and equipment in government
communication centers and other facil itiesrequired for emergency response; garages
and shel ters for emergency vehicl es and emergency aircraft;aviation control towers;
standby power generating equipment for essential facil ities.b Incl udes occupancies
and structures therein housing or supporting toxic or expl osive chemical s or
substances. Nonbuil ding structures housing, supporting, or containing quantities of
toxic or expl osivesubstances which, if contained in a buil ding, woul d cause that
buil ding to be cl assified as a hazardousfacil ity.C Occupancies with a capacity> 300
persons; buil dings for school s through secondary or day-care centers capacity>250
students; occupancies used for col l eges or adul t education school s - capacity> 500
students;medical facil ities with 50 or more resident incapacitated patients, but
not incl uded above; jail s and detentionfacil ities, al l structures with occupancy>
5,000 persons; structures and equipment in power-generatingstations and other
publ ic util ity facil ities not incl uded above, and required for continued operation.
d Al l structures having occupancies or functions not l isted above, and occupancy
towers.eGroup U occupancies except for towers.under most conditions, is set at
0.075. The fundamental period of the structure can be determinedby different
methods; in the simpl est it is approximated byT = Ct(hn ) 3/ 4 (8.3)where C, has
val ues of 0.035 for steel moment-resisting frames, 0.030 for reinforced concrete
moment-resisting frames and eccentrical l y braced steel frames, and 0.020 for al l
otherstructures, and hn is the height (in feet) of the uppermost l evel of the main
portion of thestructure above the base.Dynamic Approach. The dynamic approach of
the VBC al l ows the responseof the structure to be determined by response spectrum
anal ysis or by time-history anal ysis.Hence design ground motions can be specified
in terms of design response spectra or designground motion time histories. In both
cases, the VBC requires that the design ground motioncorrespond to a 10%
probabil ity of exceedance in a 50-year period (475-year return period).Design
response spectra can be determined in one of two ways: from site-specificground
response anal yses of the type described in Section 8.4.1, or from smooth,
normal izedspectral shapes (Figure 8.18). The normal ized spectral shapes fol l ow from
the resul tsof Seed et al . (1976) and Newmark and Hal l (1982) (see Figure 8.12), and
account for thefrequency-dependent ampl ification of ground motion by different
l ocal site conditions.These normal ized spectra are presented for three subsurface
profil es; as woul d be expected,greater l ong-period spectral accel erations are
associated with softer and deeper soil profil es.Sec. 8.4 Devel opment of Design
Parameters 333Tabl e 8-4 Val ues of USC reduction factor, Rw• Basic Structural System
Bearing Wal l SystemBuil ding Frame SystemMoment-Resisting FrameSystemDual Systems
Lateral Load Resisting SystemLight-framed wal l s with shear panel sa. Wood structural
panel wal l s for structures three stories or l essb. Al l other l ight-framed wal l s
Shear wal l sa. Concreteb. MasonryLight steel -framed bearing wal l s with tension-onl y
bracingBraced frames where bracing carries gravity l oadsa. Steel b. Concretec. Heavy
timberSteel eccentrical l y braced frameLight-framed wal l s with shear panel sa. Wood
structural panel wal l s for structures three stories or l essb. Al l other l ight-
framed wal l sShear wal l sa. Concreteb. MasonryOrdinary braced framesa. Steel b.
Concretec. Heavy timberSpecial concentrical l y braced framesa. Steel Special moment-
resisting frames (SMRF)a. Steel b. ConcreteConcrete intermediate moment-resisting
frames (IMRF)Ordinary moment-resisting frames (OMRF)a. Steel b. ConcreteMasonry
moment-resisting wal l frameShear wal l sa. Concrete with SMRFb. Concrete with steel
OMRFc. Concrete with concrete IMRFd. Masonry with SMRFe. Masonry with OMRFf.
Masonry with concrete IMRFSteel eccentrical l y braced framea. With steel SMRFb. With
steel OMRFOrdinary braced framesa. Steel with steel SMRFb. Steel with steel OMRFc.
Concrete with concrete SMRFd. Concrete with concrete IMRFSpecial concentrical l y
braced framesa. Steel with steel SMRFb. Steel with steel OMRF8666464410978888891212
8659126986712610696II6334 Local Site Effects and Design Ground Motions Chap. 8Tabl e
8-5 USC Soil Coefficient, S.Type Description sSl S2S3S4A soil profil e with either:
(a) A rock-l ike material characterized by a shear wave vel ocity greater than 2,500
feet per second or by other suitabl e means of cl assification, or(b) Medium-dense to
dense or medium-stiff to stiff soil condition where the soil depth is l ess than 200
feet.A soil profil e with predominantl y medium dense to dense or medium-stiff to
stiffsoil conditions, where the soil depth exceeds 200 feet or more.A soil profil e
containing more than 20 feet of soft to medium stiff cl ay but not morethan 40 feet
of soft cl ay.A soil profil e containing more than 40 feet of soft cl ay characterized
by a shearwave vel ocity l ess than 500 feet per second.1.01.21.52.04r---~--~--~--~--
~--~Rock and stiff soil s(soil type 1)---1.0 1.5 2.0 2.5 3.0Period, T (sec)0.5OL-
__-'---__-L.-_~-'------'-----'-----.JoFigure 8.18 Site-dependent normal ized
response spectra. (Reproduced from the 1994 edition of theUniform Buil ding Code TM,
copyright © 1994, with permission ofthe publ isher, the International Conference of
Buil ding Official s.)The design response spectrum is obtained by mul tipl ying the
ordinates of the normal izedspectrum by the effective peak ground accel eration,
which can be taken as the val ue of theseismic zone factor, Z, expressed as a
fraction of gravity. Note that the shape of the designspectrum is constant-the
spectral ordinates are l inearl y scal ed by the peak accel erational one. Since
important ground motion characteristics such as frequency content and durationare
infl uenced by earthquake magnitude and distance, the probabil ities of exceedanceof
a constant-shape design spectrum may be different at l ow periods than at high
periods.The use of such a design spectrum may produce structural designs with
different probabil itiesof fail ure. Uniform risk spectra (Section 8.3.2) minimize
this possibil ity.The VBe requires that design ground motion time histories be
devel oped on a sitespecificbasis. The use of several time histories is recommended.
Response spectra computedfrom the time histories must, either individual l y or in
combination, approximate thesite-specific design response spectrum. Site-specific
anal yses are required for fl exibl estructures (fundamental periods greater than 0.7
seconds) l ocated on soil profil e S4.Sec. 8.4 Devel opment of Design Parameters 335
Other Code Provisions. The UBC al so sets forth requirements deal ing withsite
grading, foundations, and retaining structures. In seismic zones 3 and 4, buil ding
official smay require that "the potential for soil l iquefaction and soil strength
l oss during earthquakesshal l be eval uated during the geotechnical investigation.
The geotechnical reportshal l assess potential consequences of any l iquefaction and
soil strength l oss, incl uding estimationof differential settl ement, l ateral
movement or reduction in foundation soil -bearingcapacity, and discuss mitigating
measures." The topic of soil l iquefaction is discussed indetail in Chapter 9.
8.4.2.3 NEHRP ProvisionsThe NEHRP Provisions is not a buil ding code; rather, it is
a source document intendedto aid in the devel opment of buil ding codes in areas of
seismic exposure. Their purpose is to"present criteria for the design and
construction of buil dings and nonbuil ding structures subjectto earthquake ground
motions. Their purposes are to minimize the hazard to l ife for al l buil dings and
nonbuil ding structures, to increase the expected performance of higher occupancy
structures as compared to ordinary structures, and to improve the capabil ity of
essential facil ities to function during and after an earthquake ... (they provide)
the minimumcriteria considered to be prudent and economical l y justified for the
protection of l ife safetyin buil dings subject to earthquakes at any l ocation in the
United States... The 'design' earthquakeground motion l evel s specified ... may
resul t in both structural and non-structural damage. For most structures designed
and constructed according to these provisions, it isexpected that structural damage
from a major earthquake may be repairabl e but it may not beeconomicaL .. For
motions l arger than the design l evel s, the intent of these provisions is thatthere
be a l ow l ikel ihood of buil ding col l apse." (BSSC, 1991b).Ground Motion Parameters.
The NEHRP Provisions use the effective peakaccel eration, EPA, and effective peak
vel ocity, EPV (Section 3.3.4), to describe groundmotions. These parameters can be
thought of as normal izing factors for the devel opment ofsmooth response spectra;
the EPA is proportional to the average spectral accel eration at l owperiods (0.1 to
0.5 sec) and the EPV is proportional to the spectral vel ocity at l onger periods
(about 1 sec). The EPA is usual l y somewhat l ower than the peak accel eration of a
specificground motion and can be substantial l y l ower for ground motions with very
high frequencies.The EPV is usual l y greater than the peak vel ocity, particul arl y at
l arge distances fromstrong earthquakes. To compute various design coefficients, the
EPA and EPV are repl acedby the dimensionl ess accel eration coefficients, Aa and Avo
The effective peak accel erationcoefficient, Aa, is numerical l y equal to the EPA
when expressed as a decimal fraction ofgravity (i.e., Aa =0.2 when EPA =0.2g). The
effective peak vel ocity-rel ated accel erationcoefficient, Av' is numerical l y equal
to EPV/30 when the EPV is expressed in iri.!sec (i.e.,Av =0.2 when EPV =6 in.!sec).
Note that Av is an accel eration coefficient, even though itis obtained from the
spectral vel ocity; it provides a useful measure of the l onger-period(l ower-
frequency) components of a ground motion. At any particul ar l ocation, the design of
a buil ding may be governed by Aa or Av. The NEHRP provisions contain maps, based on
probabil istic seismic hazard anal yses with a 10% probabil ity of exceedance in a 50-
yearperiod, that divide the United States into seven seismic l oading zones (Figure
8.19). Thecoefficients Aa and Av can be determined from these maps and Tabl e 8-6.
10II10II e-Figure 8.19 Maps ofNEHRP seismic l oading zones: (a) map I for Aa; and
(b) map 2 for A,. (AfterBSSe,1994.)337338 Local Site Effects and Design Ground
MotionsTabl e 8-6 NEHRP Coefficients Aa and AvChap. 8Map Area from Map 1 (for Aal or
Map 2 (for Avl 7654321Val ue of Aa and Av0.400.300.200.150.100.05<
 0.05'(8.5a)(8.5b)a For equations or expressions incorporating the terms Aa or Av,a
val ue of 0.05 shal l be used.Equival ent Lateral Force Procedure. Like the VBC, the
NEHRP Provisionsaccount for l ocal soil conditions in the determination of design
l oading. The NEHRPProvisions, however, address the effects of l ocal soil conditions
in greater detail than theVBe. The remainder of this section focuses on the
differences between this aspect of theVBC and NEHRP Provisions, as il l ustrated by
their static (or equival ent l ateral force)approaches to seismic design.The
equival ent l ateral force procedure of the NEHRP Provisions requires that theseismic
base shear that is proportional to the weight of the structure.v = CsW (8.4)where
Wincl udes the total dead l oad and appl icabl e portions of other l oads (e.g.,
permanentequipment, partitions, l ive l oads) and the seismic response coefficient,
Cs' is the smal l er ofC = 1.2Cvs RT2I3orC = 2.5Cas RIn Equations 8.5, R is a
response modification factor (anal ogous but not identical to the R;factor in the
VBe) and Tis the fundamental period of the structure. The l ong-period seismic
coefficient, Cv' and the short-period seismic coefficient, Cm refl ect both l ocal
soil conditionsand site seismicity. These coefficients are defined as(8.6a)(8.6b)
where Faand F; depend on soil type and A, and A, depend on site seismicity. The
effectivepeak accel eration, Aa , and effective peak vel ocity-rel ated accel eration,
Av' are obtainedfrom Tabl e 8-6 and Figure 8.20. The site coefficients, Faand Fv'
are shown in Tabl e 8-7.Sec. 8.4 Devel opment of Design Parameters 339Tabl e 8·7
Val ues of Fa and Fv for different site conditions and shaking intensities.Val ues of
FaSoil Shaking IntensityProfil e Type Aa S 0.1 Aa=0.2 Aa=0.3 Au =0.4 Au ~0.5A 0.8 0.8
0.8 0.8 0.8B 1.0 1.0 1.0 1.0 1.0C 1.2 1.2 l .l 1.0 1.0D 1.6 1.4 1.2 l .l 1.0E 2.5 1.7
1.2 0.9 SSl F SS SS SS SS SSVal ues of F;Soil Shaking IntensityProfil e Type Av S 0.1
Av=0.2 Av=0.3 Av=0.4 Av ~ 0.5A 0.8 0.8 0.8 0.8 0.8B 1.0 1.0 1.0 1.0 1.0C 1.7 1.6
1.5 1.4 1.3D 2.4 2.0 1.8 1.6 1.5E 3.5 3.2 2.8 2.4 SSl F SS SS SS SS SS1 Site-
specific geotechnical investigation and dynamic site response anal yses shal l be
performed.Soil Profil e TypeABCDEFDescriptionHard rock with measured shear wave
vel ocity, \Is > 5000 ft/sec (1500 m1sec)Rock with 2500 ft/sec < \Is S 5000 ft/sec
(760 m1sec < \Is S 1500 m1sec)Very dense soil and soft rock with 1200 ft/sec < \Is
S 2500 ft/sec (360 m1sec < \IsS 760 m1sec) or with either Fh 50 or su~ 2000 psf
(100 kPa)Stiff soil with 600 ft/sec < \Is S 1200 ft/sec (180 m1sec < Vs S 360
m/sec) or with either15 S NS 50 or 1000 psf S suS 2000 psf (50 kPa S suS 100 kPa)A
soil profil e with Vs < 600 ft/sec (180 m1sec) or any profil e with more than l Oft (3
m)of soft cl ay defined as soil with PI> 20, w ~ 40%, and Su < 500 psf (25 kPa)Soil
requiring site-specific eval uations: I. Soil s vul nerabl e to potential fail ure or
col l apse under seismic l oading such asl iquefiabl e soil s, quick and highl y sensitive
cl ays, col l apsibl e weakl y cemented soil s.2. Peats and/or highl y organic cl ays (H>
10 ft (3 m) of peat and/or highl y organic cl aywhere H =thickness of soil )3. Very
high pl asticity cl ays (H> 25 ft (8 m) with PI> 75)4. Very thick soft/medium stiff
cl ays (H> 120 ft (36 m))Exception: When the soil properties are not shown in
sufficient detail to determine theSoil Profil e Type, Type D shal l be used. Soil
Profil e Types E or F need not beassumed unl ess the regul atory agency determines
that Types E or F may bepresent at the site or in the event that Types E or F are
establ ished by thegeotechnical data.340 Local Site Effects and Design Ground
Motions Chap. 88.5 DEVELOPMENT OF GROUND MOTION TIME HISTORIESOn many occasions,
ground motion parameters al one do not adequatel y describe the effectsof ground
shaking. For anal ysis of nonl inear probl ems such as the response of inel astic
structuresor the permanent deformation of an unstabl e sl ope, time histories of
motion are required.Time histories can al so be required in the devel opment of site-
specific design groundmotions, as il l ustrated in Figure 8.19. In these cases, time
histories that match target groundmotion parameters such as peak accel erations,
vel ocities, or spectral ordinates are required.In some cases, the l ocal and
regional geol ogic and tectonic conditions of the site ofinterest may be so simil ar
to those of sites where actual strong motions have previousl y beenmeasured that
those strong motion records can be used directl y. Usual l y, this is not the case,and
artificial ground motions must be devel oped. Artificial ground motions can be
devel opedin a number of different ways. The main chal l enges in their devel opment
are to ensurethat they are consistent with the target parameters and that they are
real istic (i.e., that theircharacteristics are consistent with those of actual
earthquakes). This is not as easy as itmight appear; many motions that appear
reasonabl e in the time domain may not whenexamined in the frequency domain, and
vice versa. Many reasonabl e-l ooking time historiesof accel eration produce, after
integration, unreasonabl e time histories of vel ocity and/ordispl acement. The
qual ity of an artificial ground motion is very difficul t to discern by eye.The most
commonl y used methods for generation of artificial ground motions fal l into four
main categories: (a) modification of actual ground motion records, (b) generationof
artificial motions in the time domain, (c) generation of artificial motions in the
frequencydomain, and (d) generation of artificial motions using Green's function
techniques.8.5.1 Modification of Actual Ground Motion RecordsPerhaps the simpl est
approach to the generation of artificial ground motions is the modificationof
actual recorded ground motions. Maximum motion l evel s, such as peak accel erationand
peak vel ocity, have been used to rescal e actual strong motion records to higher or
l ower l evel s of shaking (Figure 8.20). Krinitszky and Chang (1979) recommended that
thescal ing factor (the ratio of the target ampl itude to the ampl itude of the record
being scal ed)shoul d be kept as cl ose to 1 as possibl e, and al ways between 0.25 and
4.0, and that anal ysesbe conducted with several scal ed records. Vanmarcke (1979),
noting that simpl e ampl itudescal ing fail s to account for differences in important
characteristics such as frequency contentand duration, suggested that l imits on the
scal ing factor shoul d be rel ated to the type ofprobl em to which the resul ting
motion is to be appl ied. For anal ysis of l inear el astic structures,the l imits of
Krinitszky and Chang (1979) were considered suitabl e, but for l iquefactiona scal ing
factor range of 0.5 to 2.0 was recommended.This type of rescal ing procedure
requires careful sel ection of the actual motion that isto be used. A desirabl e
ground motion record wil l not onl y have a peak accel eration orvel ocity cl ose to the
target val ue, but wil l have magnitude, distance, and l ocal site characteristicsthat
are simil ar to those of the target motion. Such a record is most l ikel y to have a
simil ar frequency content and duration to the target motion. Computer programs
(e.g., Dussomet aI., 1991; Ferritto, 1992) that contain, or at l east interact with,
strong motion databasesare avail abl e to aid in the sel ection of actual ground
motions for rescal ing.Rescal ing of the time scal e has been used to modify the
frequency content of an actual ground motion record. This is usual l y accompl ished by
mul tipl ying the time step of a digitizedSec. 8.5 Devel opment of Ground Motion Time
Histories 341a(t)a(t)(a)(b)Figure 8.20 (a) Original accel erogram from actual
earthquake; (b) rescal ed version of original accel erogram in which accel erations
were scal ed upward by a factor of 1.5 to match target peakaccel eration.actual
record by the ratio of the predominant period (Section 3.3.2.2) of the target
motion tothe predominant period of the actual motion (Figure 8.21). Since this
approach changes thefrequency content over the entire spectrum as wel l as the
duration of the rescal ed record, itshoul d be used careful l y to avoid unintended
consequences.To generate artificial ground motions ofl ong duration without
significantl y changingthe frequency content, some (e.g., Seed and Idriss, 1969)
have spl iced parts of actual groundmotion records together. Procedures of this type
must al so be used with caution. Careful examination of the reasonabl eness of spl iced
motions in both the time and frequencydomains is advised.8.5.2 Time-Domain
GenerationThe resembl ance of ground motion time histories to transient stochastic
processes wasnoted years ago (Housner, 1947). Since then, a number of procedures
that treat groundmotions as stochastic processes have been devel oped. Many of these
operate entirel y in thetime domain.A stationary stochastic process is one whose
statistics remain constant with time. Astationary accel erogram, for exampl e, woul d
have a constant mean accel eration, constantstandard deviation of accel eration, and
a constant frequency content-the accel erationswoul d continue indefinitel y. The fact
that the accel eration ampl itude of actual groundmotions varies with time (ground
motions have a beginning and an end, after al l ) renderstheir ampl itudes
nonstationary. Studies have al so shown that the frequency content of a typical
ground motion is al so nonstationary-it changes over the duration of shaking.342
Local Site Effects and Design Ground Motions Chap. 8a(t)a(t)(a)(b)Figure 8.21 (a)
Original accel erogram from actual earthquake; (b) rescal ed version of original
accel erogram in which time scal e was scal ed upward by a factor of 1.3 to match
target predominantperiod. Note that the duration has al so been increased by a
factor of 1.3.Generation of an artificial ground motion time
history in the time domain typical l yinvol ves mul tipl ying a stationary, fil tered
white noise (or fil tered Poisson process) signal byan envel ope function that
describes the buil dup and subsequent decay (nonstationarity) ofground motion
ampl itude (Shinozuka and Deodatis, 1988), as il l ustrated in Figure 8.22.More
recentl y, model s that consider the nonstationarity of both ampl itude and frequency
content (e.g., Sharma and Shah, 1986; Shinozuka and Deodatis, 1988) have been
devel oped.The use of autoregressive moving average (ARMA) model s (e.g., Chang et
al ., 1982;Kozin, 1988) has al so increased in recent years.(a)(b)(c)~ : > ..(d)Figure
8.22 Exampl e of time-domain generation of synthetic time history: (a) timehistory
of white noise is fil tered in the time domain to produce (b) time history of
fil teredwhite noise. Fil tered white noise is mul tipl ied by envel ope function in (c)
to produce theartificial ground motion shown in (d).Sec. 8.5 Devel opment of Ground
Motion Time Histories 3438.5.3 Frequency-Domain GenerationGround motions can be
generated quite convenientl y in the frequency domain by combininga Fourier
ampl itude spectrum with a Fourier phase spectrum. The ampl itude spectrum may be
computed from an actual ground motion spectrum or may be represented by some
theoretical means, such as a Brune spectrum or a power spectral density function
(Section 3.3.2.1). Thephase spectrum may be obtained from an actual ground motion
or may be computed from atime history given by the product of white noise and an
envel ope function (Figure 8.23). Someinvestigators (e.g., Ohsaki, 1979; Abrahamson
and Singh, 1986) have used phase differencedistributions as an indicator of phase
structure to devel op nonrandom, artificial phase spectra.Frequency-domain methods
are particul arl y useful for generating motions that areconsistent with target
response spectra. Computer programs such as EQGEN (Chang et aI.,1986b) and RASCAL
(Sil va, 1987) assume initial Fourier ampl itude and phase spectra, andthen
iterativel y adjust the ordinates of the Fourier ampl itude spectrum until a motion
consistentwith the target response spectrum is produced. The origin of the target
response spectrummust be kept in mind when generating spectrum-compatibl e motions.
Constant riskspectra (Chapter 4), for exampl e, represent the aggregate effect of
potential earthquakes ofmany different magnitudes occurring at many different
distances. Because a constant riskspectrum does not correspond to any particul ar
seismic event, a motion generated from aconstant risk target spectrum shoul d not be
expected to correspond to a particul ar seismicevent (Naiem and Lew, 1995).(a)(b)
Time.. Timeen n,u..(c) Frequency--v-~JUrt"f'".,,~• • ,rrv~ --(e)(d) Frequency.. Time
Figure 8.23 Exampl e of frequency-domain generation of synthetic time history: (a)
time history ofwhite noise is shaped by envel ope function to produce (b) time
history of envel oped white noise.Fourier transform of envel oped white noise is
performed to obtain (c) phase spectrum. Phasespectrum is combined with (d)
ampl itude spectrum to produce (e) synthetic time history.111.5.4 Green's Function
TechniquesThe Green's function approach to ground motion model ing is based on the
idea that the total motion at a particul ar site is equal to the sum of the motions
produced by a series of individual ruptures of many smal l patches on the causative
faul t. Obtaining the site motion344 Local Site Effects and Design Ground Motions
Chap. 8requires defining the geometry of the earthquake source, dividing the source
into a finitenumber of patches, defining the sequence in which the patches rupture,
defining the sl ipfunctions (functions describing the variation of sl ip displ acement
with time for each patch)across the source, and defining Green's functions
(functions that describe the motion at thesite due to an instantaneous unit sl ip at
the source; see Figure 8.24) across the source. Combiningthe Green's function with
the sl ip function gives the motion at the site due to sl ip ofeach individual patch.
Summing the effects of the sl ips of each patch whil e accounting forthe order in
which they rupture produces the overal l ground motion at the site. Obviousl y,the
summation procedure assumes that al l material s remain l inear.Cal cul ation of Green's
functions requires knowl edge of the vel ocity structure of thecrustal material s
between the source and site. However, estimation of the vel ocity structure,
particul arl y with respect to heterogeneities responsibl e for the scattering that
produces l atearrivingcoda waves, is a very difficul t probl em. Considerabl e
computational effort is al sorequired to cal cul ate Green's function; finite-el ement,
finite-difference, and ray theorytechniques are usual l y used for this purpose.
Hartzel l (1978) bypassed these computationsby using the weak motions of smal l
earthquakes as empirical Green's functions to simul atethe strong motion of l arge
earthquakes. Empirical Green's functions have the benefit ofautomatical l y retaining
the effects of the crustal vel ocity structure.The Green's function approach is
particul arl y useful for generating near-fiel dmotions, that is, motions at sites
cl ose enough to the faul t that the faul t dimensions becomesignificant (for far-
fiel d sites, the faul t can be treated as a point source without undue l ossof
accuracy). The nature of the rupture pattern, incl uding the general direction in
which ruptureprogresses, and the site azimuth (rel ative to the faul t) can strongl y
infl uence groundmotion in the near fiel d. The Green's function approach al l ows
phenomena such as directivityand fl ing (Section 2.5) to be refl ected in artificial
motions.SiteFigure 8.24 Schematic of Green's functions for a faul t divided into N
patches.Differences in the Green's functions for the different patches are due to
differences infocal depth, epicentral distance, and geol ogic structure al ong the
source-site path. OnceGreen's functions have been determined, site motions can
easil y be simul ated for avariety of faul t rupture patterns and sl ip functions.Sec.
8.6 Summary 3458.5.5 Limitations of Artificial Ground MotionsAs. discussed in
Chapter 3, actual ground motions are compl icated-they are infl uenced by,and
consequentl y refl ect, characteristics of the seismic source, the rupture process,
thesource-site travel path, and l ocal site conditions. Al though it is convenient to
characterizethem with a smal l number of parameters, it is important to remember
that such characterizationscan never be compl ete.Artificial motions that match a
smal l number of target parameters are not unique;many different motions can produce
the same target parameters. If such a set of motions areused to anal yze probl ems
for which damage correl ates wel l to the target parameters, the predicteddamage is
l ikel y to be consistent. For exampl e, a set of different motions with thesame peak
accel eration wil l produce simil ar base shears in a stiff, l inear el astic structure
founded on rock. The same set of motions, however, might produce a broad range of
baseshears in a fl exibl e and/or inel astic structure or in a structure founded on
soft soil . Theycoul d al so produce significantl y different estimates of permanent
sl ope movement or l iquefactionpotential . When using artificial motions, the
eventual use(s) of the motions mustal ways be reconcil ed with the criteria from
which they were devel oped.8.6 SUMMARY1. Historical references to the correl ation
between earthquake damage and l ocal siteconditions extend back nearl y 200 years.
Provisions that specifical l y accounted forl ocal site conditions did not appear in
buil ding codes, however, until the earl y 1970s.2. Evidence for the existence of
l ocal site effects is compel l ing. In addition to theoretical evidence, ampl ification
functions computed from measurements of surface and bedrockmotions at the same
l ocation, and comparisons of surface motion characteristicsfrom nearby sites with
different subsurface conditions, al l confirm the importanteffects of l ocal site
conditions on earthquake ground motions.3. Geometric effects can al so infl uence
gound motions. Though topographic irregul aritiesscatter seismic waves to produce
compl icated patterns of ampl ification anddeampl ification, motions at the crests of
ridges are general l y ampl ified. Al l uvial basins fil l ed with soft sediments can,
depending on their shapes, trap body waves andproduce surface waves within the
al l uvium. Ground motion in an al l uvial basin maybe considerabl y different than
those that woul d be predicted by one-dimensional ground response anal yses,
particul arl y near the edges of the basin.4. Earthquake-resistant design of new
structures and eval uation of the seismic vul nerabil ityof existing structures
invol ves prediction of their response to design groundmotions. Design ground
motions are obtained from design parameters devel opedfrom a prescribed design
earthquake or from a seismic hazard anal ysis.5. Design parameters may be devel oped
on a site-specific basis or they may be obtainedfrom buil ding codes. For a typical
site, parameters based on site-specific anal yses arel ikel y to be more accurate than
code-based parameters. They are al so l ikel y to resul tin more economical designs.346
Local Site Effects and Design Ground Motions Chap. 86. Buil ding codes have
historical l y devel oped on an incremental basis as new knowl edgeis obtained. The
seismic provisions of buil ding codes have fol l owed the sametrend with the most
significant changes occurring after major earthquakes.7. Buil ding codes are not
intended to el iminate earthquake damage compl etel y. Instead,they aim to produce
designs that wil l resist smal l ground motions without damage,moderate motions
without structural damage, and severe motions without col l apse.8. Ground motion
time histories with specified
target parameters can be devel oped byseveral different procedures. When design
parameters are used as targets, these procedureswil l produce design ground motions.
Procedures based on scal ing of recordedground motions and generation of artificial
ground motions are widel y used.HOMEWORK PROBLEMS8.1 Compute and compare the
response spectrum intensities at the State Buil ding, Al exanderBuil ding, and
Southern Pacific Company Buil ding in the 1957 San Francisco earthquake (seeFigure
8.4).8.2 An earthquake produces a peak accel eration of O.25g on a rock outcrop in
the far-fiel d. Estimatethe peak ground surface accel erations that woul d be expected
at stiff soil , deep cohesionl esssoil , and soft soil sites at the same distance from
the source of the earthquake.8.3 How woul d the frequency content of the bedrock
motion infl uence the rel ationships shown inFigures 8.10 and 8.11. Comment on the
ranges of appl icabil ity of these rel ationships.8.4 Consider the Gil roy No.1 (rock)
and Gil roy No.2 (soil ) ground motions described in Chapter3. Soil conditions at the
Gil roy No.2 (soil ) site were described in Exampl e 7.4. How wel l dothe motions at
these sites agree with the approximate rel ationships between peak accel erationson
rock and soil outcrops shown in Figure 8.12?8.5 How wel l do the shapes of the
response spectra of the Gil roy No.1 (rock) and Gil roy No.2(soil ) motions (Figure
3.15) agree with the average normal ized spectra shown in Figure 8.12?8.6 Vertical l y
propagating shear waves with a displ acement ampl itude of 1 cm at a great depth
approach the ground surface il l ustrated bel ow. Assuming that the topographic
effects can beapproximated by the triangul ar infinite wedge sol ution, estimate the
displ acement ampl itudesat points A, B, C, and D.oFigure P8.68.7 Determine and pl ot
(on arithmetic scal es) Newmark-Hal l design spectra (5% damping) for apeak
accel eration of 0.442g. Label the accel eration-control l ed, vel ocity-control l ed, and
displ acement-control l ed portions of the spectrum. How wel l does this design
spectrum correspondto the response spectrum of the Gil roy No.1 (rock) ground
motion?Chap. 8 Homework Probl ems 3478.8 Using the provisions of the Uniform
Buil ding Code, eval uate design base shear coefficients(the ratio of the design base
shear to the weight of the structure) for 8-story hospital buil dingsin Denver,
Seattl e, and San Francisco. Assume that the hospital s are constructed as ordinary
moment-resisting steel frames. Bedrock outcrops at the ground surface at the Denver
site. TheSeattl e site is l ocated on 700 feet of dense, gl acial l y overconsol idated
soil s. The San Franciscohospital site is underl ain by 45 feet of soft San Francisco
Bay Mud.8.9 Use the provisions of the Uniform Buil ding Code to devel op design
response spectra for theDenver and Seattl e hospital sites from Probl em 8.8. Briefl y
describe the steps that woul d berequired to devel op a design response spectrum for
the San Francisco site.9Liquefaction9.1 INTRODUCTIONLiquefaction is one of the most
important, interesting, compl ex, and controversial topics ingeotechnical earthquake
engineering. Its devastating effects sprang to the attention of geotechnical
engineers in a three-month period in 1964 when the Good Friday earthquake(Mw=9.2)
in Al aska was fol l owed by the Niigata earthquake (Ms =7.5) in Japan. Both
earthquakes produced spectacul ar exampl es of l iquefaction-induced damage, incl uding
sl ope fail ures, bridge and buil ding foundation fail ures, and fl otation of buried
structures. Inthe 30 years since these earthquakes, l iquefaction has been studied
extensivel y by hundredsof researchers around the worl d. Much has been l earned, but
the road has not been smooth.Different terminol ogies, procedures, and methods of
anal ysis have been proposed, and aprevail ing approach has been sl ow to emerge.
Inrecent years, many of these differences have been reconcil ed by the real ization
thattheir causes were due, in l arge part, to semantics. The term l iquefaction has
been used todescribe a number of different, though rel ated phenomena. Rather than
try to trace the convol uteddevel opment of the current state of knowl edge regarding
l iquefaction, this chapterwil l present a basic framework for the conceptual
understanding of l iquefaction-rel ated soil behavior and use it to describe the
various methods by which l iquefaction hazards can beeval uated. To do this, the
chapter introduces some new terminol ogy to distinguish between348Sec: . 9.2
Liquefaction-Rel ated Phenomena 349phenomena that have frequentl y been l umped
together under the heading of l iquefaction.The new terminol ogy al l ows these
phenomena to be il l ustrated in a way that simpl ifiesunderstanding of their
mechanics and the manner in which they contribute to earthquakedamage.9.2
LIQUEFACTION-RELATED PHENOMENAThe term l iquefaction, original l y coined by Mogami
and Kubo (1953), has historical l y beenused in conjunction with a variety of
phenomena that invol ve soil deformations caused bymonotonic, transient, or repeated
disturbance of saturated cohesionl ess soil s under undrainedconditions. The
generation of excess pore pressure under undrained l oading conditionsis a hal l mark
of al l l iquefaction phenomena. The tendency for dry cohesionl ess soil sto density
under both static and cycl ic l oading is wel l known. When cohensionl ess soil s are
saturated, however, rapid l oading occurs under undrained conditions, so the
tendency fordensification causes excess pore pressures to increase and effective
stresses to decrease.Liquefaction phenomena that resul t from this process can be
divided into two main groups: fl ow l iquefaction and cycl ic mobil ity.Both fl ow
l iquefaction and cycl ic mobil ity are very important, and any eval uation of
l iquefaction hazards shoul d careful l y consider both. In the fiel d, fl ow
l iquefaction occursmuch l ess frequentl y than cycl ic mobil ity but its effects are
usual l y far more severe. Cycl icmobil ity, on the other hand, can occur under a much
broader range of soil and site conditionsthan fl ow l iquefaction; its effects can
range from insignificant to highl y damaging. Inthis book, the generic term
l iquefaction wil l be taken to incl ude both fl ow l iquefaction andcycl ic mobil ity.
Fl ow l iquefaction and cycl ic mobil ity wil l be identified individual l y when
necessary.9.2.1 Fl ow LiquefactionFJl ow l iquefaction produces the most dramatic
effects of al l the l iquefaction-rel ated phenomena-tremendous instabil ities known as
fl ow fail ures. Fl ow l iquefaction can occurwhen the shear stress required for static
equil ibrium of a soil mass (the static shear stress)is greater than the shear
strength of the soil in its l iquefied state. Once triggered the l argedeformations
produced by fl ow l iquefaction are actual l y driven by static shear stresses. The
cycl ic stresses may simpl y bring the soil to an unstabl e state at which its
strength drops sufficientl yto al l ow the static stresses to produce the fl ow
fail ure. Fl ow l iquefaction fail uresare characterized by the sudden nature of their
origin, the speed with which they devel op,and the l arge distance over which the
l iquefied material s often move. The fl ow sl ide fail uresof Sheffiel d Dam (Figure
1.5) and Lower San Fernando Dam (Figure 1.7) are exampl es offl ow l iquefaction. The
fl uid nature of l iquefied soil is il l ustrated in Figure 9.1.9'.2.2 Cycl ic Mobil ity
Cycl ic mobil ity is another phenomenon that can al so produce unacceptabl y l arge
permanentdeformations during earthquake shaking. In contrast to fl ow l iquefaction,
cycl ic mobil ityoccurs when the static shear stress is l ess than the shear strength
of the l iquefied soil . Thedeformations produced by cycl ic mobil ity fail ures devel op
incremental l y during earthquake350 Liquefaction Chap. 9Figure 9.1 A smal l fl ow
sl ide al ong the shore of Lake Merced in San Francisco in1957 (photo by M. Bonil l a;
conrtesy of USGS).shaking. In contrast to fl ow l iquefaction, the deformations
produced by cycl ic mobil ity aredriven by both cycl ic and static shear stresses.
These deformations, termed l ateral spreading,can occur on very gentl y sl oping
ground or on virtual l y fl at ground adjacent to bodiesof water (Figure 9.2). When
structures are present, l ateral spreading can cause significantdamage (Figure 1.8
and 1.9).A special case of cycl ic mobil ity is l evel -ground l iquefaction. Because
static horizontal shear stresses that coul d drive l ateral deformations do not exist,
l evel -ground l iquefactioncan produce l arge, chaotic movement known as ground
oscil l ation during earthquake shaking,but produces l ittl e permanent l ateral soil
movement. Level -ground l iquefaction fail uresare caused by the upward fl ow of water
that occurs when seismical l y induced excess porepressures dissipate. Depending on
the l ength of time required to reach hydraul ic equil ibrium,l evel -ground
l iquefaction fail ure may occur wel l after ground shaking has ceased.Excessive
vertical settl ement and consequent fl ooding of l ow-l ying l and and the devel opmentof
sand boil s (Figure 9.3) are characteristic of l evel -ground l iquefaction fail ure.9.3
EVALUATION OF LIQUEFACTION HAZARDSBoth fl ow l iquefaction and cycl ic mobil ity can
produce damage at a particul ar site, and acompl ete eval uation of l iquefaction
hazards requires that the potential for each beaddressed. When faced with such a
probl em, the geotechnical earthquake engineer can systematical l yeval uate potential
l iquefaction hazards by addressing the fol l owing questions: Sec 9.4 Liquefaction
Susceptibil ity 351Figure 9.2 Lateral spreading of very fl at ground toward the
Motagua River fol l owingthe 1976 Guatemal a earthquake. Note orientation of ground
surface cracks paral l el toriver bank (photo by G. Pl afker; courtesy of USGS).1. Is
the soil susceptibl e to l iquefaction?2. If the soil is susceptibl e, wil l
l iquefaction be triggered?3. If l iquefaction is triggered, wil l damage occur?If
the answer to the first question is no, the l iquefaction hazard eval uation can be
terminatedwith the concl usion that l iquefaction hazards do not exist. Ifthe answer
is yes, the next questionmust be addressed. In some cases it may be more efficient
to reverse the order of thesecond and third questions, particul arl y when damage
appears unl ikel y. If the answers to al l three are yes, a probl em exists; if the
anticipated l evel of damage is unacceptabl e, the sitemust be abandoned or improved
(Chapter 12) or on-site structures strengthened. Thesequestions pertain to the
three most critical aspects of l iquefaction hazard eval uation: susceptibil ity,
initiation, and effects. Al l three must be considered in a comprehensive eval uation
of l iquefaction hazards.9.4 Ll Cl UEFACTION SUSCEPTIBILITYNot al l soil s are
susceptibl e to l iquefaction; consequentl y, the first step in a l iquefaction hazard
eval uation is usual l y the eval uation of l iquefaction susceptibil ity. If the soil at
a particul arsite is not susceptibl e, l iquefaction hazards do not exist and the
l iquefaction hazard eval uationcan be ended. If the soil is susceptibl e, however,
the matters ofl iquefaction initiation and352 Liquefaction Chap. 9Figure 9.3 Sand
boil s near Niigata, Japanfol l owing the 1964 Niigata earthquake.Sand boil s are often
al igned al ong cracks inthe ground (photo by K. Steinbrugge;courtesy of Earthquake
EngineeringResearch Center, University of Cal ifornia).effects must be addressed.
There are several criteria by which l iquefaction susceptibil ity canbe judged, and
some are different for fl ow l iquefaction and cycl ic mobil ity. These incl ude
historical , geol ogic, compositional , and state criteria.9.4.1 Historical CriteriaA
great deal of information on l iquefaction behavior has come from postearthquake
fiel dinvestigations, which have shown that l iquefaction often recurs at the same
l ocation whensoil and groundwater conditions have remained unchanged (Youd, 1984a).
Thus l iquefactioncase histories can be used to identify specific sites, or more
general site conditions, thatmay be susceptibl e to l iquefaction in future
earthquakes. Youd (1991) described a numberof instances where historical evidence
of l iquefaction has been used to map l iquefactionsusceptibil ity.Postearthquake
fiel d investigations have al so shown that l iquefaction effects havehistorical l y
been confined to a zone within a particul ar distance of the seismic source.
Ambraseys (1988) compil ed worl dwide data from shal l ow earthquakes to estimate a
l imitingepicentral distance beyond which l iquefaction has not been observed in
earthquakes ofdifferent magnitudes (Figure 9.4). The distance to which l iquefaction
can be expectedSec. 9.4 Liquefaction Susceptibil ity 353Figure 9.4 Rel ationship
between l imitingepicentral distance of sites at whichl iquefaction has been observed
and momentmagnitude for shal l ow earthquakes. Deepearthquakes (focal depths> 50 km)
haveproduced l iquefaction at greater distances.After Ambraseys (1988).5 10 50 100
500Epicentral distance (km): 21 I1 • I1 .1- - - - - .!. - - 1- 1 11II1I I I I I I.
1-.e.- -I - - - I - - 1- - -1- - - I - - I. - "10;-1 I I 1 I I·· • rI I I I I
_....: • • : I I I I. • • .Ie.·I I I • >,: ".: ~o 10 I- -I - - - 1" - - r- - -1+ -; -
eY.: -" - -t - - - 1"" - -I I I • I ._.: • • 1. I 0 I II I I. I : : : .... I I II 1 I.·
~: .... • 00 I I1 1 I .1· • 1 1 I 1- -·1- - - -I- - ............ - - - +- - -1- -
-I- - - - +- --" 0: • • ,z.... 1 I 1 1 1 1 1 1° I 1 11 1 1 I I I II 1 I I I I__ J. __
L __1 L __ 1__ J. L __I 1 I 1 1 1 1I 1 1 1 1 1 15698Q)-02'cOJtil E 7EQ)Eo: 2increases
dramatical l y with increasing magnitude. Whil e rel ationships of the type shown in
Figure 9.4 offer no guarantee that l iquefaction cannot occur at greater distances,
they arehel pful for estimation of regional l iquefaction hazard scenarios.9.4.2
Geol ogic CriteriaSoil deposits that are susceptibl e to l iquefaction are formed
within a rel ativel y narrow rangeof geol ogical environments (Youd, 1991). The
depositional environment, hydrol ogical environment, and age of a soil deposit al l
contribute to its l iquefaction susceptibil ity (Youdand Hoose, 1977).Geol ogic
processes that sort soil s into uniform grain size distributions and depositthem in
l oose states produce soil deposits with high l iquefaction susceptibil ity.
Consequentl y,fl uvial deposits, and col l uvial and aeol ian deposits when saturated,
are l ikel y to besusceptibl e to l iquefaction. Liquefaction has al so been observed in
al l uvial -fan, al l uvial pl ain,beach, terrace, pl aya, and estuarine deposits, but not
as consistentl y as in those l istedpreviousl y. The susceptibil ity of ol der soil
deposits to l iquefaction is general l y l ower thanthat of newer deposits. Soil s of
Hol ocene age are more susceptibl e than soil s of Pl eistoceneage, al though
susceptibil ity decreases with age within the Hol ocene. Liquefaction of
prePl eistocenedeposits is rare.Liquefaction occurs onl y in saturated soil s, so the
depth to groundwater (either free orperched) infl uences l iquefaction
susceptibil ity. Liquefaction susceptibil ity decreases withincreasing groundwater
depth; the effects of l iquefaction are most commonl y observed at354 Liquefaction
Chap. 9sites where groundwater is within a few meters of the ground surface. At
sites wheregroundwater l evel s fl uctuate significantl y, l iquefaction hazards may
al so fl uctuate.Human-made soil deposits al so deserve attention. Loose fil l s, such
as those pl aced withoutcompaction, are very l ikel y to be susceptibl e to
l iquefaction. The stabil ity of hydraul ic fil l dams and mine tail ings pil es, in which
soil particl es are l oosel y deposited by settl ing throughwater, remains an important
contemporary seismic hazard. Wel l -compacted fil l s, on the otherhand, are unl ikel y
to satisfy state criteria (Section 9.4.4) for l iquefaction susceptibil ity.9.4.3
Compositional CriteriaSince l iquefaction requires the devel opment of excess pore
pressure, l iquefaction susceptibil ityis infl uenced by the compositional
characteristics that infl uence vol ume changebehavior. Compositional characteristics
associated with high vol ume change potential tendto be associated with high
l iquefaction susceptibil ity. These characteristics incl ude particl esize, shape, and
gradation.For many years, l iquefaction-rel ated phenomena were thought to be l imited
to sands.Finer-grained soil s were considered incapabl e of generating the high pore
pressures commonl yassociated with l iquefaction, and coarser-grained soil s were
considered too permeabl eto sustain any generated pore pressure l ong enough for
l iquefaction to devel op. Morerecentl y, the bounds on gradation criteria for
l iquefaction susceptibil ity have broadened.Liquefaction of nonpl astic sil ts has
been observed (Ishihara, 1984, 1985) in the l aboratoryand the fiel d, indicating
that pl asticity characteristics rather than grain size al oneinfl uence the
l iquefaction susceptibil ity of fine-grained soil s. Coarse sil ts with bul ky particl e
shape, which are nonpl astic and cohesionl ess, are ful l y susceptibl e to l iquefaction
(Ishihara,1993); finer sil ts with fl aky or pl atel ike particl es general l y exhibit
sufficient cohesionto inhibit l iquefaction. Cl ays remain nonsusceptibl e to
l iquefaction, al though sensitivecl ays can exhibit strain-softening behavior simil ar
to that of l iquefied soil . Fine-grainedsoil s that satisfy each of the fol l owing
four Chinese criteria (Wang, 1979) may be consideredsusceptibl e to significant
strength l oss: Fraction finer than 0.005 mm z; 15%Liquid l imit, LL: s; 35%Natural
water content z 0.9 LLLiquidity index s; 0.75To account for differences in Chinese
and U.S. practice, the U.S. Army Corps of Engineersmodified the measured index
properties (by decreasing the fines content by 5%, increasingthe l iquid l imit by
1%, and increasing the natural water content by 2%) before appl ying theChinese
criteria to a cl ayey sil t in the foundation of Sardis Dam (Finn et aI., 1994).At
the other end of the grain size spectrum, l iquefaction of gravel s has been observed
in the fiel d (Coul ter and Migl iaccio, 1966; Chang, 1978; Wong, 1984; Youd et al .,
1985,Yegian et aI., 1994) and in the l aboratory (Wong et aI., 1975; Evans and Seed,
1987). Theeffects of membrane penetration (Section 6.3.2.3) are now thought to be
responsibl e for thehigh l iquefaction resistance observed in earl y l aboratory
investigations of gravel l y soil s.When pore pressure dissipation is impeded by the
presence of impermeabl e l ayers so thattrul y undrained conditions exist, gravel l y
soil s can al so be susceptibl e to l iquefaction.Sec. 9.4 Liquefaction Susceptibil ity
355Liquefaction susceptibil ity is infl uenced by gradation. Wel l -graded soil s are
general l yl ess susceptibl e to l iquefaction than poorl y graded soil s; the fil l ing of
voids betweenl arger particl es by smal l er particl es in a wel l -graded soil resul ts in
l ower vol ume changepotential under drained conditions and, consequentl y, l ower
excess pore pressures underundrained conditions. Fiel d evidence indicates that most
l iquefaction fail ures have invol veduniforml y graded soil s.Particl e shape can al so
infl uence l iquefaction susceptibil ity. Soil s with rounded particl eshapes are known
to densify more easil y than soil s with angul ar grains. Consequentl y,they are
usual l y more susceptibl e to l iquefaction than angul ar-grained soil s. Particl e
roundingfrequentl y occurs in the fl uvial and al l uvial environments where l oosel y
deposited saturatedsoil s are frequentl y found, and l iquefaction susceptibil ity is
often high in those areas.9.4.4 State CriteriaEven if a soil meets al l of the
preceding criteria for l iquefaction susceptibil ity, it stil l mayormay not be
susceptibl e to l iquefaction. Liquefaction susceptibil ity al so depends on the
initial state of the soil (i.e., its stress and density characteristics at the time
of the earthquake).Since
the tendency to generate excess pore pressure of a particul ar soil is strongl y
infl uencedby both density and initial stress conditions, l iquefaction
susceptibil ity dependsstrongl y on the initial state of the soil . These l iquefaction
susceptibil ity criteria, unl ike thosediscussed previousl y, are different for fl ow
l iquefaction and cycl ic mobil ity.To introduce contemporary methods for eval uating
state criteria (and to provide abackground for eval uating the effects of
l iquefaction), a brief historical review of somebasic concepts of cohesionl ess soil
behavior is required.9.4.4.1 Critical Void RatioIn his pioneering work on the shear
strength of soil s, Casagrande (1936) performeddrained, strain-control l ed triaxial
tests on initial l y l oose and initial l y dense sand specimens.The resul ts (Figure
9.5), which form the cornerstone of modern understanding of soil Axial strain(a)eL
ec eoVoid ratio(b)Figure 9.5 (a) Stress-strain and (b) stress-void ratio curves for
l oose and dense sandsat the same effective confining pressure. Loose sand exhibits
contractive behavior(decreasing void ratio) and dense sand exhibits dil ative
behavior (increasing void ratio)during shearing. By the time l arge strains have
devel oped, both specimens have reachedthe critical void ratio and mobil ize the same
l arge-strain shearing resistance.356 Liquefaction Chap. 9Figure 9.6 Use of the CVR
l ine as aboundary between l oose contractive statesand dense dil ative states.Loose
(contractive)Dense(dil ative)strength behavior, showed that al l specimens tested at
the same effective confining pressureapproached the same density when sheared to
l arge strains. Initial l y l oose specimens contracted,or densified, during shearing
and initial l y dense specimens first contracted, but thenvery quickl y began to
dil ate. At l arge strains, al l specimens approached the same densityand continued to
shear with constant shearing resistance. The void ratio corresponding tothis
constant density was termed the critical void ratio, e.. By performing tests at
differenteffective confining pressures, Casagrande found that the critical void
ratio was uniquel yrel ated to the effective confining pressure, and cal l ed the l ocus
the critical void ratio (CVR)l ine (Figure 9.6). By defining the state of the soil
in terms of void ratio and effective confiningpressure, the CVR l ine coul d be used
to mark the boundary between l oose (contractive)and dense (dil ative) states.The
equipment needed to measure pore pressure was not avail abl e at the time, but
Casagrande hypothesized that strain-control l ed undrained testing woul d produce
positiveexcess pore pressure (due to the tendency for contraction) in l oose
specimens, and negativeexcess pore pressure (due to the tendency for dil ation) in
dense specimens (Figure 9.7), until the CVR l ine was reached. This hypothesis was
subsequentl y verified experimental l y. TheCVR l ine therefore described the state
toward which any soil specimen woul d migrate atl arge strains, whether by vol ume
changes under drained conditions, changes in effectiveconfining pressure under
undrained conditions, or some combination under partial l ydrained conditions.eoDense
e(a) (b)Figure 9.7 Behavior of initial l y l oose and dense specimens under drained
andundrained conditions for (a) arithmetic and (b) l ogarithmic effective confining
pressurescal es.Since the CVR l ine marked the boundary between contractive and
dil ative behavior, itwas considered to mark the boundary between states in which a
particul ar soil was or was notsusceptibl e to fl ow l iquefaction (Figure 9.8).
Saturated soil s with initial void ratios highSec. 9.4 Liquefaction Susceptibil ity
357eNonsusceptibl eSusceptibl el og a'3eFigure 9.8 Use of CVR l ine as aboundary
between initial states that are andare not susceptibl e to fl ow l iquefaction.enough
to pl ot above the CVR l ine were considered susceptibl e to fl ow l iquefaction, and
soil swith initial states pl otting bel ow the CVR l ine were considered
nonsusceptibl e, However,when Fort Peck Dam in Montana suffered a static fl ow
l iquefaction fail ure of its upstreamsl ope during construction in 1938
(Middl ebrooks, 1942), a postfail ure investigation showedthat the initial state of
soil s that had obviousl y l iquefied actual l y pl otted somewhat bel ow theCVR l ine
(i.e., in the nonsusceptibl e region). Casagrande attributed this discrepancy to the
inabil ity of strain-control l ed drained tests to repl icate al l of the phenomena that
infl uence soil behavior under the stress-control l ed undrained conditions of an
actual fl ow l iquefaction fail ure,Over the years, Casagrande devel oped the
hypothesis that a fl owing l iquefied sand has a"fl ow structure" in which grains
continuousl y rotate to orient themsel ves in a structure ofminimum frictional
resistance (Casagrande, 1976). Casagrande was unabl e to achieve a fl owstructure in
the l aboratory until the l ate 1960s, when one of his students performed an
importantseries of stress-control l ed undrained triaxial tests (Castro, 1969).
9.4.4.2 Steady State of DeformationCastro (1969) performed static and cycl ic
triaxial tests on isotropical l y consol idatedspecimens and several static tests on
anisotropical l y consol idated specimens. Three differenttypes of stress-strain
behavior, il l ustrated for anisotropical l y consol idated specimens inFigure 9.9 were
observed. Very l oose specimens (such as specimen A in Figure 9.9) exhibiteda peak
undrained strength at a smal l shear strain and then "col l apsed" to fl ow rapidl y to
l arge strains at l ow effective confining pressure and l ow l arge-strain strength.
This type ofbehavior now recognized as fl ow l iquefaction was described at that time
as "l iquefaction."Dense specimens (specimen B) initial l y contracted but then
dil ated until a rel ativel y highconstant effective confining pressure and l arge-
strain strength was reached. At intermediatedensities (specimen C) the exceedance
of a peak strength at l ow strain was fol l owed by al imited period of strain-
softening behavior, which ended Withthe onset of dil ation at intermediatestrains.
[This reversal from contractive to dil ative behavior occurs at the phase
transformation point (Ishihara et aI., 1975).] Further l oading produced continued
dil ation tohigher effective confining pressures and, consequentl y, higher l arge-
strain strengths. Thistype of behavior was termed l imited l iquefaction.The testing
program showed a unique rel ationship between void ratio and effectiveconfining
pressure at l arge strains. Graphical l y, this rel ationship pl otted bel ow and roughl y
paral l el to the CVR l ine obtained from drained strain-control l ed tests; the
difference beingattributed to devel opment of the fl ow structure under stress-
control l ed conditions. The statein which the soil fl owed continuousl y under
constant shear stress and constant effectiveconfining pressure at constant vol ume
and constant vel ocity was l ater defined (Castro and358 Liquefaction Chap. 9q!'1uq
~----cLimitedl iquefaction~-------ALiquefactionLiquefaction__----""----- ALimited
l iquefactionB Dil ationBLimitedl iquefactionP'Figure 9.9 Liquefaction, l imited
l iquefaction, and dil ation in monotonic l oading tests.Poul os, 1977; Poul os, 1981)
as the steady state of deformation. Since the steady state ofdeformation is reached
onl y at l arge strains (after the effects of initial conditions such as soil fabric,
stress and strain history, and l oading conditions have been obscured), the
effectiveconfining pressure in an el ement of soil in the steady state of
deformation was considered todepend onl y on the density of the soil . Rel ativel y
recentl y it has been shown that the steadystateconditions are different for
compressive and extensional stress paths (e.g., Vaid et al .,1990; Reimer and Seed,
1992; Vaid and Thomas, 1995), particul arl y when the soil is depositedwith
inherentl y anisotropic structure. Specifical l y, pl uviated sands exhibit contractive
behavior over a wider range of densities when l oaded in extension than in
compression; aparticul ar el ement of sand at an intermediate density may exhibit
dil ative behavior in compressionbut contractive behavior when l oaded in extension.
As a resul t, the depositional conditions, stress conditions, and l oading conditions
that exist in the fiel d shoul d bematched as cl osel y as possibl e in l aboratory
investigations of steady-state behavior.The l ocus of points describing the
rel ationship between void ratio and effective confiningpressure in the steady state
of deformation is cal l ed the steady-state l ine (SSL). In itsmost general form, the
SSL can be viewed as a three-dimensional curve in e-cr'---r (Figure9.10) or e-r-«
space. The SSL shown in Figure 9.7a therefore represents the projection ofthe
three-dimensional SSL onto a pl ane of constant t, The SSL can al so be projected
ontopl anes of constant effective confining pressure (cr' =constant) and constant
density (e =constant). The SSL can al so be expressed in terms of the steady-state
strength, Ssu; since theshearing resistance of the soil in the steady state of
deformation is proportional to the effectiveconfining pressure, the strength-based
SSL is paral l el to the effective confining pressure-based SSL when both are pl otted
on l ogarithmic scal es (Figure 9.11).Sec. 9.4 Liquefaction Susceptibil ity.-.--...-
SSLProjection on r-o ' pl anea'359Figure 9.10 Three-dimensional steadystatel ine
showing projections on e-'t pl ane,e-a' pl ane, and 't-a' pl ane. A simil ar pl otcan be
devel oped using the stress pathparameters q and p' instead of 't and a'.e e(a)l og
8su(b)Figure 9.11 Proportional ity of Ssuto a3c produces strength-based and
effectiveconfining pressure-based steady-state l ines with identical sl opes.The SSL
is useful for identifying the conditions under which a particul ar soil mayormay not
be susceptibl e to fl ow l iquefaction (Figure 9.12). Soil s whose
state pl ots bel ow theSSL are not susceptibl e to fl ow l iquefaction. A soil whose
state l ies above the SSL wil l besusceptibl e to fl ow l iquefaction onl y if the static
shear stress exceeds its steady state (orresidual ) strength. Since the SSL can be
used to eval uate the shearing resistance ofl iquefiedsoil s, it is al so useful for
eval uating the potential effects of l iquefaction (Section 9.6).Al though
determination of the position of the SSL can be difficul t in practice (Section
9.6.4.1), the SSL is very useful for understanding the basic concepts of
l iquefaction.Cycl ic mobil ity, on the other hand, can occur in soil s whose state
pl ot above or bel owthe SSL. In other words, cycl ic mobil ity can occur in both l oose
and dense soil s.e Soil is susceptibl e to fl owl iquefaction if static stressis
greater than 8suSoil is not susceptibl eto fl ow l iquefactionl og a' or l og 8suFigure
9.12 State criteria for fl owl iquefaction susceptibil ity. Soil s withcombinations of
initial density and stressconditions that pl ot above the SSL aresusceptibl e to fl ow
l iquefaction when thestatic shear strength is greater than thesteady-state
strength. Initial conditions thatpl ot bel ow the SSL are not susceptibl e tofl ow
l iquefaction.360 Liquefaction Chap. 9The l ocation of the SSL is sensitive to the
compositional characteristics of the soil itsvertical position is strongl y
infl uenced by gradation and its sl ope by particl e angul arity.Soil s with rounded
particl es usual l y have fl at SSLs-a characteristic that often l eads to difficul tyin
the estimation of in situ steady-state strength.9.4.4.3 State Para~eterThe nature
of the steady-state l ine il l ustrates the l imited appl icabil ity of absol ute measures
of density, such as void ratio and rel ative density, for characterization of a
potential l yl iquefiabl e soil . As il l ustrated in Figure 9.12, an el ement of soil at a
particul ar void ratio(hence a particul ar density and rel ative density) can be
susceptibl e to fl ow l iquefactionunder a high effective confining pressure but
nonsusceptibl e at a l ow effective confiningpressure.Using concepts of critical -
state soil mechanics, the behavior of a cohesionl ess soil shoul d be more cl osel y
rel ated to the proximity of its initial state to the steady-state l ine thanto
absol ute measures of density (Roscoe and Pooroshasb, 1963). In other words, soil s
instates l ocated at the same distance from the steady-state l ine shoul d exhibit
simil ar behavior.Using this l ogic, a state parameter (Been and Jeffries, 1985) can
be defined as(9.1)where ess is the void ratio of the steady-state l ine at the
effective confining pressure of interest(Figure 9.13). When the state parameter is
positive, the soil exhibits contractive behaviorand may be susceptibl e to fl ow
l iquefaction. When it is negative, dil ative behavior wil l occur and the soil is not
susceptibl e to fl ow l iquefaction. The state parameter has beenrel ated to friction
angl e, dil ation angl e, CPT resistance (Been et al ., 1986, 1987; Sl aden,1985), PMT
resul ts (Yu, 1994), and DMT resul ts (Konrad, 1988). Ishihara (1993) showedthat the
abil ity of the state parameter to characterize soil behavior of very l oose sands
underl ow effective confining pressures may be l imited and proposed an anal ogous
parameter (thestate index) based on the rel ative distance between the initial state
and the quasi-steadystatel ine (a l ine anal ogous to and l ocated sl ightl y bel ow the
SSL which corresponds to thestress and density conditions at the phase
transformation points observed in cases of l imitedl iquefaction).The concept of the
state parameter is very useful and the possibil ity of determining itsval ue from in
situ tests is appeal ing. The accuracy with which the state parameter can be
determined, however, is infl uenced by the accuracy with which the position of the
SSL canbe determined (Section 9.6.3.1).eInitial state ~• Steady statel og o' 3c
Figure 9.13 State parameter.Sec. 9.5 Initiation of Liquefaction 3619.5 INITIATION
OF LIQUEFACTIONThe fact that a soil deposit is susceptibl e to l iquefaction does not
mean that l iquefaction wil l necessaril y occur in a given earthquake. Its occurrence
requires a disturbance that is strongenough to initiate, or trigger, it. Eval uation
of the nature of that disturbance is one of themost critical parts of a
l iquefaction hazard eval uation. Any discussion of the initiation ofl iquefaction
must specify which l iquefaction-rel ated phenomena are being considered.Many
previous studies of l iquefaction initiation have impl icitl y l umped fl ow
l iquefactionand cycl ic mobil ity together, but since they are distinctl y different
phenomena, it is moreappropriate to consider each separatel y.Al though cycl ic
mobil ity is an earthquake rel ated phenomenon, fl ow l iquefaction canbe initiated in
a variety of ways. Fl ow sl ides triggered by monotonic l oading (static l iquefaction)
have been observed in natural soil deposits (Koppejan et al ., 1948; Andersen and
Bjerrum, 1968; Bjerrum, 1971; Kramer, 1988), man-madefil l s (Middl ebrooks, 1942;
Cornforthet aI., 1975; Mitchel l , 1984), and mine tail ings pil es (Kl einer, 1976;
Jennings, 1979;Eckersl ey, 1985). Fl ow l iquefaction has al so been triggered by
nonseismic sources of vibration,such as pil e driving (Jakobsen, 1952; Broms and
Bennermark, 1967), train traffic (Fel l enius,1953), geophysical expl oration (Hryciw
et al ., 1990), and bl asting (Conl on, 1966;Carter and Seed, 1988). Perhaps somewhat
ironical l y, the study of static l iquefaction overthe past 10 to 15 years has
contributed greatl y to improved understanding of seismical l yinduced l iquefaction by
identifying the effective stress conditions at which l iquefactionphenomena are
initiated.Understanding the initiation of l iquefaction requires identification of
the state of thesoil when l iquefaction is triggered. In the fol l owing sections,
these conditions wil l be presentedin a framework that al l ows the mechanics of both
fl ow l iquefaction and cycl ic mobil ityto be cl earl y understood. Subsequentl y,
practical and commonl y used procedures fordetermining the nature of the disturbance
required to move from initial state to the state atwhich l iquefaction is triggered
wil l be presented.9.5.1 Fl ow Liquefaction SurfaceThe conditions at which fl ow
l iquefaction is initiated are most easil y il l ustrated with the aidof the stress
path (Section 6.2.2). Hanzawa et al . (1979) first showed that the effective stress
conditions at which strain-softening behavior occurred in l oose, saturated sands
coul d bedescribed very simpl y in stress path space. As discussed in the fol l owing
sections, the effectivestress conditions at the initiation of fl ow l iquefaction can
be described in stress pathspace by a three-dimensional surface that wil l be
referred to hereafter as the fl ow l iquefactionsurface (FLS). Whil e some practical
difficul ties in the measurement ofthe FLS for general stress paths remain, it
provides (in conjunction with steady-state concepts) a very useful framework for
conceptual understanding of the rel ationships between the various l iquefaction
phenomena. This conceptual understanding is vital for proper eval uation of the
behaviorof l iquefiabl e soil s both during and after earthquake shaking.9.5.1.1
Monotonic LoadingThe conditions at the initiation of fl ow l iquefaction can be seen
most easil y when thesoil is subjected to monotonical l y increasing stresses.
Consider, for exampl e, the responseof an isotropical l y consol idated specimen of
very l oose, saturated sand in undrained,362 Liquefaction Chap. 9stress-control l ed
triaxial compression (Figure 9.14). Immediatel y prior to undrainedshearing (point
A), the specimen is in drained equil ibrium under an initial effective confining
pressure, 0"3c, with zero shear stress (Figure 9.14a,b) and zero excess pore
pressure(Figure 9.14c). Since its initial state is wel l above the SSL (Figure
9.14d), the sand wil l exhibit contractive behavior. When undrained shearing begins,
the contractive specimengenerates positive excess pore pressure as it mobil izes
shearing resistance up to a peakval ue (point B) that occurs at a rel ativel y smal l
strain. The excess pore pressure at point Bis al so rel ativel y smal l ; the pore
pressure ratio, ru = uexcessl 0"3c , is wel l bel ow 1.00. Atpoint B, however, the
specimen becomes unstabl e, and because it is l oaded under stresscontrol l ed
conditions, col l apses (the axial strain may increase from l ess than 1% to morethan
20% in a fraction of a second). As the specimen strains from point B to point C,
theexcess pore pressure increases dramatical l y. At and beyond point C, the specimen
is in thesteady state of deformation and the effective confining pressure is onl y a
smal l fraction ofthe initial effective confining pressure. This specimen has
exhibited fl ow l iquefactionbehavior; the static shear stresses required for
equil ibrium (at point B) were greater thanthe avail abl e shear strength (at point C)
of the l iquefied soil . Fl ow l iquefaction was initiatedat the instant it became
irreversibl y unstabl e (i.e., at point B).B...................(d)Figure 9.14
Response of isotropic al l y consol idated specimen of l oose, saturated sand: (a)
stress-strain curve; (b) effective stress path; (c) excess pore pressure; (d)
effectiveconfining pressure.A• Ap'Now consider the response of a series of triaxial
specimens initial l y consol idated tothe same void ratio at different effective
confining pressures. Since al l of the specimenshave the same void ratio, they wil l
al l reach the same effective stress conditions at the steadystate, but they wil l
get there by different stress paths. Figure 9.15 il l ustrates the response ofeach
specimen under monotonic l oading. The initial states of specimens A and B are bel ow
the SSL, so they exhibit dil ative behavior upon shearing. Specimens C, D, and E al l
exhibitcontractive behavior; each reaches a peak undrained
strength after which they strain rapidl ytoward the steady state. For specimens C,
D, and E, fl ow l iquefaction is initiated at the peakof each stress path (at the
points marked with an x), Hanzawa et al . (1979), Vaid and Chern(1983), and a number
of more recent investigations have shown that the l ocus of pointsSec. 9.5
Initiation of Liquefaction 363qp'eSSLp'Figure 9.15 Response of five specimens
isotropic al l y consol idated to the same initial void ratio at different initial
effectiveconfining pressures. Fl ow l iqnefaction inspecimens C, D, and E is
initiated at thepoints marked with an x. The dotted l inepassing through these
points is a l ine ofconstant principal effective stress ratio, KLdescribingthe
effective stress conditions at the initiation of fl ow l iquefaction is a straight
l ine (the dotted l ine in Figure 9.15) that projects through the origin of the
stress path. Graphical l y,these points may be used to define the fl ow l iquefaction
surface (FLS) in stress pathspace; since fl ow l iquefaction cannot occur if the
stress path is bel ow the steady-state point,the FLS is truncated at that l evel
(Figure 9.16). This form of the FLS was first proposed(with a different name) by
Vaid and Chern (1985). It shoul d be noted that Sl aden et aI.(1985) proposed an
anal ogous surface (cal l ed the col l apse surface) that was assumed toproject l inearl y
through the steady-state point; since the preponderance of current experimental
evidence appears to support projection through the origin, the term FLS is used in
this book. For very l oose sampl es, the steady-state point may be so cl ose to the
origin thatthe practical difference between the FLS and the col l apse surface is
negl igibl e.The FLS marks the boundary between stabl e and unstabl e states in
undrained shear. Ifthe stress conditions in an el ement of soil reach the FLS under
undrained conditions,whether by monotonic or cycl ic l oading, fl ow l iquefaction wil l
be triggered and the shearingqFl owl iquefactionsurfaceFigure 9.16 Orientation of the
fl owp' l iquefaction surface in stress path space.364 Liquefaction Chap. 9q
resistance wil l be reduced to the steady-state strength. Therefore, the FLS
describes the conditionsat which fl ow l iquefaction is initiated.For isotropic
initial conditions, the sl ope of the FLS is often about two-thirds the sl opeof the
drained fail ure envel ope for cl ean sands. Specimens tested under anisotropic
initial conditions, however, indicate that the FLS is steeper for soil s with high
initial (drained)shear stress compared to soil s with l ower initial shear stress at
the same void ratio (Figure9.17). The FLS may be very cl ose to the initial stress
point when initial shear stresses arel arge, in which case fl ow l iquefaction may be
initiated by onl y a very smal l undrained disturbance(Kramer and Seed, 1988). Case
histories that have been attributed to spontaneousl iquefaction probabl y invol ved
initial shear stresses that were high enough that the smal l undrained disturbance
required to initiate fl ow l iquefaction was not observed.The l imited l iquefaction
behavior exhibited by specimens C and D (Figure 9.16) issignificant for cases in
which the static shear stress increases (as in the case of monotonicl oading
described in this section). In such cases the shearing resistance may drop to
val uesat the point of phase transformation (or the quasi -steady state) that are
l ower than the steadystatestrength. This temporary drop in shearing resistance may
produce shear strains of 5%to 20% (Ishihara, 1993) and resul t in unacceptabl y l arge
permanent deformations. Becausethe effects of initial conditions are not erased
compl etel y at these strain l evel s, they infl uencethe quasi-steady-state strength.
Procedures for estimation of quasi-steady-statestrength are given by Ishihara
(1993). Because the static component of shear stress general l yremains constant or
decreases during earthquakes, the quasi-steady state is l ess l ikel y tobe reached as
a resul t of earthquake shaking.Kc=KfKc=2.5Kc=2Kc =1Figure 9.17 Variation of fl ow
l iquefaction surface incl ination with initial principal effective stress ratio for
constantp' void ratio.9.5.1.2 Cycl ic LoadingVaid and Chern (1983) first showed that
the FLS appl ied to both cycl ic and monotonicl oading, and a considerabl e amount of
independent experimental evidence supports thatobservation. Other experimental
evidence (e.g., Al arcon-Guzman et al ., 1988) suggests thatthe effective stress path
can move somewhat beyond the FLS before l iquefaction is initiatedby cycl ic l oading.
Whether l iquefaction is initiated precisel y at the FLS under cycl ic as wel l as
monotonic l oading is not currentl y known with certainty. Because the FLS is used as
partof a conceptual model of l iquefaction behavior in this book, and because it is
sl ightl y moreconservative to do so, the FLS wil l be assumed to appl y to both cycl ic
and monotonic l oading.Sec. 9.5 Initiation of Liquefaction 365qBAcqtT1: static tp'
Figure 9.18 Initiation of fl ow l iquefaction by cycl ic and monotonic l oading.
Al thoughthe stress conditions at the initiation of l iquefaction are different for
the two types ofl oading (points B and D), both l ie on the FLS.Consider the
responses of two identical , anisotropic al l y consol idated, triaxial specimensof
l oose, saturated sand (Figure 9.18). Initial l y, the specimens are in drained
equil ibrium(point A) under a static shear stress, 'tstatic, that is greater than
the steady-state strength,SSU" The first specimen is l oaded monotonical l y (under
undrained conditions): the shearingresistance buil ds up to a peak val ue when the
stress path reaches the FLS (point B). At thatpoint the specimen becomes unstabl e
and strains rapidl y toward the steady state (point C).The second specimen is l oaded
cycl ical l y (al so under undrained conditions): the effectivestress path moves to the
l eft as positive excess pore pressures devel op and permanent srraina.,accumul ate.
When the effective stress path reaches the FLS (at point D), the specimenbecomes
unstabl e and strains toward the steady state of deformation (point C). Al though the
effective stress conditions at the initiation of l iquefaction (points B and D) were
different,they fel l in both cases on the FLS. The FLS, therefore, marks the onset
of the instabil ity thatproduces fl ow l iquefaction. Lade (1992) provided a detail ed
description of this instabil ityfrom a continuum mechanics standpoint.9.5.1.3
Devel opment of Fl ow LiquefactionFl ow l iquefaction occurs in two stages. The first
stage, which takes pl ace at smal l strain l evel s, invol ves the generation of
sufficient excess pore pressure to move the stresspath from its initial position to
the FLS. This excess pore pressure may be generated by undrainedmonotonic or cycl ic
l oading. When the effective stress path reaches the FLS, the soil becomes inherentl y
unstabl e and the second stage begins. The second stage invol ves strainsoftening(and
additional excess pore pressure generation) that is driven by the shearstresses
required for static equil ibrium. These shear stresses are the driving stresses-they
must be distinguished from the l ocked-in stresses that devel op during deposition
and consol idationof the soil (Castro, 1991). Locked-in shear stresses, such as
those that existbeneath l evel ground when Ko: : f: . 1, cannot drive a fl ow
l iquefaction fail ure. Large strainsdevel op in the second stage as the effective
stress path moves from the FLS to the steadystate. If the first stage takes the
soil to the FLS under undrained, stress-control l ed conditions,the second stage is
inevitabl e.3669.5.2 Infl uence of Excess Pore PressureLiquefaction Chap. 9qThe
generation of excess pore pressure is the key to the initiation of l iquefaction.
Withoutchanges in pore pressure, hence changes in effective stress, neither fl ow
l iquefaction norcycl ic mobil ity can occur. The different phenomena can, however,
require different l evel sof pore pressure to occur.9.5.2.1 Fl ow LiquefactionFl ow
l iquefaction can be initiated by cycl ic l oading onl y when the shear stressrequired
for static equil ibrium is greater than the steady-state strength. In the fiel d,
theseshear stresses are caused by gravity and remain essential l y constant until
l arge deformationsdevel op. Therefore, initial states that pl ot in the shaded region
of Figure 9.19 are susceptibl eto fl ow l iquefaction. The occurrence of fl ow
l iquefaction, however, requires an undraineddisturbance strong enough to move the
effective stress path from its initial point to the FLS.rSteadystate
point~--,: ;..",o__p'Figure 9.19 Zone of susceptibil ity to fl owl iquefaction. If
initial conditions fal l withinthe shaded zone, fl ow l iquefaction wil l occur if an
undrained disturbance brings theeffective stress path from the pointdescribing the
initial conditions to the FLS.If the initial stress conditions pl ot near the FLS,
as they woul d in an el ement of soil subjectedto l arge shear stresses under drained
conditions, fl ow l iquefaction can be triggered bysmal l excess pore pressures
(Kramer and Seed, 1988). The l iquefaction resistance wil l begreater if the initial
stress conditions are farther from the fl ow l iquefaction surface. The FLScan be
used to estimate the pore pressure ratio at the initiation of fl ow l iquefaction; it
decreases substantial l y with increasing initial stress ratio (Figure 9.20) for
soil s at a particul arvoid ratio. At high initial stress ratios, fl ow l iquefaction
can be triggered by very smal l static or dynamic disturbances.
1.0 ..------------------------------,,-" 0.51.5 2.0 2.5 3.0 3.5Pricipal effective
stress ratio, KcFigure 9.20 Variation of pore pressure ratio (ru,t= u/cr£,)
required to trigger fl owl iquefaction in triaxial specimens of Sacramento River Fine
Sand with initial principal effective stress ratio.Sec. 9.5 Initiation of
Liquefaction 3679.5.2.2
Cycl ic Mobil ityAl though fl ow l iquefaction cannot occur, cycl ic mobil ity can
devel op when the staticshear stress is smal l er than the steady-state shear
strength. Therefore, initial states that pl otin the shaded region of Figure 9.21
are susceptibl e to cycl ic mobil ity. Note that cycl icmobil ity can occur in both
l oose and dense soil s (the shaded region of Figure 9.21 extendsfrom very l ow to
very high effective confining pressures and corresponds to states thatwoul d pl ot
both above and bel ow the SSL). The devel opment of cycl ic mobil ity can be
il l ustratedby the response of soil s in cycl ic triaxial tests. Three combinations of
initial conditionsand cycl ic l oading conditions general l y produce cycl ic mobil ity.q
p'Figure 9.21 Zone of susceptibil ity tocycl ic mobil ity. If initial conditions pl ot
within shaded zone, cycl ic mobil ity canoccur.The first, il l ustrated in Figure
9.22a, occurs when 'tstatic - 't cyc > 0 (i.e., no shear stressreversal ) and
'tstatic + 'tcyc < Ssu (no exceedance of steady-state strength). In this case the
effective stress path moves to the l eft until it reaches the drained fail ure
envel ope. Since itcannot cross the drained fail ure envel ope, additional l oading
cycl es simpl y cause it to moveup and down al ong the envel ope. As a resul t, the
effective stress conditions stabil ize. Fl owtypedeformations cannot devel op because
any unidirectional straining woul d induce dil ation.The effective confining pressure
has decreased significantl y, and the resul ting l owstiffness can al l ow significant
permanent strains to devel op within each l oading cycl e.The second condition (Figure
9.22b) occurs when 'tstatic - 'tcyc > 0 (no stress reversal )and 'tstatic + 'tcyc >
Ssu (steady-state strength is exceeded momentaril y). Again, cycl ic l oadingwil l
cause the effective stress path to move to the l eft. When it touches the FLS,
momentaryperiods of instabil ity wil l occur. Significant permanent strain may
devel op during theseq(a)p'qFigure 9.22 Three cases of cycl ic mobil ity: (a) no
stress reversal and no exceedance ofthe steady-state strength; (b) no stress
reversal with momentary periods of steady-statestrength exceedance; (c) stress
reversal with no exceedance of steady-state strength.368 Liquefaction Chap. 9
periods, particul arl y if'Tstatic is greater than the quasi-static shear strength,
but the strainingwil l general l y cease at the end of cycl ic l oading when the shear
stress returns to 'Tstatic'The final condition is that in which 'Tstatic - 'Tcyc <
0 (stress reversal occurs) and 'Tstatic+ 'Tcyc < Ssu (steady-state strength is not
exceeded). In this case (Figure 9.22c) the directionof the shear stress changes so
that each cycl e incl udes both compressional and extensional l oading. Experimental
evidence (e.g., Dobry et al ., 1982; Mohamad and Dobry, 1986) hasshown that the rate
of pore pressure generation increases with increasing degree of stressreversal .
Hence the effective stress path moves rel ativel y quickl y to the l eft (because
excesspore pressure buil ds up quickl y) and eventual l y oscil l ates al ong the
compression and extensionportions of the drained fail ure envel ope. Each time the
effective stress path passesthrough the origin (it does so twice during each
l oading cycl e), the specimen is in an instantaneousstate of zero effective stress
(ru =100%). Al though this state of zero effective stressis referred to as initial
l iquefaction (Seed and Lee, 1966), it shoul d not be taken to impl y thatthe soil has
no shear strength. If monotonic l oading is appl ied at the state of initial
l iquefaction,the specimen wil l dil ate until the steady-state strength is mobil ized
(Figure 9.23).Significant permanent strains may accumul ate during cycl ic l oading,
but fl ow fail ure cannotoccur. Note that initial l iquefaction can onl y occur when
stress reversal s occur.In contrast to fl ow l iquefaction, there is no cl earcut point
at which cycl ic mobil ity is initiated.Permanent strains, and the permanent
deformations they produce, accumul ate incremental l y.Their magnitude depends on the
static shear stress and the duration of the groundmotion. For ground motions of
short duration at nearl y l evel sites, permanent deformationsmay be smal l . For
moderatel y sl oping sites or gentl y sl oping sites subjected to groundmotions of l ong
duration, cycl ic mobil ity can produce damaging l evel s of soil deformation.q qSteady
statepoint-----,/?--_...-Figure 9.23 Dil ative behavior of specimen l oaded
monotonical l y after occurrence ofcycl ic mobil ity. Cycl ic l oading with stress
reversal causes the effective confiningpressure to decrease rapidl y, eventual l y
reaching momentary val ues of zero. Subsequentmonotonic l oading, however, causes
dil ation as the steady-state strength is mobil ized.FLSp'9.5.3 Eval uation of
Initiation of LiquefactionThe combination of steady-state and fl ow l iquefaction
surface concepts described in Section9.5.2 provides a framework in which the basic
mechanisms of l iquefaction can be understood.This framework integrates l iquefaction
susceptibil ity with l iquefaction initiation andSec. 9.5 Initiation of Liquefaction
369l iquefaction effects. It al so il l ustrates the important infl uence of excess pore
pressure generationon the extent of l iquefaction-rel ated hazards.A number of
approaches to eval uation of the potential for initiation of l iquefactionhave
devel oped over the years. In the fol l owing sections, the most common of these-the
cycl ic stress approach and a useful al ternative, the cycl ic strain approach-are
presented.Each has advantages and l imitations, and each is preferred by different
groups of engineers.For particul arl y important projects, it is not unusual to use
more than one approach in a l iquefactionhazard eval uation.9.5.3.1 Cycl ic Stress
ApproachIn the 1960s and 1970s, many advances in the state of knowl edge of
l iquefaction phenomenaresul ted from the pioneering work of H. B. Seed and his
col l eagues at the Universityof Cal ifornia at Berkel ey. This research was directed
l argel y toward eval uation of thel oading conditions required to trigger
l iquefaction. This l oading was described in terms ofcycl ic shear stresses, and
l iquefaction potential was eval uated on the basis of the ampl itudeand number of
cycl es of earthquake-induced shear stress. The general approach has come tobe known
as the cycl ic stress approach.Seed and Lee (1966) defined initial l iquefaction as
the point at which the increase inpore pressure is equal to the initial effective
confining pressure [i.e., when Uexcess =<J3c (orwhen ru = 100%)]. Because most of
the earl y l aboratory testing investigations were basedon cycl ic triaxial tests on
isotropical l y consol idated specimens (consequentl y, with compl etestress reversal ),
initial l iquefaction coul d be produced in both l oose and dense specimens.According
to the definitions of Section 9.2, this behavior woul d now be cl assified(since the
static shear stress was zero) as cycl ic mobil ity. The use of the term
initial l iquefactionl ed many to the erroneous bel ief that fl ow l iquefaction coul d be
initiated in any l ooseor dense cohesionl ess soil .The cycl ic stress approach is
conceptual l y quite simpl e: the earthquake-induced l oading,expressed in terms of
cycl ic shear stresses, is compared with the l iquefaction resistanceof the soil ,
al so expressed in terms of cycl ic shear stresses. At l ocations where the l oading
exceeds the resistance, l iquefaction is expected to occur. Appl ication of the
cycl ic stressapproach, however, requires careful attention to the manner in which
the l oading conditionsand l iquefaction resistance are characterized.
Characterization of Earthquake Loading. The l evel of excess pore pressurerequired
to initiate l iquefaction is rel ated to the ampl itude and duration of
earthquakeinducedcycl ic l oading. The cycl ic stress approach is based on the
assumption that excess porepressure generation is fundamental l y rel ated to the
cycl ic shear stresses, hence seismic l oadingis expressed in terms of cycl ic shear
stresses. The l oading can be predicted in two ways: bya detail ed ground response
anal ysis or by the use of a simpl ified approach.Ground response anal yses (Chapter
7) can be used to predict time histories of shearstress at various depths within a
soil deposit. Such anal yses produce time histories with thetransient, irregul ar
characteristics of actual earthquake motions. However, the l aboratorydata from
which l iquefaction resistance can be estimated are usual l y obtained from tests in
which the cycl ic shear stresses have uniform ampl itudes. Therefore, comparison of
earthquake-induced l oading with l aboratory-determined resistance requires
conversion of anirregul ar time history of shear stress to an equival ent series of
uniform stress cycl es. Seed et370 Liquefaction Chap. 9a1. (197 5a) appl ied a
weighting procedure to a set of shear stress time histories fromrecorded strong
ground motions to determine the number of uniform stress cycl es, Neq (at an
ampl itude of 65% of the peak: cycl ic shear stress, i.e., 'tcyc =O.65'tmax) that
woul d produce anincrease in pore pressure equival ent to that of the irregul ar time
history (Figure 9.24). Simil arrel ationships have been devel oped for other stress
l evel s (e.g., Ha1dar and Tang, 1981)but the 65% l evel is most commonl y used. In al l
cases, the equival ent number of uniformstress cycl es increases with increasing
earthquake magnitude (just as strong-motion durationincreases with increasing
earthquake magnitude).40 Mean +1 Standarddeviation7 8 9Earthquake magnitude, M6
Ol .---...L----'------'-----''-------'-------'------L._----'_--'5xesEpL{)(0~ 30 co
(/)Q)~'01l 20E: JCCQ)~.~ 10wFigure 9.24 Numberof equival ent uniform stress cycl es,
Neq, for earthquakes ofdifferent magnitude. (AfterSeed,et aI., 1975a.)Exampl e 9.1
Figure E9.1 shows a typical irregul ar time history of shear stress that was
produced by aMw =7.0 earthquake. Estimate the ampl itude and number of l oading
cycl es of an equival ent series ofuniform stress cycl es.Sol ution Figure E9.1 shows
that the maximum shear stress is 780 Ib/ft2 (37.4 kPa). Consequentl y,the ampl itude
of the equival ent series of uniform stress cycl es is'teye = 0.65'tmax = (0.65)
(780Ib/ft2) = 507 l b/ft? (24.3 kPa)Referring to Figure 9.28, the corresponding
equival ent number of cycl es is approximatel y 10(al though val ues ranging from 5 to
14 woul d fal l within one standard deviation of that val ue).Sec. 9.5 Initiation of
Liquefaction 371800600~ 400-Sop 200en C/) ~ 0Ci5Cii 200(]) ..c(f) 4006008000 6 12
18Time, (sec)24 30Figure E9.1 Typical irregul ar time history of shear stress. For
l iquefaction purposes,10 cycl es nf harmonic shear stress at 510 Ib/ft2 woul d be
considered equival ent to thistime history.The uniform cycl ic shear stress ampl itude
for l evel (or gentl y sl oping) sites can al sobe estimated from a simpl ified
procedure (Seed and Idriss, 1971) as(9.2)where amax is the peak ground surface
accel eration, g the accel eration of gravity, o, the total vertical stress, and rd
the val ue of a stress reduction factor (Figure 9.25) at the depth of interest.This
uniform cycl ic shear stress is assumed to be appl ied for the equival ent number of
cycl es shown in Figure 9.24.Regardl ess of whether a detail ed ground response
anal ysis or the simpl ified procedureis used, the earthquake-induced l oading is
characterized by a l evel of uniform cycl icshear stress that is appl ied for an
equival ent number of cycl es.Exampl e 9.2The site shown in Figure E9.2a is subjected
to earthquake shaking that produces a peak groundaccel eration of O.22g. Estimate
and pl ot the variation of maximum shear stress with depth.Compute and pl ot the
variation of equival ent uniform cycl ic shear stress with depth.6ftt14 ft~Yd =112
Ib/ft3Yb =71 Ib/ft3l Il l l l 7l l 771l 7l l l l l 7l 1l l 7l l 7771l 7l J7777777177777777777Figure
E9.2a372 Liquefaction Chap. 9O.--r---,-----,---...,---.---,--~-__,...-_____,__-
____.30 -,,,,,,- - _1- _ ,,,,,- - - - -1,- ~ ---,,,,,,-----1----------1"-----1-----
: Ave: rage val ues, ,___~ ~ ~ ~ L- .L" ," , I I I II I I I---~-----+-----~---
----~-----I I I II I I I[ I I II I I I II I I I I I - - - - -'-Range-for-diffe-
rentr - -- --!- ----: ---,soil pr,ofil es ': , ----------1--,,,, ,r------~-----~-----
~--- ,, ,, , ,, ,,-----I,,-----~,,-----+,,--- -,, ,, ,, , , ,- - - - -1- - - -
- ...,- - - - - T - - , , , ,, ,, ,, ,, ,, ,,-----1-----1-----1, , , ,, ,, ,,,,,10
-----: -----,,20405s: 50 E.OJ060708090100 L-_-'----_-'---_...&--: >.-"-'>..l >--'>--
>...c'-"'--"--'--'>.""--"-'>..L'-_-'-_-------'-_--'o 0.1 0.2 0.3 0.4 0.5 0.6 0.7
0.8 0.9 1.0Stress reduction factor, rdFigure 9.25 Reduction factor to estimatethe
variationof cycl ic shear stress with depthbel owl evel or
gentl ysl opinggroundsurfaces. (AfterSeed and Idriss, 1971.)Sol ution Using the
simpl ified procedure of Seed and Idriss (1971), the maximum shear stresscan be
estimated fromEstimation of the variation of'tmax with depth requires eval uation of
the variation of total vertical stress, O'v' and stress reduction factor, rd' with
depth. At depths of 5ft and 25 ft, for exampl e,the total vertical stresses areO'v(Z
= 5 ft)O'v(z = 25 ft)(5 ft)(105 l b/ft') = 525 l b/ft?(10 ft)(105 l b/ft") + (6 ft)
( 112 l b/ft ') + (9 ft)(71 l b/ft ' + 62.4 Ib/ft 3)2923Ib/ft2At the same depths,
Figure 9.29 indicates that the stress reduction factor has val ues ofr d (z = 5 ft)r
d (z = 25 ft)0.9920.947The maximum shear stresses at depths of 5 ft and 25 ft can
then be estimated as'tmax(z = 5 ft)'tmax(z = 25 ft)(0.22)(525 Ib/ft2)(0.992) = 115
Ib/ft2(0.22)(2923 Ib/ft2)(0.947) = 6091b/ft2Sec: . 9.5 Initiation of Liquefaction
373The equival ent uniform cycl ic shear stresses are simpl y taken as 65% of the
maximum shearstresses, i.e.'teye = (Z =5 ft) = 0.65't max(z =5 ft) = (0.65)(115
l b/ft-) = 75 l b/ft?'teye = (z =5 ft) = 0.65't max(z =5 ft) = (0.65)(609 Ib/ft2) =
3961b/ft2By repeating this process for other depths, the variations of'tmax and
'teye can be determined andpl otted as in Figure E9.2b.1000J: : : l i(]) o50030
'--------"-----"--------' Figure E9.2bCharacterization of Liquefaction Resistance.
The l iquefaction resistanceof an el ement of soil depends on how cl ose the initial
state of the soil is to the state correspondingto "fail ure" and on the nature of
the l oading required to move it from the initial state to the fail ure state. As the
preceding sections have shown, however, the fail ure state isdifferent for fl ow
l iquefaction and cycl ic mobil ity. The fail ure state for fl ow l iquefaction iseasil y
defined using the FLS, and its initiation is easil y recognized in the fiel d. The
definitionof fail ure for cycl ic mobil ity is imprecise-a certain l evel of
deformation caused bycycl ic mobil ity may be excessive at some sites and acceptabl e
at others. In contrast to fl owl iquefaction, there is no distinct point at which
cycl ic mobil ity "fail ure" can be defined.Cycl ic mobil ity fail ure is general l y
considered to occur when pore pressures become l argeenough to produce ground
oscil l ation, l ateral spreading, or other evidence of damage at theground surface.
This definition of fail ure is imprecise; in practice the presence of sand boil sis
frequentl y taken as evidence of cycl ic mobil ity. The devel opment of sand boil s,
however,depends not onl y on the characteristics of the l iquefiabl e sand but al so on
the characteristics(e.g., thickness, permeabil ity, and intactness) of any overl ying
soil s (Section 9.6.2).In the fiel d, where stresses and pore pressures are sel dom
measured, it is often difficul tto distinguish between different l iquefaction
phenomena after an earthquake hasoccurred. When the cycl ic stress approach was
devel oped, l ittl e distinction was madebetween the different l iquefaction phenomena-
cases of fl ow l iquefaction and cycl icmobil ity were l umped together under the
general heading of "l iquefaction." Characterizationof l iquefaction resistance
devel oped al ong two l ines: methods based on the resul ts ofl aboratory tests, and
methods based on in situ tests and observations of l iquefaction behaviorin past
earthquakes.Characterization Based on Laboratory Tests. The earl y devel opment of
thecycl ic stress approach emphasized l aboratory testing for characterization of
l iquefaction resistance.To create an initial condition of zero driving stress (to
simul ate the stress conditions on374 Liquefaction Chap. 9horizontal pl anes beneath
l evel ground), most l aboratory tests were performed on isotropical l yconsol idated
triaxial specimens or on Ko-consol idated simpl e shear specimens. In these tests,
"l iquefaction fail ure" was usual l y defined as the point at which initial
l iquefaction was reachedor at which some l imiting cycl ic strain ampl itude (commonl y
5% for dense specimens) wasreached.Laboratory tests show that the number of l oading
cycl es required to produce l iquefactionfail ure, NL , decreases with increasing
shear stress ampl itude and with decreasingdensity (Figure 9.26). Whil e l iquefaction
fail ure can occur in onl y a few cycl es in a l oosespecimen subjected to l arge cycl ic
shear stresses, thousands of cycl es of l ow-ampl itudeshear stresses may be required
to cause l iquefaction fail ure of a dense specimen. The rel ationshipbetween density,
cycl ic stress ampl itude, and number of cycl es to l iquefaction fail urecan be
expressed graphical l y by l aboratory cycl ic strength curves, such as those shownin
Figure 9.27. Cycl ic strength curves are frequentl y normal ized by the initial
effectiveoverburden pressure to produce a cycl ic stress ratio (CSR). The CSR must
be defined differentl yfor different types of tests. For the cycl ic simpl e shear
test, the CSR is taken as theratio of the cycl ic shear stress to the initial
vertical effective stress [i.e., (CSR)ss= 't cy/ Ci~o ].For the cycl ic triaxial
test, it is taken as the ratio of the maximum cycl ic shear stress to theinitial
effective confining pressure [i.e., (CSR)tx =Cid/2Ci3c]'As discussed in Chapter 6,
thecycl ic simpl e shear and cycl ic triaxial tests impose quite different l oading,
and their cycl icstress ratios are not equival ent. For l iquefaction testing, the two
are usual l y rel ated by(9.3)where the correction factor, c., is estimated from Tabl e
9-1.In contrast to l aboratory cycl ic simpl e shear and cycl ic triaxial tests,
earthquakes produceshear stresses in different directions. Mul tidirectional shaking
has been shown (Pyke,et al ., 1975) to cause pore pressures to increase more rapidl y
than does unidirectional shaking.Seed, et al . (1975b) suggested that the CSR
required to produce initial l iquefaction in20 '& 10~ 0~ 10... 20~ 10rft g-;: 1~1.0
,: : : I 0.5onV uv.vt~Time(a)60 '& 30~ 0 f-t-I++++I-H++++t+-I'+t++rrf+t-++++H-I~ 30...
60~ 3I"'oCVtr'cA-f-'ri-h4+++t+-I++++t-+t-H-t-H?- 51.0
f-------.,....,----,-;--: -;-;-;-;--;-;-7'<,...,-;-;-;-.-d\...: : : 1 0.5 ~~Jl Jl !
VVVV\}VVWUV\JVVVVIJVl IVVVV\JUVV~0"'------------------'Time(b)Figure 9.26 Resul ts of
torsional shear tests on isotropical l y consol idated (00 =98 kPa)specimens of (a)
l oose sand (47% rel ative density) and (b) dense sand (75% rel ativedensity). Loose
specimen reached initial l iquefaction tr; = 1.00) on 10th l oading cycl e.Desite much
higher l oading, dense specimen has not quite reached initial l iquefaction after17
cycl es. (After Ishihara, 1985; used by permission ofKl uwer Academic Publ ishers.)
Sec. 9.5 Initiation of Liquefaction 375160Cil (L 140~0 "0 120\: )'<"Jl 100~Ci5 80~.s;
Q)"'0.~U>. o01e;=void ratiocr'3C= 100 kPa__________ e;=0.61_____________ e;=0.71e;=
0.78------ e;= 0.8710 100 1000 10,000 100,000 1,000,000Number of cycl es, NFigure
9.27 Cycl ic stressesrequiredto produceinitial l iquefaction and 20%axial strain in
isotropical l y
consol idated Sacramento River Sandtriaxial specimens. (AfterSeedand Lee, 1965.)
Tabl e 9-1 Val ues of CSR Correction Factor, c,c; for: ReferenceFinn et al . (1971)Seed
and Peacock(1971)Castro (1975)EquationCr = (I + Ko)/2VariesCr =2(1+ 2Ko)/3-f3Ko=0.4
0.70.55--0.720.69Ko= 1.01.01.01.15the fiel d was about 10% l ess than that required
in unidirectional cycl ic simpl e shear tests.Therefore, the l iquefaction resistance
of an el ement of soil in the fiel d is given by the cycl icstress ratio(CSRhiel d ~{~
= 0.9(CSR)ss = 0.9tr(CSR )tx (9.4)Exampl e 9.3A 2-m-thick l ayer of Sacramento River
Sand (e = 0.87; <1>' = 33°) is overl ain by 4 m of compactedfil l (p, = 2.1 Mg/rrr').
The water tabl e is at the bottom of the fil l . Using the cycl ic triaxial test resul ts
shown in Figure 9.27, estimate the maximum cycl ic shear stress required to initiate
l iquefaction in the sand in a magnitude 7.5 earthquake.Sol ution The sand is under
an average effective vertical stress ofO"~o = [(4 m)(2.1 Mg/m') + (1.0 m)(0.91
Mg/m3) ] (9.81~) = 91.3 kPasecwhich is cl ose to the effective confining pressures
of the cycl ic triaxial test data shown in Figure9.27.Using Figure 9.24, a magnitude
7.5 earthquake woul d be expected to produce about 14uniform stress cycl es (at 65%
of the maximum shear stress). From Figure 9.27, the cycl ic deviator376 Liquefaction
Chap. 9stress that woul d cause initial l iquefaction in 14 cycl es woul d be about 39
kPa. Then the triaxial cycl ic stress ratio is« : 39 kPa(CSR),x = 2---'-- = = 0.195
(J3c (2)( 100 kPa)The corresponding fiel d cycl ic stress ratio can be determined
using equation (9.4). LettingKo= 1 - sin <1>' = 0.46, we have1+0.46 (CSR)fiel d =
0.9c r(CSR),x = 0.9-2-(0.195)= 0.128Then't cyc = (CSR)fiel d(J~O = (0.128)(91.3 kPa)
'Ecyc 11.7 kPa'tm• x = 0.65 = 0.65 = 18.0 kPa11.7 kPaTherefore, a peak shear stress
of 18 kPa woul d be required to initiate l iquefaction in the SacramentoRiver Sand in
a magnitude 7.5 earthquake.Laboratory tests can al so reveal the manner in which
excess pore pressure is generated.For stress-control l ed cycl ic tests with uniform
l oading, Lee and Al baisa (1974) andDeAl ba et al . (1975) found that the pore
pressure ratio, ru' is rel ated to the number of l oadingcycl es byr, = ~+~sin-
l [2(: Jl Ia-1J (9.5)where NL is the number of cycl es required to produce initial
l iquefaction (ru =1.00) and ais a function of the soil properties and test
conditions. As il l ustrated in Figure 9.28, excesspore pressures increase quickl y in
the first and l ast l oading cycl es. Equation (9.5) can beused to estimate the excess
pore pressure generated when initial l iquefaction does not occur(i.e., when Neq <
NL) . In an approach that coul d address irregul ar l oading, Martin et al .1.0
,---------r-----,--------r-------,-----: ;;;;o0.80.60.40.20.2 0.4 0.6 0.8 1.0Figure
9.28 Rate of pore pressure generation in cycl ic simpl e shear tests. The dashedl ine
was generated from equation (9.5) with ex = 0.7. (After De Al ba et aI., 1975.)Sec.
9.5 Initiation of Liquefaction 377(1975) devel oped a fundamental model that
combined data on the rate of settl ement ofcycl ical l y l oaded dry sand with data on
the rebound and stress-deformation characteristicsof the soil to predict pore
pressure generation. The model has been incorporated into nonl inearground response
anal yses to provide an effective stress-based approach to l iquefactionanal ysis
(Section 9.5.3.3).For a number of years, l iquefaction resistance was commonl y
characterized by cycl icstresses determined from l aboratory tests. However,
subsequent work showed that cycl icstress-based measures of l iquefaction resistance
are infl uenced by factors other than the initial density and stress conditions. For
exampl e, l iquefaction resistance is infl uenced by differencesin the structure of
the soil (or soil fabric) produced by different methods ofspecimen preparation
(Ladd, 1974; Mul il is et al ., 1975; Toki et al ., 1986; Tatsuoka et al .,1986). The
history of prior seismic straining al so infl uences l iquefaction resistance [i.e.,
thel iquefaction resistance of a specimen that has been subjected to prior seismic
straining isgreater than that of a specimen of the same density that has not (Finn
et al ., 1970; Seed et al .,1975b)]. Al so, l iquefaction resistance increases with
increasing overconsol idation ratio andl ateral earth pressure coefficient (Seed and
Peacock, 1971). Final l y, the l ength of time undersustained pressure has been shown
(Ohsaki, 1969; Seed, 1979; Yoshimi et al ., 1989) toincrease the l iquefaction
resistance. These additional parameters are al l functions of thedepositional and
historical environment of a soil deposit, and they tend to infl uence soil behavior
primaril y at the l ow strain l evel s associated with the initiation of l iquefaction.
These l ow-strain effects are easil y destroyed by sampl ing disturbance, and are very
difficul tto repl icate in reconstituted specimens. Because of these factors,
characterization of l iquefactionresistance by l aboratory testing is extremel y
difficul t and has been suppl anted bymethods based on in situ test resul ts for many
projects. Trul y undisturbed sampl ing (e.g., bycareful ground freezing and coring)
is required for l aboratory tests to be abl e to characterizel iquefaction resistance
rel iabl y.Cycl ic triaxial tests of l iquefaction resistance can al so be compl icated
by specimennonuniformity. As high pore pressures devel op in a cycl ic triaxial test
specimen, the soil grains tend to settl e causing densification of the l ower part and
l oosening of the upper partof the specimen. The nonuniform density l eads to
nonuniform strain, and eventual l y to thinningor necking of the upper portion of the
specimen. This non uniformity can cause considerabl euncertainty in the appl ication
of cycl ic triaxial test resul ts to fiel d conditions.Characterization Based on In
Situ Tests. An al ternative approach, first describedby Whitman (1971), is to use
l iquefaction case histories to characterize l iquefaction resistancein terms of
measured in situ test parameters. Previous case histories can be characterized by
thecombination of a l oading parameter, £, and a l iquefaction resistance parameter,
CR, which canbe pl otted with a symbol that indicates whether l iquefaction was or
was not observed (Figure9.29). A boundary can then be drawn between the £-CR
combinations that have and have notproduced l iquefaction in past earthquakes. The
boundary is usual l y drawn conservativel y suchthat al l cases in which l iquefaction
has been observed l ie above it. In this approach, the cycl icstress ratio is usual l y
used as the l oading parameter, and in situ test parameters that refl ect thedensity
and pore pressure generation characteristics of the soil are used as l iquefaction
resistanceparameters.378 Liquefaction Chap. 9o• • Liquefactionobserved Boundary5£ •
• oNo Liquefaction o observedFigure 9.29 Typical pl ot showingcombinations of l oading
parameter, 5£, andl iquefaction resistance parameter, (l k, forcases where
l iquefaction has been observed(sol id circl es) and not observed (opencircl es).
Boundary indicates minimum val ueof l iquefaction resistance parameterrequired to
prevent l iquefaction.1. Standard Penetration Resistance. In the United States and
most other countries,the standard penetration test (SPT) has been the most commonl y
used in situ test for characterizationof l iquefaction resistance; factors that tend
to increase l iquefaction resistance(e.g. density, prior seismic straining,
overconsol idation ratio, l ateral earth pressures, andtime under sustained pressure)
al so tend to increase SPT resistance. Seed et al . (1983) comparedthe corrected SPT
resistance (Section 6.3.1.2) and cycl ic stress ratio for cl ean sand(Figure 9.30)
and sil ty sand (Figure 9.31) sites at which l iquefaction was or was notobserved in
earthquakes of M =7.5 to determine the minimum cycl ic stress ratio at which
l iquefaction coul d be expected in a cl ean sand of a given SPT resistance.The
presence of fines can affect SPT resistance and therefore must be accounted forin
the eval uation ofl iquefaction resistance (Seed et aI., 1985; Ishihara and Kosecki,
1989;Koester, 1994). Examination of Figures 9.30 and 9.31 shows that the
l iquefaction resistanceof sands is not infl uenced by fines unl ess the fines
comprise more than 5% of the soil . Athigher fines contents, the fines tend to
inhibit l iquefaction [i.e., the CSR required to initiatel iquefaction (for a given
(N j ) 60 val ue)]. The pl asticity of the fines can al so infl uence l iquefaction
resistance; the adhesion of pl astic fines tends to resist the rel ative movement of
individual soil particl es and thereby reduce the generation of excess pore pressure
duringearthquakes. Laboratory tests (Ishihara and Koseki, 1989) indicate l ittl e
infl uence at pl asticityindices bel ow 10, and a gradual increase in l iquefaction
resistance at pl asticity indicesgreater than 10. Ishihara (1993) suggested that the
effects of pl asticity coul d be accountedfor by mul tipl ying the CSR by the factor{
1.0p-- 1.0 + 0.022(PI - 10)PI: S;10PI> 10(9.6)Since most sandy soil s in al l uvial
deposits and man-made fil l s have pl asticity indices l essthan about 15, the effect
of fines pl asticity is usual l y smal l .Because strong-motion duration (hence
equival ent number of uniform stress cycl es)increases with earthquake magnitude, the
minimum cycl ic stress ratio required to initiatel iquefaction decreases with
increasing magnitude. The minimum cycl ic stress ratio for othermagnitudes may be
obtained by mul tipl ying the cycl ic stress ratio for M = 7.5 earthquakesby the
factors shown in Tabl e 9-2.Sec. 9.5 Initiation of Liquefaction 379
0.6 ..-----,----------,,----------,-----,--------,• Marginal NoLiquefaction
Liquefaction LiguefactionPan-American data • 0 0Japanese data • 0 0Chinese data ...
L: >.[0o---'O'-j----------------------ooc9
0: Fines content ,;5%Chinese buil ding code (cl ay content = 0)~O-----------------
-----i-II~Ol i"""·: ---------------- -----~·--~: : ------ .' • .: • • • • • II ./• • • •
.-, • : 0.20.10.40.3 ----------0.5 ---------10 20 30 40 50o L
__---.L_---'=====================~oFigure 9.30 Rel ationship between cycl ic stress
ratios causing l iquefaction and (N j )60val ues for cl ean sands in M = 7.5
earthquakes. (After Seed et al . (1975). Infl uence ofSPT procedures in soil
l iquefaction resistance eval uations, Journal ofGeotechnical Engineering, Vol . 111,
No. 12. Reprinted by permission of ASCE.)380 Liquefaction Chap. 9
0.6 .------,---------.-..----------------~• qFines content 2: 5%Modified Chinese
code proposai (cl ay content = 5%)Marginal NoLiquefaction Liguefaction Liguefaction
50oo1: : : ..: : ;;5Ti---I,Ii~IiI ~I ~-----------r--' ------------------IIIIIII15Pan-
American data • Japanese data • <>Chinese data ...20 30 40(N1)60~ercent fines = ~5
-------,~" ~I : ',I!• 10• • 0.50.4 ----0.3--a--• 0.10.2 ---CSRM = 7.5Figure 9.31
Rel ationship between cycl ic stress ratios causing l iquefaction and (N j )60val ues
for sil ty sands in M =7.5 earthquakes. (After Seed et al . (1975)_Infl uence of SPT
procedures in soil l iquefaction resistance eval uations, Journal ofGeotechnical
Engineering, Vol . 111, No. 12. Reprinted by permission of ASCE.)Sec. 9.5 Initiation
of LiquefactionTabl e 9-2 Magnitude CorrectionFactors for Cycl ic Stress Approach381
Magnitude, M1.501.321.131.000.891.5Ka 1.0The data from which Figures 9.30 and 9.31
were devel oped correspond primaril y tol evel -ground sites with rel ativel y shal l ow
deposits of potential l y l iquefiabl e soil . At siteswith sl oping ground conditions or
at sites that support heavy structures, the presence of initial static shear
stresses wil l infl uence l iquefaction resistance. For conditions in which thestatic
shear stress is greater than the steady-state strength, the initial conditions are
cl oser tothe FLS and the l iquefaction resistance is reduced. Laboratory tests show
that the cycl icshear stress required to trigger l iquefaction increases at high
effective confining pressures(greater than those ofthe fiel d performance database).
Seed (1983) proposed that the effectsof initial shear stress and high effective
confining pressures be accounted for by modifyingthe cycl ic stress ratio as
fol l ows: (CSRfie1d)a, o = (CSRfiel d)a= 0, (j < 1tOnl ft2KaK(j (9.7)where ex = 'th,
staticl (j~o , and K a and K(j are correction factors for initial shear stress
(Figure9.32) and effective overburden pressure (Figure 9.33), respectivel y. The
val ues of Ka and K(jvary for different soil s and shoul d be eval uated on a site-
specific basis whenever possibl e.Dr z55 -70%0.5 ..0.1 0.2a0.3 0.4600 700 800<> o<>o
Fol som foundation gravel (Dr = 40%)s500oFol som embankment gravel (Dr=64%) TEstimated
average curve for sand400o.................. .... 0 .... ....-- ~-_-____ 0
~: : : : : : : : : : ~---------------300• • " ...EJo200Legend1001\ .q\\\ .\\0\ \·Cb \\\ \\,\ \
\\\\\,",~", ,, """"" "~""" 8" "" " " " " " "• Fairmont Dam• Lake Arrowhead Damo
Upper San Leandro Dam'V Lower San Fernando Dam Shel l .6. Upper San Fernando Dam
Shel l T Los Angel es Dam Shel l ... Perris Dam Shel l , RC = 90,95,100%o Sardis Dam Shel l
<> Sardis Dam Foundationo Thermal ito Afterbay Dam Foundation• Thermal ito Forebay
Dam FoundationEJ Antel ope Dam Impervious Material o Aswan Dam Dune Sando Sacramento
River Sand, Dr = 38, 60, 78, 100%• Monterey 0 Sand, Dr = 40, 60%o Reid Bedford
Sand, Dr = 40, 60%<> New Jersey Backfil l , FPI RC= 95%0.0 O~'----~)---roD---
30~----;~------;: : ~-----.L-----'-----------'0.20.40.81.01.2Kcr 0.6wcoNEffective
overburden pressure (kPa)Figure 9.33 Variation of correction factor, Kcr , with
effective overburden pressure. (AfterMarcuson et al ., 1990. Used by permission of
EERL)Sec. 9.5 Initiation of Liquefaction 3832. Cone Penetration Resistance. The tip
resistance from the cone penetration test(CPT) can al so be used as a measure of
l iquefaction resistance; indeed, it has a pronouncedadvantage over the SPT in its
abil ity to detect thin seams of l oose soil . The database of sitesat which CPT
resistance has been measured and where the occurrence or nonoccurrence of
l iquefaction has been documented, al though growing rapidl y, remains fairl y smal l .
By suppl ernentingthese data with correl ations between CPT and SPT resistances, the
minimumcycl ic stress ratio at which l iquefaction can be expected in a cl ean sand of
a given CPT resistancecan be determined (e.g., Robertson and Campanel l a, 1985; Seed
and DeAl ba, 1986).Since the CPT-SPT correl ation depends on grain size, CPT-based
l iquefaction curves havebeen devel oped for different mean grain sizes (Figure
9.34a). Mitchel l and Tseng (1990)devel oped curves based on l aboratory tests and
theoretical l y derived val ues of CPT resistance(Figure 9.34b). CPT-SPT correl ation
that make use of the friction ratio as wel l as thetip resistance (Dougl as et al .,
1981; Martin, 1992) el iminate the need for measurement ofmean grain size (and the
dril l ing and sampl ing required to do so). In CPT-based l iquefactioneval uations, the
tip resistance is normal ized to a standard effective overburden pressure of1
ton/fr' (96 kPa) by0.7 ,------------------., 0.5 r----------------,0.60.1 ...Seed
and De Al ba(1986)Oso = 0.40 mm0.30.40.20.1S 5% fines~D50 (mm) 0.25 0.4 0.8IIII
I /, ;'.", Robertson and Campanel l a (1985)..". 0 50 > 0.25 mmIshihara (1985)0.25 s
D50 S 0.55 mm ) 0.20.5co 0.4"" >;a: CJ)o 0.340 80 120 160 200 240Normal ized cone
resistance, Qc1 tst0.0 '-----~-'-~----'-'------'---'----'----'-----'-~o 10 20
Normal ized cone resistance, qct (MPa)30(a) (b)Figure 9.34 CPT-based l iquefaction
curves: (a) based on correl ations with SPT data;(b) based on
theoretical /experimental resul ts. (After Mitchel l and Tseng, 1990, H. Bol tonSeed
Memorial Symposium Proceedings, Vol . 2, p. 347. Used by permission of BiTech
Publ ishers, Ltd.)384orLiquefaction Chap. 9(9.8a)qc1 = 081.8 , qc (9.8b). + <Jvo
where <J~o is in tons/fr' (Kayen et al ., 1992). Adjustment for magnitudes other
than 7.5 canbe made using the CSR correction factors presented in Tabl e 9-2. Kayen
et aI. (1992) foundthat l iquefaction observations in the 1989 Lorna Prieta
earthquake agreed wel l with thecurves of Robertson and Campanel l a (1985) and
Mitchel l and Tseng (1990).For sil ty sands (> 5% fines), the effects of fines can be
estimated by adding the fol l owingtip resistance increments to the measured tip
resistance to obtain an equival ent cl eansand tip resistance (Ishihara, 1993)Fines
Content(%)~ 5-10- 15- 35Tip ResistanceIncrement (tons/ft2)o1222403. Shear Wave
Vel ocity. Improved methods of in situ shear wave vel ocity measurementsand studies
rel ated to devel opment of the cycl ic strain approach (Section 9.5.3.2)have
contributed to the recognition of shear wave vel ocity as a useful measure of
l iquefactionresistance. Measured shear wave vel ocities can be normal ized to a
standard effectiveoverburden pressure of 1 ton/ft'' (96 kPa) byvsl = vA <J~o)-l In
(9.9)where <J~o is in tons/fr' and n has been taken as 3 (Tokimatsu et al ., 1991)
or 4 (Finn, 1991;Kayen et al ., 1992). Stokoe et aI. (1988) used the cycl ic strain
approach and equival ent l inearground response anal yses to expl ore the rel ationship
between peak ground surfaceaccel eration (for stiff soil site conditions) and shear
wave vel ocity. The resul ts were used todevel op bounds for the conditions under
which l iquefaction coul d be expected; the resul tsagreed wel l with observed behavior
in two earthquakes in the Imperial Val l ey of Cal ifornia(Figure 9.35). Tokimatsu et
al . (1991) used the resul ts ofl aboratory tests to devel op curvesshowing the CSR
required to produce a cycl ic strain ampl itude of 2.5% in various numbersof cycl es
as a function of corrected shear wave vel ocity (Figure 9.36).The observation that
the shear wave vel ocity of sand is insensitive to factors (e.g., soil fabric,
overconsol idation ratio, prior cycl ic straining) that are known to infl uence
l iquefactionresistance suggests that shear wave vel ocity measurements al one may not
be sufficientto eval uate the l iquefaction potential of al l soil deposits
(Jamiol kowsky and LoPresti, 1992;Verdugo, 1992).4. Dil atometer Index. Correl ations
for l iquefaction potential with the horizontal stressindex of the dil atometer test
(DMT) have al so been proposed (Marchetti, 1982; RobertsonSec. 9.5 Initiation of
Liquefaction 385500 r------------r<---..,---..,---..,---..,.,300 L..-_-'-
_....L-"'-.....: ...: .L-_-'--_~_---l 0.0 0.1 0.2 0.3 0.4 0.5 0.6amax at stiff soil
site (g)uCD i 450-;,.'";: .'0~ 400>CD~;;: (ij 350CDs: (f)NoLiquefactionFigure 9.35
Chart for eval uation ofl iquefaction potential from shear wavevel ocity and peak
ground accel eration (10cycl es). (After Stokoe et aI., 1988.)1.5
,----~----,--------,------Figure 9.36 Correl ations between cycl icstress ratio
required to produce cycl ic strainampl itude of 2.5% in cl ean sand and shearwave
vel ocity. (After Tokimatsu et aI.,1991.)150 200 250 300Normal ized shear wave
vel ocity, vs1 (m/sec)Number of cycl es-----: : : : : -OL---------'----------'1001.00.5CSR
and Camponel l a, 1986; Reyna and Chameau, 1991). Figure 9.37 il l ustrates the
performanceof three such correl ations when appl ied to data from several sites in
the Imperial Val l ey ofCal ifornia. New correl ations for the DMT shoul d be expected
as additional experience withthis rel ativel y new in situ test is obtained.5. Use
ofIn Situ Test Resul ts. SPT resistance is by far the most commonl y used in situtest
parameter for characterization of l iquefaction resistance. The SPT al l ows a sampl e
to beretrieved (for identification, measurement of fines content, etc.) and has the
l argest
case historydatabase of any in situ test. However, the CPT is becoming much more
commonl y usedfor characterization of l iquefaction resistance. The CPT provides a
continuous record of penetrationresistance (an important benefit when thin l ayers
of seams of potential l y l iquefiabl esoil may exist) and is much faster and l ess
expensive than the SPT. Because CPT-based l iquefactionresistance is infl uenced by
grain size characteristics (Figure 9.34), compl ementaryborings with sampl ing may be
required.386 Liquefaction Chap. 920(1) 0.55 9(2) mean ace.(3) mean +1a15(1)o5 10
,, ,,','Robertson and,.. Campanel l a~eTf-+---1~,-,---1,,/( 1986 0(3)(2)oo Heber Road
- Deposit A1,---------,-~_r_--_,___1. Heber Road - Deposit A2o Heber Road - Deposit
A3... Wil dl ife - Deposit B• Kornbl oom - Deposit B0.70.60.50.4CSR0.30.20.10.00
Horizontal stress index, KdFigure 9.37 Suggested rel ationships between cycl ic
stress ratio required to producel iquefaction and horizontal stress index from
dil atometer test. (After Reyna andChameau, 1991.)Regardl ess of which is used, the
in situ test parameters al l ow estimation of CSRLtheCSR required to initiate
l iquefaction. Using the definition of cycl ic stress ratio, thecycl ic shear stress
required to initiate l iquefaction is given by't"cyc,L = CSRL(J~O (9.10)Eval uation
of Initiation of Liquefaction. Once the cycl ic l oadingimposed by an earthquake and
the l iquefaction resistance of the soil s have been characterized,l iquefaction
potential can be eval uated. The cycl ic stress approach characterizes earthquake
l oading by the ampl itude of an equival ent uniform cycl ic stress and l iquefaction
resistance by the ampl itude of the uniform cycl ic stress required to produce
l iquefaction inthe same number of cycl es. The eval uation of l iquefaction potential
is thus reduced to acomparison of l oading and resistance throughout the soil
deposit of interest.The eval uation is easil y performed graphical l y. First, the
variation of equival entcycl ic shear stress (earthquake l oading, 't"cyc), with depth
is pl otted as in Figure 9.38 (thenumber of equival ent cycl es, Neq, corresponding to
the earthquake magnitude must bedetermined if l iquefaction resistance is to be
characterized using l aboratory test resul ts).The variation of the cycl ic shear
stress required to cause l iquefaction (l iquefaction resistance,'t"cyc,L) with depth
is then pl otted on the same graph (the val ues of 't"cyc,L must correspondto the
same earthquake magnitude, or same number of equival ent cycl es, as 't"cyc)'
Liquefaction can be expected at depths where the l oading exceeds the resistance or
when thefactor of safety against l iquefaction, expressed asFS = cycl ic shear stress
required to cause l iquefactionL equival ent cycl ic shear stress induced by
earthquakeis l ess than 1.'t"cyc,L't"cycCSRL (9.11)CSRSec. 9.5DepthInitiation of
LiquefactionShear stress-1 Zone of~iqUefactionCycl ic shear stress requiredto cause
l iquefactionEquival entcycl ic shearstress inducedby earthquake387Figure 9.38
Processby whichzone ofl iquefaction is identified.It shoul d be noted that
significant excess pore pressure can devel op even if the computedfactor of safety
is greater than 1. At l evel -ground sites, for exampl e, the magnitude ofthis excess
pore pressure can be estimated from Figure 9.39. The reduction in effectivestress
associated with such excess pore pressures can reduce the stiffness of the soil ,
and significantsettl ement can occur as the excess pore pressures dissipate (Section
9.6.3.2).1.0~ Gravel 0.8 ...• .2> Sand ~0.6ru0.40.20.01.0 1.2 1.4 1.6 1.8 2.0 2.2 2.4
2.6FSLFigure 9.39 Rel ationship betweenresidual excesspore pressureand factor of
safetyagainstl iquefaction for l evel -ground sites. (AfterMarcusonand Hynes, 1990.)
Exampl e 9.4A site in Japan underl ain by cl ean, potential l y l iquefiabl e sand was
described in Exampl e 6.6.SPT tests were performed at the site using standard
Japanese SPT techniques; the process bywhich the measured SPT resistances are
corrected to obtain (Nj )60 val ues was described inExampl e 6.6. Using the
information avail abl e, determine the extent to which l iquefactionwoul d have been
expected in the 1964 Niigata earthquake (M= 7.5) if the peak horizontal
accel erationat the ground surface was O.16g.388 Liquefaction Chap. 9Sol ution
Eval uation of l iquefaction potential by the cycl ic stress approach invol ves
comparisonof the l evel of cycl ic stress induced by the earthquake with the l evel of
cycl ic stressrequired to initiate l iquefaction. The l evel of cycl ic stress induced
by the earthquake can be estimatedfrom the information avail abl e using the
simpl ified procedure of Seed and Idriss (1971).The variation of total vertical
stress with depth is easil y computed (col umn 2 bel ow). For exampl e,the total
vertical stress at a depth of 6.2 m is given byavo = (1.5 m)( 1.874 ~;) + (4.7 m) (
2.180 ~; )(9.81 S;2) = 128.1 kPaThe val ue of the stress reduction factor can be
obtained graphical l y from Figure 9.25 (col umn3 bel ow). At a depth of 6.2 m, rd =
0.960. Knowing the peak ground surface accel eration, thecycl ic shear stress is
computed (col umn 4 bel ow) by equation (9.2). At a depth of 6.2 m,0.16g't eye = 0.65
-g- (128.1 kPa) (0.960) = 12.8 kPaThe l evel of cycl ic shear stress required to
initiate l iquefaction depends on the l iquefactionresistance of the soil . For this
exampl e, l iquefaction resistance can be characterized by the(N1)60 val ues (col umn 5
bel ow-from col umn 6 in Exampl e 6.6). Using the chart for l ess than5% fines in
Figure 9.31, the cycl ic stress ratios required to initiate l iquefaction can be
foundgraphical l y (col umn 6 bel ow). At a depth of 6.2 m, the val ue (N 1) 60 =11.7
corresponds toCSRM =7.5 = 0.130. Because the magnitude of the Niigata earthquake
was 7.5, the magnitudecorrection factor (Tabl e 9-2) has a val ue of 1.0. ThenCSRL =
(CSRM = 7.5)(1.0) = 0.130Factor of safetyo 1 2 30,-------,---,-----,80 100Shear
stress (kPa)o 20 40 60O,-------,-----,-----,------r----,gs: 15.Q) o1020Due to
earthquake~ Required for l iquefaction .-----1020(a) (b)Figure E9.4Sec. 9.5
Initiation of Liquefaction 389Knowing CSRL and the variation of vertical effective
stress with depth (col umn 7 bel ow-fromcol umn 3 of Exampl e 6.6), the cycl ic stress
required to initiate l iquefaction can be computed(col umn 8 bel ow and Figure E9.4a).
At a depth of 6.2 m,"teye,L = CSRL(j~O = (0.130)(82.0 kPa) = 10.7 kPaFinal l y, the
cycl ic stress induced by the earthquake can be compared to the cycl ic stress
required to initiate l iquefaction. Using the concept of a factor of safety against
l iquefaction[equation (9. II)], the variation of factor of safety with depth can be
determined (col umn 9bel ow and Figure E9.4b). At a depth of 6.21l l ,10.7 kPa12.8 kPa
0.84As il l ustrated in Figure E9.4b, the factor of safety against l iquefaction is
greater than one wherethe cycl ic stress induced by the earthquake is l ower than the
cycl ic stress required to initiate l iquefaction("teye < "teye,L)' At this site,
extensive l iquefaction woul d have been expected in theupper 8 to 10m and at some
greater depths as wel l .The conditions of this exampl e are representative of those
in the Kawagishi-Cho area ofNiigata, Japan (Figure 1.6) though the actual peak
accel erations may have been on the order of0.2 to 0.3g. The extensive l iquefaction
predicted in this exampl e is consistent with what wasactual l y observed in that area
in the 1964 Niigata earthquake.(1) (2) (3) (4) (5) (6) (7) (8) (9)D(m) O"vo rd
'teye (N1)60 CSRL (j~o "teye,L FSL1.2 22.1 0.994 2.28 17.3 22.12.2 42.6 0.989 4.38
7.9 0.087 36,4 3.17 0.723.2 64.0 0.982 6.53 5.1 0.056 47.3 2.65 0.414.2 85.3 0.976
8.66 4.6 0.051 58.8 3.00 0.355.2 i06.7 0,968 10.8 7.0 0.077 70.4 5.42 0.506.2 128.1
0.960 12.8 11.7 0.130 82.0 10.7 0.847.2 149.5 0.951 14.8 14.6 0,161 93.6 15,1 1.02
8.2 170.9 0.940 16,7 13.7 0.151 105.3 15,9 0,959.2 192.3 0.926 18.5 15.2 0.167
116.8 19.5 1.0510.2 213,8 0.910 20.2 9.3 0.102 128.4 13.1 0.6511.2 235.1 0.886 21.7
22.8 0.250 140.0 35.0 1.6112,2 256.5 0.864 23.1 12.4 0.136 151.5 20.6 0.8913.2
277.9 0.835 24,1 10.1 0.111 163.1 18.1 0,7514.2 299.3 0.813 25.3 9.8 0.108 174,7
18.9 0,7515.2 320.7 0.767 25.6 20.7 0,228 186.3 42.5 1.6616.2 342.1 0.739 26.3 22.5
0.245 197.9 48,5 1.8417.2 363,5 0.704 26.6 4,1 0.045 209.4 9.42 0.3518.2 384.8
0.671 26.9 4,7 0.052 221.1 11.5 0.4319.2 406.2 0.642 27.1 3.1 0.034 232.7 7.91 0.29
20.2 427,7 0.625 27.8 28.5 0.350 244.3 85.5 3.08Because of their physical
significance, the initiation of l iquefaction has been presentedin terms of cycl ic
stresses in this section. After devel oping an understanding of therel ationship
between the cycl ic shear stresses induced by an earthquake and the cycl ic shear
stresses required to initiate l iquefaction, many engineers wil l find it more
convenient tocharacterize earthquake l oading and l iquefaction resistance in terms
of the cycl ic stress390 Liquefaction Chap. 9ratios, CSR and CSRL, both of which
vary over a much smal l er range than the cycl ic shearstresses themsel ves.9.5.3.2
Cycl ic Strain ApproachThe l arge number of factors that infl uence the cycl ic
stresses required to produce l iquefactioncan make l aboratory eval uation of
l iquefaction resistance ih the cycl ic stressapproach difficul t. As Seed (1976)
pointed out, "the l iquefaction characteristics of in situsand deposits are
determined by a number of compl ex factors, of which rel ative density isonl y one,
and careful eval uations of al l of these factors is required in sel ecting soil
characteristicsfor use in design."In an effort to devel op a more robust approach to
the l iquefaction probl em, Dobry andLadd (1980) and Dobry et al . (1982) described an
approach that used cycl ic strains ratherthan cycl ic stresses to characterize
earthquake-induced l oading and l iquefaction resistance.The
approach is based on experimental evidence that shows densification of dry sands
to becontrol l ed by cycl ic strains rather than cycl ic stresses (e.g., Sil ver and
Seed, 1971; Youd,1972) and the existence of the threshol d vol umetric shear strain
(Section 6.4.1) bel ow whichdensification does not occur. Since the tendency for a
sand to densify when dry is directl yrel ated to its tendency to devel op excess pore
pressure when saturated, it fol l ows that porepressure generation shoul d be more
fundamental l y rel ated to cycl ic strains than cycl icstresses.Characterization of
Loading Conditions. In the cycl ic strain approach,earthquake-induced l oading is
expressed in terms of cycl ic strains. The time history ofcycl ic strain in an actual
earthquake is transient and irregul ar. To compare the l oading withl aboratory-
measured l iquefaction resistance, it must be represented by an equival ent seriesof
uniform strain cycl es. The conversion procedure is anal ogous to that used in the
cycl icstress approach.The time history of cycl ic shear strain may be computed in a
ground response anal ysis.This is perhaps the weakest l ink in the cycl ic strain
approach since cycl ic strains are considerabl ymore difficul t to predict accuratel y
than cycl ic stresses (Seed, 1980). Dobry et al .(1982) proposed a simpl ified method
for estimating the ampl itude of the uniform cycl icstrain from the ampl itude of the
uniform cycl ic stress of equation (9.2): a max (JvrdYcyc = 0.65--G() (9.12)g Ycyc
where G(ycyc) is the shear modul us of the soil at Y= Ycyc' Since Ycyc infl uences
both sides ofequation (9.12), the val ue of G(ycyc) must be obtained iterativel y
from a measured Gmax profil eand appropriate modul us reduction curves (Section
6.4.2.1). The equival ent number of straincycl es, Neg, depends on the earthquake
magnitude, and can be estimated from Figure 9.24.Once Ycyc is determined, it can be
compared with the threshol d shear strain, Yt·l fycyc <Yt,no pore pressure wil l be
generated and, consequentl y, l iquefaction cannot be initiated. The l iquefaction
hazard eval uation woul d end at that point. Ifycyc >Yt, l iquefaction is possibl e and
thel iquefaction resistance of the soil must be eval uated.Characterization of
Liquefaction Resistance. The cycl ic strainapproach simpl ifies the interpretation of
l iquefaction resistance from l aboratory tests. Experimental evidence indicates that
factors that increase the cycl ic stresses required to initiateSec. 9.5 Initiation
of Liquefaction 3911.00.8ru 0.60.40.2l iquefaction (e.g., density, soil fabric,
strain history, overconsol idation ratio, l ength of timeunder sustained pressure)
al so increase the shear modul us of the soil . Because these factorsinfl uence both
'Ccyc and G simil arl y, their infl uence on the ratio Ycyc ='Ccyc IG is muchsmal l er.
Consequentl y, they have l ittl e infl uence on pore pressure generation when
interpretedin terms of cycl ic strains. Dobry and Ladd (1980) provided striking
evidence of thisresul t; Figure 9.40 shows the pore pressure ratio produced by 10
strain-control l ed cycl es ofl oading on two different sands prepared by three
different methods at three different initial effective confining pressures. The
insensitivity of the generated pore pressure to factorsother than cycl ic strain
ampl itude il l ustrated in Figure 9.40 is a hal l mark of the cycl ic strainapproach.
Soil SpecimenPreparation a3c (psi) Symbol Crystal sil ica No. 20 sand Dry vibration 14
0 Fresh sampl eWet rodding 14 J: : : ,. 0 Fresh sampl e.... Dry vibration 14 .. 0 J: : : ,.
Staged sampl eSand NO.1 Moist tamping 10 VStaged sampl e20 T0 J: : : ,.~0Shear strain,
Ycyc (%)Figure 9.40 Measured pore pressure ratio after 10 cycl es of l oading in
straincontrol l edcycl ic triaxial tests. After Dobry and Ladd, (1980). Discussion,
Journal of theGeotechnical Engineering Division, Vol . 106, No. GT6. Reprinted by
permission ofASCE.Dobry et al . (1984) devel oped an innovative axial /torsional
triaxial test cal l ed theCyT-CAD test for measurement of l iquefaction resistance. In
this test a sol id cyl indrical triaxial specimen is anisotropical l y consol idated
under a stress-control l ed static deviatorstress. Cycl ic strains are then imposed
under undrained conditions by strain-control l edcycl ic torsion whil e the stress-
control l ed deviator stress remains constant. The cycl ic shearstrains induce excess
pore pressure in the specimen. If the effective stress conditions reachthe FLS and
the steady-state strength is l ess than the static shear stress, fl ow l iquefaction
wil l occur. This test cl osel y model s the behavior of soil s that experience fl ow
l iquefaction392 Liquefaction Chap. 9in the fiel d: excess pore pressures are
generated by cycl ic shear strains, but fl ow fail ure isdriven by static shear
stresses.Vasquez-Herrera and Dobry (1988) used theCyT-CAU test to investigate the
generationof excess pore pressure and initiation of fl ow l iquefaction. Confirming
the existenceof a FLS for several sands, Vasquez-Herrera and Dobry al so observed
that the excess porepressure required to trigger l iquefaction decreased with
increasing initial principal effectivestress ratio (see Figure 9.20), and a
corresponding decrease in the number of strain cycl esrequired to trigger
l iquefaction (Figure 9.41). The cycl ic strain ampl itude required to trigger
l iquefaction in Neq cycl es can be expressed in the form"(eye = 0.01 + [aNeq~;' ~
ru, t)] 1113 (9.13)where a and ~ are experimental l y determined functions of K; used
to characterize l iquefactionresistance in the cycl ic strain approach. The rate at
which excess pore pressures devel opis al so infl uenced by Kc ; normal ized pore
pressures increase more quickl y in the first cycl esfor specimens at l ower K; val ues
(Figure 9.42).10,---------------------------,zc,.,."s.C.~1ii"iiic0.~.8(ijQ).I: : 0.1
C/).2ts>-010 100 1000Number of cycl es to trigger fl ow l iquefaction, n,Figure 9.41
Liquefaction resistance curves for a sand obtained from an actual l iquefaction
fail ure. For this sand, IX = 4.78 to L91Kc andB = 2.96 to O.78Kc- Note the
significant reduction in l iquefaction resistance with increasing l evel of initial
shearstress. (After Vasquez-Herrera and Dobry, 1988.)10,000Eval uation of
Liquefaction Potential . Liquefaction potential may beeval uated in the cycl ic strain
approach in a manner simil ar to that used in the cycl ic stressapproach. The cycl ic
l oading imposed by the earthquake, characterized by the ampl itude ofSec. 9.5
Initiation of Liquefaction 3931.0 r----------------------"0.2 0.4 0.6 0.8 1.00.0
"------'------'---------'----'--------'0.00.20.60.40.8Figure 9.42 Normal ized rate
of pore pressure generation for sands subjected to differentl evel s of initial shear
stress. (After Vasquez-Herrera and Dobry, 1988.)a series of Neq uniform strain
cycl es, is compared with the l iquefaction resistance, which isexpressed in terms of
the cycl ic strain ampl itude required to initiate l iquefaction in the samenumber of
cycl es. Liquefaction can be expected at depths where the cycl ic l oading exceedsthe
l iquefaction resistance (Figure 9.43). Since l oading and resistance are
characterized interms of strains rather than stresses, the cycl ic strain approach
does not yiel d a factor ofsafety against l iquefaction.Shear strain ampl itudeDepth
--------rZone of_______l l qustactionycl ic shear strain requiredto cause l iquefaction
in N cycl esEquival entcycl ic shearstrain inducedby earthquakefor Ncycl esFigure 9.43
Process by which the zone ofl iquefaction is identified in the cycl ic strain
approach.394 Liquefaction Chap. 9(9.14)The primary advantage of the cycl ic strain
approach derives from the strong rel ationshipbetween pore pressure generation and
cycl ic strain ampl itude. For a given soil , excesspore pressure can be predicted
more accuratel y from cycl ic strains than from cycl ic stresses.However, cycl ic
strains are considerabl y more difficul t to predict accuratel y than cycl icstresses.
The cycl ic strain approach is not used as commonl y as the cycl ic stress approachin
geotechnical earthquake engineering practice.9.5.3.3 Other Approaches to the
Initiationof LiquefactionAl though the cycl ic stress and cycl ic strain approaches
are the most commonl y usedin geotechnical earthquake engineering practice, other
approaches have al so been devel oped.Two that offer considerabl e promise are the
dissipated energy approach and the effectivestress-based response anal ysis
approach. As these approaches are refined, and asadditional data with which to
cal ibrate them become avail abl e, their use is l ikel y to increase.Energy Dissipation
Approach. The use of dissipated energy as a measureof l iquefaction resistance
offers a number of advantages; it is rel ated to both cycl ic stressesand cycl ic
strains, and it can be rel ated to inherentl y stochastic earthquake ground motionsin
a way that methods based on peak ground motion parameters al one cannot, it is a
scal arquantity, and it can be rel ated to fundamental earthquake parameters.The
densification of dry soil invol ves rearrangement of grains and hence the
expenditureof energy. As a cycl ical l y l oaded dry soil densifies and approaches its
minimum voidratio, the amount of energy required to rearrange individual soil
grains increases. For a saturatedsoil , however, the tendency for densification
causes the pore pressure to increase andthe interparticl e contact forces to
decrease. As these contact forces decrease, the amount ofenergy needed to rearrange
soil grains decreases. By combining these observations, NematNasserand Shokooh
(1979) devel oped a simpl e, unified theory that rel ated densificationunder drained
conditions and pore pressure generation under undrained conditions to dissipated
energy. Others have since attempted to characterize the rel ationship between excess
pore pressure and dissipated energy experimental l y (e.g., Simcock et al .,
1983; Law et al .,1990; Figueroa and Dahisaria, 1991). Al though some of the data
are quite scattered, theexcess pore pressure can be predicted by a rel ationship of
the formL1,u = <Xwtavowhere WN is a dimensionl ess energy term (Law et al ., 1990)
and <X and ~ are coefficientsdetermined from l aboratory tests.Davis and Berril l
(1982) demonstrated the potential for using energy to rel ate l iquefactionbehavior
to fundamental earthquake parameters such as magnitude and distance. Bycombining an
estimate of the energy content of seismic waves at a distance R from an earthquake
of magnitudeM with a simpl e energy dissipation function cal ibrated by experimental
resul ts and fiel d observations of l iquefaction behavior, an expression for excess
pore pressurewas devel oped.(9.15)Sec. 9.5 Initiation of Liquefaction 395(9.16)where
R is in meters, NI is the SPT resistance corrected to an effective overburden
pressureof 1 ton/ftZ (96 kPa), and Au and a~o are in kPa. Law et al . (1990) used
energy principl esto devel op the fol l owing criterion for l iquefaction fail ure in
sands: l O1.5M-----: : -: ,...--,~___: _: : : ~ 1.02.28 X 1O-IONj1.5R4.3where R is the
hypocentral distance in kil ometers.Effective Stress-Based Response Anal ysis
Approach. Stress-strainmodel ing of soil behavior has been a subject of intense
research activity for many years, andthe task of repl icating the compl ex behavior
of potential l y l iquefiabl e soil s with simpl e constitutivemodel s has proven
chal l enging. As discussed in Section 6.4, the nonl inear stressstrainbehavior of
soil s can be described by cycl ic nonl inear stress-strain model s and byadvanced
constitutive model s.Cycl ic nonl inear stress-strain model s (Section 6.4.3) use an
empirical backbonecurve and a series of unl oading-rel oading rul es that govern
cycl ic behavior. Pore pressureprediction is accompl ished by pore pressure model s
(e.g., Martin et al ., 1975; Ishihara andTowhata, 1980; Finn and Bhatia, 1981) that
can predict the generation of pore pressureunder irregul ar cycl ic l oading
conditions. The computed pore pressure is used to degrade, orsoften, the backbone
curve as the effective stress (and soil stiffness) decreases.In the Martin et al .
(1975) model , for exampl e, the pore pressure generated in an incrementof undrained
l oading is rel ated to the vol umetric strain that woul d have occurred in thesame
l oading increment under drained conditions by(9.17)(9.18)(9.19)where E, is the
rebound modul us and Atvd is the incremental vol umetric strain underdrained
conditions. The rebound modul us can be expressed as(a: )I-mwhere a: and a~o are the
current and initial vertical effective stresses and m, n, and Kz areexperimental l y
determined from a rebound test in a consol idometer. The incremental vol umetric
strain is computed asC3t~dACvd = CI(y- Cztvd) + --"-=-''''----Y+ C4 t vdwhere yand
tvd are the cycl ic shear and vol umetric strains, respectivel y, and CI-C4 are
constantsdetermined from the resul ts of drained cycl ic simpl e shear tests. Martin
et al . (1981)devel oped a procedure for estimation of these constants without
l aboratory test resul ts.When incorporated into nonl inear ground response anal yses,
cycl ic nonl inear and porepressure model s al l ow computation of the generation,
redistribution, and dissipation of porepressures. Hence the effective stress
conditions throughout a soil deposit can be monitoredthroughout and even after an
earthquake to eval uate l iquefaction hazards.396 Liquefaction Chap. 9(9.20)Advanced
constitutive model s (Section 6.4.4) provide a more rigorous approach toprediction
of soil behavior under a wide variety of l oading conditions. Such model s describe
the increments of vol umetric and deviatoric strain produced by increments of
vol umetric anddeviatoric stress. By setting the incremental vol umetric strain to
zero to represent undrainedconditions, changes in effective stresses can be
computed. Such model s can be incorporatedinto nonl inear ground response and dynamic
response anal yses. The one-dimensional nonl inearground response program DYNAID
(Prevost, 1989), for exampl e, uses a nested yiel dsurface constitutive model to
account for nonl inear, anisotropic, hysteretic soil behavior topredict the
generation, redistribution, and dissipation of excess pore pressure during and
afterearthquake shaking.Probabil istic Approach. There are many potential sources of
uncertainty inboth the l oading and resistance aspects of l iquefaction probl ems, and
probabil istic approacheshave been devel oped to deal with them. Uncertainties in
cycl ic l oading can be eval uated usingthe standard probabil istic seismic hazard
anal yses described in Section 4.4. Uncertainties inl iquefaction resistance can be
treated in one of two general ways.One group of methods is based on probabil istic
characterization of the parametersshown by l aboratory tests to infl uence pore
pressure generation. Hal dar and Tang (1979)characterized uncertainty in the
parameters of the simpl ified cycl ic stress approachdescribed in Section 9.5.3.1.
Fardis and Veneziano (1982) used a simil ar approach withtotal stress and effective
stress model s. Chameau and Cl ough (1983) described pore pressuregeneration
probabil istical l y using experimental data and an effective stress model .Each of
these methods can compute the probabil ity of l iquefaction due to a particul ar set
ofl oading conditions. Their accuracy depends on the accuracy of the underl ying
l iquefaction/pore pressure model and on how accuratel y the uncertainty of the model
parameterscan be determined.An al ternative group of methods are based on in situ
test-based characterization ofl iquefactionresistance (e.g., Christian and Swiger,
1975; Yegian and Whitman, 1978; Venezianoand Liao, 1984; Liao et aI., 1988). These
methods use various statistical cl assificationand regression anal yses to assign
probabil ities of l iquefaction to different combinations ofl oading and resistance
parameters. Liao et al . (1988), for exampl e, anal yzed 278 case studiesto produce
the fol l owing expression for the probabil ity of l iquefaction: 1PL =1 + exp [-(~o +
~l In(CSR) + ~2(Nl )60)]where the parameters ~O-~2 are shown in Tabl e 9-3.
Liquefaction probabil ity curves for thecl ean and sil ty sand cases are shown
graphical l y in Figure 9.44.Exampl e 9.5Use the approach of Liao et al . (1988) to
estimate the probabil ities of l iquefaction at the sitedescribed in Exampl e 9.4.
Sol ution The (j~o and 't cyc data from Exampl e 9.4 can be used to compute the
cycl ic stressratio (col umn 4 bel ow). Using this cycl ic stress ratio and the (N j )
60 val ues eval uated original l yin Exampl e 6.6, probabil ities of l iquefaction at each
depth in the soil profil e using equation(9.20) (with the cl ean sand regression
coefficients).Al though the resul ts of this anal ysis are expressed in terms of a
probabil ity of l iquefactionrather than a factor of safety against l iquefaction (as
in Exampl e 9.4), both indicate thatSec. 9.6 Effects of Liquefaction 397D(m) 'Leye
(kPa) cr~o (kPa) CSR (N j )60 PL1.2 2.28 22.1 0.103 17.32.2 4.38 36.4 0.120 7.9
0.4033.2 6.53 47.3 0.138 5.1 0.8354.2 8.66 58.8 0.147 4.6 0.9035.2 10.8 70.4 0.153
7.0 0.8236.2 12.8 82.0 0.156 11.7 0.4697.2 14.8 93.6 0.158 14.6 0.2188.2 16.7 105.3
0.159 13.7 0.2939.2 18.5 116.8 0.158 15.2 0.18010.2 20.2 128.4 0.157 9.3 0.68711.2
21.7 140.0 0.155 22.8 0.00912.2 23.1 151.5 0.152 12.4 0.34213.2 24.1 163.1 0.148
10.1 0.52214.2 25.3 174.7 0.145 9.8 0.51915.2 25.6 186.3 0.137 20.7 0.01016.2 26.3
197.9 0.133 22.5 0.00317.2 26.6 209.4 0.127 4.1 0.81518.2 26.9 221.1 0.122 4.7
0.72919.2 27.1 232.7 0.116 3.1 0.78520.2 27.8 244.3 0.114 28.5 0.0001Tabl e 9-3
Regression Parameters for Cal cul ating Probabil ity ofLiquefactionNumberData" of
Cases ~o ~j ~2Al l cases 278 10.167 4.1933 -0.24375Cl ean sand 182 16.447 6.4603
-0.39760cases onl ySil ty sand 96 6.4831 2.6854 -0.18190cases onl ySource: After Liao
et al . (1988).a A fines content of 12% is used as the bouudary between cl ean and
sil tysands.extensive l iquefaction woul d have been expected at this particul ar site
in the 1964 Niigataearthquake. Widespread l iquefaction was indeed observed at sites
with soil conditions simil arto these in that earthquake.9.6 EFFI: CTS OF
LIQUEFACTIONLiquefaction phenomena can affect buil dings, bridges, buried pipel ines,
and other constructedfacil ities in many different ways. Liquefaction can al so
infl uence the nature ofground surface motions. Fl ow l iquefaction can produce
massive fl ow sl ides and contributeto the sinking or til ting of heavy structures,
the fl oating of l ight buried structures, and to thefail ure of retaining structures.
Cycl ic mobil ity can cause sl umping of sl opes, settl ement ofLiquefaction Chap. 90.6
0.5 09 0.10.4"r-: 'rl " 0.3(/) o0.20.10.00 0 0 0 0 0 0(N1)60(b)0.01o o0.0 '-----
_---'--''-----_L-_---'---__'-----_-'o0.13980.40.50.6
r----,------,--------,---,------,"r-: 'r1" 0.3(/) oFigure 9.44 Contours of equal
probabil ity of l iquefaction for (a) cl ean sand (l ess than12% fines), and (b) sil ty
sands (greater than 12% fines). (After Liao et aI., 1988.Regression model s for
eval uating l iquefaction probabil ity, Journal of Geotechnical Engineering, Vol . 114,
No.4. Reprinted by permission of ASCE.)buil dings, l ateral spreading, and retaining
wal l fail ure. Substantial ground oscil l ation,ground surface settl ement, sand boil s,
and postearthquake stabil ity fail ures can devel op atl evel -ground sites.The effects
of l iquefaction can be better appreciated by studying wel l -documentedcase
histories. In addition to the regul arl y publ ished journal s that deal with
geotechnical earthquake engineering, an excel l ent compil ation of case histories can
be found in Hamadaand O'Rourke (1992).9.6.1 Al teration of Ground MotionThe
infl uence of the shear modul us and damping characteristics of soil s on ground
responseis wel l establ ished.
Several exampl es of the effects of these characteristics presented inChapter 8
showed that soft soil deposits respond differentl y than stiff soil deposits to the
same motion.The devel opment of positive excess pore pressures causes soil stiffness
to decreaseduring an earthquake. A deposit of l iquefiabl e soil that is rel ativel y
stiff at the beginning ofthe earthquake may be much softer by the end of the
motion. As a resul t, the ampl itude andfrequency content of the surface motion may
change considerabl y throughout the earthquake.In the most extreme case, the
devel opment of very high pore pressures can cause thestiffness (and strength) of
even a thin l ayer to be so l ow that the high-frequency componentsof a bedrock
motion cannot be transmitted to the ground surface. An exampl e of this effectis
shown in Figure 9.45. It is not difficul t to identify the point at which
l iquefaction-inducedreduction of the stiffness of the underl ying soil took pl ace-
the accel eration ampl itude andfrequency content both changed dramatical l y about 7
sec after the motion began. The factSec. 9.6 Effects of Liquefaction 399200'"oQ)~
100E~c 0 0~Q) Cii 100 oo<t: 2000 5 10 15 20 25 30Time (sec)Figure 9.45 Accel erogram
from site near apartment buil ding resting on l iquefiabl e soil (shown in Figure 1.6)
in 1964 Niigata earthquake. (After Aki, 1988.)that surface accel eration ampl itudes
decrease when pore pressures become l arge does notmean that damage potential is
necessaril y reduced because l ow accel eration ampl itudes atl ow frequencies can stil l
produce l arge displ acements. These displ acements may be of particul arconcern for
buried structures, util ities, and structures supported on pil e foundationsthat
extend through l iquefied soil s (Figure 9.46).The occurrence of l iquefaction at
depth beneath a fl at ground surface can decoupl e thel iquefied soil s from the
surficial soil s and produce l arge, transient ground oscil l ations. Thesurficial
soil s are often broken into bl ocks (Figure 9.47) separated by fissures that can
openand cl ose during the earthquake. Ground waves with ampl itudes of up to several
feet havebeen observed during ground oscil l ation, but permanent displ acements are
usual l y smal l .Before earthquakePil e (a).....--Nonl iquefiedLiquefiedNonl iquefied
After earthq~aked ~~ Sand boil s~uqumed.. ..(b)Figure 9.46 Potential effects of
subsurfacel iquefaction on pil e foundations. The l arge strainsthat may devel op in a
l iquefied l ayer can induce highbending moments in pil es that extend through it.
Figure 9.47 Ground oscil l ation (a) before and (b) afterearthquake. (After Youd,
1984b.)400 Liquefaction Chap. 9Youd (1993) attributed to ground oscil l ation most of
the chaotic ground movements thatfractured and buckl ed pavements in the Marina
District of San Francisco during the 1989Loma Prieta earthquake. Prediction of the
ampl itude of ground oscil l ation at a particul ar siteis very difficul t; even
detail ed nonl inear ground response anal yses can provide onl y crudeestimates.9.6.2
Devel opment of Sand Boil sLiquefaction is often accompanied by the devel opment of
sand boil s. During and fol l owingearthquake shaking, seismical l y induced excess pore
pressures are dissipated predominantl yby the upward fl ow of porewater. This fl ow
produces upward-acting forces on soil particl es [these forces can l oosen the upper
portion of the deposit and l eave it in a state susceptibl eto l iquefaction in a
future earthquake (Youd, 1984a)]. If the hydraul ic gradientdriving the fl ow reaches
a critical val ue, the vertical effective stress wil l drop to zero and thesoil wil l
be in a quick condition. In such cases, the water vel ocities may be sufficient to
carrysoil particl es to the surface. In the fiel d, soil conditions are rarel y
uniform so the escapingporewater tends to fl ow at high vel ocity through l ocal ized
cracks or channel s. Sand particl escan be carried through these channel s and ejected
at the ground surface to form sand boil s.The devel opment of sand boil s is a
compl icated and somewhat random process; it dependson the magnitude of the excess
pore pressure; the thickness, density, and depth of the zoneof excess pore
pressure; and the thickness, permeabil ity, and intactness of any soil l ayersthat
overl ay the zone of high excess pore pressure. There are pitfal l s to rel iance upon
thepresence of sand boil s for evidence of l iquefaction-rel ated phenomena;
l iquefaction at greatdepths or in thin l ayers may not produce sand boil s, but l ower
excess pore pressures in thickl ayers at shal l ow depths may. Al so, the l ow
permeabil ity of sil ty sand may prevent porewaterfrom fl owing quickl y enough to
produce sand boil s, even if high excess pore pressuresdevel op. Ishihara (1985)
examined the soil conditions associated with variousl iquefaction-rel ated damage
reports from the 1983 Nihonkai-chubu earthquake (M =7.7)and 1976 Tangshan (M =7.7)
earthquakes (Gao et al . 1983) and produced estimates of thethickness of the
overl ying l ayer required to prevent l evel -ground l iquefaction-rel ated damage(Figure
9.48). These estimates have been val idated against a much l arger database forsites
not susceptibl e to ground oscil l ation or l ateral spreading, but are insufficient
for predictingdamage at other sites (Youd and Garris, 1995).Sand boil s are of
l ittl e engineering significance by themsel ves, but they are useful indicators of
high excess pore pressure generation. Shaking tabl e (Liu and Qiao, 1984) and
centrifuge (Fiegel and Kutter, 1992) tests have shown that porewater draining from
thevoids of the l oose l ayers can accumul ate beneath the l ess pervious l ayers and
form waterinterl ayers (Figure 9.49). Sand boil s can devel op when the water
interl ayers break throughto the ground surface. Some redistribution of soil grains
is al so l ikel y to accompany the formationof water interl ayers; specifical l y the
sand immediatel y beneath the water interl ayermay be l oosened by the upward fl ow of
water toward the interl ayer. If such conditionsdevel op beneath an incl ined ground
surface, the presence of the water interl ayer and thereduced steady state strength
of the l oosened sand immediatel y beneath it can contribute tol arge fl ow
deformations (Section 9.6.4).Ground surface settl ement resul ts from the vol umetric
strain that devel ops as seismical l yinduced pore pressures dissipate. The phenomenon
is il l ustrated for an el ement of soil Sec. 9.6 Effects of Liquefaction 401E 12;£ 11
Q5 10~ 9-g 8<tl (f)(1) 715 .s 6Q;5- 5: 2 4 ogJ 3(1)~ 2(J~ 1o 1 2 3 4 5 6 7 8 9 10
Thickness of surface l ayer, H1 (m)(a)C'<.'.: : : .·.·..','.s.'a·.n·: d·.·.· ....
'.: --: --"'. '...v.ii,;...·~, H1(b)Figure 9.48 (a) Rel ationship between thickness of
l iquefiabl e l ayer and thickness ofoverl ying l ayer at sites for which surface
manifestation of l evel -ground l iquefaction hasbeen observed, and (b) guides to
eval uation of respective l ayer thicknesses. (AfterIshihara, 1985; used by
permission of Kl uwer Academic Publ ishers.)Boil inghol e HfffftH /- Model foundation
I------~~~Q~'r- ~~~--)rl t: ------I------ - : -: -I'~'-: --------- : .: ...'....-...
-"': : - ..'......-... ---_----./ /._---- "'"",--: : : '; ~---_____ /-;-.: : : : - /1......
- II ......--- -: L - --Lj---g Waterinterl ayer -I---
=: : : i;: '~-'~-";~: : '=': : ~....I._"';;;;~~=-=Initial ----------------stratification
===============Figure 9.49 Formation of water interl ayers in shaking tabl e tests of
Liu and Qiao(1984).in Figure 9.50. Initial l y, the el ement is in drained equil ibrium
(zero excess pore pressure) atpoint A. Earthquake shaking causes excess pore
pressure to buil d up under undrained conditions,thereby reducing the effective
stress to that shown at point B. The excess pore pressureproduces a hydraul ic
gradient that drives the porewater out of the voids. The fl ow ofwater reduces the
hydraul ic gradient until the excess pore pressure has compl etel y dissipated(point
C). As the water fl ows from the voids, the vol ume of the el ement decreases. As
Figure 9.50 cl earl y il l ustrates, the magnitude of the vol ume change increases with
the magnitudeof the seismical l y induced excess pore pressure. Even smal l excess
pore pressures,402 Liquefaction Chap. 9(j'10-4 10-3Yeff(Geff / Gmax)cGeneration of
excess pore pressureConsol idation curveSeismical l y inducedexcessporepressureeA 1 eo
Di;;i;;~t;;'~ -oieI\AJl \l 1J1l f\)1,l .... LI eexcessporepressureFigure 9.50 Process
of earthquake-induced settl ement fromdissipation of seismical l y induced excess pore
pressure.Figure 9.51 Pl ot for determinationof effective cycl ic shear strain insand
deposits. (After Tokimatsu andSeed, 1987. Eval uation ofsettl ements in sand due to
earthquakeshaking, Journal ofGeotechnical Engineering, Vol . 113, No.8.Reprinted by
permission of ASCE.)which may not be sufficient to produce fl ow l iquefaction or
cycl ic mobil ity, can producesome postearthquake settl ement.9.6.3 Settl ementThe
tendency of sands to densify when subjected to earthquake shaking is wel l
documented.Subsurface densification is manifested at the ground surface in the form
of settl ement.Earthquake-induced settl ement frequentl y causes distress to
structures supported on shal l owfoundations, damage to util ities that serve pil e-
supported structures, and damage to l ifel inesthat are commonl y buried at shal l ow
depths.Dry sands densify very quickl y; settl ement of a dry sand deposit is usual l y
compl eteby the end of an earthquake. The settl ement of a saturated sand deposit
requires moretime-settl ement can occur onl y as earthquake-induced pore pressures
dissipate. The timerequired for this settl ement to occur depends on the
permeabil ity and compressibil ity of thesoil , and on the l ength of the drainage
path-it can range from a few minutes up to about aday.Estimation of earthquake-
induced settl ements of sands is difficul t. Errors of 25 to50% are common in static
settl ement predictions;
even l ess accuracy shoul d be expected forthe more compl icated case of seismic
l oading. Neverthel ess, the fol l owing procedures havebeen shown to produce resul ts
that agree reasonabl y wel l with many cases of observed fiel dbehavior.9.6.3.1
Settl ement of Dry SandsThe densification of dry sands subjected to earthquake
l oading depends on the densityof the sand, the ampl itude of the cycl ic shear strain
induced in the sand, and the number ofcycl es of shear strain appl ied during the
earthquake (Sil ver and Seed, 1971). Settl ementsSec. 9.6 Effects of Liquefaction 403
can be estimated using detail ed ground response anal yses with corrections for the
effects of.mul tidirectional shaking (Seed and Sil ver, 1972; Pyke et al ., 1975), or
by simpl ified procedures(Tokimatsu and Seed, 1987). In the simpl ified procedure,
the effective cycl ic shearstrain, "feye, is estimated by a procedure simil ar to
that proposed for the cycl ic strainapproach to the initiation of l iquefaction
(Section 9.5.3.2): a max (Jvrd (921)"feye = 0.65--G() .g "feyeSince the shear
modul us varies with "feye' several iterations may be required to cal cul ate aval ue
of "feye that is consistent with the shear modul us. Figure 9.51 may be used to
estimate"feye when Gmax is known. The effective cycl ic shear strain can then be
used, al ong with therel ative density or SPT resistance of the sand, to estimate the
vol umetric strain due to compaction,Co from Figure 9.52. The data from Figure 9.52
are based on 15 cycl es of strain,which corresponds to aM = 7.5 earthquake. For
earthquakes of other magnitudes, the vol umetricstrain due to compaction can be
determined from the vol umetric strain ratio presentedin Tabl e 9-4.Fol l owing the
experimental resul ts of Pyke et a1. (1975), the effect of mul tidirectional shaking
is accounted for by doubl ing cc,M' Pyke et a1. (1975) al so suggested that the
vertical component of ground motion coul d cause an additional 50% increase in
settl ement (abovethat caused by the horizontal components); the effect of vertical
motion is not expl icitl yaccounted for in the simpl ified procedure of Tokimatsu and
Seed (1987). Since the stiffness,density, and shear strain ampl itudes typical l y
vary with depth, a given soil deposit is usual l ydivided into subl ayers, with the
settl ement of each subl ayer computed and summed toobtain the total ground surface
settl ement.-.N,~40', . 1'5 cycl es- ~30 "~20 ",~15 "~10 ",~5" ""
, , , , , , , , , , , , , , , , , ,Cycl ic shear strain, Ycyc (%)10.3 10.2 10.1 1w"
10.3Co~0-~ 10.2o.8Q)-@ 10.1c.~1i5o~E~o>...Figure 9.52 Rel ationship between
vol umetric shear strain, Ec.M ~ 7.5 and cycl ic shearstrain, Yeye, in terms of (a)
rel ative density and (b) standard penetration resistance. (AfterTokimatsu and Seed,
1987. Eval uation of settl ements in sand due to earthquake shaking,Journal
afGeotechnical Engineering, VoL 113, No.8. Reprinted by permission ofASCE.)404Tabl e
9-4 Infl uence ofEarthquake Magnitudeon Vol umetric Strain forDry SandsEarthquake
MagnitudeLiquefaction Chap. 90.40.60.851.01.25Source: After TokimatsuandSeed, 1987
Exampl e 9.6The settl ement of a 50-ft-thick deposit ofl oose sand (Dr = 45%) with an
average corrected SPTresistance of9 was estimated for an earthquake of magnitude
6.6 by Seed and Sil ver (1972) andTokimatsu and Seed (1987).Sol ution The deposit can
be divided into a series of l ayers; six l ayers are used in this exampl eas
il l ustrated in tabul ar form bel ow. Based on the given (N j )60val ues, Gmax can be
estimated usingthe rel ationship of Ohto and Goto (1976) given in Tabl e 6-6 (col umn
5 bel ow). This al l ows estimationofYeye (col umn 6 bel ow) using Figure 9.51. The
resul ting vol umetric strain (for a magnitude7.5 earthquake) can then be obtained
from Figure 9.52 (col umn 7 bel ow). Using Tabl e 9-5,the vol umetric strain in aM= 6.6
earthquake shoul d be about 80% of that in a magnitude 7.5 earthquake(col umn 8
bel ow). Accounting for mul tidirectional shaking, the vol umetric strains are then
doubl ed (col umn 9 bel ow). Final l y, the settl ement of each l ayer can be estimated as
the product ofthe vol umetric strain and the thickness ofthe l ayer (col umn 10
bel ow).(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)Layer Thickness Dr Gmax f.c,M=7.5
f.c,M=6.6 2f. c.M=6 .6 Settl ementNumber (ft) (%) (N j )60 (ksf) Yeye (%) (%) (%)
(in.)I 5 45 9 520 0.0005 0.14 0.11 0.22 0.132 5 45 9 900 0.0008 0.23 0.18 0.36 0.22
3 10 45 9 1270 0.0012 0.35 0.28 0.56 0.674 10 45 9 1630 0.0014 0.40 0.32 0.64 0.775
10 45 9 1930 0.0015 0.45 0.36 0.72 0.866 10 45 9 2190 0.0013 0.38 0.30 0.60 0.72
3.37These settl ements are consistent with those that were observed at a site with
these conditions inthe 1971 San Fernando earthquake (Tokimatsu and Seed, 1987). The
number of significant figuresin this exampl e is intended to facil itate
understanding of the cal cul ations, not to suggestthat the resul ts are accurate to
0.01 in.9.6.3.2 Settl ement of Saturated SandsThe postearthquake densification of
saturated sand is infl uenced by the density of thesand, the maximum shear strain
induced in the sand, and the amount of excess pore pressureSec. 9.6 Effects of
Liquefaction 405generated by the earthquake. Laboratory experiments have shown that
the vol umetric strainafter initial l iquefaction varies with rel ative density and
maximum shear strain. Tokimatsuand Seed (1987) used a correl ation between (N\)60
and rel ative density and an estimate of theshear strain potential ofl iquefied soil
from (Nj ) 60 and cycl ic stress ratio (Seed et al ., 1984)to produce a chart (Figure
9.53) that al l ows the vol umetric strain after l iquefaction in aM =7.: 5 earthquake
to be estimated directl y from the cycl ic stress ratio and SPT resistance. For
earthquakes of other magnitudes, an equival ent cycl ic stress ratio, CSRM, can be
determinedfrom equation (9.2) and Tabl e 9-2. Note that the vol umetric strain after
l iquefaction can beas high as 2 to 3% for l oose to medium dense sands and higher
for very l oose sands; a 5-mthick (16 ft) l ayer of very l oose sand produced
settl ements of 50 to 70 em (20 to 28 in.) inHachinohe, Japan fol l owing the
Tokachioki earthquake (M =7.9) in 1968 (Ohsaki, 1970);settl ements of 50 to 100 em
(20 to 39 in.) were observed on Port Isl and and Rokko Isl and inKobe, Japan
fol l owing the 1995 Hyogo-ken Nanbu earthquake. If sand boil s are produced,
postearthquake settl ements are l ikel y to be irregul ar. The resul ting differential
movementcan be damaging to structures, pavements, or pipel ines on or near the
ground surface.0.6 ,-------,-------,----,-----,---------,Figure 9.53 Chart for
estimation ofvol umetric strain in saturated sands fromcycl ic stress ratio and
standard penetrationresistance. (After Tokimatsu and Seed,1987. Eval uation of
settl ements in sand due50 to earthquake shaking, Journal ofGeotechnical
Engineering, Vol . 113, No.8.Reprinted by permission of ASCE.)40Vol umetric strain
(%)10543 2 1 0.5Til iO.2.4#"/ ..../ • • • • • • • • • : : >...0.50.4 ..to r-: IIE 0.3 ..a: (J)0
0.20.1 ..00Exampl e 9.7Significant settl ement was observed in the Marina District of
San Francisco fol l owing the 1989Loma Prieta earthquake. Much of this settl ement was
determined to have resul ted from densificationof hydraul ic fil l s that were pl aced
to recl aim the area from San Francisco Bay in the1890s. Subsurface conditions at
one l ocation in the Marina District are shown in Figure E9.7(O'Rourke et aI.,
1991). Estimate the settl ement that woul d have been expected at this l ocationin the
Loma Prieta earthquake. A peak accel eration of 0.20g was measured in the vicinity
of theMarina District.Sol ution The subsurface profil e shows measured (uncorrected)
SPT resistances. Assumingthat the SPT resistances were measured using standard
procedures (Em = 0.60Eff) , correctedSPT resistances can be computed using equation
(6.30) (col umn 4 bel ow). The simpl ified procedureof Seed and Idriss (1971) can be
used to estimate the cycl ic stress ratio (col umn 5 bel ow).406 Liquefaction Chap. 9
Sqil TypeFil l Loose sandy fil l Recent bay mudN598195o102030Dense sandFigure E9.7(1)
(2) (3) (4) (5) (6) (7)Layer Depth Range (m) N (Nj )60 CSR tv (%) Mf(m)1 2.5-3.5 5
7.5 0.141 3.0 0.0302 3.5-5.3 9 11.4 0.165 2.4 0.0433 5.3-6.8 8 8.6 0.180 2.8 0.0424
6.8-8.5 19 18.7 0.186 0.2 0.0030.118Depth (m)The vol umetric strain can then be
estimated using Figure 9.53. The settl ement of each l ayer isthen computed as the
product of the vol umetric strain and the l ayer thickness.The actual settl ements in
the vicinity of this site were on the order of 12 to 15 ern. Some oftheactual
settl ements resul ted from densification of the dry soil above the water tabl e; that
settl ementis not accounted for in this exampl e.In an al ternative approach, either
the factor of safety against l iquefaction [equation(9.14)] or the maximum cycl ic
shear strain, and the rel ative density, SPT resistance, or CPTtip resistance, can
be used to estimate postl iquefaction vol umetric strain (Figure 9.54). Integrationof
these vol umetric strains over the thickness of the l iquefied l ayer produces the
ground surface settl ement.Sec. 9.6 Effects of Liquefaction2.0 : 1.8 .1.6 .1.44070.2
--1.20.80.60.4~""'''''------'''': : : ''''''-----''''''''',: ,: ''.: .: ''---''''''''''~ ~
- ~ ~ ~~~>~ - -1 In .... .... ,.'"~ ,.. ---.; ,," 0,= 40 0,= 30, 0, = 50 [N1= 6)
[N1 = 3 i0,=60 [N1=10) qc1=45 qC1=33)/ [N1 = 14) qc1 = 60/ Qc1 = 80/ / 0,=708% /
(N, = 20,Qc1 =110)Ymax=10% 0,=80/ (N1 =25,Qc1 =147)0,=90%[N, = 30 )Qc1 = 200kg I
em2o 1.0 2.0 3.0 4.0 5.0Post-Iiquefication vol umetric strain, Ev(%)Figure 9.54
Chart for estimating postl iquefaction vol umetric strain of cl ean sand asfunction of
factor of safety against l iquefaction or maximum shear strain. (After Ishiharaand
Yoshimine, 1992; used by permission of JSSMFE.)Exampl e 9.8 ,Repeat Exampl e 9.7
using
the Ishihara-Yoshimine approach,Sol ution The Ishihara-Yoshimine approach requires
eval uation of the factor of safety againstl iquefaction. Using the (N])60 val ues
from Exampl e 9.7 and Figure 9.31, the (magnitude corrected)critical stress ratio
required to initiate l iquefaction can be determined (col umn 4 bel ow).Using the CSR
val ues from Exampl e 9.7 and equation (9.11), the factor of safety against
l iquefactioncan be computed for each l ayer. Then the (Nj )60val ues must be
converted to N] val ues[recal l that Japanese SPT procedures typical l y transmit 20%
more energy to the SPT sampl er;hence N] = 0.833(N1)60]. Using the resul ting val ues
of and FSL, Figure 9.54 can be used to estimatethe vol umetric strain in each l ayer
(col umn 8 bel ow). Again, the settl ement of each l ayer(col umn 9 bel ow) is given by
the product of the vol umetric strain and l ayer thickness.The total settl ement is
cal cul ated as 17.7 em by the Ishihara-Yoshimine procedure.Hence the Tokimatsu-Seed
and Ishihara-Yoshimine procedures both provided good estimates408 Liquefaction
Chap. 9(1) (2) (3) (4) (5) (6) (7) (8) (9)Layer Depth Range (m) (N j )60 CSRL CSR
FSL Nj tv (%) til l (m)1 2.5-3.5 7.5 0.088 0.141 0.62 6.3 4.4 0.0442 3.5-5.3 11.4
0.134 0.165 0.81 9.5 3.5 0.0633 5.3-6.8 8.6 0.102 0.180 0.57 7.2 4.2 0.0634 6.8-8.5
18.7 0.220 0.186 1.18 15.6 0.4 0.0070.177of the observed postearthquake
settl ementat this site. Ishihara(1993)il l ustrated the use of theIshihara-Yoshimine
procedure for a nearby site (with a smal l er thickness of hydraul ic fil l )usingCPT
data; a settl ement of 7.3 ern was predicted.In many cases, strong ground motion
produces excess pore pressures that are not sufficientto produce initial
l iquefaction. These pore pressures wil l dissipate, however, andmay produce some
vol ume change. Tokimatsu and Seed (1987) combined a rel ationshipbetween normal ized
stress ratio, CSR/CSRL , and pore pressure ratio (Tokimatsu andYoshimi, 1983) and a
rel ationship between pore pressure ratio and vol umetric strain (Leeand Al baisa,
1974) to devel op the rel ationship shown in Figure 9.55. Figure 9.55 can beused to
estimate the vol umetric strain for cases in which the maximum pore pressure ratiois
l ess than 100%. Note that postearthquake vol ume changes are quite smal l unl ess the
CSRis greater than 70 to 80% of that required to cause initial l iquefaction tr, =
100%).The rate at which settl ement devel ops depends on the reconsol idation
characteristicsof the l iquefied soil . Reconsol idation general l y begins at the
bottom of the l iquefied soil (Fl orin and Ivanov, 1961; Heidari and James, 1982) and
proceeds upward. Because effectivestresses can vary over several orders of
magnitude during reconsol idation, nonl inear consol idationmodel s are required to
predict settl ement rates (Whitman et al ., 1982; Scott, 1986).1.21.0 --~.c§ 0.8en o
0.6 --'C1il E 0.4 : > g0.200 0.2 0.4 0.6CSRCSRLLoosesandS~!Dense : sands~: /ii0.8 1.0
Figure 9.55 Postearthquake vol umetricstrains for pore pressure ratios l ess than
100%. (After Tokimatsu and Seed, 1987.Eval uation of settl ements in sand due to
earthquake shaking, Journal ofGeotechnical Engineering, Vol . 113, No.8.Reprinted by
permission of ASCE.)9.6.4 Instabil ityLiquefaction-induced instabil ities are among
the most damaging of al l earthquake hazards.Their effects have been observed in the
form of fl ow sl ides, l ateral spreads, retaining wal l Sec. 9.6 Effects of
Liquefaction 409fail ures, and foundation fail ures in countl ess earthquakes
throughout the worl d. Instabil ityfail ures can be produced by different l iquefaction
phenomena, and it is not al ways cl earexactl y which is responsibl e for a given
fail ure. Al l are strongl y infl uenced, however, by theshear strength of the
l iquefied soil .9.6.4.1 Shear Strength of Liquefied Soil Instabil ity occurs when the
shear stresses required to maintain equil ibrium of a soil deposit exceed the shear
strength of that deposit. The soil then deforms until it reaches aconfiguration in
which the shear stresses do not exceed the shear strength. The amount ofdeformation
required to reach a stabl e configuration is strongl y infl uenced by the difference
between the shear stresses required for equil ibrium and the shear strength of the
l iquefiedsoil . If the shear strength of the l iquefied soil is onl y sl ightl y l ower
than the shear stressrequired for equil ibrium, permanent deformations are l ikel y to
be smal l . If the differencebetween the shear strength and shear stress is l arge,
very l arge deformations may devel op.Accurate eval uation of the effects of
l iquefaction-induced instabil ity requires accurate estimationof the shear strength
of the l iquefied soil .Three different approaches have been devel oped to estimate
the shear strength of l iquefiedsoil s. One is based on a program of careful
undisturbed sampl ing and l aboratory testing.Another is based on in situ test
parameters and interpretation of known l iquefactioncase histories. The third treats
the shear strength of l iquefied soil as a normal ized strength.Laboratory Testing
Approach: Steady-State Strength. The steadystatestrength (defined in Section
9.4.4.2) governs the behavior of l iquefied soil . In concept,it is a function sol el y
of the density of the soil , al though some experimental studies suggestthat it may
be infl uenced by stress path and other factors. For many soil s, the steady-state
strength is extremel y sensitive to density (i.e., smal l changes in density can
resul t in l argechanges in steady-state strength).If cohesionl ess soil s coul d be
sampl ed "perfectl y" (i.e., without any disturbance whatsoever),the steady-state
strength, Ssu, of a given specimen coul d be measured in a singl e triaxial test by
consol idating the specimen to the in situ stress conditions and then shearing it
under undrained conditions fol l owing the most appropriate stress path.
Unfortunatel y, theacts of sampl ing, transportation, handl ing, and consol idation
cause changes in the density ofa l aboratory specimen by the time it is tested.
Given the sensitivity of Ssu to density, thesteady-state strength of a l aboratory
specimen must be corrected to correspond to the in situvoid ratio. Poul os et al .
(1985) proposed the fol l owing four-step procedure for measurementof in situ steady-
state strength. The procedure is il l ustrated graphical l y in Figure 9.56.1.
Determine the insitu void ratio from one or more "undisturbed" specimens. This is
themost critical , and most difficul t, step in the entire procedure. Poul os et al .
(1985) suggestedthree satisfactory sampl ing procedures: (1) fixed-piston sampl ing,
(2) freezingof the ground and coring, and (3) sampl ing in test pits. The importance
of careful sampl ingto the success of this procedure cannot be overemphasized.2.
Locate the steady-state l ine using reconstituted specimens. Since the sl ope of the
SSLis chiefl y infl uenced by grain shape, the SSL from a series of reconstituted
specimensshoul d have the same sl ope as that of the in situ soil . Tests on five or
six specimensconsol idated to initial states wel l above the SSL (to ensure
contractive behavior) aresuggested.410 Liquefaction Chap. 9100,000Void ratio after
sampl ing, extrusionand consol idationUndrained steady strengthfor "undisturbed"
specimenAssumed steadystate strength l ine for"undisturbed"specimen1,000 10,000
Steady state strength, Ib!ft2Steady state l inefor the compactedspecimens0.7 .In-
situ void ratiofor the "undisturbed"../ specimen0.5- ~stimated in-si: I.': : : :
0.4 ..· undrained steadystate strength0.3 '--_~~~~~-LL"""~_-'--4l "---'-'-'-
i...i."---_-'-_'---'-'--'--'...i...i.J100o~0.6""C~Figure 9.56 Procedure for
estimating steady-state shear strength. (After Poul os et aI.,1985. Liquefaction
eval uation procedure, Journal ofGeotechnical Engineering, Vol . III,No.6. Reprinted
by permission of ASCE.)3. Determine the steady-state strength of an undisturbed
specimen. A consol idatedundrainedtriaxial test is performed on a specimen obtained
from step 1. To ensurethat the steady state is reached at reasonabl e strain l evel s,
consol idation to el evatedeffective confining pressures (again, to produce
contractive behavior) is suggested.4. Correct the measured steady state strength to
the in situ void ratio. By assuming thatthe SSL of the undisturbed specimen has the
same sl ope as that of the reconstitutedspecimens, the in situ steady-state strength
can be found by projecting a l ine throughthe point describing the undisturbed
specimen (from step 3) paral l el to the SSL (fromstep 2) and back to the in situ
void ratio (from step 1). The resul ting strength is takenas the in situ steady-
state strength.The corrected in situ steady-state strength can then be used to
eval uate potential l iquefactionhazards. Al though this procedure is rational , the
user must remain aware of thesensitivity of the resul ts to uncertainties in the
input parameters, particul arl y the in situ voidratio. Kramer (1989) presented a
procedure for estimating the uncertainty in SSU'In Situ Testing Approach: Residual
Strength. As an al ternative to l aboratory-based procedures, Seed (1986) devel oped a
correl ation between SPT resistance andthe apparent shear strength back-cal cul ated
from observed fl ow sl ides. Since there is noguarantee that al l the conditions of
the steady state of deformation were satisfied in the casehistories-indeed, Stark
and Mesri (1992) suggested that partial drainage occurred beforesome of the fl ow
sl ides came to rest-the back-cal cul ated strength is termed the residual strength.
Seed and Harder (1990) reanal yzed many of the fl ow sl ides and added new data toSec.
9.6 Effects of Liquefaction 411devel op a rel ationship between residual strength and
an equival ent cl ean-sand SPT resistance(Figure 9.57). For sands with more than 10%
fines, the equival ent cl ean-sand SPTresistance is obtained from(Nj )60- cs = (Nj
)6o+ Ncorrwhere Ncorr is obtained from Tabl e 9-5.(9.22)Tabl e 9-5 Recommeded Fines
Correction forEstimation of Residual Undrained Strength bySeed-Harder and Stark-
Mesri ProceduresNCOIT (bl ows/ft)Percent Fines Seed-Harder Stark-Mesri0 0 010 I 2.5
15 420 525 2 630 6.535 750 4 775 5 7200024D Construction - induced l iquefaction and
sl iding case histories.• Earthquake - induced l iquefaction and sl iding case
histories whereSPT data and residual strength parameters have been measured.o
Earthquake - induced l iquefaction and siiding case histories whereSPT data and
residual strength parameters have been measured.OL-_~: : : ""-_....o;",=-__-,-__---,-
__----,-__---.JoCcii..--: 1600s: "5>c~U;til 1200(]) .r: (J)""0(])c 800 _ .~""0 c: : : : l
""iii-6 400'w(]) a: Figure 9.57 Rel ationship between residual strength and corrected
SPT resistance.(After Seed and Harder, 1990. H. Bol ton Seed Memorial Symposium
Proceedings, Vol .2, p. 371. Used by permission of BiTech Publ ishers, Ltd.)412
Liquefaction Chap. 9The back-cal cul ation procedures used to obtain the residual
strength are not simpl emanyassumptions and approximations are invol ved. Combined
with the variabil ity of theSPT and the difficul ty in sel ecting a singl e SPT val ue
to represent each case history, theseresidual strengths must be considered
approximate, as suggested by the wide band in Figure9.57. As more data become
avail abl e and as back-cal cul ation procedures improve, thisuncertainty is l ikel y to
decrease.Exampl e 9.9In the 1971 San Fernandoearthquake, soil within the
upstreamsl ope of Lower San FernandoDaml iquefiedandproduceda fl ow sl ide that
nearl ybreachedthedam (Figure 1.7).Subsurfaceinvestigations indicatedthat the
hydraul ical l y pl aced sil ty sandin the upstreamsl ope containedabout 25% fines and
had an average measured (N 1)60 of about 11.5. Estimate the residual strengthof the
sil ty sand.Sol ution For a sand with 25% fines,Ncorr = 2, so(N')60-cs = 11.5+2 =
13.5Referringto Figure9.57,the estimatedresidual strengthshoul dbe betweenabout300
l b/fr' and750 l b/fr' (the val ues given by the upper and l ower curves). Considering
that the data fromactual case histories of earthquake-induced l iquefactionfal l in
the l ower portion of this range,the residual strength appears most l ikel y to be on
the order of 300 to 500 l b/fr',Normal ized Strength Approach: Residual Strength
Ratio. The conceptof normal ized strength is widel y accepted in geotechnical
engineering practice for cohesivesoil s (Ladd and Foott, 1974). The general concept
can al so be appl ied to the residual strength of l iquefied soil s. The significant
advantage of this approach is that it al l ows estimationof residual strength from
initial effective stresses, which are much easier to determineaccuratel y than
initial void ratios.If the consol idation curve and steady-state l ine of a
l iquefiabl e soil are paral l el , thesteady state strength shoul d be proportional to
the consol idation stress (i.e., Ssu/O"l c =constant).Appl ication of this concept is
compl icated by the fact that sandy soil s do not exhibitunique consol idation curves;
hence the ratio Ssu/ 0"1 c is not unique for a given soil . However,if l aboratory
specimens are prepared with the same void ratio and soil fabric that existin the
fiel d at comparabl e initial effective stresses, measured residual strength ratios
can berepresentative offiel d conditions (Vasquez-Herrera et al ., 1990; Baziar et
al ., 1992; Ishihara,1993). Vasquez-Herrera et al . (1990) found that S/O"~o '" 0.12
for reconstituted specimensof sil ty sand from Lower San Fernando Dam. Byrne et al .
(1993) used specimens obtained byfreezing and coring at Duncan Dam to measure
S/O"~o '" 0.21 . The normal ized strengthapproach has al so been used with residual
strength measured in fiel d vane shear tests. Castroand Troncoso (1989) used vane
shear tests to measure S/O"~o val ues of 0.07, 0.11, and 0.08in very l oose sl imes
(PI =20 to 22) at three different tail ings dams in Chil e. For a cl ayey sil tin the
foundation of Sardis Dam (Finn et al ., 1991), vane shear tests indicated thatS/O"~o
'" 0.075 . The l imited avail abl e data indicate that the normal ized strength
approachhas promise as a practical tool for estimation of the residual strength of
l iquefied soil s. Thedata al so cl earl y indicate that residual strength ratios can
vary over a considerabl e range andthat their eval uation requires careful , site-
specific investigation and testing.The basic normal ized strength approach requires
site-specific measurement of the residual strength ratio. It is l ogical to expect
that different soil s woul d have different residual Sec. 9.6 Effects of Liquefaction
413strength ratios; for exampl e, the residual strength ratio of a wel l -graded
angul ar sand shoul d begreater than that of a uniform rounded sand. Assuming that
the factors that infl uence the residual strength ratio al so infl uence in situ test
resul ts, it shoul d be possibl e to rel ate the residual strength ratio to in situ test
parameters. Stark and Mesri (1992) used back-cal cul ated and l aboratory-measured
residual strengths from fiel d case histories to correl ate val ues of residual
strength ratio to SPT resistance, that is,S;" = 0.0055(N j ) 6o _ cs (9.23)avowhere
(N1)60-cs is computed using the val ues of Ncarr given in Tabl e 9-5. Al though this
procedurehas the important advantage of al l owing the residual strength ratio to be
estimatedfrom SPT resistance, equation (9.23) must be recognized as a reasonabl y
conservativeapproximation to a set of scattered empirical data.Exampl e 9.10
According to Stark and Mesri (1992), the effective overburden pressure at the
center of the l iquefiedportion of the upstream sl ope of Lower San Fernando Dam (see
Exampl e 9.9) was 3930Ib/ft2. Estimate the residual strength of these soil s using
the Stark-Mesri approach.Sol ution According to the Stark-Mesri approach, a sil ty
sand with 25% fines woul d indicate(from Tabl e 9-6) Ncarr = 6, so(Nd60-cs = 11.5 + 6
= 17.5Then, using equation (9.23), the residual strength can be estimated asS, =
0.0055(Nl )60-cs (")~o = (0.0055)(17.5)(3930 Ib/ft2) = 3781b/ft2This val ue is
consistent with the range of val ues obtained in Exampl e 9.9.Discussion. Eval uation
of the residual strength of a l iquefied sand is one of themost difficul t probl ems
in contemporary geotechnical earthquake engineering practice.Steady-state concepts
are very useful for understanding the behavior of l iquefiabl e soil , butsteady-state
strengths can be stress path dependent. In the fiel d, fail ure general l y invol ves
different stress paths, and consequentl y different steady-state strengths, on
different parts ofthe fail ure surface. Furthermore, some drainage may occur during
fail ure. Given these factors,it is not surprising that residual strengths are
difficul t to determine.The l aboratory testing approach is rational and based on
tests in which the shearstresses are accuratel y known, but the sensitivity of its
resul ts to uncertain input parametersintroduces considerabl e uncertainty into
steady-state strength estimates. The in situ testingapproach is al so rational , but
the variabil ity of soil and stress conditions in the case historieson which it is
based introduces considerabl e uncertainty into residual strength estimates.Upon
reviewing the appl ication of these approaches to the Lower San Fernando Dam case
history, Marcuson et al . (1990) preferred the in situ approach but suggested that
both beused where hazards are high. The normal ized strength approach offers some
advantagesover the other approaches, but avail abl e data suggest that residual
strength ratios vary considerabl yfor different soil s.9.6.4.2 Fl ow Fail ures
Liquefaction-induced fl ow fail ures occur when the shear stresses required to
maintainstatic equil ibrium are greater than the shear strength of a l iquefied soil .
This situation can arise414 Liquefaction Chap. 9in several different ways; the
National Research Council (1985) identified four differentmechanisms of fl ow
fail ure. Estimation of the deformations produced by l iquefaction-inducedfl ow
fail ures is extremel y difficul t; avail abl e procedures are described in Section
10.6.2.1.Fl ow Liquefaction Fail ures (NRC Mechanism AJ. Fl ow l iquefactionrepresents
an important fl ow fail ure mechanism. Fl ow l iquefaction occurs under total l y
undrained conditions-no redistribution of pore water (or change in void ratio) is
invol ved.As described in Section 9.5.2.1, fl ow l iquefaction is initiated when
sufficient pore pressureis generated to move the effective stress path of an
el ement of soil from its initial positionto the fl ow l iquefaction surface. When
that occurs, the el ement becomes unstabl e and fl owl iquefaction fail ure begins in
that el ement of soil . In other words, a fl ow l iquefaction fail ureoccurs at the
l ocations where l iquefaction is initiated by earthquake shaking. These
characteristicscan be used to distinguish between fl ow l iquefaction fail ures and
other types offl ow fail ures. Fl ow l iquefaction fail ures often occur very quickl y
and produce l arge soil movements (Figure 9.58).Before ~fl ow sl ide ).-----1
_---: 7'/~A."- - - - - - _- - - - - - - - - fl ow sl ide(a)qFLSB(b)ApFigure 9.58 (a)
Typical cross section through a fl ow sl ide showing how l iquefiedmaterial s can
travel l arge distances and come to rest on very fl at surfaces; (b) stressconditions
at typical el ement of soil within fail ed mass. Prior to earthquake, el ement is in
equil ibrium at point A under static shear stress greater than steady state (or
residual )strength. Cycl ic l oading brings stress path to FLS (point B) after which
strain softeningreduces shearing resistance to steady-state (or residual ) strength
(point C).Fl ow l iquefaction fail ures can devel op progressivel y (i.e., the
initiation of fl ow l iquefactionin a smal l vol ume of soil may spread to produce a
l arge fl ow fail ure).
When fl owl iquefaction is initiated at a particul ar l ocation, the shearing
resistance drops to the steadystatestrength. The static shear stresses that were
resisted at that l ocation must then be transferredto the surrounding soil , where
they may initiate further fl ow l iquefaction. As theredistribution of stresses
proceeds, the zone of l iquefaction grows. Eventual l y, a massivefl ow sl ide may
devel op.Local Loosening Fl ow Fail ure (NRC Mechanism BJ. Since the steadystate
strength is very sensitive to the density of many soil s, a smal l amount of
l oosening canreduce the steady-state strength substantial l y. In some cases,
l oosening may reduce thesteady-state strength to a val ue smal l er than the shear
stress required for equil ibrium, therebyproducing a fl ow fail ure.If a sand l ayer is
overl ain by a l ess permeabl e material that does not permit drainageduring the
earthquake itsel f, the total vol ume of the sand wil l remain constant. If a
conditionSec. 9.6 Effects of Liquefaction(a)qInitial steadystate pointSteady state
I point after ': r'o, B... ___(b)ApFLS415Figure 9.59 Fl ow fail ure due to l oosening:
(a) redistribution of grains in sand l ayercauses l ocal vol ume change even though
total vol ume remains constant; (b) stressconditions at point marked with x. Prior
to earthquake, soil is in equil ibrium at point Awith steady-state strength (point
B) that is greater than static shear stress. As effectivestresses are reduced
during earthquake, l oosening of soil reduces steady-state strength tol ower val ue
(point C), thereby al l owing fl ow fail ure to occur.of initial l iquefaction (zero
effective stress) is reached, however, the sand particl es mayrearrange under the
action of gravity so that the l ower part of the l ayer becomes denser andthe upper
part l ooser. If the upper part l oosens sufficientl y to reduce the steady-state
strength to a val ue smal l er than the static shear stress, a l ocal l oosening fl ow
fail ure canoccur, as il l ustrated in Figure 9.59. In extreme cases a water
interl ayer may form beneath thel ess permeabl e material . Since the water interl ayer
woul d have zero shear strength, a fl owfail ure coul d easil y be produced.Gl obal
Loosening Fl ow Fail ure (NRC Mechanism CJ. High excess porepressures generated at
depth wil l cause porewater to fl ow toward drainage boundaries duringand after an
earthquake. As il l ustrated in Figure 9.60a, most of the fl ow is usual l y directed
toward the ground surface. Shal l ow soil s may be l oosened by this fl ow to the extent
that theirsteady-state strength drops bel ow the shear stress required to maintain
equil ibrium. In contrastwith the l ocal l oosening case, this l oosening is not
compensated for by densification at a differentl ocation. The process is il l ustrated
schematical l y in Figure 9.60b. Since the steady-statestrength is not reduced until
water fl ows into the shal l ower soil , fail ure may not occur until wel l after the
earthquake. Cracking of the surficial soil s may al so contribute to the fail ure.
Interface Fl ow Fail ure (NRC Mechanism DJ. Fl ow-type fail ures canal so occur when the
shear strength of the interface between a l iquefiabl e soil and a structurebecomes
smal l er than the shear stress required for equil ibrium. Pl unging fail ure of
frictionpil es (DeAl ba, 1983) is an exampl e of an interface fl ow fail ure. If the
interface is smooth, aswith steel or precast concrete pil es, interface fl ow fail ure
does not require vol ume change ofthe soil and therefore can occur in contractive or
dil ative sands.9.6.4.3 Deformation Fail uresNot al l l iquefaction-rel ated fail ures
invol ve fl ow and l arge displ acements. Cycl icmobil ity can produce smal l ,
incremental , permanent deformations that, by the end of anearthquake, may be
sufficient to produce extensive damage. Lateral spreading is an exampl eof
deformation fail ure. As il l ustrated in Figure 9.61, l ateral spreading causes
surficial Steady state pointafter l oosening416Sand l oosenedby outward fl ow(a)Zone of
high porepressureqcLiquefactionInitial steadystate pointEnd of JshakingI B~ _______
tAp(b)Chap. 9FLSFigure 9.60 (a) Exampl e of gl obal l oosening fl ow fail ure (after
National ResearchCouncil , 1985) where earthquake-induced pore pressure at depth
causes fl ow thatl oosens surficial soil s; (b) possibl e effective stress path for
el ement of surficial soil . Priorto earthquake, static shear stress (point A) is
l ess than steady-state strength (point B).Loosening due to outward fl ow of
porewater reduces steady-state strength (from point Bto point C), al l owing fl ow
fail ure to occur.(a)(b)---------t~"--I-------------- (Initial sectionBl ocks~R;'"~
D,jocmooYFigure 9.61 Lateral spreading adjacent to a river channel (a) before and
(b) afterearthquake. Lateral movement of l iquefied soil (shaded zone) breaks
surface l ayer intobl ocks separated by fissures. Bl ocks may til t and settl e
differential l y, and sand boil s mayerupt at fissures. (After Youd, 1984b.)l ayers to
break into bl ocks that progressivel y move downsl ope or toward a free face (Figure
9.2) during earthquake shaking. The ground surface may exhibit fissures and scarps
at thehead of the l ateral spread, shear zones al ong its l ateral margins, and
compressed or buckl edSec. 9.7 Summary 417soil at the toe. The surficial bl ocks
usual l y move irregul arl y in both horizontal and vertical directions; buil dings and
pipel ines extending across or through the head of a l ateral spreadmay be pul l ed
apart, pipel ines crossing the l ateral margins may be sheared, and bridges or
pipel ines near the toe may be buckl ed (Figure 1.8). Lateral displ acements usual l y
rangefrom a few centimeters to a meter or two, but may be l arger if shaking is
particul arl y strongor of l ong duration. Procedures for estimating l ateral spreading
deformations are describedin Section 10.6.2.2.9.7 SUIVl IMARY1. The term
l iquefaction has been used to describe a number of rel ated but different phenomena
observed in l oose, saturated soil s. For engineering purposes, these phenomenacan be
divided into three main groups: fl ow l iquefaction, cycl ic mobil ity, andl evel -ground
l iquefaction.2. Fl ow l iquefaction can occur when the static shear stress in a
l iquefiabl e soil deposit isgreater than the steady-state strength of the soil . It
can produce devastating fl ow sl idefail ures during or after earthquake shaking. Fl ow
l iquefaction can occur onl y in l oosesoil s.3. Cycl ic mobil ity can occur when the
static shear stress is l ess than the steady-statestrength and the cycl ic shear
stress is l arge enough that the steady-state strength isexceeded momentaril y.
Deformations produced by cycl ic mobil ity devel op incremental l ybut can become
substantial by the end of a strong and/or l ong-duration earthquake.Cycl ic mobil ity
can occur in both l oose and dense soil s but deformation decreasesmarkedl y with
increased density.4. Level -ground l iquefaction can occur when cycl ic l oading is
sufficient to produce highexcess pore pressures, even when static driving stresses
are absent. Its occurrence isgeneral l y manifested by ground oscil l ation,
postearthquake settl ement, and/or thedevel opment of sand boil s. Permanent l ateral
displ acements due to l evel -ground l iquefactionare usual l y smal l . Level -ground
l iquefaction can occur in l oose and densesoil s.5. Liquefaction hazard eval uation
requires that questions of l iquefaction susceptibil ity,initiation, and effects be
addressed. For a site to be considered free from l iquefactionhazards, the soil s
must be nonsusceptibl e to l iquefaction, the anticipated l oading mustbe insufficient
to initiate l iquefaction, or the effects of l iquefaction must be tol erabl e.6.
Liquefaction susceptibil ity can be judged on the basis of historical , geol ogic,
compositional ,and state considerations. Geol ogic, compositional , and state criteria
must bemet for a soil to be susceptibl e to l iquefaction; if any of these criteria
are not met, thesoil is nonsusceptibl e to l iquefaction.7. Liquefaction
susceptibil ity is different for different l iquefaction phenomena. A soil that is
susceptibl e to cycl ic mobil ity or l evel -ground l iquefaction may not be susceptibl e
to fl ow l iquefaction. Susceptibil ity to various l iquefaction phenomena depends
primaril y on the state (stress and density conditions) of the soil at the time of
the earthquake.418 Liquefaction Chap. 98. Under given l oading conditions, any sand
wil l reach a unique combination of effectiveconfining pressure, shear strength, and
density at l arge strains. The combinationcan be described graphical l y by a steady-
state l ine. The position of the steady-statel ine is most strongl y infl uenced by
grain size and grain shape characteristics. Thebehavior of a sand is strongl y
rel ated to its position rel ative to the steady-state l ine.9. Fl ow l iquefaction is
initiated when the principal effective stress ratio reaches a critical val ue under
undrained, stress-control l ed conditions. The stress state at the initiationof fl ow
l iquefaction can be described graphical l y in stress path space by the fl ow
l iquefactionsurface. Once the effective stress path of an el ement of soil reaches
the fl owl iquefaction surface, additional straining wil l induce additional excess
pore pressureand the avail abl e shearing resistance wil l drop to the steady-state
strength.10. Because the state of a l oose sand in equil ibrium under high initial
shear stress is cl oserto the fl ow l iquefaction surface than that of a simil ar soil
subjected to l ower initial shear stress, l ess excess pore pressure is required to
initiate fl ow l iquefaction. Fl owl iquefaction can be triggered by smal l undrained
disturbances in soil s subjected tohigh initial shear stresses. Such soil s may
represent a high l iquefaction hazard.11. Cycl ic mobil ity can produce high excess
pore pressures and l ow effective stresses, butunidirectional movement wil l cause
the soil to dil ate. The increased shearing resistanceproduced
by dil ation wil l arrest soil movement so that fl ow sl ides cannot devel op.12. The
existence of sand boil s is often taken as evidence of l evel -ground l iquefaction.
Sand boil formation, however, depends on factors such as the depth, thickness, and
void vol ume of the l iquefied l ayer and on the characteristics of overl ying soil s.
Sincel evel -ground l iquefaction of a thin and/or sil ty l ayer at depth may not be
expressed atthe ground surface, the absence of sand boil s does not necessaril y
indicate that l evel groundl iquefaction has not occurred.13. The cycl ic stress
approach to eval uation of l iquefaction potential characterizes bothearthquake
l oading and soil l iquefaction resistance in terms of cycl ic stresses. A transient
earthquake motion is converted to an equival ent series of uniform cycl es ofshear
stress. The number of equival ent cycl es, a function of the duration of themotion,
is correl ated with the magnitude of the earthquake. Liquefaction resistance is
obtained from l aboratory or in situ tests. Cycl ic triaxial and cycl ic simpl e shear
testsare usual l y used in the l aboratory; l iquefaction resistance is expressed in
terms of thenumber of cycl es required to produce fail ure of a soil of given density
subjected to aparticul ar l evel of cycl ic shear stress. The cycl ic stress-based
l iquefaction resistance,however, is infl uenced by factors such as soil fabric,
stress and strain history, and agethat may be destroyed by sampl ing and are
difficul t to repl icate in the l aboratory. Insitu test-based procedures characterize
l iquefaction resistance in terms of in situ testparameters associated with soil s
that have l iquefied in past earthquakes; the SPTresistance is most commonl y used,
but other insitu parameters, incl uding CPT resistanceand shear wave vel ocity, are
gaining acceptance. The cycl ic stress approachal l ows estimation of a factor of
safety against l iquefaction.14. In the cycl ic strain approach, earthquake l oading
and l iquefaction resistance are characterizedby cycl ic strains. Since the factors
that infl uence the cycl ic shear stressesSec. 9.7 Summary 419required to initiate
l iquefaction have a simil ar effect on the shear modul us, the cycl icstrain (the
ratio of cycl ic shear stress to shear modul us) is l ess sensitive to them. Theshear
modul us is an important l iquefaction resistance parameter in the cycl ic strain
approach. Liquefaction is expected at l ocations where the cycl ic strain ampl itude
induced for a particul ar number of cycl es by an earthquake is greater than the
cycl icstrain ampl itude required to initiate l iquefaction in the same number of
cycl es. Thecycl ic strain approach does not produce a factor of safety against
l iquefaction.15. Other approaches to the eval uation of l iquefaction potential have
been devel oped.Dissipated energy has been used as a measure of l iquefaction
resistance; its comparisonwith the energy content of a ground motion al l ows
l iquefaction potential to beeval uated. Effective stress ground response anal yses,
with cycl ic stress-strain andpore pressure model s or advanced constitutive model s,
can be used to predict the generationof excess pore pressure and its redistribution
both during and after earthquakeshaking. Probabil istic approaches, based both on
l aboratory tests resul ts and on fiel dperformance observations, al l ow estimation of
the l ikel ihood of l iquefaction.16. The effects of l iquefaction are different for
different l iquefaction phenomena.Al though fl ow l iquefaction is capabl e of producing
the most spectacul ar effects,cycl ic mobil ity and l evel -ground l iquefaction can al so
produce extensive damage.17. Liquefaction can dramatical l y al ter the ampl itude and
frequency content of groundsurface motions. As the buil dup of excess pore pressure
causes a l ayer of l iquefiabl esoil to soften, ground surface displ acements may
increase even when ground surfaceaccel erations decrease. Ground oscil l ation may
produce chaotic permanent movementof fractured bl ocks of surficial soil .18. Ground
surface settl ement can devel op during and/or after earthquakes due to densification
of dry or saturated sands. Settl ement of dry sand occurs al most immediatel y,but
settl ement of saturated sands may not devel op until wel l after earthquake shaking
has ended. The magnitude of postearthquake settl ement depends on the density of the
sand, and on the ampl itude and duration of shaking.19. When earthquake-induced
shear stresses exceed the shear strength of a l iquefied soil ,instabil ity fail ures
can occur. The shear strength ofl iquefied soil may be eval uated bycareful
undisturbed sampl ing and l aboratory testing or by comparison with in situ test
parameters and back-cal cul ated strengths from l iquefaction case histories. Al l
avail abl eapproaches produce strength estimates with considerabl e uncertainty.20.
Liquefaction fl ow fail ures occur when static shear stresses exceed the shear
strengthof a l iquefied soil . This situation can arise during and/or after an
earthquake. Theeffects of soil l oosening due to porewater fl ow on shear strength
must be recognizedand accounted for in an eval uation of possibl e fl ow fail ure.21.
Deformation fail ures, such as l ateral spreading, devel op incremental l y during the
period of earthquake shaking. For strong l evel s and/or l ong durations of shaking,
deformation fail ures can produce l arge displ acements and cause significant damage.
Procedures have been devel oped to estimate displ acements caused by deformation
fail ures.420HOMEWORK PROBLEMSLiquefaction Chap. 90.8Ql ,g 0.7e: !"tl '0> 0.60.59.1 A
rounded sand subjected to a series of monotonic triaxial compression tests exhibits
thesteady-state l ine (SSL) shown bel ow. Assuming that the sand can mobil ize a
friction angl e of33° in the steady state, estimate the steady-state strength that
woul d be mobil ized by the fol l owingtest specimens: Specimen Initial Void Ratio
Initial 0"3c Test ConditionA 0.75 100kPa DrainedB 0.75 l OOkPa UndrainedC 0.60 50kPa
DrainedD 0.60 50kPa Undrained10 100 1000Effective confining pressure - kPa Figure
P9.19.2 Consider an anisotropical l y consol idated direct simpl e shear test specimen
with the initial conditionsshown bel ow. Show graphical l y how the pore pressure
ratio at the initiation of l iquefaction,ru, t , varies with the ampl itude of the
cycl ic shear stress.q- kpa755025o50 100Figure P9.2• , Initial conditions150 p'-kPa
Chap. 9 Homework Probl ems 421~).3 A 10 m thick deposit ofl oose sand is saturated
bel ow a depth of 3 m, The soil bel ow the watertabl e is highl y susceptibl e to
l iquefaction. Estimate the ground surface accel eration that woul dbe required to
produce sand boil s in aM = 7.7 earthquake.~}.4 Estimate the variation of uniform
cycl ic shear stress ampl itude with depth for the upper 50 ftofsoil (assumey=
l 20pcf) at the Gil roy No.2 (soil ) station in the Loma Prieta earthquake (seeFigure
3.1). Estimate the number of equival ent uniform stress cycl es.l }.5 A l evel deposit
of saturated cl ean sand has an average (Nl )60-val ue of i 8 and an average dry unit
weight of 105 pcf. Pl ot the variation of cycl ic shear stress required to produce
l iquefaction inM = 5.5, 6.5, 7.5, and 8.5 earthquakes.'~.6 Repeat Probl em 9.5
assuming that the sand has 15% fines.'~.7 The surface of a thick deposit of sil ty
sand (30% fines) sl opes at an angl e of 10°. Standard penetrationtests indicate an
average (Nl )60-val ue of 12; density tests indicate an average density of1.80 Mg/m'.
Compute and pl ot the cycl ic shear stress required to produce l iquefaction in the
upper 30 ft of the sil ty sand in aM = 7.5 earthquake.'~.8 A l evel deposit of cl ean
sand with an average void ratio of 0.70 has an uncorrected CPT tipresistance of 80
tsf at a depth of 20 ft. The mean grain size of the sand is 0.40 mm and the
groundwater tabl e is at a depth of 4 ft. Compute the cycl ic shear stress required
to produce l iquefactionat a depth of 20 ft in aM = 6.5 earthquake.9.9 Prior to the
1989 Loma Prieta earthquake, a site investigation at the Hunters Point Navy Basein
San Francisco showed 13 to 15 m of cl ean, hydraul ical l y fil l ed sand with Dso= 0.29
mm. Thewater tabl e was at a depth of about 2.5 m. CPT tests produced the fol l owing
average tipresistances: Depth Interval (m)3-44--55--{i6-77-88-99-1010-11Average qc
(MPa)6.55.75.24.95.16.36.97.6The Loma Prieta earthquake produced peak ground
accel erations on the order of0.15g to 0.20gat simil ar sites in the Bay Area.
Assuming that the sand has an average dry density of1.60 Mg/rrr', compute and pl ot
the factor of safety against l iquefaction that woul d have beenexpected in the Loma
Prieta earthquake.9.10 Pl ot the estimated porewater pressure distributionat the end
of the Loma Prieta earthquake atthe Hunters Point site described in Probl em 9.9.
9.11 Repeat Probl em 9.9 for aM = 7.6 earthquake that produces a peak ground
accel eration of 0.30g.9.12 Using the energy-based criterion of Law et al . (1990)
[(equation (9.16)], pl ot the l imiting hypocentral distance beyond which l iquefaction
woul d not be expected in N1 = 10 soil s as a functionof earthquake magnitude.
Compare the resul t with Figure 9.4 and comment on the differences.422 Liquefaction
Chap. 99.13 Estimate the vertical settl ement that woul d have been expected at
Hunters Point (see Probl em9.9) fol l owing the 1989 Loma Prieta earthquake.9.14
Estimate and pl ot the variation of the post-l iquefaction residual strength of the
sand at the sitedescribed in Exampl e 9.4 using (a) the Seed-Harder approach, and
(b) the Stark-Mesriapproach.10Seismic Sl ope Stabil ity10.1 INTRODUCTIONLandsl ides
occur on a regul ar basis throughout the worl d as part of the ongoing evol ution of
l andscapes. Many l andsl ides occur in natural sl opes, but sl ides al so occur in man-
madesl opes from time to time. At any point in time, then, sl opes exist in states
ranging from verystabl e to marginal l y stabl e. When
an earthquake occurs, the effects of earthquake-inducedground shaking is often
sufficient to cause fail ure of sl opes that were marginal l y to moderal tel ystabl e
before the earthquake. The resul ting damage can range from insignificant to
catastrophicdepending on the geometric and material characteristics of the sl ope.
Earthquake-induced l andsl ides, which have been documented from as earl y as 1789B.C.
(Li, 1990), have caused tremendous amounts of damage throughout history. In many
earthquakes, l andsl ides have been responsibl e for as much or more damage than al l
other seismichazards combined. In the 1964 Al aska earthquake, for exampl e, an
estimated 56% of thetotal cost of damage was caused by earthquake-induced
l andsl ides (Youd, 1978; Wil son andKeefer, 1985). Kobayashi (1981) found that more
than hal f of al l deaths in l arge (M> 6.9)earthquakes in Japan between 1964 and 1980
were caused by l andsl ides. The 1920 Haiyuanearthquake (M= 8.5) in the Ningxia
Province of China produced hundreds of l arge l andsl idesthat caused more than
100,000 deaths (Cl ose and McCormick, 1922). Eval uation of seismicsl ope stabil ity is
one of the most important activities of the geotechnical earthquake engineer.423424
Seismic Sl ope Stabil ity Chap. 10This chapter we describes different types of
earthquake-induced l andsl ides and theconditions under which they occur. It al so
reviews the basic principl es of sl ope stabil ityeval uation, incl uding static
stabil ity anal ysis, and then presents several methods for seismicsl ope stabil ity
anal ysis.10.2 TYPES OF EARTHQUAKE-INDUCED LANDSLIDESMany factors, incl uding
geol ogic and hydrol ogic conditions, topography, cl imate, weathering,and l and use,
infl uence the stabil ity of sl opes and the characteristics of l andsl ides. A numberof
procedures for cl assification of l andsl ides have been proposed; that of Vames
(1978)is perhaps most widel y used in the United States. Simil ar principl es and
terminol ogy can beused to cl assify earthquake-induced l andsl ides (Tabl e 10-1) on
the basis of material type (soil Tabl e 10·1 Types and Characteristics of Earthquake-
Induced Landsl idesWater content"Type of Internal Name Movement Disruption" D U PS S
Vel ocity" Depth"Disrupted Sl ides and Fal l sRock fal l s Bounding, rol l - High or very x
x x x Extremel y Shal l owing, free fal l high rapidRock sl ides Transl ational sl id-
High x x x x Rapid to Shal l owing on basal extremel yshear surface rapidRock ava-
Compl ex, invol v- Very high x x x x Extremel y Deepl anches ing sl iding rapidand/or
fl ow, asstream of rockfragmentsSoil fal l s Bounding, rol l - High or very x x x x
Extremel y Shal l owing, free fal l high rapidDisrupted soil Transl ational sl id- High x
x x x Moderate to Shal l owing on basal rapidshear surfaceor zone ofweakened,
sensitivecl aySoil aval anches Transl ational sl id- Very high x x x x Very rapid
Shal l owing with subsid- to exiaryfl ow tremel yrapidCoherent Sl idesRock sl umps
Sl iding on basal Sl ight or ? x x x Sl ow to rapid Deepshear shear sur- moderateface
with componentof headwardrotationSec. 10.2 Types of Earthquake-Induced Landsl ides
425Tabl e 10-1 Types and Characteristics of Earthquake-Induced Landsl ides
(continued)Water content"Type of Internal Name Movement Disruption" D U PS S
Vel ocity" Depth"Rock bl ock Transl ational sl id- Sl ight or ? x x x Sl ow to rapid Deep
sl ides ing on basal moderateshear surfaceSoil sl umps Sl iding on basal Sl ight or ? x
x x Sl ow to rapid Deepshear shear sur- moderateface witb componentof headward
rotationSoil bl ock sl ides Transl ational sl id- Sl ight or ? ? x x Sl ow to rapid Deep
ing on basal moderateshear surfaceSl ow earth Transl ational sl id- Sl ight x x Very
sl ow General l yfl ows ing on basal to mod- shal l ow,shear surface erate with occaswitb
minor very rapid ional l yinternal fl ow surges deepVariabl eShal l owVariabl eVery rapid
to extremel yrapidGeneral l yrapid toextremel yrapid, occasional l ysl ow tomoderatex Very
rapidxx?xx? ?General l yhigh orvery high,occasional l ymoderateorsl ightLateral Spreads
and Fl owsGeneral l ymoderate,occasional l ysl ight,occasional l yhighVery highCompl ex,
general l yinvol vingl ateral spreading,and/orfl ow; occasional l yinvol vingsl umping
and/orbl ock sl idingTransl ation onbasal zone ofl iquefied sand,or sil t or weakened,
sensitivecl ayFl owSubaqueousl andsl idesRapid soil fl owsSoil l ateral spreadsSource:
Keefer (1984)."l ntemal disruption: "sl ight" signifies l andsl ide consists of one or
a few coherent bl ocks; "moderate" signifies several coherent bl ocks; "high"
signifies numerous smal l bl ocks and individual soil grains and rock fragments;
"very high"signifies nearl y compl ete disaggregation into individual soil grains or
smal l rock fragments.bWater content: D, dry; U, moist, but unsaturated; PS, partl y
saturated; S, saturated.'Vel ocity: 0.6 m/yr 1.5 m/yr 1.5 m/month 1.5 m/day 0.3 m/min
3 m/secextremel y sl ow very sl ow sl ow moderate rapiddDepth: "shal l ow" signifies
thickness general l y < 3 m; "deep" general l y> 3 m.very rapid extremel y rapid426
Seismic Sl ope Stabil ity Chap. 10or rock), character of movement (disrupted or
coherent), and other attributes, such as vel ocity,depth, and water content.
Earthquake-induced l andsl ides can be divided into three maincategories: disrupted
sl ides and fal l s, coherent sl ides, and l ateral spreads and fl ows.Disrupted sl ides
and fal l s incl ude rock fal l s, rock sl ides, rock aval anches, soil fal l s,disrupted
soil sl ides, and soil aval anches. The earth material s invol ved in such fail ures are
sheared, broken, and disturbed into a nearl y random order. These types of fail ures,
usual l yfound in steep terrain, can produce extremel y rapid movements and
devastating damage;rock aval anches and rock fal l s have historical l y been among the
l eading causes of deathfrom earthquake-induced l andsl ides.Coherent sl ides, such as
rock and soil sl umps, rock and soil bl ock sl ides, and sl owearth fl ows, general l y
consist of a few coherent bl ocks that transl ate or rotate on somewhatdeeper fail ure
surfaces in moderate to steepl y sl oping terrain. Most coherent sl ides occur atl ower
vel ocities than disrupted sl ides and fal l s.Lateral spreads and fl ows general l y
invol ve l iquefiabl e soil s, al though sensitive cl ayscan produce l andsl ides with very
simil ar characteristics. Due to the l ow residual strength ofthese material s,
sl iding can occur on remarkabl y fl at sl opes and produce very high vel ocities.
Liquefaction-induced spreads and fl ow sl ides were discussed in detail in Chapter 9.
The different types of earthquake-induced l andsl ides occur with different
frequencies.Rock fal l s, disrupted soil sl ides, and rock sl ides appear to be the
most common typesofl andsl ides observed in historical earthquakes (Tabl e 10-2).
Subaqueous l andsl ides, sl owearth fl ows, rock bl ock sl ides, and rock aval anches are
l east common, al though the difficul tyof observing subaqueous sl ides may contribute
to their apparent rarity.Tabl e 10-2 Rel ative Abundance of Earthquake-Induced
Landsl ides from Study of 40Historical Earthquakes Ranging from Ms = 5.2 to Mw = 9.5
AbundanceVery abundant(> 100,000 in the 40 earthquakes)Abundant(10,000to 100,000 in
the 40 earthquakes)Moderatel y common(1000to 10,000in the 40 earthquakes)Uncommon
Source: Keefer (1984).DescriptionRock fal l s, disrupted soil sl ides, rock sl idesSoil
l ateral spreads, soil sl umps, soil bl ock sl ides,soil aval anchesSoil fal l s, rapid
soil fl ows, rock sl umpsSubaqueous l andsl ides, sl ow earth fl ows, rock bl ocksl ides,
rock aval anches10.3 EARTHQUAI{E-INDUCED LANDSLIDE ACTIVITYFor prel iminary stabil ity
eval uations, knowl edge of the conditions under which earthquakeinducedl andsl ides
have occurred in past earthquakes is useful . It is l ogical to expect that theextent
of earthquake-induced l andsl ide activity shoul d increase with increasing earthquake
magnitude and that there coul d be a minimum magnitude bel ow which earthquake-
inducedSec. 10.3 Earthquake-Induced Landsl ide Activity 427l andsl iding woul d rarel y
occur. It is equal l y l ogical to expect that the extent of earthquakeinduced
l andsl ide activity shoul d decrease with increasing source-to-site distance and that
there coul d be a distance beyond which l andsl ides woul d not be expected in
earthquakes ofa given size.A study of 300 U.S. earthquakes between 1958 and 1977
showed that the smal l estearthquakes noted to have produced l andsl ides had l ocal
magnitudes of about 4.0 (Keefer,1984). Minimum magnitudes for different types of
l andsl ides were estimated as shown inTabl e 10-3. Where magnitudes were not
avail abl e, minimum Modified Mercal l i Intensity(MMI) val ues of IV and V have been
observed for disrupted sl ides or fal l s and other typesof sl ides, respectivel y.
Al though these empirical l y based l imits are useful , their approximatenature must be
recognized; fail ure of sl opes that are near the brink of fail ure understatic
conditions coul d be produced by quite weak earthquake shaking.Tabl e 10-3 Estimates
of the Smal l est Earthquakes Likel y to CauseLandsl idesML Description4.0 Rock fal l s,
rock sl ides, soil fal l s, disrupted soil sl ides4.5 Soil sl umps, soil bl ock sl ides5.0
Rock sl umps, rock bl ock sl ides, sl ow earth fl ows, soil l ateral spreads,rapid soil
fl ows, and subaqueous l andsl ides6.0 Rock aval anches6.5 Soil aval anchesSource: After
Keefer (1984).The maximum source-to-site distance at which l andsl ides have been
produced in historical earthquakes are different for different types of l andsl ides
(Figure 10.1). Disruptedsl ides or fal l s, for exampl e, have rarel y been found beyond
epicentral distances of about 15km for M =5 events but have been observed as far as
about 200 km (124 mi.) inM =7 earthquakes.Note that the curve for l ateral spreads
and fl ows correl ates reasonabl y wel l with themagnitude-distance curve for
l iquefaction shown in Figure 9.5. Simil arl y, the area overwhich earthquake-induced
l andsl iding
can be expected al so increases with increasing earthquakemagnitude (Figure 10.2).
Regional differences in attenuation behavior have l ittl eapparent infl uence on the
area of earthquake-induced l andsl iding.Exampl e 10.1Estimate the maximum distances
at which rock aval anches, soil sl umps, and soil l ateral spreadswoul d be expected in
aM =6.5 earthquake.Sol ution Rock aval anches, soil sl umps, and soil l ateral spreads
fal l under the headings of disruptedfal l s and sl ides, coherent sl ides, and l ateral
spreads and fl ows, respectivel y. From Figure10.1, the maximum distances of these
types of sl ides from the faul t rupture zone woul d beRock aval ancheSoil sl umpsSoil
l ateral spreads61km22km20km428 Seismic Sl ope Stabil ity Chap. 1010100OJoc .sen ~'5 E
- -" <l l ~~c en OJOJ""C.>2 ~Q.""COJ CE.!!1: : J- E 0.~~..#--.0.: /;'.: /'" ,.: /"" /,' "
' " h I ,I !JI • I" • !JI /I • .,'" .I!,: . I0.1 t,~'------+~~~~------'-~~~~------'-
~~~~---'4.0 4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5Magnitude, MFigure 10.1
Maximum epicentral distance for different types of l andsl ides. Dashed l ine is for
disrupted fal l s and sl ides, dash-dot l ine is for coherent sl ides, and sol id l ine is
for l ateral spreadsand fl ows. (After Keefer, 1984. Used by permission of the
Geol ogical Society of America.)4.5 5.0 5.5 6.0 6.5 7.0 7.5 8.0 8.5 9.0 9.5
Magnitude, M• • • •
500,000 ,------,------,----------,-----,-----------,-----,-----------,-------,-----
-----,------c7--------,-------,200,000100,00050,00020,00010,0005,0002,0001,000500
2001005020105214.0Figure 10.2 Area affected by l andsl ides for earthquakes of
different magnitude. (AfterKeefer, 1984. Used by permission of the Geol ogical
Society of America.)Sec. 10.4 Eval uation Of Sl ope Stabil ity 42910.4 EVALUATION OF
SLOPE STABILITYThe stabil ity of sl opes is infl uenced by many factors, and a
compl ete sl ope stabil ity eval uationmust consider the effects of each. Geol ogical ,
hydrol ogical , topographical , geometrical ,and material characteristics al l infl uence
the stabil ity of a particul ar sl ope. Informationon these characteristics is needed
to rel iabl y perform and interpret the resul ts of both staticand seismic sl ope
stabil ity anal yses. Review of avail abl e documents, fiel d reconnaissance,fiel d
monitoring, subsurface investigation, and material testing can al l be used to
obtain thisinformation.For many sites, considerabl e useful information can be
obtained from previousl y publ isheddocuments such as geol ogic maps, soil survey
and/or agricul tural maps, topographicmaps, natural hazard maps, and geol ogic and
geotechnical engineering reports. Additional information may be obtained from aerial
photographs (particul arl y stereo-paired aerial photographs)and other forms of
remote sensing.Fiel d reconnaissance invol ves careful observation and detail ed
mapping of a varietyof site characteristics associated with existing or potential
sl ope instabil ity. Features such asscarps; tension cracks; bul ges; hummocky
terrain; displ aced ditches, channel s, and fences;cracked foundations, wal l s, or
pavements; and l eaning trees or pol es can be identified andmapped as evidence of
instabil ity. The l ocations of streams, springs, seeps, ponds, andmoist areas, and
differences in vegetative cover, can provide evidence of al tered or disruptedwater
fl ow caused by sl ope instabil ity.If time permits, sl ope movement can be monitored.
Surface monuments can beinstal l ed at points on and near the sl ope and surveyed
periodical l y to identify the magnitudeand direction of surface movement.
Photogrammetric methods can be used to determine rel ativemovements from sets of
stereo-paired aerial photographs taken at different times.Incl inometers are very
useful for monitoring l ateral deformation patterns bel ow the groundsurface. In many
cases, crack gauges, til tmeters, and extensometers can al so be used toobserve the
effects of sl ope movement. When, as is commonl y the case, pore pressures are
important, piezometers and/or observation wel l s can provide important information
on porepressures and their variation with time.Subsurface investigation can incl ude
excavation and mapping of test pits andtrenches, boring and sampl ing, in situ
testing, and geophysical testing. Such investigationscan reveal the depth,
thickness, density, strength, and deformation characteristics of subsurfaceunits,
and the depth and variation of the groundwater tabl e. In situ and geophysical tests
are particul arl y useful for determining the l ocation of an existing fail ure
surface.Laboratory tests are often used to quantify the physical characteristics of
the varioussubsurface material s for input into a numerical sl ope stabil ity
anal ysis. Soil density,strength, and stress-strain behavior are of prime
importance; other characteristics, such asgrain size distribution, pl asticity,
permeabil ity, and compressibil ity, are al so useful .Onl y after this information is
obtained can a stabil ity anal ysis be performed. Al thoughin the remainder of this
chapter we focus on methods of sl ope stabil ity anal ysis, it is importantto remember
that the anal ysis itsel f is but a singl e part of a compl ete sl ope stabil ity
eval uationand that its accuracy wil l be reduced if careful attention is not given
to the otheraspects of the eval uation.43010.5 STATIC SLOPE STABILITY ANALYSIS
Seismic Sl ope Stabil ity Chap. 10Sl opes become unstabl e when the shear stresses
required to maintain equil ibrium reach orexceed the avail abl e shearing resistance
on some potential fail ure surface. For sl opes inwhich the shear stresses required
to maintain equil ibrium under static gravitational l oadingare high, the additional
dynamic stresses needed to produce instabil ity may be l ow. Hencethe seismic
stabil ity of a sl ope is strongl y infl uenced by its static stabil ity. Because of
thisand the fact that the most commonl y used methods of seismic stabil ity anal ysis
rel y on staticstabil ity anal yses, a brief summary of static sl ope stabil ity
anal ysis is presented.The procedures for anal ysis of sl ope stabil ity under static
conditions are wel l establ ished.An excel l ent, concise review of the state of the
art for static anal ysis ~as presentedby Duncan (1992). Detail ed descriptions of
specific methods of anal ysis can be found instandard references such as National
Research Council (1976), Chowdhury (1978), andHuang (1983). Currentl y, the most
commonl y used methods of static sl ope stabil ity anal ysisare l imit equil ibrium
anal yses and stress-deformation anal yses.10.5.1 Limit Equil ibrium Anal ysisLimit
equil ibrium anal yses consider force and/or moment equil ibrium of a mass of soil
above a potential fail ure surface. The soil above the potential fail ure surface is
assumed tobe rigid (i.e., shearing can occur onl y on the potential fail ure
surface). The avail abl e shearstrength is assumed to be mobil ized at the same rate
at al l points on the potential fail ure surface.As a resul t, the factor of safety is
constant over the entire fail ure surface. Because thesoil on the potential fail ure
surface is assumed to be rigid-perfectl y pl astic (Figure 10.3),l imit equil ibrium
anal yses provide no information on sl ope deformations.Sl ope stabil ity is usual l y
expressed in terms of an index, most commonl y the factor ofsafety, which is usual l y
defined asFS = avail abl e shear strength (10.1)shear stress required to maintain
equil ibriumThus the factor of safety is a ratio of capacity (the shear strength of
the soil ) to demand (theshear stress induced on the potential fail ure surface). The
factor of safety can al so be viewedas the factor by which the strength of the soil
woul d have to be divided to bring the sl ope tothe brink of instabil ity. In contrast
to the assumptions of l imit equil ibrium anal ysis, thestrength of the soil in actual
sl opes is not reached at the same time at al l points on the fail uresurface (i.e.,
the l ocal factor of safety is not constant).Shear strain, yFigure 10.3 Stress-
strain curve for arigid-perfectl y pl astic material . No shearstrain occurs until the
strength of thematerial is reached, after which the material strains at constant
shear stress.Sec. J0.5 Static Sl ope Stabil ity Anal ysis 431A variety of l imit
equil ibrium procedures have been devel oped to anal yze the staticstabil ity of
sl opes. Sl opes that fail by transl ation on a pl anar fail ure surface (Figure 10Aa)
such as a bedding pl ane, rock joint, or seam of weak material can be anal yzed quite
easil yby the Cul mann method (Tayl or, 1948). Sl opes in which fail ure is l ikel y to
occur on two orthree pl anes (Figure l OAb) can be anal yzed by wedge methods (e.g.,
Perl off and Baron,1976; Lambe and Whitman, 1969). Inhomogeneous sl opes, the
critical fail ure surface usual l yhas a circul ar (Figure WAc) or l og-spiral shape.
Since the minimum factors of safetyfor circul ar and l og-spiral fail ure surfaces are
very cl ose, homogeneous sl opes are usual l yanal yzed by methods such as the ordinary
method of sl ices (Fel l enius, 1927) or Bishop'smodified method (Bishop, 1955), which
assume circul ar fail ure surfaces. When subsurfaceconditions are not homogeneous
(e.g., when l ayers with significantl y different strength,highl y anisotropic
strength, or discontinuities exist), fail ure surfaces are l ikel y to be noncircul ar
(Figure l OAd). In such cases, methods l ike those of Morgenstern and Price (1965),
Spencer (1967), and Janbu (1968) may be used. Nearl y al l l imit equil ibrium methods
aresusceptibl e to numerical probl ems under certain conditions. These conditions
vary for differentmethods but are most commonl y encountered where soil s with high
cohesive strengthare present at the top of a sl ope and/or when fail ure surfaces
emerge steepl y at the base ofsl opes in soil s with high frictional strength (Duncan,
1992).In concept, any sl ope with a factor of safety above 1.0 shoul d be stabl e. In
practice,however, the l evel of
stabil ity is sel dom considered acceptabl e unl ess the factor of safety is
significantl y greater than 1.0. Criteria for acceptabl e factors of safety recognize
(1) uncertaintyin the accuracy with which the sl ope stabil ity anal ysis represents
the actual mechanismof fail ure, (2) uncertainty in the accuracy with which the
input parameters (shearstrength, groundwater conditions, sl ope geometry, etc.) are
known, (3) the l ikel ihood andduration of exposure to various types of external
l oading, and (4) the potential consequencesof sl ope fail ure. Typical minimum
factors of safety used in sl ope design are about1.5 for normal l ong-term l oading
conditions and about 1.3 for temporary sl opes or end-ofconstructionconditions in
permanent sl opes (when dissipation of pore pressure increasesstabil ity with time).
When the minimum factor of safety of a sl ope reaches a val ue of 1.0, the avail abl e
shear strength of the soil is ful l y mobil ized on some potential fail ure surface and
the sl ope(a)--~Figure l OA Common fail ure surface geometries: (a) pl anar; (b)
mul tipl anar; (c)circul ar; (d) noncircul ar.432 Seismic Sl ope Stabil ity Chap. J0is at
the point of incipient fail ure. Any additional l oading wil l cause the sl ope to fail
(i.e., todeform until it reaches a configuration in which the shear stresses
required for equil ibriumare l ess than or equal to the avail abl e shear strength of
the soil ). The l imit equil ibriumassumption of rigid-perfectl y pl astic behavior
suggests that the required deformation wil l occur in a ductil e manner. Many soil s,
however, exhibit brittl e, strain-softening stressstrainbehavior. In such cases the
peak shear strength may not be mobil ized simul taneousl yat al l points on the fail ure
surface. When the peak strength of a strain-softening soil isreached, such as point
A in Figure 1O.5a, the avail abl e shearing resistance wil l drop from thepeak to the
residual strength. As it does so, shear stresses rel ated to the difference between
the peak and residual strength of the soil at point A are transferred to the
surrounding soil .These redistributed shear stresses may cause the peak strengths in
the surrounding soil to bereached (Figure 1O.5b) and exceeded, thereby reducing
their avail abl e shearing resistancesto residual val ues. As the stress
redistribution process continues, the zone of fail ure maygrow until the entire
sl ope becomes unstabl e. Many instances of such progressive fail urehave been
observed in strain-softening soil s, even when the l imit equil ibrium factor ofsafety
(based on peak strength) is wel l above 1.0. Within the constraints of l imit
equil ibriumanal ysis, the stabil ity of sl opes with strain-softening material s can be
anal yzed rel iabl y onl yby using residual shear strengths.Limit equil ibrium anal yses
must be formul ated with great care. Since the avail abl eshearing resistance of the
soil depends on porewater drainage conditions, those conditionsmust be considered
careful l y in the sel ection of shear strengths and pore pressure conditionsfor the
anal ysis. Duncan (1992) provided guidel ines for the sel ection of input parameters
forl imit equil ibrium sl ope stabil ity anal yses.1: Peak strength mobil izedat point A
(a)y(b)1: BShearing resistance atpoint A drops from peakA to residual strengthy~~
Fail ure at point A causes shearstresses to be redistrributed to points BFigure 10.5
Devel opment of progressive fail ure in sl ope comprised of strain-softeningmaterial s:
(a) exceedance of peak strength at any point (A) reduces strength at that pointto
residual val ue; (b) redistribution of shear stresses from fail ure zone to
surroundingarea produces fail ure in surrounding zone (points B). Continued
redistribution of stressescan eventual l y l ead to fail ure of the entire sl ope
(points C and beyond).Sec. 10.6 Seismic Sl ope Stabil ity Anal ysis 43310.5.2 Stress-
Deformation Anal ysesStress-deformation anal yses al l ow consideration of the stress-
strain behavior of soil androck and are most commonl y performed using the finite-
el ement method. When appl ied tosl opes, stress-deformation anal yses can predict the
magnitudes and patterns of stresses,movements, and pore pressures in sl opes during
and after construction/deposition. Nonl inearstress-strain behavior, compl ex
boundary conditions, irregul ar geometries, and avariety of construction operations
can al l be considered in modern finite-el ement anal yses.For static sl ope stabil ity
anal ysis, stress-deformation anal yses offer the advantages ofbeing abl e to identify
the most l ikel y mode of fail ure by predicting sl ope deformations up to(and in some
cases beyond) the point of fail ure, of l ocating the most critical l y stressed zones
within a sl ope, and of predicting the effects of sl ope fail ures. These advantages
come at thecost of increased engineering time for probl em formul ation,
characterization of material properties and interpretation of resul ts, and increased
computational effort.The accuracy of stress-deformation anal yses is strongl y
infl uenced by the accuracywith which the stress-strain model represents actual
material behavior. Many differentstress-strain model s have been used for stress-
deformation anal ysis of sl opes; each hasadvantages and l imitations. The accuracy of
simpl e model s is usual l y l imited to certainranges of strain and/or certain stress
paths. Model s that can be appl ied to more general stressand strain conditions are
often quite compl ex and may require a l arge number of inputparameters whose val ues
can be difficul t to determine. For many probl ems, the hyperbol icmodel (Kondner,
1963; Kondner and Zel asko, 1963; Duncan and Chang, 1970; Duncan etal ., 1980) offers
an appropriate compromise between simpl icity and accuracy.10.6 SE: l SMIC SLOPE
STABILITY ANALYSISThe previousl y described procedures for static sl ope stabil ity
anal ysis have been used formany years and cal ibrated against many actual sl ope
fail ures. The database against whichseismic sl ope stabil ity anal yses can be
cal ibrated is much smal l er. Anal ysis of the seismicstabil ity of sl opes is further
compl icated by the need to consider the effects of (1) dynamicstresses induced by
earthquake shaking, and (2) the effects of those stresses on the strengthand
stress-strain behavior of the sl ope material s.Seismic sl ope instabil ities may be
grouped into two categories on the basis of whichof these effects is predominant in
a given sl ope. In inertial instabil ities, the shear strength ofthe soil remains
rel ativel y constant, but sl ope deformations are produced by temporaryexceedances of
the strength by dynamic earthquake stresses. Weakening instabil ities arethose in
which the earthquake serves to weaken the soil sufficientl y that it cannot remain
stabl e under earthquake-induced stresses. Fl ow l iquefaction and cycl ic mobil ity
(Chapter 9)are the most common causes of weakening instabil ity. A number of
anal ytical techniques,based on both l imit equil ibrium and stress-deformation
anal yses, are avail abl e for both categoriesof seismic instabil ity.l '0.6.1 Anal ysis
of Inertial Instabil ityEarthquake motions can induce significant horizontal and
vertical dynamic stresses insl opes. These stresses produce dynamic normal and shear
stresses al ong potential fail uresurfaces within a sl ope. When superimposed upon the
previousl y existing static shear434 Seismic Sl ope Stabil ity Chap. 10stresses, the
dynamic shear stresses may exceed the avail abl e shear strength of the soil and
produce inertial instabil ity of the sl ope. A number of techniques for the anal ysis
of inertial instabil ity have been proposed. These techniques differ primaril y in the
accuracy withwhich the earthquake motion and the dynamic response of the sl ope are
represented. Thefol l owing sections describe several common approaches to the
anal ysis of inertial instabil ity.The first, pseudostatic anal ysis, produces a
factor of safety against seismic sl ope fail urein much the same way that static
l imit equil ibrium anal yses produce factors of safety againststatic sl ope fail ure.
Al l the other approaches attempt to eval uate permanent sl ope displ acementsproduced
by earthquake shaking.10.6.1.1 Pseudostatic Anal ysisBeginning in the 1920s, the
seismic stabil ity of earth structures has been anal yzed bya pseudostatic approach
in which the effects of an earthquake are represented by constanthorizontal and/or
vertical accel erations. The first expl icit appl ication of the pseudostaticapproach
to the anal ysis of seismic sl ope stabil ity has been attributed to Terzaghi (1950).
In their most common form, pseudostatic anal yses represent the effects of
earthquakeshaking by pseudostatic accel erations that produce inertial forces, Fh
and F; which act throughthe centroid of the fail ure mass (Figure 10.6). The
magnitudes of the pseudostatic forces arer, ahW = khW (l 0.2a)gF - avWv - kvW
(l 0.2b) gwhere ahand av are horizontal and vertical pseudostatic accel erations, kh
and k; are dimensionl esshorizontal and vertical pseudostatic coefficients, and Wis
the weight of the fail uremass. The magnitudes of the pseudo static accel erations
shoul d be rel ated to the severity ofthe anticipated ground motion; sel ection of
pseudostatic accel erations for design is, as discussedin the next section, not a
simpl e matter. Resol ving the forces on the potential fail uremass in a direction
paral l el to the fail ure surface,FS = resisting force = dab + [(W - Fv) cos ~ - Fh
sin~] tan<l > (l 0.3)driving force (W - Fv) sinf + Fh cosjiwhere c and <I> are the
Mohr-Coul omb strength parameters that describe the shear strengthon the fail ure
pl ane and l ab is the l ength of the fail ure pl ane. The horizontal pseudostaticab
Figure 10.6 Forces acting on triangul ar wedge of soil above pl anar fail ure surface
inpseudostatic sl ope stabil ity anal ysis.Sec. 10.6 Seismic Sl ope Stabil ity Anal ysis
435force cl earl y decreases the factor of safety-it reduces the resisting
force (for <p > 0) andincreases the driving force. The vertical pseudostatic force
typical l y has l ess infl uence onthe factor of safety since it reduces (or increases,
depending on its direction) both the drivingforce and the resisting force-as a
resul t, the effects of vertical accel erations are frequentl ynegl ected in
pseudostatic anal yses. The pseudostatic approach can be used toeval uate
pseudostatic factors of safety for pl anar, circul ar, and noncircul ar fail ure
surfaces.Many commercial l y avail abl e computer programs for l imit equil ibrium sl ope
stabil ity anal ysishave the option of performing pseudostatic anal yses.Exampl e 10.2
Assuming kh =0.1 and k; =0.0, compute the static and pseudo static factors of
safety for the30-ft-high 2: 1 (H: V) sl ope shown in Figure El O.2.c =600 Ibl ft2<I> =0;
y=110 Ibl ft3--d'''----': : ..----O-------+---c=1000 Ib/ft2<1>=0y = 125 Ib/ft3Figure
EIO.2Sol ution Using a simpl e moment equil ibrium anal ysis, the factor of safety can
be defined asthe ratio ofthe moment that resists rotation of a potential fail ure
mass about the center of a circul arpotential fail ure surface to the moment that is
driving the rotation. The critical fail ure surface,defined as that which has the
l owest factor of safety, is identified by anal yzing a numberof potential fail ure
surfaces. Shown bel ow are the factor-of-safety cal cul ations for one potential
fail ure surface which mayor may not be the critical fail ure surface.Computation of
the factor of safety requires eval uation of the overturning and resistingmoments
for both static and pseudostatic conditions. The overturning moment for static
conditionsresul ts from the weight of the soil above the potential fail ure surface.
The overturningmoment for pseudostatic conditions is equal to the sum of the
overturning moment for staticconditions and the overturning moment produced by the
pseudostatic forces. The horizontal pseudostatic forces are assumed to act in
directions that produce positive (cl ockwise, in thiscase) driving moments. In the
cal cul ations shown in tabul ar form bel ow, the soil above thepotential fail ure mass
is divided into two sections.Overturning moments: Static Pseudostatic Total Area
Moment Moment khW Moment Moment MomentSection {ft2) t (l b/ft3) W(kips/ft) Arm (ft)
(kip-ft/ft) (kips/ft) Arm (ft) (kip-ft/ft) (kip-ft/ft)A 11360 110 149.6 30 4488.1
15.0 38 570.0 5058.0B 2300 125 287.5 5 1437.5 28.8 62 1785.6 3223.15925.5 8281.1436
Resisting moment: Seismic Sl ope Stabil ity Chap. J0SectionABLength (ft)11.5129.3600
1000Force (kips)6.9129.3Moment Arm (ft)7878Moment (kip-ftl ft)538.210,085.410,623.6
Factor of safety: Static FSPseudostatic FSresisting moment 10,623.6static
overturning moment = 5925.5resisting momentstatic + pseudostatic overturning
moments1.79= 10,623.6 = 1 288281.1 .Sel ection of Pseudostatic Coefficient. The
resul ts of pseudostatic anal ysesare critical l y dependent on the val ue of the
seismic coefficient, kh. Sel ection of an appropriatepseudostatic coefficient is the
most important, and most difficul t, aspect of apseudostatic stabil ity anal ysis. The
seismic coefficient control s the pseudostatic force on thefail ure mass, so its
val ue shoul d be rel ated to some measure of the ampl itude of the inertial force
induced in the potential l y unstabl e material . If the sl ope material was rigid, the
inertial force induced on a potential sl ide woul d be equal to the product of the
actual horizontal accel erationand the mass of the unstabl e material . This inertial
force woul d reach its maximumval ue when the horizontal accel eration reached its
maximum val ue. In recognition of the factthat actual sl opes are not rigid and that
the peak accel eration exists for onl y a very short time,the pseudostatic
coefficients used in practice general l y correspond to accel eration val ues wel l bel ow
amax. Terzaghi (1950) original l y suggested the use of kh = 0.1 for "severe"
earthquakes(Rossi-Forel IX), kh =0.2 for "viol ent, destructive" earthquakes (Rossi-
Forel X), and kh =0.5for "catastrophic" earthquakes. Seed (1979) l isted
pseudostatic design criteria for 14 dams in10 seismical l y active countries; 12
required minimum factors of safety of 1.0 to 1.5 with pseudostaticcoefficients of
0.10 to 0.12. Marcuson (1981) suggested that appropriate pseudostaticcoefficients
for dams shoul d correspond to one-third to one-hal f of the maximum accel eration,
incl uding ampl ification or deampl ification effects, to which the dam is subjected.
Using shearbeam model s, Seed and Martin (1966) and Dakoul as and Gazetas (1986)
showed that the inertial force on a potential l y unstabl e sl ope in an earth dam
depends on the response of the damand that the average seismic coefficient for a
deep fail ure surface is substantial l y smal l er thanthat of a fail ure surface that
does not extend far bel ow the crest. Seed (1979) al so indicated thatdeformations of
earth dams constructed of ductil e soil s (defined as those that do not generatehigh
pore pressures or show more than 15% strength l oss upon cycl ic l oading) with crest
accel erationsl ess than 0.75g woul d be acceptabl y smal l for pseudostatic factors of
safety of at l eastU5 with kh=0.10 (M=6.5) to kh =0.15 (M=8.25). This criteria woul d
al l ow the use of pseudostaticaccel erations as smal l as 13 to 20% of the peak crest
accel eration. Hynes-Griffin andFrankl in (1984) appl ied the Newmark sl iding bl ock
anal ysis described in the fol l owing sectionto over 350 accel erograms and concl uded
that earth dams with pseudostatic factors of safetygreater than 1.0 using kh =
0.5amaxl g woul d not devel op "dangerousl y l arge" deformations.Sec. 10.6 Seismic
Sl ope Stabil ity Anal ysis 437As the preceding discussion indicates, there are no
hard and fast rul es for sel ection ofa pseudostatic coefficient for design. It seems
cl ear, however, that the pseudostatic coefficientshoul d be based on the actual
anticipated l evel of accel eration in the fail ure mass (incl udingany ampl ification
or deampl ification effects) and that it shoul d correspond to some fraction ofthe
anticipated peak accel eration. Al though engineering judgment is required for al l
cases,the criteria of Hynes-Griffin and Frankl in (1984) shoul d be appropriate for
most sl opes.Limitations of the Pseudostatic Approach. Representation of the
compl ex,transient, dynamic effects of earthquake shaking by a singl e constant
unidirectional pseudostatic accel eration is obviousl y quite crude. Even in its
infancy, the l imitations of thepseudostatic approach were cl earl y recognized.
Terzaghi (1950) stated that "the concept itconveys of earthquake effects on sl opes
is very inaccurate, to say the l east," and that a sl opecoul d be unstabl e even if
the computed pseudostatic factor of safety was greater than 1.Detail ed anal yses of
historical and recent earthquake-induced l andsl ides (e.g., Seed et aI.,1969, 1975;
Marcuson et 311., 1979) have il l ustrated significant shortcomings of the
pseudostaticapproach. Experience has cl earl y shown, for exampl e, that pseudostatic
anal ysescan be unrel iabl e for soil s that buil d up l arge pore pressures or show more
than about 15%degradation of strength due to earthquake shaking. As il l ustrated in
Tabl e l OA, pseudostaticanal yses produced factors of safety wel l above 1 for a
number of dams that l ater fail edduring earthquakes. These cases il l ustrate the
inabil ity of the pseudostatic method to rel iabl yeval uate the stabil ity of sl opes
susceptibl e to weakening instabil ity. Neverthel ess, thepseudostatic approach can
provide at l east a crude index of rel ative, if not absol ute, stabil ity.Discussion.
The pseudostatic approach has a number of attractive features. Theanal ysis is
rel ativel y simpl e and straightforward; indeed, its simil arity to the static l imit
equil ibrium anal yses routinel y conducted by geotechnical engineers makes its
computationseasy to understand and perform. It produces a scal ar index of stabil ity
(the factor ofsafety) that is anal ogous to that produced by static stabil ity
anal yses. It must al ways be recognized,however, that the accuracy of the
pseudostatic approach is governed by the accuracywith which the simpl e pseudostatic
inertial forces represent the compl ex dynamicinertial forces that actual l y exist in
an earthquake. Difficul ty in the assignment of appropriatepseudo static
coefficients and in interpretation of pseudostatic factors of safety, coupl edwith
the devel opment of more real istic methods of anal ysis, have reduced the use of the
pseudostatic approach for seismic sl ope stabil ity anal yses. Methods based on
eval uation ofpermanent sl ope deformation, such as those described in the fol l owing
sections, are beingused increasingl y for seismic sl ope stabil ity anal ysis.Tabl e 10-
4 Resul ts of Pseudostatic Anal yses of Earth Dams That Fail ed duringEarthquakesDam
kh FS Effect of EarthquakeSheffiel d Dam 0.10 1.2 Compl ete fail ureLower San Fernando
Dam 0.15 1.3 Upstream sl ope fail ureUpper San Fernando Dam 0.15 -2-2.5 Downstream
shel l , incl uding crestsl ipped about 6 ft downstreamTail ings dam (Japan) 0.20 -1.3
Fail ure of dam with rel ease of tail ingsSource: After Seed (1979).438 Seismic Sl ope
Stabil ity Chap. J010.6.1.2 Newmark Sl iding Bl ock Anal ysisThe pseudostatic method of
anal ysis, l ike al l l imit equil ibrium methods, provides anindex of stabil ity (the
factor of safety) but no information on deformations associated withsl ope fail ure.
Since the serviceabil ity of a sl ope after an earthquake is control l ed by
deformations,anal yses that predict sl ope displ acements provide a more useful
indication of seismicsl ope stabil ity. Since earthquake-induced accel erations vary
with time, the pseudostaticfactor of safety wil l vary throughout an earthquake. If
the inertial forces acting on a potential fail ure mass become l arge enough that the
total (static pl us dynamic) driving forcesexceed the avail abl e resisting forces,
the
factor of safety wil l drop bel ow 1.0. Newmark(1965) considered the behavior of a
sl ope under such conditions. When the factor of safetyis l ess than 1.0, the
potential fail ure mass is no l onger in equil ibrium; consequentl y, it wil l be
accel erated by the unbal anced force. The situation is anal ogous to that of a bl ock
restingon an incl ined pl ane (Figure 10.7). Newmark used this anal ogy to devel op a
method for predictionof the permanent displ acement of a sl ope subjected to any
ground motion.Sl iding~'Fail uresurface... ...(a)... ...(b)Figure 10.7 Anal ogy
between (a) potential l andsl ide and (b) bl ock resting on incl inedpl ane.Consider the
bl ock in stabl e, static equil ibrium on the incl ined pl ace of Figure 1O.7b.Under
static conditions, equil ibrium of the bl ock (in the direction paral l el to the
pl ane)requires that the avail abl e static resisting force, Rs , exceed the static
driving force, D,(Figure 1O.8a). Assuming that the bl ock's resistance to sl iding is
purel y frictional (c = 0)FS = avail abl e resisting forcestatic driving forceR, =
Wcos~tan<j>o, Wsin~tan otan~(l OA)Ns = Wcos J3(a) (b)Figure 10.8 Forces acting on a
bl ock resting on an incl ined pl ane: (a) static conditions;(b) dynamic conditions.
Sec. J 0.6 Seismic Sl ope Stabil ity Anal ysis 439where <I> is the angl e of friction
between the bl ock and the pl ane. Now consider the effect ofinertial forces
transmitted to the bl ock by horizontal vibration of the incl ined pl ane with
accel eration, ah(t) = kh(t)g (the effects of vertical accel erations wil l be
negl ected for simpl icity).At a particul ar instant of time, horizontal accel eration
of the bl ock wil l induce ahorizontal inertial force, khW(Figure 10.8b). When the
inertial force acts in the downsl opedirection, resol ving forces perpendicul ar to
the incl ined pl ane givesFS ( ) = avail abl e resisting force = Rd(t) = [cos ~ - kh(t)
sin~] tan<l > (10.5)d t pseudostatic driving force Dd(t) sin~ + kh(t)COS~Obviousl y,
the dynamic factor of safety decreases ~s kh increases and there wil l be (for a
statical l y stabl e bl ock) some positive val ue of kh that wil l produce a factor of
safety of 1.0(Figure 10.9). This coefficient, termed the yiel d coefficient, ky,
corresponds to the yiel daccel eration, ay = kyg. The yiel d accel eration is the
minimum pseudostatic accel erationrequired to produce instabil ity of the bl ock. For
the bl ock of Figure 10.8,(10.6)for sl iding in the downsl ope direction. For sl iding
in the uphil l direction (which can occurwhen ~ and <I> are smal l ),tan<l > + tan~1+
tano tanf(10.7)- - - - - - - - - FS = 13-f 2-f/l '000uc.o.<I> = 20°0--0.0 0.1 0.2
0.3 0.4 0.5 0.6Figure 10.9 Variation of pseudostatic factor of safety with
horizontal pseudostaticcoefficient for bl ock on pl ane incl ined at 20°. For <I>
=20°, bl ock is at the point of fail ure(FS =1) under static conditions, so the yiel d
coefficient is zero. For <I> =30° and <I> =40°,yiel d coefficients are 0.17 and
0.36, respectivel y.Exampl e 10.3Compute the yiel d accel eration for the sl ope
described in Exampl e 10.2.Sol ution The yiel d accel eration can be computed by trial
and error, or computed directl y forrel ativel y simpl e sl opes. Reviewing Exampl e
10.2, it is apparent that the total moment is equal toM, : = 4488 k-ft/ft + kh(5685
k-ftl ft) + 1438 k-ftl ft + kh(17825 k-ftl ft): = 5926 k-ftl ft + kh(23510 k-ftl ft)440
Seismic Sl ope Stabil ity Chap. 10The yiel d coefficient is the val ue of kh that
produces a pseudostatic factor of safety of I.Because the resisting moment is equal
to the overturning moment when FS = I,5926 k-ft/ft + kh(23510 k-ft/ft) = 10624 k-
ft/ftork = 10624 k-ft/ft - 5926 k-ft/ft = 0.20h 23510 k-ft/ftTherefore, the yiel d
accel eration is 0.20g.When a bl ock on an incl ined pl ane is subjected to a pul se of
accel eration that exceedsthe yiel d accel eration, the bl ock wil l move rel ative to
the pl ane. To il l ustrate the procedureby which the resul ting permanent
displ acements can be cal cul ated, consider the case inwhich an incl ined pl ane is
subhected to asingl e rectangul ar accel eration pul se of ampl itudeA and
duration !!.t. If the yiel d accel eration, a y, is l ess than A (Figure 10.l Oa), the
accel erationof the bl ock rel ative to the pl ane during the period from to to
to+ !': !.t isarel (t) = ab(t) - ay = A - a; to -;, t -;, to + !': !.t (1O.8a)where
abet)is the accel eration of the incl ined pl ane. The rel ative movement or the bl ock
duringthis period can be obtained by integrating the rel ative accel eration twice,
that is,Vrel (t) =rarel (t) dt = [A - ay](t - to)todre1(t) rVrel (t) dt HA - ay](t -
to)2to(l O.8b)(l O.8c)..Time..Time• I TimeII I: ~ I~'I I IIII JII : : I II ~II I II II
dr(t) 1(c)_____~~_L_ ---'- ___raj "~: t~ ~~ ~~ ~~D,----------------v.(t) 1(b)-----
~jL____---t-------"'_: __-------...Figure 10.10 Variation of rel ative vel ocity and
rel ative displ acement between sl idingbl ock and pl ane due to rectangul ar pul se that
exceeds yiel d accel eration between t =toand t = to+ St.Sec. 10.6 Seismic Sl ope
Stabil ity Anal ysis 441At t = to+ /: 1t, the rel ative vel ocity reaches its maximum
val ue. At that timeVrel (to + M) [A - ay]/: 1t (1O.9a)(l O.l Oc)dre1(to + /: 1t) ~[A -
ay]f: 1t2 (l 0.9b)After the base accel eration drops to zero (at t = to+ Az), the
sl iding bl ock is decel erated bythe friction force acting on its base. The bl ock
wil l continue to sl ide on the pl ane, but at adecreasing vel ocity which eventual l y
reaches zero. The accel eration during this time isgiven byarel (t) = ab(t)-ay = O-ay
= -ay to+M~t~tl (l O.l Oa)where tl is the time at which the rel ative vel ocity becomes
zero (note that the bl ock undergoesnegative accel eration, or decel eration, during
this period). Between to+ /: 1t and tJ, therel ative vel ocity wil l decrease with time
according toVrel (t) = Vrel (to+/: 1t)+ftarel dt = A/: 1t- ai t- to)to + I: !.tSetting
the rel ative vel ocity equal to zero at t = tl givest] = to+ : !./: 1tayThendrel (t) =
ft Vrel (t)dt =A /: 1t (t - to - M)to + !!.t- ~ [t2- (to + /: 1t)2] to+ M ~ t ~ tl
After time tJ, the bl ock and incl ined pl ane move together. During the total period
of timebetween t = to and t = tJ, the rel ative movement of the bl ock is as shown in
Figure 10.10.Between to and to+ /: 1t, the rel ative vel ocity increases l inearl y and
the rel ative displ acementquadratical l y. At to + /: : ,.t, the rel ative vel ocity has
reached its maximum val ue, after whichit decreases l inearl y. The rel ative
displ acement continues to increase (but at a decreasingrate) until t = tl ' Note
that the total rel ative displ acement(l 0.11)depends strongl y on both the amount by
which and the l ength of time during which the yiel daccel eration is exceeded. This
suggests that the rel ative displ acement caused by a singl epul se of strong ground
motion shoul d be rel ated to both the ampl itude and frequency contentofthat pul se.
An earthquake motion, however, can exceed the yiel d accel eration a numberof times
and produce a number of increments of displ acement (Figure 10.11). Thus thetotal
displ acement wil l be infl uenced by strong-motion duration as wel l as ampl itude and
frequency content. Indeed, appl ication of this approach to a variety of simpl e
waveforms(e.g., Sarma, 1975; Yegian et aI., 1991) have shown that the permanent
displ acement of asl iding bl ock subjected to rectangul ar, sinusoidal , and triangul ar
periodic base motions isproportional to the square of the period of the base
motion.442 Seismic Sl ope Stabil ity Chap. 1055c'0o~50jO~: ~: ______ _ _(a) 1il 0
- .~ ~ ¥.1"\.. ,,/\ 4iq. W~, A/V • WI \FEo _ I I VV~05 : I I. I II II III I~ 50~E
(b) ~ 0 1------l Lj-J...I--.L-------(e) r: 1-. ~OL_ _~ 0.-£5 10 : Time (s)Figure 10.11
Devel opment of permanent sl ope displ acements for actual earthquakeground motion.
(After Wil son and Keefer, 1985.)Infl uence of Yiel d Accel eration on Sl ope
Displ acements. Obviousl y,the sl iding bl ock model wil l predict zero permanent sl ope
displ acement if earthquakeinducedaccel erations never exceed the yiel d accel eration
(a!amax ~ 1.0) as il l ustrated in Figure1O.12a. Since the permanent displ acement is
obtained by doubl e integration of the excessaccel eration, the computed
displ acements for a sl ope with a rel ativel y l ow yiel d accel eration(smal l a!amax )
wil l be greater than that of a sl ope with a higher yiel d accel eration (Figure
10.12b, c). The rel ationship between sl ope displ acement and a!amax has been
investigated bya number of researchers.a(~ay f-----,,-------aCt)ay f---
ttn----,-------(a) (b) (c)Figure 10.12 Permanent sl ope displ acements depend on the
rel ationship between theyiel d accel eration and the maximum accel eration. (a) If the
yiel d accel eration of a sl opeis greater than the maximum accel eration of a
particul ar ground motion, no displ acementwil l occur. As yiel d accel erations
decrease, as in (b) and (c), sl ope displ acementsincrease quickl y.Sec. 10.6 Seismic
Sl ope Stabil ity Anal ysis 443(10.13)Using the rectangul ar pul se sol ution devel oped
in Section 10.6.1.1, Newmark (1965)rel ated singl e-pul se sl ope displ acement to peak
base vel ocity, Vmax' byd = V~ax(l -ay) (10.12)reI Za; A[note that equation (10.12)
is equival ent to equation (10.11) with Vmax =A M)]. Anal ysis ofseveral earthquake
motions normal ized to peak accel erations of 0.5g and peak vel ocities of30 in.l sec
(76 ern/sec) suggested that the effective number of pul ses in an earthquake motion
coul d be approximated by Al a.; Newmark found that a reasonabl e upper bound to the
permanentdispl acements produced by these earthquake motions was given by2d = Vmax
amaxmax La; aywhere a/amax;: : : : 0.17. Sarma (1975) and Yegian et al . (1988) derived
cl osed-form sol utionsfor the permanent displ acements produced by simpl e periodic
(triangul ar, sinusoidal , andrectangul ar) input motions (Figure 10.13). Studies of
permanent displ acements predictedby the sl iding bl ock method for actual earthquake
motions
(e.g., Sarma, 1975; Frankl in andChang, 1977; Makdisi and Seed, 1978; Ambraseys and
Menu, 1988) show shapes that aresimil ar to those of the sinusoidal and triangul ar
waves at a/amax val ues greater than about0.5. Ambraseys and Menu (1988) found that
the shape at smal l er a/amax val ues was infl uencedby whether or not upsl ope
movements were considered; for the case in which theywere not, permanent
displ acements (in centimeters) caused by actual ground motions weregiven by[(a )
2'53( a )-1.09] l ogu = 0.90 + l og 1--y -y a max a maxO"l ogu = 0.30 (10.14)10CQl EQl
otil 0. l fJ 10'1'5CQl c 10'2til Eiii c.. 10'3 ""0Ql .!: : : !""iiiE 10'4<5Z10'50 0.2 0.4
0.6 0.8 1.0~BmaxFigure 10.13 Variation of normal izedpermanent displ acement with
ratio of yiel daccel eration to maximum accel eration forsimpl e waveforms. The
normal izedpermanent displ acement is defined inequation (10.15). (After Yegian et
al ., 1991.)444 Seismic Sl ope Stabil ity Chap. 10for 0.1 : s; ayiamax: S;
0.9,6.6: S;Ms: S; 7.3, and ay computed using residual soil strength. To al l owmeasures
of frequency content and duration to be considered expl icitl y, Yegian et al . (1991)
used the database of Frankl in and Chang (1977) to devel op the fol l owing expression
for themedian permanent normal ized displ acement: (10.15)l ogu* = l og( u 2) = 0.22-
1O.12"s: '-' + 16.38(..S: .-.)2 _11.48(..S: .-.)3amaxNeqT amax amax amaxO"l ogu* = 0.45
where Neq is an equival ent number of cycl es and T is the predominant period of the
inputmotion. Considering onl y this source of uncertainty (i.e., negl ecting
uncertainty in amax' ay,Neq , and T), probabil ities of exceeding various
displ acements can be determined(Figure 10.14). Al ternative approaches to the
probabil istic anal ysis of sl ope displ acementshave been presented by Constantinou
and Gazetas (1984) and Lin and Whitman (1986).Recognition of the l imitations of
peak accel eration as a sol e descriptor of strongground motion has l ed to the use of
other ground motion parameters in sl ope-displ acementprediction. Sl iding bl ock
displ acements have been correl ated with Arias intensity: l ogu = 1.460 l ogI, -
6.642ay+ 1.546 O"l ogu = 0.409 (10.16)where u is in em, I a is in m/sec, and ayis in
g's (Jibson, 1994) and used to predict areal l imitsof earthquake-induced
l andsl iding (Wil son and Keefer, 1985).Two aspects of seismic sl ope stabil ity are
cl earl y il l ustrated by the studies described inthe preceding paragraphs. First,
earthquake-induced sl ope displ acements are very sensitiveto the val ue of the yiel d
accel eration. Consequentl y, smal l differences in yiel d accel erationcan produce
l arge variations in predicted sl ope displ acement. Second, the great variabil ity in
distributions of accel eration pul se ampl itudes between different ground motions
producesgreat variabil ity in predicted sl ope displ acements. Even ground motions
with simil ar ampl itudes,frequency contents, and durations can produce significantl y
different predicted sl ope10C(])E(]) oCl l Q.(J)'6C(])<: Cl l EID c.'0(]).t: !"iiiE 10-4 :
0Z10-50 02 OA OB OB 1~~8 maxFigure 10.14 Contours of equal probabil ity of exceedance
of normal izedpermanent displ acement. (After Yegian etal ., 1991.)Sec. 10.6 Seismic
Sl ope Stabil ity Anal ysis 445displ acements. This uncertainty must be recognized in
the prediction of earthquake-inducedsl ope deformations.Exampl e 10.4Estimate the
expected permanent displ acement of the sl ope described in Exampl e 10.3 if subjected
to a ground motion equival ent to the Gil roy NO.1 (rock) earthquake motion. Use the
proceduresof Newmark (Equation 10.13) and Jibson (Equation 10.16).Sol ution From
Exampl e 3.1, the peak accel eration and vel ocity of the Gil roy No.1 (rock)motion are
amax = 0.442Vmax = 33.7 cml secThe yiel d accel eration was computed as 0.20g in
Exampl e 10.3. Then, using the Newmark procedure(Equation 10.13), an upper bound
estimate of the permanent displ acement woul d be(33.7 cml secf 0.442g = 6.4 cmdmax =
2(0.20g)(981 cm/secv'g) 0.200gThe Arias Intensity of the Gil roy No.1 (rock) motion
was computed as fa = 167.7 ern/sec inExampl e 3.6. Using the Jibson procedure
(Equation 10.16), the average permanent displ acementwoul d be given byl ogu =
1.460l 0g(1.677) - 6.642(0.20) + 1.546 = 0.545sou = 10°·545 = 3.5 ernInput Motions.
The accuracy of a sl iding bl ock anal ysis depends on the accuracyof the input motion
appl ied to the incl ined pl ane. As original l y proposed, the sl idingbl ock method
assumes the potential fail ure mass to be rigid, in which case the appropriateinput
motion woul d be the ground motion at the l evel of the fail ure surface. Actual
sl opes,however, are compl iant-they deform during earthquake shaking. Their dynamic
responsedepends on their geometry and stiffness and on the ampl itude and frequency
content of themotion of the underl ying ground. For sl opes composed of very stiff
soil s and/or sl opes subjectedto l ow-frequency motion (a combination that
producesl ong wavel engths), l ateral displ acementsthroughout the potential fail ure
mass wil l be nearl y in phase (Figure 1O.15a)andthe rigid bl ock assumption wil l be
at l east approximatel y satisfied. Lateral displ acements inu uzLow frequencyLong
wavel ength(a)High frequencyShort wavel ength(b)\Figure 10.15 Infl uence of frequency
on motions induced in sl opes. Long wavel engthassociated with l ow-frequency motion
(a) causes soil above fail ure surface to moveessential l y in phase. For higher-
frequency motion (b), portions of soil above fail uresurface may be moving in
opposite directions.446 Seismic Sl ope Stabil ity Chap. 10potential fail ure masses of
sl opes in softer soil s (and/or sl opes subjected to higherfrequencymotion), however,
may be out of phase (Figure l O.15b). When this occurs, theinertial forces at
different points within the potential fail ure mass may be acting in opposite
directions and the resul tant inertial force may be significantl y smal l er than that
impl ied bythe rigid-bl ock assumption.The effects of sl ope response on the inertial
force acting on a potential fail ure masscan be computed using dynamic stress-
deformation anal yses (Chopra, 1966). Using adynamic finite-el ement anal ysis (Figure
l O.16a), the horizontal components of the dynamicstresses acting on a potential
fail ure surface (Figure 10.16b) are integrated over the fail uresurface to produce
the time-varying resul tant force that acts on the potential fail ure surface.This
resul tant force can then be divided by the mass of the soil above the potential
fail uresurface to produce the average accel eration of the potential fail ure mass.
Al though the procedurewas devel oped original l y for dams, the basic concept can be
appl ied to any type ofsl ope. The average accel eration time history, which may be of
greater or smal l er ampl itudethan the base accel eration time history (depending on
the input motions and the ampl ificationcharacteristics of the sl ope), provides the
most real istic input motion for a sl iding bl ockanal ysis of the potential fail ure
mass.(a) (b)Figure 10.16 Eval uation of average accel eration for sl ope in
embankment. Finiteel ementanal ysis predicts variation of shear and normal stresses
on potential fail ure pl anewith time. Integration of horizontal components of
stresses over potential fail ure surfacegives resul tant horizontal force acting on
potential l y unstabl e soil . Time history ofaverage accel eration is obtained by
dividing resul tant force by mass of potential l yunstabl e soil .Other Factors
Infl uencing Sl ope Displ acement. The standard sl idingbl ock anal ysis is based on the
assumption of rigid-perfectl y pl astic stress-strain behavioron a pl anar fail ure
surface. Conditions for actual sl opes may vary from these assumptions ina number of
ways.The shear strength of some soil s is rate dependent. Since earthquake-induced
shearstresses are appl ied at different rates, the shear strength (and hence the
yiel d accel eration) canvary with time throughout an earthquake (e.g., Hungr and
Morgenstern, 1984; Lemos et al .,1985). Consideration of rate-dependent strength in
a sl iding bl ock anal ysis is compl icated bydifferences between strain rates in the
fiel d and in the l aboratory tests used to measure thestrength. Lemos and Coel ho
(1991) and Tika-Vassil ikos et al . (1993) suggested proceduresfor incorporating
rate-dependent fiel d strengths into numerical sl iding bl ock anal yses.In the fiel d,
soil s rarel y behave as perfectl y pl astic material s. Instead, they usual l yexhibit
strain-hardening or strain-softening stress-strain behavior (Figure 10.17) after
yiel ding. The yiel d accel erations of sl opes comprised of strain-hardening or
strain-softeningsoil s wil l vary with sl ope displ acement. Consequentl y, the
permanent displ acement ofSec. 10.6 Seismic Sl ope Stabil ity Anal ysisRigid, strain-
hardening447Rigid, strain-softeningCfJCfJ~Ciil -E: ;: : .----------------mOJs: CIJShear
strainRigid, perfectl y pl asticFigure 10.17 Stress-strain behavior of rigid-
perfectl y pl astic, rigid-strain hardening,and rigid-strain softening material s.a
sl ope in strain-hardening material s wil l be smal l er than predicted by a
conventional sl idingbl ock anal ysis; the reverse wil l be true for strain-softening
soil s. Modification of sl idingbl ock anal yses for consideration of displ acement-
dependent strength is fairl ystraightforward.Many sl opes fail by mechanisms that
differ from the pl anar fail ure mechanismassumed in sl iding bl ock anal yses (see
Figure l OA). Negl ecting the effects ofrate- and displ acement-dependent strength,
the stabil ity of a bl ock on a pl ane wil l be the same bothbefore and after a pul se
of displ acement-because the geometry of the bl ock rel ative to thepl ane is
unchanged. Movement of a sl ope on a nonpl anar fail ure surface, however, tends to
fl atten the sl ope, thereby reducing the driving forces. As a resul t, the yiel d
accel erationshoul d increase due to changes in the geometry of the
unstabl e soil . For most sl opes, however,this effect does not become significant
until l arge displ acements have occurred.10.6.1.3 Mal (disi-Seed Anal ysisMakdisi and
Seed (1978) used average accel erations computed by the procedure ofChopra (1966)
and sl iding bl ock anal yses to compute earthquake-induced permanent deformationsof
earth dams and embankments. By making simpl ifying assumptions about theresul ts of
dynamic finite el ement and shear beam anal yses of such structures, a simpl ified
procedure for prediction of permanent displ acements was devel oped.In the simpl ified
procedure, the yiel d accel eration for a particul ar potential fail uresurface is
computed using the dynamic yiel d strength [80% of the undrained strength (Section
6.5.2)] of the soil . The dynamic response of the dam/embankment is accounted for by
an accel eration ratio that varies with the depth of the potential fail ure surface
rel ative to theheight of the dam/embankment (Figure 10.18).By subjecting several
real and hypothetical dams to several actual and syntheticground motions scal ed to
represent different earthquake magnitudes, Makdisi and Seedcomputed the variation
of permanent displ acement with ayl amax and magnitude. Scatter inthe predicted
displ acements was reduced by normal izing the displ acement with respect tothe peak
base accel eration and the fundamental period of the dam/embankment (note thatthe
normal ized displ acement has units of seconds). Prediction of permanent
displ acementsby the Makdisi-Seed procedure is accompl ished with the charts shown in
Figure 10.19.448 Seismic Sl ope Stabil ity Chap. J00.20.4zR0.60.81.0 '-
~"""""'''''''''-----'----'-------'o 0.2 0.4 0.6 0.8 1.0amax,aveamax,z=OFigure 10.18
Variation of average maximum accel eration with depth of potential fail ure surface
for dams and embankments. (After Makdisi and Seed (1978). Simpl ifiedprocedure for
estimating dam and embankment earthquake-induced deformations,Journal of the
Geotechnical Engineering Division, Vol . 104, No. GT7. Reprinted bypermission of
ASCE.)0.0010.4 0.6 0.8 1.0~8 maxM - 8 1/ : 40.2------·~···_------·i····-------
+·····------". : : 1 : -, ~ : M - 72. : -, : " : /. : -_: : ~'--~~~'-" .________.__!. V-
_-~_, __M ' 6 1 /: i ---, : - 2. I : ---1-~----·------r------·---r--------..·!
·--------··r·~;~-----10 .------~---------,0.10.0001o0.001uamaxTo(sec) 0.010.4 0.6
0.8 1.0~8 max0.2"",',..""····,··-----f"·····-----~ ..·····----+······-----: ':
("'''''''''11"",1, _M ~ 7 ~ j100.10.0001ouamaxTo(sec) 0.01Figure 10.19 Variation of
normal ized permanent displ acement with yiel d accel erationfor earthquakes of
different magnitudes: (a) summary for several earthquakes anddams/embankments; (b)
average val ues. (After Makdisi and Seed (1978). Simpl ifiedprocedure for estimating
dam and embankment earthquake-induced deformations,Journal of the Geotechnical
Engineering Division, Vol . 104, No. GT7. Reprinted bypermission of ASCE.)Sec. 10.6
Seismic Sl ope Stabil ity Anal ysis 449Exampl e 10.5Assume that a fail ure surface that
extends over the upper two-thirds of the earth dam shown inExampl e 7.6 has a yiel d
accel eration of 0.24g. Estimate the permanent displ acement that woul doccur if the
base of the dam was subjected to the Gil roy No.1 (rock) motion.Sol ution The Gil roy
No.1 motion was recorded in the 1989 Loma Prieta earthquake whichhad a magnitude of
7.1. The peak accel eration was 0.442g. From Exampl e 7.6, the fundamental period of
the dam isTo = 2n = 0.33 sec19.2 rad/secUsing Figure l O.l 9b with a/amax
=0.24g/0.442g =0.54 and M =7.1, the average normal izeddispl acement is about 0.04.
Therefore,u = 0.04 amax To = 0.04(0.442g)(32.2 ft/sec 2/g)(0.33 sec) = 0.19 ft =
2.3 inThe Makdisi-Seed simpl ified procedure is widel y used for estimation of
permanentdispl acements in dams and embankments. Because the procedure is based on
the dynamicresponse characteristics of dams and embankments, its resul ts must be
interpreted with cautionwhen appl ied to other types of sl opes.10.6.1.4 Stress-
Deformation Anal ysisJust as stress-deformation anal yses of static sl ope stabil ity
are usual l y performedusing static finite-el ement anal yses, stress-deformation
anal yses of seismic sl ope stabil ityare usual l y performed using dynamic finite-
el ement anal yses (Section 7.3.1). In such anal ysesthe seismical l y induced permanent
strains in each el ement of the finite-el ement meshare integrated to obtain the
permanent deformation of the sl ope. Permanent strains withinindividual el ements can
be estimated in different ways. The strain potential and stiffnessreduction
approaches estimate permanent strains using l aboratory test resul ts to determinethe
"stiffness" of soil s subjected to earthquake l oading. Nonl inear anal ysis approaches
usethe nonl inear inel astic stress-strain behavior of the soil to compute the
devel opment of permanentstrains throughout an earthquake.Strain Potential Approach.
In their l andmark investigation ofthe sl ides thatoccurred in the Upper and Lower
San Fernando dams during the 1971 San Fernando earthquake,Seed et al . (1973)
devel oped a procedure for estimating earthquake-induced sl opedeformation from the
resul ts of l inear or equival ent l inear anal yses. In this procedure thecycl ic shear
stresses are computed in each el ement of a dynamic finite-el ement anal ysis.Using
the resul ts of cycl ic l aboratory tests, the computed cycl ic shear stresses are used
topredict the strain potential , expressed as a shear strain, for each el ement.
Deformations arethen estimated as the product of the average strain potential al ong
a vertical section throughthe sl ope and the height of that section. The method
impl icitl y assumes that the strains thatdevel op in the fiel d wil l be the same as
those that devel op in a simil arl y l oaded l aboratorytest specimen and that the
maximum shear stress acts in the horizontal direction in al l el ements.Consequentl y,
the strain potential approach estimates onl y horizontal displ acements.Anal yses
based on the strain potential approach are cl earl y very approximate, andtheir
resul ts shoul d al ways be interpreted with that fact in mind.450 Seismic Sl ope
Stabil ity Chap. 10Stiffness Reduction Approach. Another method for estimation of
permanentsl ope displ acement was devel oped by Lee (1974) and Serff et al . (1976). In
thisapproach, computed strain potential s are used to reduce the stiffness of the
soil as il l ustratedin Figure 10.20. Earthquake-induced sl ope displ acements are then
taken as the differencebetween the nodal point displ acements from two static
finite-el ement anal yses: one usingthe initial shear modul i and the other using the
reduced shear modul i. The technique can beused with l inear or nonl inear model s.
Unl ike the strain potential approach, the stiffnessreduction approach can estimate
vertical as wel l as horizontal movements. It is a veryapproximate procedure,
however, and is subject to many of the l imitations of the strainpotential approach.
Work-energy principl es can be used to provide a more fundamental procedurefor
stiffness reduction (Byrne, 1991; Byrne et al ., 1992).Yt Shear strainFigure 10.20
Procedure used to reducestiffness from initial val ue, Ci , to final val ue, C/o in
stiffness reduction approach toestimation of permanent sl ope deformation.(After
Serff et aI., 1976.)Nonl inear Anal ysis Approach. Permanent sl ope deformations can
al so becomputed by finite-el ement anal yses that empl oy nonl inear inel astic soil
model s. The basicprocedures of nonl inear finite-el ement anal ysis of earth
structures were introduced in Section7.2.3. The seismic performance of sl opes has
been anal yzed with two- and threedimensional finite-el ement anal yses using both
cycl ic stress-strain model s (e.g., Finn et aI.,1986) and advanced constitutive
model s (e.g., Prevost, 1981; Mizuno and Chen, 1982;Kawai, 1985; Daddazio et aI.,
1987). The most common appl ication of these techniques, todate, has been the
anal ysis of earth dams. Exampl es of such anal yses can be found in Prevostet al .
(1985), Griffiths and Prevost (1988), Finn (1990), El gamal et al . (1990), and
Suecariehet al . (1991). As discussed in Chapter 7, the accuracy of nonl inear
finite-el ementanal yses depends primaril y on the accuracy of the stress-strain or
constitutive model s onwhich they are based.10.6.2 Anal ysis of Weakening Instabil ity
Through a process of pore pressure generation and/or structural disturbance,
earthquakeinducedstresses and strains can reduce the shear strength of a soil .
Weakening instabil itiescan occur when the reduced strength drops bel ow the static
and dynamic shear stressesinduced in the sl ope. Weakening instabil ities are usual l y
associated with l iquefactionSec. 10.6 Seismic Sl ope Stabil ity Anal ysis 451phenomena
and can be divided into two main categories, fl ow fail ures (Section 9.6.3.2) and
deformation fail ures (Section 9.6.3.3). Fl ow fail ures occur when the avail abl e
shearstrength becomes smal l er than the static shear stress required to maintain
equil ibrium of asl ope. Fl ow fail ures, therefore, are actual l y driven by static
stresses. They can produce veryl arge deformations that occur quickl y and without
warning. Deformation fail ures occurwhen the shear strength of a soil is reduced to
the point where it is temporaril y exceeded byearthquake-induced shear stresses.
Much l ike inertial fail ures, deformation fail ures occur asa series of "pul ses" of
permanent displ acement that cease at the end of earthquake shaking.Different
procedures are avail abl e for the anal ysis of fl ow fail ures and deformation
fail ures.10.6.2.1 Fl ow Fail ure Anal ysisBecause they usual l y invol ve significant
reduction in soil strength, fl ow fail ures usual l yproduce l arge deformations and
severe damage. The first step in their anal ysis is general l yto determine whether or
not one wil l occur. To estimate the extent of the damageproduced by fl ow fail ures,
procedures for estimation of fl ow fail ure deformations have al sobeen devel oped.
Anal ysis of Stabil ity. Potential fl ow sl ide instabil ity is most commonl y eval uated
by conventional static sl ope stabil ity anal yses using soil strengths based on end-
ofearthquakeconditions (Marcuson et aI., 1990). In a typical anal ysis, the factors
of safetyagainst l iquefaction at al l points on a potential fail ure surface is first
computed. Residual strengths are then assigned to those portions of the fail ure
surface on which the factor ofsafety against l iquefaction is l ess than 1. At
l ocations where the factor of safety against l iquefactionis greater than 1,
strength val ues are based on the effective stresses at the end ofthe earthquake
(i.e., considering pore pressures generated during the earthquake). Withthese
strengths, conventional l imit equil ibrium sl ope stabil ity anal yses are used to
cal cul atean overal l factor of safety against fl ow sl iding. If the overal l factor of
safety is l ess than 1,fl ow sl iding is expected. The possibil ity of progressive
fail ure (Section 10.5.1) must be consideredin stabil ity eval uations of this type-
the redistribution of stresses invol ved in progressivefail ure are not accounted for
directl y in l imit equil ibrium anal yses.Anal ysis of Deformations. If stabil ity
anal yses indicate that fl ow fail ure isl ikel y, the extent of the zone infl uenced by
the fail ure can be determined from an anal ysisof fl ow fail ure deformations. By
negl ecting the smal l deformations that precede the triggeringof fl ow sl iding, rough
estimates of fl ow sl iding deformations can be obtained from proceduresbased on
l imit equil ibrium, fl uid mechanics, and stress-deformation anal yses.Simpl e pl ane
strain, l imit equil ibrium procedures can be used to estimate the distance a
l iquefied soil woul d fl ow over a gentl e « 3 to 4°) sl ope (Lucia et aI., 1981). By
assuming thatthe l iquefied soil woul d eventual l y come to rest with a l inear
surface, a postfail ure geometrythat satisfies equil ibrium and vol umetric
constraints can be identified. With reference to thenotation of Figure 10.21, the
procedure can be impl emented in the fol l owing steps: 1. Using Figure 1O.22a, compute
val ues of the height of the sl ope at the end of fl ow(when the static factor of
safety reaches 1.0) based on strength considerations usingSrHT,s = Noy (10.17)452
Seismic Sl ope Stabil ity Chap. 10/ Before fl ow ___H_1'": -+-Zl H'I" -I t LFigure 10.21
Geometric notation for estimation of fl ow fail ure distance by procedureof Lucia et
al . (1981).1,000.0..'------------,I-------=---;\1/2 I100.0 : : - 13 = 00Strength
curveNoHT /vol ume curve: 1l Conditions that10.0 --- .s satisfy strength andJ: : .'"
vol ume contraintssimul taneousl y1.00.1 1.0 2.0 4.0 10.020.0 astabl e aSl ope angl e, a
(a) (b)Figure 10.22 Charts for estimation of fl ow fail ure distance: (a) stabil ity
number chartsfor computing strength curve; (b) determination of HT and a val ues
that simul taneousl ysatisfy strength and vol ume constraints. After Lucia et al .
(1981).for various assumed val ues of the sl ope angl e, a. Pl ot the data in the form
of a "strengthcurve" as in Figure 1O.22b.2. For various assumed val ues of a,
cal cul ate the height of the sl ope after fl ow based onconstant-vol ume conditions
using(10.18)whereAl tan a Az2 tanZatana- tan~ = tanrz - tanfA3tanfHe4Srtana- tanji y
Sec. 10.6 Seismic Sl ope Stabil ity Anal ysis 453(10.19)and Vf is the estimated vol ume
of soil invol ved in the fl ow sl ide. Pl ot the resul ting datain the form of a "vol ume
curve" as in Figure 1O.22b.3. The strength and vol ume curves intersect where Hy,s==
Hy, v. The resul ting Hstabl e andastabl e val ues satisfy both strength and vol ume
requirements with a factor of safetyequal to 1. The horizontal distance covered by
the fl ow sl ide can then be computed asL == Hstabl e - Hetan astabl eAl though the
procedure invol ves several simpl ifying assumptions and requires an estimateof the
strength of the l iquefied soil (Section 9.6.4.1), it can provide at l east a crude
estimateof the deformations invol ved in certain fl ow sl ides.The fl uidl ike behavior
of l iquefied soil s has motivated fl uid mechanics approaches tothe model ing of fl ow
sl ide behavior. Most of this work has been directed toward debris fl ows(e.g.,
Johnson, 1970; Iverson and Denl inger, 1987) and tail ings dam fail ures (e.g.,
Jeyapal anet aI., 1981). Rheol ogical model ing of l iquefied soil s is quite difficul t.
The Bingham model (strength == 'Ly + IIpy, where t, and IIp are the Bingham yiel d
strength and pl astic hscosity,respectivel y, and y is the shear strain rate) is most
commonl y used (Johnson, 1970; Jeyapal an,1980; O'Brien and Jul ien, 1988; Phil l ips
and Davies, 1991), al though its ~bil ity torepresent the frictional nature of
l iquefied soil is l imited (Iverson and LaHusen, 1P93).The devel opment of advanced
nonl inear dynamic anal yses have made an al ternativeapproach possibl e. The finite-
el ement program TARA-3FL (Finn and Yogendtakumar,1989), for exampl e, can reduce the
strength of any el ement in the sl ope to the,residual strength at the time
l iquefaction of the el ement is initiated. The program periodical l yupdates the
finite-el ement mesh at each time step to al l ow computation of l arge deformations
(Figure 10.23). Finn (1990) described its appl ication to Sardis Dam in Mississippi,
where l iquefaction of the core and a thin seam of cl ayey sil t was expected (Figure
10.23).Anal yses of this type not onl y indicate whether fl ow sl iding wil l occur but
al so provide anestimate of the distribution and magnitude of any resul ting
deformations.10.6.2.2 Deformation Fail ure Anal ysisAl though deformation fail ures
general l y invol ve smal l er deformations than fl ow fail ures,they are capabl e of
causing considerabl e damage. Lateral spreading is the most Commontype of
deformation fail ure. In recent years a number of investigators have devel oped
methods to estimate permanent displ acements produced by deformation fail ures.
Becausethe mechanisms that produce deformation fail ures are so compl icated,
procedures for predictionof the resul ting displ acements are l argel y empirical in
nature.Hamada et al . Approach. Hamada et al . (1986) considered the effects of
geotechnical and topographic conditions on permanent ground displ acements observed
inuniform sands of medium grain size in the 1964 Niigata (M == 7.5), 1971 San
Fernando(M == 7.1), and 1983 Nihonkai-Chubu (M == 7.7) earthquakes. Permanent
displ acementswere found to be most strongl y infl uenced by the thickness of the
l iquefied l ayer and thesl opes of the ground surface and l ower boundary of the
l iquefied zone. Permanent horizontal ground displ acement, D, was found to vary
according to the empirical rel ationship(10.20)454 Seismic Sl ope Stabil ity Chap. 10.
200 feeto 150 feet. . , ---t _I/~Ia~r~e~e~ft ~.~./rJ 1\ _I-,L' v '~r: -r.-;;". "1/
t- If X ' """.- ~. " 1/ t .'.'" ."I I.. 150oQ)Ooi(J)~ .. 100~"'50Horizontal scal e
Figure 10.23 Initial (dashed) and postl iquefaction (sol id) configurations of Sardis
Damin Mississippi from TARA-3FL anal yses. Note the l arge strains due to
l iquefaction incore and thin seam bel ow the upstream shel l . (After Finn, 1990.)
Liquefaction hazardswere reduced by driving compaction pil es into the upstream
embankment (see Figure12.8).where H is the thickness of the l iquefied l ayer in
meters and 8 is the l arger of the ground surfacesl ope or the sl ope of the l ower
boundary of the l iquefied zone in percent. For case historiesfrom the three l isted
earthquakes, 80% of the observed displ acements were within afactor of 2 of those
predicted by equation (10.20). Note that equation (10.20) does notaccount for the
strength of the l iquefied soil ; l ike al l such empirical approaches, it must be
appl ied cautiousl y when conditions vary from those on which it is based.Youd and
Perkins (Liquefaction Severity Index) Approach. Basedon observed l ateral
displ acements from a number of case histories in the western UnitedStates, Youd and
Perkins (1987) defined the l iquefaction severity index (LSI) as "the general maximum
d-val ue (in inches) for l ateral spreads generated on wide active fl ood pl ains,
del tas, or other areas of gentl y-sl oping Late Hol ocene fl uvial deposits." As
defined, the LSIrepresents a conservative estimate of ground displ acement in a
given area; fail ures withsmal l er displ acements woul d al so be expected in the area.
An anal ysis of the case historydatabase indicated that LSI coul d be predicted byl og
(LSI) = - 3.49 - 1.85 10gR + 0.98Mw s: ; 100 (10.21)where R is the horizontal
distance from the seismic energy source in kil ometers. The variationof LSI with M
and R is shown in Figure 10.24. Qual itative descriptions of the natureof
deformation fail ures for different LSI val ues are presented in Tabl e 10-5. The
dependenceof LSI on magnitude and distance l ends itsel f to incorporation into a
probabil isticseismic hazard anal ysis. Youd and Perkins (1987) used this approach to
produce probabil isticLSI maps for southern Cal ifornia.Sec. 10.6 Seismic Sl ope
Stabil ity Anal ysis 4551000Data from: o 1906 San Francisco+ 1964 Al aska• 1973 PI.
Mugu• 1978 Santa BarbaraD 1979 Imperial Val l eyX 1981 Westmorl and... 1983 Borah Peak
1i5100--'~"0£~.~>Q) enc: 0t5.l l ! Q) 10: : J0-: : J1000Horizontal distance from surface
projection of energy source (km)Figure 10.24 Variation of LSI with distance and
earthquake magnitude. (After Youdand Perkins, 1986. Mapping of l iquefaction
severity index, Joumal of Geotechnical Engineering, Vol . 103, No.1 I. Reprinted by
permission of ASCE.)(1O.22a)(1O.22b)Byrne Approach. Model ing a sl ope as a crust of
intact soil resting on a l ayer ofl iquefied soil (Figure 10.25), Byrne (1991) used
work-energy principl es with an el asticperfectl ypl astic model of l iquefied soil to
devel op expressions for estimation of permanentsl ope displ acement. In this
approach,
the permanent displ acement, D, is obtained fromD3Sr 1 2----'-----: : 2 - D'"Cst -
2mvo = 03(Yl imh)D = 3mv~ + 4SrYl imh6(Sr - '"Cst)where S, is the residual strength
of the l iquefied soil (Figure 9.56), Yl im the l imiting shear strain,~c the
thickness ofthe l iquefied l ayer, '"Cst the average shear stress required for static
equil ibrium(on a fail ure surface passing through the middepth of the l iquefied
l ayer), m the mass of the soil above the fail ure surface, and v0 the vel ocity of the
mass at the instant of l iquefaction. Typical 456 Seismic Sl ope Stabil ity Chap. 10
Tabl e 10-5 Abundance and General Character of Liquefaction Effects for Different
LSI Val ues in Areaswith Widespread Liquefiabl e DepositsLSI Description5 Very
sparsel y distributed minor ground effects incl ude sand boil s with sand aprons up to
0.5 m (1.5 ft) indiameter, minor ground fissures with openings up to 0.1 m (0.3 ft)
wide, ground settl ements of up to 25 mm(I in.). Effects l ie primaril y in areas of
recent deposition and shal l ow groundwater tabl e such as exposedstreambeds, active
fl oodpl ains, mudfl ats, shorel ines, etc.10 Sparsel y distributed ground effects
incl ude sand boil s with aprons up to I m (3 ft) in diameter, ground fissureswith
openings up to 0.3 m (I ft) wide, ground settl ements of a few inches over l oose
deposits such astrenches or channel s fil l ed with l oose sand. Sl umps with up to a
few tenths of a meter displ acement al ongsteep banks. Effects l ie primaril y in areas
of recent deposition with a groundwater tabl e l ess than 3 m (l Oft)deep.30 General l y
sparse but l ocal l y abundant ground effects incl ude sand boil s with aprons up to 2 m
(6 ft) diameter,ground fissures up to several tenths of a meter wide, some fences
and roadways noticeabl y offset, sporadicground settl ements of as much as 0.3 m (I
ft), sl umps with 0.3 m (I ft) of displ acements common al ongsteep stream banks.
Larger effects l ie primaril y in areas of recent deposition with a groundwater tabl e
l essthan 3 m (10 ft) deep.50 Abundant effects incl ude sand boil s with aprons up to
3 m (10 ft) in diameter that commonl y coal esce intobands al ong fissures, fissures
with widths up to 1.5 m (4.5 ft), fissures general l y paral l el or curve toward
streams or depressions and commonl y break in mul tipl e strands, fences and roadways
are offset or pul l edapart as much as 1.5 m (4.5 ft) in some pl aces, ground
settl ements of more than I ft (0.3 m) occur l ocal l y,sl umps with a meter of
displ acement are common in steep stream banks.70 Abundant effects incl ude many
l arge sand boil s [some with aprons exceeding 6 m (20 ft) in diameter thatcommonl y
coal esce al ong fissures], l ong fissures paral l el to rivers or shorel ines, usual l y
in mul tipl e strandswith many openings as wide as 2 m (6 ft), many l arge sl umps
al ong streams and other steep banks, someintact masses of ground between fissures
displ aced I to 2 m (3 to 6 ft) down gentl e sl opes, frequent groundsettl ements of
more than 0.3 m (I ft).90 Very abundant ground effects incl ude numerous sand boil s
with l arge aprons, 30% or more of some areascovered with freshl y deposited sand,
many l ong fissures with mul tipl e paral l el streams and shorel ines withopenings as
wide as 2 m or more, some intact masses of ground between fissures are horizontal l y
displ aced acoupl e of meters down gentl e sl opes, l arge sl umps are common in stream
and other steep banks, groundsettl ements of more than 0.3 m (I ft) are common.
Source: After Youd and Perkins (1987).rl im-------~ SsuFigure 10.25 Stress, strain,
and geometric notation for deformation estimation model of Byrne (1991). El astic
perfectl y pl astic approximation to stress-strain behaviorassumes that residual
strength is mobil ized at l imiting shear strain.Sec. 10.6 Seismic Sl ope Stabil ity
Anal ysis 457val ues OfYl im are presented in Tabl e 10-6. Displ acements predicted by
equations (10.22) agreewel l with those of equation (10.20) for sl opes fl atter than
about 3% and (N\)60= 4. For higher(N\)60 val ues, equations (10.22) predict
considerabl y smal l er displ acements.Tabl e 10-6 AverageVal ues of Limiting ShearStrain
for Cl ean Sand"Yl im468101216203040501.000.800.630.500.400.250.160.050.015oSource:
Seed et al . (1985).Byrne et al . (1992) extended this approach to determine factors
by which the initial stiffness of a soil shoul d be reduced for finite-el ement
anal ysis of deformation fail ures.Deformations predicted by this approach were in
good agreement with those observed in the1971 fail ure of Upper San Fernando Dam
(Figure 10.26).Liquefiedsoil 'V ~T - - - - • """,- - - -IIII IIIOriginal
configuration(a) Final configuration- v - - - - - -~-~ 'r- ---- - -fO': >.(b)Figure
10.26 (a) Finite-el ement mesh for anal ysis of Upper San Fernando Dam withel ements
determined to have l iquefied by Serff et al . (1976) shaded; (b) positions of
original and final meshes (displ acements exaggerated by factor of 2) by procedure
ofByrne et al . (1992). Note l arge shear strains in l iquefied zones.458 Seismic
Sl ope Stabil ity Chap. 10Exampl e 10.6The gentl y sl oping site shown bel ow consists of
a 2 m thick l ayer of sil ty cl ay overl ying a 4 mthick l ayer of l oose, saturated
sand. The sand has an average fines content of about 3% and anaverage Dso=0.22 mm.
Subsurface investigations indicate that the corrected SPT resistance ofthe sand is
quite consistent with an average val ue of II. Estimate the permanent displ acementof
the sl ope when subjected to earthquake shaking sufficient to cause l iquefaction of
the sand.=- 2m4mFigure EIO.6Sil ty cl ay Pd 1.92 Mg/m3SandPsat 1.85 Mg/m3(10.23)
Sol ution The static shear stress at the center of the l iquefiabl e l ayer is'Cst = o,
sino; = [(2 m)(1.92~~)~.8Is: 2)+ (2 m)(1.85~~)(9.81 S: 2)J= 5.2 kPaFrom Tabl e 10.6,
'Yl im =0.45 and from Figure 9.57, Ssu = 300 psf =14.4 kPa. Assuming that thesl ope
has no initial vel ocity (vo = 0), the direct sol ution of Equation 1O.22bgivesD = 0
+ 4(14.4 kPa)(0.45)(4 m) = 1.88 m6( 14.4 kPa - 5.2 kPa)Because this displ acement is
l ess than 'Yl im Tv the permanent displ acement must be determinedusing the cubic
equation of Equation 1O.22aD3( 14.4 kPa) _3[(0.45)(4 m)]-D(5.2 kPa)-O - 0from which
D = 1.87 mThus, the estimated permanent displ acement woul d be about 1.9 m. Note
that this estimate isbased on an average val ue ofthe residual strength of the
l iquefied soil ; considering the range ofuncertainty of that strength (Figure 9.58),
the actual permanent displ acement coul d be considerabl ysmal l er or l arger.Baziar et
al . Approach. Using a sl iding bl ock anal ysis to describe fundamental aspects of
seismic sl ope stabil ity, Baziar et a1. (1992) devel oped a general expressionfor
permanent l ateral displ acementd = Nv'fnax f(~)arnax arnaxwhere N is the equival ent
number of cycl es of harmonic l oading, Vrnax is the peak horizontal vel ocity, arnax
is the peak horizontal accel eration, and ayis the yiel d accel eration. The functionf
(ayl arnax) was obtainedby assuming harmonic accel erations (Figure 10.27).
Cal ibrationvSec. 10.6 Seismic Sl ope Stabil ity Anal ysis1 --0.10.001 __0.0001 L-
_-'-----_--'-----_-'-_~_ ___'0.00 0.20 0.40 0.60 0.80 1.00..!!La maxFigure 10.27
Variation off(ay/ama, ) with a/amax. (After Baziar et al ., 1992.)459against case
histories from the western United States suggested the use of N = 2 for 5.0: 5: Mw: 5:
7.7. By assuming a yiel d accel eration representative of those associated with the
case historydatabase ofYoud and Perkins (1987), Baziar et al . (1992) were abl e to
compare displ acementspredicted by equation (10.23) with the corresponding LSI
val ues. As shown inFigure 10.28, the two approaches are quite consistent at l onger
site distances, but l ess so atshorter distances. Until additional near-source data
becomes avail abl e, the physical basis ofequation (10.23) appears to provide a
stronger basis than LSI for estimation of displ acements.1000 : -c: : ;: : ;.10Baziar et
al . (1992)Youd and Perkins(1987)100 1000Source-site distance (km)Figure 10.28
Comparison of permanent displ acements with LSI. (After Baziar et al ., 1992.)460
Seismic Sl ope Stabil ity Chap. 10Bartl ett and Youd Approach. Bartl ett and Youd
(1992) used a l arge databaseof l ateral spreading case histories to devel op
empirical expressions rel ating l ateral ground displ acement to a number of source and
site parameters. The database incl uded sitesfrom the western United States and
Japan at source-site distances up to 90 km subjected toearthquakes ranging from Mw
= 6.4 to Mw = 9.2. Regression anal yses were used to identifythe factors that most
strongl y infl uenced l ateral ground displ acements, so that the empirical model coul d
be based on those factors.Two empirical model s were devel oped: afree-face model for
sites near steep banksand a ground-sl ope model for gentl y sl oping sites. For free-
face sites, displ acements can beobtained froml ogDH = -16.3658 + 1.1782Mw -
0.92751ogR - 0.0133R + 0.6572l ogW(10.24)+ 0.34831ogT15 + 4.572Ol og(100 - F15 ) -
0.9224(D50 ) 15where DH is the estimated l ateral ground displ acement in meters, Mw
the moment magnitude,R the horizontal distance from the seismic energy source in
kil ometers, W the ratio ofthe height of the free face to the horizontal distance
between the base of the free face and thepoint of interest (Figure 10.29), T15 the
cumul ative thickness of saturated granul ar l ayerswith (NI ) 60 < 15 in meters, FI 5
the average fines content for the granul ar l ayers comprisingTI 5 in percent, and
(D50) 15 the average mean grain size for the granul ar l ayers comprisingTI 5 in
mil l imeters. For gentl y sl oping sites, the ground-sl ope model predictsl ogDH =
-16.3658 + 1.1782Mw - 0.92751ogR - 0.0133R + 0.42931ogS(10.25)+ 0.34831ogT15 +
4.572Ol og(100 - F15 ) - 0.9224 (D 50 ) 15where S is the ground sl ope in percent
(Figure 10.29). Appl ication of these equations to thecase history database showed
that 90% of the observed displ acements were within a factor1/ST£--L---.~.7;»I(;
SiteL = Distance from toe of free face to siteH =Height of free face (crest el ev. -
toe el ev.)W =Free-face ratio =(H/L) (100), in percentS =Sl ope of natural ground
toward channel in percentFigure 10.29 Parameters describing sl ope geometry for
free-face and ground-sl opedeformation model s. L, distance from toe of free face to
site under consideration; H, heightof free face (crest el ev. - toe el ev.); W, free-
face ratio = (HIL)(100), n percent; S, sl ope ofnatural ground toward chaonel =I/X·
100, in percent. (After Bartl ett aod Youd, 1992.)Sec. 10.6 Seismic Sl ope Stabil ity
Anal ysis 461of 2 of the val ues predicted. The ranges of input parameters for which
predicted resul ts areverified by case history observations are shown in Tabl e 10-7.
Tabl e 10-7 Range of Parameter Val ues for Which Equations(10.24) and (10.25) Can Be
Appl iedInput ParameterMagnitudeFree-face ratioThickness ofl oose l ayerFines content
Mean grain sizeGround sl opeDepth to bottom of sectionRange of Val ues6.0<Mw< 8.0
1.0%< W<20%0.3 m < TIS < 12 mO%<F1S<SO%0.1 mm < (DSO)IS < 1.0 mm0.I%<S<6%Depth to
bottom ofl iquefied zone < IS mSource: After Bartl ett and Youd (1992).Exampl e 10.7
Estimate the permanent displ acement of the sl ope described in Exampl e 10.6 due to
Mw =6.5and Mw = 7.5 earthquakes occurring at a (horizontal source-site) distance of
30 km.Sol ution From the description in Exampl e 10.6, the ground-sl ope model
ofBartl ett and Youdis most appropriate. The rel evant parameters areS = 4TIS = 4F 1S
= 3(Dso>ts = 0.22Then, the permanent displ acement due to the Mw = 6.5 earthquake
can be estimated from equation10.25 as10gD = - 16.3658 + (1.1782)( 6.5) - 0.9275
l og(30) - 0.0133(30) + 0.4293 l oge 4)+ 0.3483 l oge 4) + 4.5270 l og (100 - 3) -
0.9224( 0.22) = -1.217soD = 10- 1 217 = 0.061 m = 6.1 emFor the Mw =7.5 earthquake,
10gD = - 16.3658 + (1.1782)(7.5) - 0.9275 l og(30) - 0.0133(30) + 0.4293 l oge 4)+
0.3483 l og (4) + 4.5270 10g(l 00 - 3) - 0.9224(0.22) = -0.039soD = 10-0.039 = 0.91 m
= 91 cmDiscussion. The preceding sections presented a variety of methods for
estimationof the permanent displ acements produced by deformation fail ures. Most of
these methodsare highl y empirical , and al l produce onl y approximate estimates of
permanent462 Seismic Sl ope Stabil ity Chap. 10displ acements. The appl icabil ity of
each method to a particul ar site depends on the simil aritybetween the conditions at
that site and those corresponding to the databases from whichthe method was
devel oped.10.7 SUMMARY1. Historical l y, earthquake-induced l andsl ides have been
among the most damaging ofal l seismic hazards. Their characteristics are infl uenced
by geol ogic, hydrol ogic,topographic, cl imatic, weathering, and l and-use conditions.
Sl ides can be cl assifiedon the basis of material type, type of movement, degree of
internal disruption, watercontent, vel ocity, and depth. Earthquake-induced
l andsl ides are usual l y divided intothree main categories: disrupted sl ides and
fal l s, coherent sl ides, and l ateral spreadsand fl ows.2. Anal ysis of historical data
al l ows estimation of the minimum earthquake magnituderequired to produce different
types of l andsl ides and of the maximum distance towhich l andsl ides can be expected
in earthquakes of different magnitudes.3. A sl ope stabil ity anal ysis is onl y one
part of a comprehensive eval uation of sl ope stabil ity.Prior to the anal ysis,
detail ed information on geol ogic, hydrol ogic, topographic,geometric, and material
characteristics must be obtained. The accuracy ofthe anal ysis wil l be onl y as good
as the accuracy of this information.4. The dynamic shear stresses produced by
earthquake shaking represent a source ofl oading and may al so infl uence the strength
and stress-strain behavior of the sl opematerial s. Seismic sl ope instabil ities may
be grouped into two categories on the basisof which of these effects is predominant
in a given sl ope. Inertial instabil ities arethose in which the shear strength of
the soil remains essential l y constant and sl opedeformations are caused by its
temporary exceedance by dynamic earthquakestresses. Weakening instabil ities occur
when the earthquake serves to weaken the soil sufficientl y that it cannot remain
stabl e under earthquake-induced stresses.5. Inertial instabil ities are most
commonl y anal yzed by pseudostatic, sl iding bl ock, orstress-deformation anal yses.
The Makdisi-Seed approach, based on the resul ts ofsl iding bl ock anal yses, is al so
used frequentl y.6. Pseudostatic anal yses represent the effects of an earthquake by
appl ying static horizontal and/or vertical accel erations to a potential l y unstabl e
mass of soil . The inertial forces induced by these pseudostatic accel erations
increase the driving forces andmay decrease the resisting forces acting on the
soil . Pseudostatic anal yses are notappropriate for soil s that buil d up l arge pore
pressures or show more than about 15%degradation of strength due to earthquake
shaking. Stabil ity is expressed in terms ofa pseudostatic factor of safety
cal cul ated by l imit equil ibrium procedures. Sel ectionof an appropriate pseudostatic
accel eration requires great care; val ues considerabl ysmal l er than the peak
accel eration of the sl iding mass are usual l y used.7. The pseudostatic accel eration
required to bring a sl ope to the point of incipient fail ureis known as the yiel d
accel eration. If earthquake-induced accel erations in a sl opeChap. 10 Homework
Probl ems 463momentaril y exceed the yiel d accel eration, the unstabl e soil wil l
momentaril yaccel erate rel ative to the material beneath it. Sl iding bl ock anal yses
can be used to cal cul atethe amount of displ acement that occurs. The total
displ acement depends on theamount by which the yiel d accel eration is exceeded (a
function of the ground motionampl itude), the time over which the yiel d accel eration
is exceeded (a function of thefrequency content of the ground motion), and the
number of times the yiel d accel erationis exceeded (a function of ground motion
duration). Given the highl y variabl enature of ground motion characteristics,
computed displ acements can be quite variabl e.8. The Makdisi-Seed procedure is based
on sl iding bl ock anal yses of earth dams andembankments. Knowing the fundamental
period of vibration of the dam/embankmentand the yiel d accel eration of the sl ope,
simpl e charts can be 'used to estimate earthquake-induced permanent displ acements.
9. Stress-deformation anal yses have been used to estimate permanent deformations
caused by inertial instabil ities. Strain potential and stiffness reduction
approachesal l ow estimation of permanent deformations from rel ativel y simpl e
anal yses; theirestimates are highl y approximate. Al though the computational effort
is dramatical l yincreased, permanent deformations can be anal yzed more rigorousl y
using nonl inearfinite-el ement techniques. As the accuracy of constitutive model s
for soil s improve,the use of nonl inear finite-el ement anal yses is l ikel y to
increase.10. Weakening instabil ities occur when earthquake-induced stresses and
strains reducethe shear strength of the soil within a sl ope. Depending on whether
the reducedstrength is greater than or l ess than the stresses required to maintain
static equil ibrium,weakening instabil ities may be cl assified as fl ow fail ures or
deformation fail ures.l l . Fl ow fail ure instabil ity is usual l y eval uated by l imit
equil ibrium anal ysis. Residual strengths are appl ied to those portions of the
fail ure surface that pass through l iquefiedsoil . A factor of safety l ess than
1suggests that fl ow fail ure is l ikel y. Simpl e l imitequil ibrium anal yses combined
with constant-vol ume constraints can be used to estimatethe distance over which
material s travel in fl ow fail ures. Fl uid mechanics model shave al so been used to
estimate fl ow fail ure deformations. Nonl inear dynamicanal yses that al l ow weakening
of l iquefied el ements and l arge strains have al so beendevel oped.12. The effects of
deformation fail ures are usual l y expressed in terms of sl ope deformations.A number
of approaches, ranging from purel y empirical to sl iding bl ock model based, have been
devel oped to estimate the displ acements produced by deformationfail ures.HOMEWORK
PROBLEMS10.1 The sl ope shown bel ow is intersected by two 6-inch-thick seams of
cl ayey material . The intactsl ope material s can be characterized by the parameters,
c = 2500 psf, <I> = 0, y = 130 pet. Thecl ayey seams exhibit c = 750 psf, <I> = 0,
and y= 120 psf. Compute the minimum static factor ofsafety for the sl ope.46425 ft25
ftSeismic Sl ope Stabil ity125 ft 25 ft 25 ft 25 It .. .+ "I" -I" -IFigure PIO.l Chap.
1010.2 Compute the minimum pseudostatic factor of safety for the sl ope of Probl em
Ia.l assuming apseudostatic coefficient of a.l g.10.3 Compute the yiel d accel eration
for the sl ope of Probl em 10.1.10.4 Using hand cal cul ations, sl ope stabil ity charts,
or a sl ope stabil ity anal ysis computer program,l ocate the critical circul ar fail ure
surface for the sl ope shown bel ow. Considering onl y this fail uresurface, compute
the yiel d accel eration for the sl ope.40 ft20 ft20 ftFigure PIO.4c= 0<I> = 35°P=1.70
Mg/m310.5 The sl ope shown in Probl em 10.4 is subjected to the time history of crest
accel eration shownbel ow. Compute the permanent displ acement of the sl ope.§ 0.3co.~
Q) 0.0(jjo~-0.3Figure PIO.5Time - sec10.6 An existing embankment is determined to
have a yiel d accel eration of a.2g. Using the rel ationshipsof Ambraseys and Menu
(equation 10.14), estimate the probabil ity that an earthquakeChap. 10 Homework
Probl ems 465that produces a peak accel eration of 0.3g woul d cause a permanent sl ope
displ acement greaterthan 2.5 ern.10.7 A sl ope in cohesive soil is determined to
have a yiel d accel eration ofO.17g. Assuming Neq = 10and using the rel ationship
ofYegian
et al . (equation 10.15), compute the expected val ue of permanentsl ope displ acement
if the sl ope was subjected to (a) the Gil roy No. I (rock) motion, and(b) the Gil roy
No.2 (soil ) motion. (Note: Peak accel erations and predominant periods of these
motions were computed in the exampl e probl ems in Chapter 3.)10.8 Using the
rel ationship of Jibson (equation 10.16), repeat Probl em 10.7. Then compute the
permanentsl ope displ acement that woul d have a 5 percent probabil ity of being
exceeded for eachground motion.10.9 The fundamental frequency of the earth dam in
Exampl e 7.6 was computed as 3.1 Hz.Pseudo-static sl ope stabil ity anal yses indicate
a yiel d accel eration of 0.2Ig. Use theMakdisi-Seed procedure to estimate the
permanent displ acement of the dam in aM= 7 earthquakethat produces a peak
accel eration of O.28g.10.10 The sl ope shown bel ow consists of 5 m of l oose, cl ean
sand overl ying very dense cl ayey gravel with a groundwater tabl e 2 m bel ow the
ground surface. Using the Byrne approach, estimatethe permanent displ acements that
woul d occur if the sl ope was subjected to earthquake shakingstrong enough to
initiate l iquefaction of the l oose sand.3°2m3me= 0.72(N1) 60 =8Figure PIO.IO10.11
Using the approach of Baziar et aI., estimate the permanent displ acements that
woul d haveoccurred in identical sl opes with yiel d accel erations of 0.26g if
subjected to (a) the Gil roy No.I (rock) motion, and (b) the Gil roy No.2 (soil )
motion. Comment on the useful ness of peakaccel eration as a sol e indicator of
potential sl ope deformations.10.12 The sl ope shown bel ow consists of a l oose sil ty
sand overl ying stiff cl ay. Estimate the l ateral spreading displ acement that woul d
devel op if a Mw =7.3 earthquake occurred at a distance (toseismic source) of 40 km.
6°1m4mFigure PIO.12Sil ty sandfines contents =14%050 = 0.10mm11Seismic Designof
Retaining Wal l s11.1 INTRODUCTIONEarth retaining structures, such as retaining
wal l s, bridge abutments, quay wal l s, anchored bul kheads,braced excavations, and
mechanical l y stabil ized wal l s, are used throughout seismical l yactive areas. They
frequentl y represent key el ements of ports and harbors, transportation systems,
l ifel ines, and other constructed facil ities. Earthquakes have caused permanent
deformationof retaining structures in many historical earthquakes. In some cases,
these deformationswere negl igibl y smal l ; in others they caused significant damage.
In some cases, retaining structureshave col l apsed during earthquakes, with
disastrous physical and economic consequences.This chapter discusses the behavior
of retaining wal l s during earthquakes and presents several of the most common
approaches to the seismic design of different types of retaining wal l s.11.2 TYPES
OF RETAINING WALLSThe probl em of retaining soil is one of the ol dest in
geotechnical engineering; some of theearl iest and most fundamental principl es of
soil mechanics were devel oped to al l ow rational design of retaining wal l s. Many
different approaches to soil retention have been devel opedand used successful l y. In
recent years, the devel opment of metal l ic, pol ymer, and geotextil ereinforcement has
l ed to the devel opment of many innovative types of mechanical l y stabil izedearth
retention systems.466Sec. J 1.3 Types of Retaining Wal l Fail ures 467Gravity wal l
.....,=1~,. .Cantil ever wal l Cantil ever wal l Reinforced soil wal l ~J ' ..8,.: .-
_.,..' -...-, : .. ' ,..'....1W"'~Basement wal l Bridge abutment wal l Anchored
bul khead Tieback wal l Figure 11.1 Common types of earth retaining structures.
Retaining wal l s are often cl assified in terms of their rel ative mass, fl exibil ity,
andanchorage conditions. Gravity wal l s (Figure 11.1) are the ol dest and simpl est
type of retainingwal l . Gravity wal l s are thick and stiff enough that they do not
bend; their movementoccurs essential l y by rigid-body transl ation and/or rotation.
Certain types of composite wal l systems, such as crib wal l s and mechanical l y
stabil ized wal l s, are thick enough that theybend very l ittl e and consequentl y are
often designed as gravity wal l s (with appropriate considerationof internal
stabil ity). Cantil ever wal l s, which bend as wel l as transl ate and rotate,rel y on
their fl exural strength to resist l ateral earth pressures. The actual distribution
of l ateral earth pressure on a cantil ever wal l is infl uenced by the rel ative
stiffness and deformationof both the wal l and the soil . Braced wal l s are
constrained against certain types ofmovement by the presence of external bracing
el ements. In the cases of basement wal l s andbridge abutment wal l s, l ateral
movements of the tops of the wal l s may be restrained by thestructures they support.
Tieback wal l s and anchored bul kheads are restrained against l ateral movement by
anchors embedded in the soil behind the wal l s. The provision of l ateral supportat
different l ocations al ong a braced wal l may keep bending moments so l ow that
rel ativel yfl exibl e structural sections can be used.11.3 TYf'ES OF RETAINING WALL
FAILURESTo design retaining wal l s, it is necessary to define "fail ure" and to know
how wal l s can fail .Under static conditions, retaining wal l s are acted upon by body
forces rel ated to the mass ofthe wal l , by soil pressures, and by external forces
such as those transmitted by braces. Aproperl y designed retaining wal l wil l achieve
equil ibrium of these forces without inducingshear stresses that approach the shear
strength of the soil . During an earthquake, however,inertial forces and changes in
soil strength may viol ate equil ibrium and cause permanentdeformation of the wal l .
Fail ure, whether by sl iding, til ting, bending, or some other mechanism,occurs when
these permanent deformations become excessive. The question of whatl evel of
deformation is excessive depends on many factors and is best addressed on a
sitespecificbasis.Gravity wal l s usual l y fail by rigid-body mechanisms such as
sl iding and/or overturningor by gross instabil ity (Figure 11.2). Sl iding occurs
when horizontal force equil ibriumis not maintained (i.e., when the l ateral
pressures on the back of the wal l produce a thrust468..-r ----IIIIIIAr-rI, I,I\
Seismic Design of Retaining Wal l s,,-, ,,,,,,- -...... ,Chap. 11(a) (b)
-----------------""--- (c)Figure 11.2 Typical fail ure mechanisms for a gravity
retaining wal l : (a) sl iding(transl ational ) fail ure; (b) overturning (rotational )
fail ure; (c) gross instabil ity fail ure.that exceeds the avail abl e sl iding
resistance on the base of the wal l ). Overturning fail uresoccur when moment
equil ibrium is not satisfied; bearing fail ures at the base of the wal l areoften
invol ved. Gravity wal l s may al so be damaged by gross instabil ity of the soil s
behindand beneath them. Such fail ures may be treated as sl ope stabil ity fail ures
that encompass thewal l . Composite wal l systems, such as crib wal l s, bin wal l s, and
mechanical l y stabil izedwal l s, can fail in the same ways or by a number of internal
mechanisms that may invol veshearing, pul l out, or tensil e fail ure of various wal l
el ements.Cantil ever wal l s are subject to the same fail ure mechanisms as gravity
wal l s, and al soto fl exural fail ure mechanisms. Soil pressures and bending moments
in cantil ever wal l sdepend on the geometry, stiffness, and strength of the wal l -soil
system (Figure 11.3a,b;pressure and moment diagrams for typical wal l ). If the
bending moments required for equil ibriumexceed the fl exural strength of the wal l ,
fl exural fail ure may occur (Figure 11.3c).The structural ductil ity of the wal l
itsel f may infl uence the l evel of deformation produced byfl exural fail ure.Figure
11.3 (a) Soil pressures, (b) bending moments, and (c) fl exural fail uremechanism for
cantil ever retaining wal l .Braced wal l s usual l y fail by gross instabil ity, til ting,
fl exural fail ure, and/or fail ure ofbracing el ements. Til ting of braced wal l s
typical l y invol ves rotation about the point atwhich the brace acts on the wal l ,
often the top of the wal l as in the cases of basement andbridge abutment wal l s
(Figure II.4a). Anchored wal l s with inadequate penetration may til tby "kicking out"
at their toes (Figure II.4b). As in the case of cantil ever wal l s, anchoredwal l s may
fail in fl exure, al though the point of fail ure (maximum bending moment) is l ikel yto
be different. Fail ure of bracing el ements can incl ude anchor pul l out, tierod
fail ure, orSec. 11.4 Static Pressures on Retaining Wal l s 469r·············: ··· ··
··· · · · r······.·...···.··.··-'-·,··.·.• J . I .. - . ..• • j .,: ,." .. I ,: ',-..
, ,~ ,(a) (b) (c)Figure 11.4 Potential modes of fail ure for braced wal l s: (a)
rotation of bridge abutmentabout top; (b) rotation of anchored bul khead due to l ack
of passive resistance ("kickout")at the toe; (c) l ack of adequate anchor capacity.
bridge buckl ing. Backfil l settl ements can al so impose additional axial and
transverse l oadingon bracing el ements such as tierods and tiebacks.11.4 ST}~TIC
PRESSURES ON RETAINING WALLSThe seismic behavior of retaining wal l s depends on the
total l ateral earth pressures thatdevel op during earthquake shaking. These total
pressures incl ude both the static gravitational pressures that exist before an
earthquake occurs, and the transient dynamic pressuresinduced by the earthquake.
Since the response of a wal l is infl uenced by both, a brief reviewof static earth
pressures is presented.Static earth pressures on retaining structures are strongl y
infl uenced by wal l and soil movements. Active earth pressures devel op as a retaining
wal l moves away from the soil behind it, inducing extensional l ateral strain in the
soil . When the wal l movement is sufficientto ful l y mobil ize the strength of the
soil behind the wal l , minimum active earth pressuresact on the wal l . Because very
l ittl e wal l movement is required to devel op minimumactive earth pressures (for the
usual case of cohesionl ess backfil l material s), free-standingretaining
wal l s are usual l y designed on the basis of minimum active earth pressures. Where
l ateral wal l movements are restrained, such as in the cases of tieback wal l s,
anchored bul kheads,basement wal l s, and bridge abutments, static earth pressures may
be greater thanminimum active. Passive earth pressures devel op as a retaining wal l
moves toward the soil ,thereby producing compressive l ateral strain in the soil .
When the strength of the soil is ful l ymobil ized, maximum passive earth pressures
act on the wal l . The stabil ity of many freestandingretaining wal l s depends on the
bal ance between active pressures acting predominantl yon one side of the wal l and
passive pressures acting on the other.Even under static conditions, prediction of
actual retaining wal l s forces and deformationsis a compl icated soil -structure
interaction probl em. Deformations are rarel y consideredexpl icitl y in design-the
typical approach is to estimate the forces acting on a wal l andthen to design the
wal l to resist those forces with a factor of safety high enough to produce
acceptabl y smal l deformations. A number of simpl ified approaches are avail abl e to
eval uatestatic l oads on retaining wal l s. The most commonl y used are described in
the fol l owingsections.11.4.1 Ranl dne TheoryRankine (1857) devel oped the simpl est
procedure for computing minimum active and maximumpassive earth pressures. By
making assumptions about the stress conditions and470 Seismic Design of Retaining
Wal l s Chap. J Jstrength envel ope of the soil behind a retaining wal l (the backfil l
soil ), Rankine was abl e torender the l ateral earth pressure probl em determinate and
directl y compute the static pressuresacting on retaining wal l s.For minimum active
conditions, Rankine expressed the pressure at a point on the backof a retaining
wal l as(11.1)where KA is the coefficient ofminimum active earth pressure, (j~ is
the vertical effectivestress at the point of interest, and c is the cohesive
strength of the soil . When the principal stress pl anes are vertical and horizontal
(as in the case of a smooth vertical wal l retaining ahorizontal backfil l ), the
coefficient of minimum active earth pressure is given byK = 1 - sin <I> = tan2(45 -
<1» (11.2)A 1 + sin <\> 2For the case of a cohesionl ess backfil l incl ined at an
angl e ~ with the horizontal , infinitesl ope sol utions can be used (Terzaghi, 1943;
Tayl or, 1948) to compute KA asK_ Rcos~-Jcos2~-cOS2<\>A - cos I-' (11.3)cos ~ +
JCOS2~ - cos2<\>for ~ ~ <\> [equation (11.3) is equival ent to equation (11.2) when
~ = 0]. The pressure distributionon the back of the wal l , as indicated by equations
(11.1), depends on the rel ativemagnitudes of the frictional and cohesive components
of the backfil l soil strength (Figure11.5). Al though the presence of cohesion
indicates that tensil e stresses wil l devel opbetween the upper portion of the wal l
and the backfil l , tensil e stresses do not actual l ydevel op in the fiel d. The creep,
stress rel axation, and l ow-permeabil ity characteristics ofcohesive soil s render
them undesirabl e as backfil l material for retaining structures, and theiruse in
that capacity is general l y avoided whenever possibl e. For dry homogeneous
cohesionl essbackfil l , Rankine theory predicts a triangul ar active pressure
distribution orientedparal l el to the backfil l surface. The active earth pressure
resul tant, PA,acts at a point l ocatedH/3 above the base of a wal l of height, H
(Figure 11.5a) with magnitudePA = ~KAYH2 (11.4)Under maximum passive conditions,
Rankine theory predicts wal l pressures given byPP = Kp(5~ + 2cJ"i<; (11.5)where Kp
is the coefficient ofmaximum passive earth pressure. For smooth, vertical wal l s
retaining horizontal backfil l s,K p = l +sin<\> = tan2(45 +<1» (11.6)1- sin <\> 2and
(11.7)Sec. 11.4 Static Pressures on Retaining Wal l sz Height of / 1 tension zone"/
y, c/ $= 0/'- Fail ure/ / surfacePA471KA =tan2 (45 -l \J/2)PA =KAyzPA =KAyH2/2(a)Zo=
2c/yPA=yz-2c2c 2PA=yH2/2 -2cH+y(b)Zo = (~}an(45 + l \J/2)PA =yztan2(45 -l \J/2) - 2c
tan(45 -l \J/2)PA =(y~2Jtan2(45 -l \J/2) - 2cH tan(45 -l \J/2)+2c2/y(c)Figure 11.5
Minimum Rankine active earth pressure distributions for backfil l s withvarious
combinations of frictional and cohesive strength: (a) frictional resistance, no
cohesion; (b) cohesive soil , no frictional resistance; (c) combined cohesion and
friction.(After NAVFAC, 1982.)for backfil l s incl ined at ~ to the horizontal .
Passive pressure distributions for various backfil l strength characteristics are
shown in Figure 11.6. For a dry homogeneous backfil l ,Rankine theory predicts a
triangul ar passive pressure distribution oriented paral l el to thebackfil l surface.
The passive earth pressure resul tant, or passive thrust, Pp , acts at a point
l ocated H/3 above the base of a wal l of height H (Figure 11.6a) with magnituder, =
~KpyH2 (11.8)The presence of water in the backfil l behind a retaining wal l
infl uences the effectivestresses and hence the l ateral earth pressure that acts on
the wal l . For wal l design the hydrostaticKp =tarr' (45 + l \J/2)Pp= KpYzPp= Kp yH2/2
(a)Pp=Yz +2cPp= 1/2yH2 + 2cH(b)Pp = yztan2(45+ l \J/2) +2ctan(45+l \J/2)pp=
(y~2Jtan2(45+l \J/2) +2cHtan(45 + l \J/2)(c)Figure 11.6 Maximum Rankine passive earth
pressure distributions for backfil l s withvarious combinations of frictional and
cohesive strength: (a) frictional resistance, nocohesion; (b) cohesive soil , no
frictional resistance; (c) combined cohesion and friction.(After NAVFAC, 1982.)472
Seismic Design of Retaining Wal l s Chap. J Jpressure due to the water must be added
to the l ateral earth pressure. Because the total l ateral thrust on a wal l retaining
a saturated backfil l is considerabl y greater than that on a wal l retainingdry
backfil l , the provision of backfil l drainage is an important part of retaining wal l
design.11.4.2 Coul omb TheoryCoul omb (1776) was the first to study the probl em of
l ateral earth pressures on retainingstructures. By assuming that the force acting
on the back of a retaining wal l resul ted fromthe weight of a wedge of soil above a
pl anar fail ure surface, Coul omb used force equil ibriumto determine the magnitude of
the soil thrust acting on the wal l for both minimumactive and maximum passive
conditions. Since the probl em is indeterminate, a number ofpotential fail ure
surfaces must be anal yzed to identify the critical fail ure surface (i.e., the
surface that produces the greatest active thrust or the smal l est passive thrust).
Under minimum active earth pressure conditions, the active thrust on a wal l with
thegeometry shown in Figure l 1.7a is obtained from force equil ibrium (Figure
l 1.7b). For thecritical fail ure surface, the active thrust on a wal l retaining a
cohesionl ess soil can beexpressed as(11.9)(11.10)wherecos28 cos(o + 8) [1 + sin(o +
<jJ)sin(<jJ - ~) ] 2cos(o + 8)cos(~ - 8)o is the angl e of interface friction
between the wal l and the soil (Tabl e 11-1), and ~ and 8 areas shown in Figure
11.7a. The critical fail ure surface is incl ined at an angl e(11.11)to the horizontal
whereC1 Jtan (<jJ - ~) [tan (<jJ - ~) + cot (<jJ - 8)] [1 + tan (0 + 8) cot (<jJ -
8)]C2 1 + {tan (0 + 8) [tan(<jJ -~) + cot (<jJ - 8)] }(a) (b)wFigure 11.7 (a)
Triangul ar active wedge bounded by pl anar backfil l surface, fail uresurface, and
wal l ; (b) force pol ygon for active Coul omb wedge. The critical fail uresurface is
that which gives the l argest val ue of PA'Sec. 11.4 Static Pressures on Retaining
Wal l s 473Tabl e 11-1 Typical Interface Friction Angl esInterface Material sInterface
Friction Angl e 0Mass concrete against: cl ean sound rockcl eau gravel , gravel -sand
mixtures, coarse sandcl ean fine to medium sand, sil ty medium to coarse sand,sil ty
or cl ayey gravel cl ean fine saud, sil ty or cl ayey fine to medium sandfine sandy
sil t, nonpl astic sil tFormed concrete against: medium-stiff aud stiff cl ay and sil ty
cl aycl ean gravel , gravel -sand mixture, wel l -graded rock fil l with spal l scl ean sand,
sil ty sand-gravel mixture, singl e-size hardrock fil l Steel sheet pil es against:
sil ty sand, gravel ,or sand mixed with sil t or cl ayfine sandy sil t, nonpl astic sil t
cl ean gravel , gravel -sand mixture, wel l -graded rock fil l with spal l scl ean sand,
sil ty sand-gravel mixture, singl e-size hardrock fil l sil ty saud, gravel , or saud
mixed with sil t or cl ayfine sandy sil t, nonpl astic sil t2529-3124-2919-2417-1917-19
22-2617-22171422171411Source: After NAVFAC(1982).Coul omb theory does not expl icitl y
predict the distribution of active pressure, but it can beshown to be triangul ar
for l inear backfil l surfaces with no surface l oads. In such cases, PAacts at a
point l ocated H/3 above the height of a wal l of height H.For maximum passive
conditions in cohesionl ess backfil l s (Figure 11.8), Coul ombtheory predicts a
passive thrust(11.12)where(11.13)cos2 (<I> + 8)cos28 cost S - 8) [1 + sin(b +
<1»sin (<I> + ~)J2cos(b - 8)cos(~ - 8)Kp = ------;: ---';=: =: ;========: : : : : ;=: ;-2"
(a) (b)Figure 11.8 (a) Triangul ar passive wedge bounded by pl anar backfil l surface,
fail uresurface, aud wal l ; (b) force pol ygon for passive Coul omb wedge. The critical
fail uresurface is that which gives the l argest val ue of Pp.474 Seismic Design of
Retaining Wal l s Chap. 11The critical fail ure surface for maximum passive earth
pressure conditions is incl ined to thehorizontal atwhere-1 [tan (<I> +~) + C3j Up =
- <I> + tan C4(11.14)C3 Jtan(<I> +~) [tan(<I> +~) + cot (<I> + 8)] [1 + tan (0- 8)
cot (<I> + 8)]C4 = 1+ {tan(0-8)[tan(<I>+~)+cot(<I>+8)]}In contrast to Rankine
theory, Coul omb theory can be used to predict soil thrusts onwal l s with irregul ar
backfil l sl opes, concentrated l oads on the backfil l surface, and seepageforces. By
considering the soil above a potential fail ure pl ane as a free body and incl uding
forces due to concentrated l oads, boundary water pressures, and so on, the
magnitude of theresul tant
thrust (PA or Pp ) can easil y be computed.11.4.3 Logarithmic Spiral MethodAl though
the major principal stress axis may be nearl y perpendicul ar to the backfil l surface
at some distance behind a rough (0 > 0) wal l , the presence of shear stresses on the
wal l -soil interface can shift its position near the back of the wal l . If the
incl ination of the principal stress axes varies within the backfil l , the incl ination
of the fail ure surface must al so vary. Inother words, the fail ure surface must be
curved. A l ogarithmic spiral function has been usedto describe such curved fail ure
surfaces for active and passive earth pressure conditions.For active earth pressure
conditions, the critical fail ure surface consists of a curvedportion near the back
of the wal l and a l inear portion that extends up to the ground surface(Figure
11.9a). The active earth pressure distribution is triangul ar (Figure 11.9b) for
wal l sretaining pl anar, cohesionl ess backfil l s. Thus the active soil thrust can be
expressed in thesame form as equation (11.4), where the l og spiral coefficients of
minimum active earthpressure for various wal l and backfil l incl inations are given
in Tabl e 11-2. The active earthpressure coefficients given by the l og spiral
approach are general l y considered to be sl ightl ymore accurate than those given by
Rankine or Coul omb theory, but the difference is sosmal l that the more convenient
Coul omb approach is usual l y used.(a) (b)Figure 11.9 (a) Logarithmic spiral
representation of the critical fail ure surface forminimum active earth pressure
conditions; (b) orientation of critical fail ure surface fornonvertical wal l with
incl ined backfil l surface.Sec. 11.4 Static Pressures on Retaining Wal l s 475Tabl e
11-2 Val ues for KA for Log-Spiral Fail ure Surface<I>0 ~ e 20° 25° 30° 35° 40° 45°0°
-15° _10° 0.37 0.30 0.24 0.19 0.14 0.110° 0.42 0.35 0.29 0.24 0.19 0.1610° 0.45
0.39 0.34 0.29 0.24 0.210° 0° -10° 0.42 0.34 0.27 0.21 0.16 0.120° 0.49 0.41 0.33
0.27 0.22 0.1710° 0.55 0.47 0.40 0.34 0.28 0.240° 15° _10° 0.55 0.41 0.32 0.23 0.17
0.130° 0.65 0.51 0.41 0.32 0.25 0.2010° 0.75 0.60 0.49 0.41 0.34 0.28-15° _10° 0.31
0.26 0.21 0.17 0.14 0.110° 0.37 0.31 0.26 0.23 0.19 0.1710° 0.41 0.36 0.31 0.27
0.25 0.230° _10° 0.37 0.30 0.24 0.19 0.15 0.120° 0.44 0.37 0.30 0.26 0.22 0.1910°
0.50 0.43 0.38 0.33 0.30 0.2615° -10° 0.50 0.37 0.29 0.22 0.17 0.140° 0.61 0.48
0.37 0.32 0.25 0.2110° 0.72 0.58 0.46 0.42 0.35 0.31Source: After Caquot and
Kerise1(1948).The effect of wal l friction on the shape of the critical fail ure
surface is more noticeabl efor passive earth pressure conditions. The passive
fail ure surface al so has curved andl inear portions (Figure 11.l Oa), but the curved
portion is much more pronounced than foractive conditions. For pl anar cohesionl ess
backfil l s, the passive earth pressure distributionis triangul ar (Figure 11.l Ob), so
the passive thrust can be expressed in the form of equation(11.8), where the l og
spiral coefficients of maximum passive earth pressure are as given inTabl e 11-3.
The passive earth pressure coefficients given by the l og spiral method are~"'~
Curved(a) (b)Figure 11.10 (a) Logarithmic spiral representation of the critical
fail ure surface formaximum passive earth pressure conditions; (b) orientation of
critical fail ure surface fornonvertical wal l with incl ined backfil l surface.476
Seismic Design of Retaining Wal l s Chap. IITabl e 11·3 Val ues for Kp for Log-Spiral
Fail ure Surface<I>s ~ e 20° 25° 30° 35° 40° 45°0° -15° _10° 1.32 1.66 2.05 2.52
3.09 3.950° 1.09 1.33 1.56 1.82 2.09 2.4810° 0.87 1.03 1.17 1.30 1.33 1.540° 0°
-10° 2.33 2.96 3.82 5.00 6.68 9.200° 2.04 2.46 3.00 3.69 4.59 5.8310° 1.74 1.89
2.33 2.70 3.14 3.690° 15° _10° 3.36 4.56 6.30 8.98 12.2 20.00° 2.99 3.86 5.04 6.72
10.4 12.810° 2.63 3.23 3.97 4.98 6.37 8.20<I> -15° -10° 1.95 2.90 4.39 6.97 U.8
22.70° 1.62 2.31 3.35 5.04 7.99 14.310° 1.29 1.79 2.50 3.58 5.09 8.860° _10° 3.45
5.17 8.17 13.8 25.5 52.90° 3.01 4.29 6.42 10.2 17.5 33.510° 2.57 3.50 4.98 7.47
12.0 21.215° _10° 4.95 7.95 13.5 24.8 50.4 U50° 4.42 6.72 10.8 18.6 39.6 73.610°
3.88 5.62 8.51 13.8 24.3 46.9Source: After Caquot and Kerise1 (1948).considerabl y
more accurate than those given by Rankine or Coul omb theory; the Rankineand Coul omb
coefficient tend to underpredict and overpredict the maximum passive earthpressure,
respectivel y. Rankine theory greatl y underpredicts actual passive earth pressures
and is rarel y used for that purpose. Coul omb theory overpredicts passive pressures
(anunconservative error) by about 11%for 8 =<1>/2 and by 100% or more for 8 =<1>.
For that reason,Coul omb theory is rarel y used to eval uate passive earth pressures
when 8 > <1>/2.11.4.4 Stress-Deformation Anal ysisSince the actual pressures that
act on retaining wal l s depend on interaction between the wal l and the surrounding
soil , it seems l ogical to expect that they coul d be estimated by stressdeformation
techniques such as the finite-el ement method. Finite-el ement anal yses are, infact,
very useful for estimating retaining wal l pressures and movements (Cl ough and
Duncan,1971; Duncan et aI., 1990). In addition, they can hel p expl ain unexpected or
anomal ousfiel d measurements of actual wal l behavior (Cl ough and Duncan, 1971;
Duncan andCl ough, 1971).The accuracy of stress-deformation anal yses, however,
depends on how wel l they areabl e to model the actual fiel d conditions. A useful
method of anal ysis shoul d be abl e todescribe the stress-strain behavior of the soil
(which is nonl inear) and wal l (usual l yassumed to remain l inear), the stress-
displ acement behavior of the soil -wal l interface, andthe sequence of wal l
construction and backfil l pl acement. Without careful attention to eachof these
factors, the resul ts of a finite-el ement anal ysis may have l imited appl icabil ity.
Sec. 11.6 Seismic Pressures on Retaining Wal l s 47711.5 DYINAMIC RESPONSE OF
RETAINING WALLSThe dynamic response of even the simpl est type of retaining wal l is
quite compl ex. Wal l movements and pressures depend on the response of the soil
underl ying the wal l , theresponse of the backfil l , the inertial and fl exural
response of the wal l itsel f, and the natureof the input motions. Since few wel l -
documented case histories invol ving fiel d measurementsof wal l response are
avail abl e, most of the current understanding of the dynamicresponse of retaining
wal l s has come from model tests and numerical anal yses. These testsand anal yses,
the majority of which have invol ved gravity wal l s, indicate that: 1. Wal l s can move
by transl ation and/or rotation. The rel ative amounts of transl ationand rotation
depend on the design of the wal l ; one or the other may predominate forsome wal l s
(Nadim and Whitman, 1984), and both may occur for others (Siddharthanet al ., 1992).
2. The magnitude and distribution of dynamic wal l pressures are infl uenced by the
modeof wal l movement (e.g., transl ation, rotation about the base, or rotation about
the top)(Sherif et aI., 1982; Sherif and Fang, 1984a,b).3. The maximum soil thrust
acting on a wal l general l y occurs when the wal l has transl atedor rotated toward the
backfil l (i.e., when the inertial force on the wal l is directedtoward the
backfil l ). The minimum soil thrust occurs when the wal l has transl ated orrotated
away from the backfil l .4. The shape of the earth pressure distribution on the back
of the wal l changes as the wal l moves. The point of appl ication of the soil thrust
therefore moves up and down al ongthe back of the wal l . The position of the soil
thrust is highest when the wal l has movedtoward the soil and l owest when the wal l
moves outward.5. Dynamic wal l pressures are infl uenced by the dynamic response of
the wal l and backfil l and can increase significantl y near the natural frequency of
the wal l -backfil l system(Steedman and Zeng, 1990). Permanent wal l displ acements
al so increase atfrequencies near the natural frequency of the wal l -backfil l system
(Nadim, 1982).Dynamic response effects can al so cause defl ections of different
parts of the wal l to beout of phase. This effect can be particul arl y significant
for wal l s that penetrate into thefoundation soil s when the backfil l soil s move out
of phase with the foundation soil s.6. Increased residual pressures may remain on
the wal l after an episode of strong shakinghas ended (Whitman, 1990).Given these
compl ex, interacting phenomena and the inherent variabil ity and uncertaintyof soil
properties, it is not currentl y possibl e to anal yze al l aspects of the seismic
response of retaining wal l s accuratel y. As a resul t, simpl ified model s that make
variousassumptions about the soil , structure, and input motion are most commonl y
used for seismicdesign of retaining wal l s.11.6 SEISMIC PRESSURES ON RETAINING WALLS
One common approach to the seismic design of retaining wal l s invol ves estimating
thel oads imposed on the wal l during earthquake shaking and then ensuring that the
wal l can478 Seismic Design of Retaining Wal l s Chap. 11resist those l oads. Because
the actual l oading on retaining wal l s during earthquakes isextremel y compl icated,
seismic pressures on retaining wal l s are usual l y estimated usingsimpl ified methods.
11.6.1 Yiel ding Wal l sRetaining wal l s that can move sufficientl y to devel op minimum
active and/or maximumpassive earth pressures are referred to as yiel ding wal l s. The
dynamic pressures acting onyiel ding wal l s are usual l y estimated by pseudostatic
procedures that share many features ofthose described for seismic sl ope stabil ity
anal ysis in Section 10.6.1.1. More recentl y, apseudodynamic procedure that
accounts, in an approximate manner, for the dynamicresponse of the backfil l has
been devel oped.11.6.1.1 Mononobe-Okabe MethodOkabe (1926) and Mononobe and Matsuo
(1929) devel oped the basis of a pseudostaticanal ysis of seismic earth pressures on
retaining structures that
has become popul arl yknown as the Mononobe-Okabe (M-O) method. The M-O method is a
direct extension ofthe static Coul omb theory to pseudostatic conditions. In a M-O
anal ysis, pseudostatic accel erationsare appl ied to a Coul omb active (or passive)
wedge. The pseudostatic soil thrust isthen obtained from force equil ibrium of the
wedge.Active Earth Pressure Conditions. The forces acting on an active wedgein a
dry, cohesionl ess backfil l are shown in Figure 11.l l a. In addition to the forces
that existunder static conditions (Figure 11.7), the wedge is al so acted upon by
horizontal and vertical pseudostatic forces whose magnitudes are rel ated to the mass
of the wedge by the pseudostaticaccel erations ah =khg and a; =kvg. The total active
thrust can be expressed in a formsimil ar to that devel oped for static conditions,
that is,PAE = ~KAEYH2(1 - kv )where the dynamic active earth pressure coefficient,
KAE, is given byK_ cos 2(<jJ - 8- \jf)AE -COS\jfCOS28COS(O+8+\jf)[I+
Sin(O+<jJ)Sin(<jJ-~-\jf)J2cos(O + 8 + \jf)cos(~ - 8)(11.15)(11.16)(a) (b)wFigure
11.11 (a) Forces acting on active wedge in Mononobe-Okabe anal ysis, (b)force
pol ygon il l ustrating equil ibrium of forces acting on active wedge.Sec. 11.6 Seismic
Pressures on Retaining Wal l s 479where <\l -~ : : 2: 'l ', Y=Yd, and 'l ' = tan-I [kh/(1
- kv) ]' The critical fail ure surface, which is fl atterthan the critical fail ure
surface for static conditions, is incl ined (Zarrabi-Kashani, 1979) atan angl ewhere
C1E =-I [- tan (<\l - 'l ' -~) + C1E]UAE = <\l - 'l ' + tan C2E(11.17)(11.19),)tan(<\l -
'l ' -~) [tan(<\l - 'l ' -~) + cot (<\l - 'l ' - 8)] [l + tan (0 + 'l ' + 8) cot (<\l -
'l ' - 8)]C2E = 1+ {tan(0+'l '+8)[tan(<\l -'l '-~)+cot(<\l -'l '-8)]}Al though the M-O
anal ysis impl ies that the total active thrust shoul d act at a point H/3 abovethe
base of a wal l of height, H, experimental resul ts suggest that it actual l y acts at
a higherpoint under dynamic l oading conditions. The total active thrust, PAE
[equation (11.15)], canbe divided into a static component, PA [equation (11.9)],
and a dynamic component, I1PAE: PAE = PA + !'1PAE (11.18)The static component is
known to act at H/3 above the base of the wal l . Seed and Whitman(1970) recommended
that the dynamic component be taken to act at approximatel y 0.6H.On this basis, the
total active thrust wil l act at a heighth = PAH/3+!'1PAE(0.6H)PAEabove the base of
the wal l . The val ue of h depends on the rel ative magnitudes of PA andPAE-it often
ends up near the midheight of the wal l . M-O anal yses show that kv' whentaken as
one-hal f to two-thirds the val ue of kh, affects PAE by l ess than 10%. Seed and
Whitman(1970) concl uded that vertical accel erations can be ignored when the M-O
method isused to estimate PAE for typical wal l designs.Exampl e 11.1Compute the
overturning moment about the base of the wal l shown bel ow for kh =0.15 andk; =
0.075.SmFigure El l .!Dry sil ty sandp = 1.76 Mg/m3<I> =34 degreesS=17 degreesSol ution
First, estimate the static active thrust on the wal l . Because the wal l is not
smooth(8) 0), Coul omb theory shoul d be used. From equations (11.9) and (11.10),cos
2(34° - 0°)cos2(00)cos( 170 + 00) [1 + sin( 17° + 34 O)sin(34° - OO)J 2cos(17° +
OO)cos(O°- 0°)480 Seismic Design of Retaining Wal l s Chap. IIand1 1PA = "2KAyH2 =
"2(0.256)(1.76 Mg/m3)(9.81 m/sec2)(5 m)? = 55.3 kN/mNow, the total active thrust
can be computed from equations (11.15) and (11.16). The angl e, \jf,is given by-
l ( kh ) .-l ( 0.15 ) 920\jf = tan 1 _ k; = tan 1 - 0.075 = .andcos2(34° - 0° - 9.2°)
[sin(l 7° + 34°)sin(34° + 0° - 9.2 0)J2cos(9.2°)cos2(00)cos(l 7° + 0° + 9.2°) I +
1---'-: -: -=: : : --: : -: : -'--: : -: : : : : ---: -: : -: : ---: : -: : -: -'-cos( 17° + 0° + 9.2°)cos(0°
- 0°)= 0.362KAE = -------------'---r=~: : : ==========================~?and1 IPAE =
"2KAEyH2(l - kv ) ="2(0.362)(1.76 Mg/m3)(9.81 m/sec 2)(5 m)2(l - 0.075)= 72.3 kN/m
The dynamic component of the total thrust isAPAE = PAE- PA = 72.3 kN/m - 55.3 kN/m
From equation (11.19), the total thrust acts at a point17 kN/m55.3 kN/m 53m + 17
kN/m(0.6)(5 m)72.3 kN/m = 1.98 mabove the base of the wal l . Because onl y the
horizontal component of the total active thrust contributesto the overturning
moment about the base, the overturning moment is given byu, = (PAth h = (72.3
kN/m)cos(17°)(1.98 m) = 137 ~mPassive Earth Pressure Conditions. The total passive
thrust on a wal l retaininga dry, cohesionl ess backfil l (Figure 11.12) is given by
PPE = ~KPEYH2(1-kv) (11.20)where the dynamic passive earth pressure coefficient,
KpE, is given byK (11.21)_ cos2( <jJ + 8 - \If)PE - 2cos\If cos28 cos(8 - 8 + \If)
[1 - sine8 + <jJ )sin( <jJ + 13 - \If)Jcos(8 - 8 + \If)cos(f3 - 8)The critical
fail ure surface for M-O passive conditions is incl ined from horizontal by anangl e
_l [tan (<jJ + \If + (3)+ C3EJapE = \If - <jJ + tan C4E(11.22)Sec. 11.6 Seismic
Pressures on Retaining Wal l s 481Pp(a)F(b)PpwFigure 11.12/ (a) Forces acting on
passive wedge in Mononobe-Okabe anal ysis; (b)force pol ygon il l ustrating equil ibrium
of forces acting on passive wedge.whereJ tan (<I> + ~ - 'l ') [tan(<I> + ~ - 'l ') +
cot (<I> + e - 'l ')] [1 + tan (3 + 'l ' - e)cot (<I> + e - 'l ')]C~E = 1+ {tan(3+'l '-
e)[tan(<I>+~-'l ')+cot(<I>+e-'l ')]}The total passive thrust can al so be divided
(Towhata and Isl am, 1987) into static anddynamic components: PpE = Pp+ M PE (11.23)
where PPE and P, are computed from equations (11.20) and (11.12), respectivel y.
Note thatthe dynamic component acts in the opposite direction of the static
component, thus reducingthe avail abl e passive resistance.Discussion. Al though
conceptual l y quite simpl e, the M-O anal ysis provides auseful means of estimating
earthquake-induced l oads on retaining wal l s. A positive horizontal accel eration
coefficient causes the total active thrust to exceed the static active thrustand
the total passive thrust to be l ess than the static passive thrust. Since the
stabil ity of aparticul ar wal l is general l y reduced by an increase in active thrust
and/or a decrease in passivethrust, the M-O method produces seismic l oads that are
more critical than the staticl oads that act prior to an earthquake. The effects of
distributed and discrete surface l oadsand irregul ar backfil l surfaces are easil y
considered by modifying the free-body diagram ofthe active or passive wedge. In
such cases, equations (11.16) and (11.21) no l onger appl ythetotal thrusts must be
obtained from the anal ysis of a number of potential fail ure pl anes.As a
pseudostatic extension of the Coul omb anal ysis, however, the M-O anal ysis issubject
to al l of the l imitations of pseudostatic anal yses as wel l as the l imitations of
Coul ombtheory. As in the case of pseudo static sl ope stabil ity anal yses (Section
10.6.1.1), determinationof the appropriate pseudostatic coefficient is difficul t
and the anal ysis is notappropriate for soil s that experience significant l oss of
strength during earthquakes (e.g.,l iquefiabl e soil s). Just as Coul omb theory does
under static conditions, the M-O anal ysiswil l overpredict the actual total passive
thrust, particul arl y for 3 > <1>/2. For these reasons theM-O method shoul d be used
and interpreted careful l y.482 Seismic Design of Retaining Wal l s Chap. 1111.6.1.2
Steedman-Zeng MethodAs a pseudostatic anal ysis, the M-O method accounts for the
dynamic nature of earthquakel oading in a very approximate way. It is possibl e,
however, to account for certaindynamic response characteristics in a rel ativel y
simpl e manner. To account for phase differenceand ampl ification effects within the
backfil l behind a retaining wal l can be consideredusing a simpl e pseudodynamic
anal ysis of seismic earth pressures (Steedman andZeng, 1990).Consider the fixed-
base cantil ever wal l shown in Figure 11.13. If the base is subjectedto harmonic
horizontal accel eration of ampl itude ah, the accel eration at a depth z, bel ow the
top of the wal l can be expressed asa(z, t) = ahsin [ o{t - Hv~ z)J (11.24)If the
seismic wal l pressures are assumed to resul t from the soil within a triangul ar
wedgeincl ined at a to the horizontal , the mass of a thin el ement of the wedge at
depth z isYH-z m(z) = ---dzgtan a(11.25)(11.28)where Yis the unit weight of the
backfil l . The total inertial force acting on the wal l can thereforebe expressed as
Qh(t) = fH m(z)a(z, t)dz = 2'A,yah [21tH cos w~ + 'A,(sin w~ - sin wt)] (11.26)o
41t g tan awhere 'A, = 21tv/w is the wavel ength of the vertical l y propagating shear
wave and ~ = t - Htv;The special case of a rigid wedge is given, in the l imit, as-
yH2ah _ ah v~i~ W - k W oo (Qh)max - 2g tan a - g - h (11.27)which is equival ent to
the pseudostatic force assumed by the M-O method. The total (staticpl us dynamic)
soil thrust can be obtained by resol ving forces on the wedge, that is,Qh(t)cos(a -
<\» + W sin(a - <\»P (t) - -----=------- AE - cos(D+<\>-a)and the total earth
pressure distribution by differentiating the total soil thrust()dPAE(t)PAE t =
dZ ..'J.£ sinea-<\» + khyz cos(a-<\» sin [w(t - !: ...)~ (11.29)tano; cos(D + <\> -
a) tano; cos(D + <\> - a) Vs ~HzFigure 11.13 Wal l geometry and notationfor
Steedman-Zeng method.Sec. 11.6 Seismic Pressures on Retaining Wal l s 483(11.30)The
first term in equation (11.29), which increases l inearl y with depth and does not
varywith time, represents the static earth pressure acting on the wal l . The
resul tant static thrustacts in accordance with static earth pressure theories at a
point hs=HI3 above the base of thewal l . The second term represents the dynamic
earth pressure. It increases as a nonl inearfunction of depth with a shape that
depends on the ratio H/A. A typical exampl e of the nonl ineardynamic pressure is
shown in Figure 11.14. Since the dynamic pressure increasesnonl inearl y with depth,
the position of the dynamic thrust varies with time according toh = H _ 2rt2H2cos
(Os
+ 2rtAH sin (Os - A2( cos (Os - cos (Of)d 2rtHcos (Os + rtA(sin (Os - sin rot)The
point of appl ication of the dynamic thrust for very l ow frequency motions (smal l
H/A,so the backfil l moves essential l y in phase) is at hd =H13. For higher-frequency
motions, thepoint of appl ication moves higher on the wal l , as indicated in Figure
11.15.Steedman and Zeng (1989) found that the soil thrusts for backfil l s of
different stiffnesseswere cl ose to those obtained when the mean shear wave
vel ocities of the backfil l were used in the pseudodynamic anal yses. Backfil l
ampl ification effects can al so be consideredby expressing ah as a function of depth
[ratherthan as a constant in equation (11.24)]and repeating the integration of
equation (11.26). Note that backfil l ampl ification wil l increase both the l oads
acting on the wal l and the height of the resul tant soil thrust. Assumingthat ah(Z,
f) varied l inearl y from the input accel eration at the base of the wal l to a val ue
twice as l arge at the top, Steedman and Zeng (1990) showed good agreement with the
resul ts of centrifuge tests.0.00zR0.20 --0.40 -- f )Steedman-Zeng0.600.80 --1.00
PaeIYH0.00 0.10 0.20 0.30 0.40PAEyHFigure 11.14 Comparison of normal ized pressure
distributions for M-O and SteedmanZengmethods assuming that kh=0.2 and H/'A =0.3.
(After Steedman and Zeng, 1990.)484 Seismic Design of Retaining Wal l s Chap. I JhHd
0.50 --OA5DAD --0.35 --AD .60HIA.200.30 '------'--------'------'-----.00Figure
11.15 Location of dynamic thrust at instant of maximum overturning momentfor kh =
0.2. (After Steedman and Zeng, 1990.)11.6.2 Nonyiel ding Wal l sSome retaining
structures, such as massive gravity wal l s founded on rock or basement wal l sbraced
at both top and bottom, do not move sufficientl y to mobil ize the shear strength of
thebackfil l soil . As a resul t, the l imiting conditions of minimum active or maximum
passiveearth pressures cannot be devel oped.Wood (1973) anal yzed the response of a
homogeneous l inear el astic soil trappedbetween two rigid wal l s connected to a rigid
base (Figure 11.16). If the two wal l s areassumed to be spaced far apart, the
pressures on one wal l wil l not be strongl y infl uenced bythe presence of the other.
Wood showed that dynamic ampl ification was negl igibl e for l owfrequencyinput motions
[i.e., motions at l ess than hal f the fundamental frequency of theunrestrained
backfil l (fa =vs/4H) ]. For this range of frequencies, in which many practical
probl ems l ie, wal l pressures can be obtained from the el astic sol ution for the case
of a uniform,constant, horizontal accel eration appl ied throughout the soil . For
smooth rigid wal l s,Wood (1973) expressed the dynamic thrust and dynamic overturning
moment (about thebase of the wal l ) in the formyH2 a h F g PYH3 a h g Fm(11.31)
(11.32)Sec. 11.6 Seismic Pressures on Retaining Wal l s 485Rigid wal l Rigid wal l Rigid
baseFigure 11.16 Wal l geometry and notation for Wood (1973) anal ysis of pressures
onnonyiel ding wal l s.iwhere ah is the ampl itude of the harmJnic base accel eration
and Fp and Fm are the dimensionl essdynamic thrust and moment factors shown in
Figures 11.17 and 11.18, respectivel y.The point of appl ication of the 1ynamic
thrust is at a height! AMeqheq ! = AP (11.33)eqabove the base of the wal l ;
typical l y, heq '" O.63H.1.00.20.4 _________ LV =0.3v =0.50.2
0.8 ,---,---------,---------,----------,------,v = 0.5 . • • v=0~4~ • _--.---'-- ---
'-~/~-=~---;---~------0.8 ~-----------;-;f1fV,'~--j v =0.3: v =0.2 :
1.2 ,-------,------,--------,------,Fp 0.6 -------4 6 8 10L/H2o
"----------'----'--------'---'----'4 6 8 10 oL/H2o L-_---'-_---'---_---'__--'-----
_-'oFigure 11.17 Dimensionl ess thrust factor forvarious geometries and soil
Poisson's ratioval ues. After Wood (1973).Figure 11.18 Dimensionl ess moment factor
for various geometries and soil Poisson's ratioval ues. After Wood (1973).486
Seismic Design of Retaining Wal l s Chap. 11Exampl e 11.2A reinforced concrete box
cul vert is used to provide an undercrossing through a rail roadembankment as shown
bel ow. Estimate the dynamic thrust on a wal l of the cul vert when subjectedto a
ground motion with kh =0.2.T115 ftFigure El l .2Sol ution Assuming they are properl y
reinforced and that the cul vert cannot sl ide on its base,the cul vert wal l s wil l not
yiel d. Consequentl y, the dynamic thrust can be estimated using equation(1l .31) and
Figure 11.17"""Peg = yH2': !: !2Fp = (130pcf)(l 5ft)2 0.2g(1.0) g g = 5850l b/ft11.6.3
Effects of Water on Wal l PressuresThe procedures for estimation of seismic l oads on
retaining wal l s described in the precedingsections have been l imited to cases of
dry backfil l s. Most retaining wal l s are designed withdrains to prevent water from
buil ding up within the backfil l . This is not possibl e, however,for retaining wal l s
in waterfront areas, where most earthquake-induced wal l fail ures havebeen observed.
The presence of water pl ays a strong rol e in determining the l oads on waterfront
retaining wal l s both during and after earthquakes. Water outboard of a retaining
wal l canexert dynamic pressures on the face of the wal l . Water within a backfil l
can al so affect thedynamic pressures that act on the back of the wal l . Proper
consideration of the effects ofwater is essential for the seismic design of
retaining structures, particul arl y in waterfrontareas. Since few waterfront
retaining structures are compl etel y impermeabl e, the waterl evel in the backfil l is
usual l y at approximatel y the same l evel as the free water outboard ofthe wal l .
Backfil l water l evel s general l y l ag behind changes in outboard water l evel -the
difference in water l evel depends on the permeabil ity of the wal l and backfil l and
on the rateat which the outboard water l evel changes. The total water pressures
that act on retainingwal l s in the absence of seepage within the backfil l can be
divided into two components: hydrostatic pressure, which increases l inearl y with
depth and acts on the wal l before, during,and after earthquake shaking, and
hydrodynamic pressure, which resul ts from thedynamic response of the water itsel f.
11.6.3.1 Water Outboard of Wal l Hydrodynamic water pressure resul ts from the dynamic
response of a body of water.For retaining wal l s, hydrodynamic pressures are usual l y
estimated from Westergaard'sSec. 11.6 Seismic Pressures on Retaining Wal l s 487
(11.34)sol ution (Westergaard, 1931) for the case of a vertical , rigid dam retaining
a semi-infinitereservoir of water that is excited by harmonic, horizontal motion of
its rigid base. Westergaardshowed that the hydrodynamic pressure ampl itude
increased with the square root ofwater depth when the motion is at appl ied at a
frequency l ower than the fundamental frequencyof the reservoir.j; = vi4H, where vp
is the p-wave vel ocity of water (about 4700ft/sec (1400 m/sec)) and H is the depth
of water in the reservoir (the natural frequency of a20-ft-deep (6.1 m) reservoir,
for exampl e, woul d be over 58 Hz, wel l above the frequenciesof interest for
earthquakes). Westergaard computed the ampl itude of the hydrodynamicpressure as7 ah
Cl JPw = 8gYw0zwHThe resul tant hydrodynamic thrust is given by7 ah 2 r; = 12 gYwH
(11.35)The total water pressure on the face of the wal l is the sum of the
hydrostatic and hydrodynamicwater pressures. Simil arl y, tae tota{ {atera{ ({u'ust
due to t{l e water rs equa{ to taesum of the hydrostatic and hydrodynamic thrusts.
Another important consideration in the design of a waterfront retaining wal l is the
potential for rapid drawdown of the water outboard of the wal l . Earthquakes
occurring nearl arge bodies of water often induce l ong-period motion of the water,
such as tsunamis orseiches (Section 1.4.7), that cause the water surface to move up
and down. Whil e theupward movement of water outboard of a retaining wal l wil l
general l y tend to stabil ize thewal l (assuming that it does not rise above the l evel
of the top of the wal l ), downward movementcan create a destabil izing rapid drawdown
condition. When l iquefiabl e soil s existunder rel ativel y high l evel s of initial
shear stress, fail ures can be triggered by very smal l changes in water l evel . Such
fail ures can originate in the soil s adjacent to or beneath theretaining structure
rather than in the backfil l .11.6.3.2 Water in Backfil l The presence of water in the
backfil l behind a retaining wal l can infl uence the seismicl oads that act on the
wal l in three ways: (1) by al tering the inertial forces within the backfil l ,(2) by
devel oping hydrodynamic pressures within the backfil l , and (3) by al l owing excess
porewater pressure generation due to cycl ic straining of the backfil l soil s.The
inertial forces in saturated soil s depend on the rel ative movement between the
backfil l soil particl es and the porewater that surrounds them. If, as is usual l y
the case, thepermeabil ity of the soil is smal l enough (typical l y k: S; 10-3 em/sec
(33 x 10-5 ft/sec) or so)that the porewater moves with the soil during earthquake
shaking (no rel ative movement ofsoil and water, or restrained porewater
conditions), the inertial forces wil l be proportional to the total unit weight of
the soil . If the permeabil ity of the backfil l soil is very high, however,the
porewater may remain essential l y stationary whil e the soil skel eton moves backand
forth (the soil particl es move through the porewater in free porewater conditions).
Insuch cases, inertial forces wil l be proportional to the buoyant (or submerged)
unit weight ofthe soil . Hydrodynamic water pressures (Section 11.5.3.1) can al so
devel op under free488 Seismic Design of Retaining Wal l s Chap. J 1pore-water
conditions and must be added to the computed soil and hydrostatic pressures to
obtain the total l oading on the wal l .For restrained porewater conditions, the M-O
method can be modified to account forthe presence of porewater within the backfil l
(Matsuzawa
et al ., 1985). Representing theexcess porewater pressure in the backfil l by the
pore pressure ratio, r u (Section 9.5.1.1), theactive soil thrust acting on a
yiel ding wal l can be computed from equation (11.15) using(11.36)\If = tan-1
[ 'Ysatkh ] (11.37)'Yb(l - ru)(l - kv)An equival ent hydrostatic thrust based on a
fl uid of unit weight 'Yeq = 'Yw + ru'Yb must be addedto the soil thrust. Note that
as ru approaches 1 (as it coul d in a l iquefiabl e backfil l ), the wal l thrust
approaches that imposed by a fl uid of equival ent unit weight, 'Yeq = 'Ysat.As
discussedin Chapter 9, subsequent unidirectional movement of a soil that devel ops
high excess porewaterpressures may, depending on its residual (or steady state)
strength, cause dil ation withaccompanying porewater pressure reduction and strength
gain.Soil thrusts from partial l y submerged backfil l s may be computed using an
averageunit weight based on the rel ative vol umes of soil within the active wedge
that are above andbel ow the phreatic surface (Figure 11.19): y = A2Ysat+(I-A2)Yd
(11.38)Again, the hydrostatic thrust (and hydrodynamic thrust, if present) must be
added to the soil thrust.Y=YsatAHHFigure 11.19 Geometry and notation for partial l y
submerged backfil l .Exampl e 11.3Compute the total thrust that woul d be expected on
the wal l shown in Exampl e 11.1 if the wal l backfil l was compl etel y saturated. Assume
that the pore pressure ratio, ruo reaches a val ue of 0.5.Sol ution The total active
thrust can be computed from equation (11.15) usingG -IY = Yb(1- ru ) = Pbg(1 - ru )
= yPdg(1- ru )s= 2'~~6; I (1.76 Mg/m') (9.81 m/sec") (I - 0.5) = 5.44 kN/m3-1
[ Ysatkh ] -I [(10.9 kN/m3 + 9.8 kN/m3) (0. 15) ] = 31.6 0\jf = tan Yb(1-ru)(l -kv )
= tan (10.9 kN/m3)(l -0.5)(l -0.075)Sec. 11.7ThenSeismic Displ acements of Retaining
Wal l s 489° ° ° [ sin(l 70+31.60)Sin(340-00-31.6o)]2cos(31.6°)cos(00)cos(l 7 + 0 +
31.6 ) 1 + cos(I?" + 00 + 31.60)cos(00 _ 00)= 1.195andThe hydrostatic thrust is
given by_ 1 2 _ 1( )H2 Pw-;{'feqH -Z'Yw+ru'Yb -_Z1[9.81kN/m3 +(05.)(I0.9 kNl m3)] (5
m)2= 190.6 kN/mTherefore, the total thrust isP,ot = PAE + Pw = 75.2 kN/m + 190.6
kN/m = 265.8 kN/m11.6.4 Finite-El ement Anal ysisEarthquake-induced pressures on
retaining wal l s can al so be eval uated using dynamicresponse-anal yses. A number of
computer programs are avail abl e for such anal yses (Section7.3). Linear or
equival ent l inear anal yses can be used to estimate wal l pressures,al though their
inabil ity to represent actual modes of fail ure can make their resul ts difficul tto
interpret. Nonl inear anal yses are capabl e of predicting permanent deformations
(Section11.6.3) as wel l as wal l pressures.11.7 SEIISMIC DISPLACEMENTS OF RETAINING
WALLSAl though the methods of anal ysis described in the preceding section provide
useful informationon the seismic l oads that act on retaining wal l s, the
postearthquake serviceabil ity ofsuch wal l s is more cl osel y rel ated to the permanent
deformations that occur during earthquakes.Whil e l arge permanent deformations may
be acceptabl e for some wal l s, others maybe considered to have fail ed at much
smal l er deformations. Anal yses that predict permanentwal l deformations may provide
a more useful indication of retaining wal l performance.Several methods have been
proposed for predicting permanent deformations of yiel dingwal l s.11 .7.1 Richards-
El ms MethodRichards and El ms (1979) proposed a method for the seismic design of
gravity wal l s basedon al l owabl e permanent wal l displ acements. The method estimates
permanent displ acementsin a manner anal ogous to the Newmark sl iding bl ock procedure
(Section 10.6.1.2)devel oped original l y for eval uation of seismic sl ope stabil ity.
490 Seismic Design of Retaining Wal l s Chap. J J: rr Figure 11.20 Gravitywal l acted
upon bygravityand pseudostatic accel erations.Appl ication of the Richards-El ms
method requires eval uation of the yiel d accel erationfor the wal l -backfil l system.
Consider the gravity wal l shown in Figure 11.20. Whenthe active wedge is subjected
to accel eration acting toward the backfil l , the resul ting inertial forces wil l act
away from the backfil l . The l evel of accel eration that is just l arge enough tocause
the wal l to sl ide on its base is the yiel d accel eration. When the accel eration is
equal to the yiel d accel eration, horizontal and vertical equil ibrium require thatT =
Fh+ (PAE)h(11.39)N = W + (PAE)vSubstituting T=Ntan <\l b' Fh= ayW/g, (PAEh =
PAECOS(O+ 8), and (PAE)v = PAEsin(o+ 8),the yiel d accel eration can be computed as[
'" PAE cos(o + 8) - PAE sin(o + 8)]ay = tan v» - W g (11.40)Richards and El ms
recommended that PAEbe cal cul ated using the M-0 method (since theM-O method
requires that ay be known, the sol ution of equation (11.40) must be obtained
iterativel y). Using the resul ts of sl iding bl ock anal yses in the same manner as
Newmark(1965) and Frankl in and Chang (1977), Richards and El ms proposed the
fol l owing expressionfor permanent bl ock displ acement: (11.41)amax ? 0.32 3d = 0087
Vmaxamaxperm' 4a ywhere Vmax is the peak ground vel ocity, a max the peak ground
accel eration, and a y the yiel daccel eration for the wal l -backfil l system. Equation
(11.41) provides displ acement estimatesthat are cl ose to the estimated maximum
displ acements [equation (10.13)] of Newmark(1965).Exampl e 11.4Estimate the
permanent displ acement of the concrete gravity wal l shown bel ow that woul d be
produced by the Gil roy NO.2 (soil ) motion. Assume k; = O.Sec. , '.7 Seismic
Displ acements of Retaining Wal l s--j3ft f--49120 ft--j 8 ft 1----Figure El l ALoose
sandc=O$= 33°y=105pcf0=17°Sol ution The weight of the wal l section isW = (20 ft)( 8
ft; 3ft)(150 Ib/ft3) = 16500l b/ftThe total active thrust can be estimated by the
Mononobe-Okabe approach. From equations(11.15) and (11.16). Assuming a trial
pseudostatic accel eration of 0.10g, \If=5.7° andK AE =(570) 2(710) (17° 710 570)[1
Sin(170+330)Sin(330-00-5.r)]2cos . cos . cos +. +. + cos(107 +7.1 ° + 5.7°)cos(00
-7. 10)= 0.385andPAE = ~(0.385)( 105 pcf)(20 ft)2 = 8085 l b/ftThen, from equation
(11.40),a = [tan(35 0) _ (80851b/ft)cos(17 + 7.1) - (8085 Ib/ft)sin(17 + 7.1)J =
045y 16500l b/ft g . gBecause the computed yiel d accel eration (0.43g) is
inconsistent with the assumed pseudostaticaccel eration (0.109), another iteration
is required. For the next iteration, assume that the pseudostaticaccel eration is
0.30g. Then \If is 16.7° andKAE =cos(16.70)cos2(7.1 0)cos(17 0 + 7.1 ° + 16.70)[1 +
sin(17° + 33°)sin(33° _0° -16.70)]2cos(17° + 7.1 ° + 16.7°)cos(0° -7.1°)= 0.574and
PAE = ~(0.574)(105 pcf)(20 ft)2 = 12054l b/ft492 Seismic Design of Retaining Wal l s
Chap. 11Then, from equation (11.40),a = [tan(35°) - (120541b/ft)cos(17 + 7.1) -
(120541b/ft)sin(17 + 7.1)J = 033y 165001b/ft g . gNow, the computed yiel d
accel eration is fairl y cl ose to the assumed pseudostatic accel eration.Using
equation (l 1.41) with the resul ts of Exampl e 3.1d - 0087(39.2 crn/sec)2 [(0.322g)
(981 crn/sec 2/g)]3 = 6.5cmperm - • [(0.30g)(981 crn/sec2/g)] 411.7.2 Whitman-Liao
MethodThe Richards-El ms method offers a rational deterministic approach to the
estimation ofgravity wal l displ acements. Its simpl icity comes, in part, from
assumptions that negl ect certainaspects of the dynamic earth pressure probl em.
Whitman and Liao (1985) identifiedseveral model ing errors that resul t from the
simpl ifying assumptions of the Richards-El msprocedure. The most important of these
are negl ect of the dynamic response of the backfil l ,negl ect of kinematic factors,
negl ect of til ting mechanisms, and negl ect of vertical accel erations.Finite-el ement
anal yses of the effects of the dynamic response of the backfil l on wal l
displ acements (Nadim, 1982), for exampl e, show that ampl ification occurs when input
motions coincide with the natural period of the backfil l and produce considerabl y
greaterpermanent displ acement than the rigid-bl ock model used by Richards and El ms.
Anal ysesin which the backfil l wedge and wal l were treated as separate bl ocks
(Zarrabi-Kashani,1979) show that the kinematic requirements of horizontal and
vertical displ acement of thebackfil l wedge cause systematical l y smal l er
displ acements than the singl e-bl ock model ofRichards and El ms. Studies of combined
til ting and sl iding (Nadim, 1980; Siddharthan etal ., 1992), indicate that til ting
mechanisms general l y increase wal l displ acements over thoseproduced by sl iding-onl y
model s such as that of Richards and El ms. Consideration of vertical accel erations
produces sl ightl y l arger displ acements than when they are negl ected, atl east for
motions with high peak horizontal accel eration (amax greater than about 0.5g) anda!
amax;: : : : 0.4 (Whitman and Liao, 1985). Whitman and Liao quantified and combined
theeffects of each of these sources of model ing error to describe ~e total model ing
error by al ognormal l y distributed random variabl e with mean val ue, M, and standard
deviation,al nM'Using the resul ts of sl iding bl ock anal yses of 14 ground motions by
Wong (1982),Whitman and Liao found that the permanent displ acements were
l ognormal l y distributedwith mean val ue- 37v~ax (-9.4a y)dperm = --- exp ---amax
amax(11.42)Uncertainty due to statistical variabil ity of ground motions ~as
characterized by a l ognormal l ydistributed random variabl e, Q, with a mean val ue of
Q and standard deviation, al n Q'The effects of uncertainty in soil properties,
specifical l y the friction angl es, on permanentdispl acement were al so investigated.
Using standard deviations of a<\> =2 to 3° forsoil friction angl es and al ) =5° for
wal l -soil interface friction angl es, the computed yiel daccel eration [the onl y term
on the right side of equation (11.40) that is a function of <I> andSec. 11.7
Seismic Displ acements of Retaining Wal l s 4938] was defined as a random variabl e
with mean val ue Gy and standard deviation aa, . Themean val ue Gy is the
yiel d accel eration computed using the mean val ues of <p and 8.Combining al l of
these sources of uncertainty, the permanent displ acement can becharacterized as a
l ognormal l y distributed random variabl e with mean val ueand varianced -37v~ax (-
9.4aY)Q-M- = --exp --amax amax2 _ (9.4 g )22 2 2GIn d - -- Gay + Gl nM + Gl nQ amax
(11.43)(11.44)Suggested val ues of the means and standard deviations of the ground
motion, soil resistance,and model error factors are shown in Tabl e 11-4.Tabl e "-4
Mean and Standard Deviation Val ues forGravity Wal l Displ acement Anal ysisFactorModel
errorSoil resistanceGround motionMeanM = 3.5a: ; = ay(jl , 3)Q=IStandard Deviation
al nM = 0.84aG, = 0.04 to 0.065al n Q = 0.58 to 1.05Source: After Whitman and Liao
(1985).Using equations (11.43) and (11.44), al ong with the CDF for the standard
normal variabl e (Tabl e C-l ), the probabil ity of exceeding any particul ar val ue of
dal l can easil y becomputed.Exampl e 11.5Estimate the expected permanent displ acement
of the gravity wal l shown in Exampl e 11.4using the Whitman-Liao approach.Sol ution
From equation (11.42), the mean or expected val ue of permanent displ acement isgiven
byd el m = 37 (39.2 cm/sec)l [( 94)(030 )/0322] 003p [ (0.322g) (981 cm/secvg)] exp
-. . g . g = . emNote that the mean permanent displ acement is considerabl y smal l er
than the displ acement predictedby the Richards-El ms method in Exampl e 11.4.111.7.3
Finite-El ement Anal ysisEarthquake-induced deformations of retaining wal l s can be
predicted by dynamic stressdeformationanal yses. Obviousl y, prediction of permanent
deformations requires the use ofa nonl inear anal ysis (Section 7.2.3). A rigorous
anal ysis shoul d be capabl e of accounting fornonl inear, inel astic behavior of the
soil and of the interfaces between the soil and wal l el ements.Among the rel ativel y
few exampl es of rigorous two-dimensional finite-el ementanal yses that predict
permanent deformations are those reported by Al ampal l i and El gamel (1990), Finn et
al . (1992), and Iai and Kameoka (1993).494 Seismic Design of Retaining Wal l s Chap.
J 111.8 SEISMIC DESIGN CONSIDERATIONSThe design of retaining wal l s for seismic
conditions is simil ar, in many respects, to designingfor static conditions. In both
cases, potential modes of fail ure are identified and the wal l designed to avoid
initiating them. Al though the response of retaining wal l s under seismicl oading
conditions is much more compl ex than under static conditions, conventional design
procedures make use of simpl ifying assumptions that render the probl em tractabl e.
Several design approaches for different types of retaining wal l s are described in
the fol l owingsections.11 .8.1 Gravity Wal l sGravity wal l s are the simpl est type of
retaining wal l , and more attention has been paid totheir design than to the design
of other types of wal l s. Gravity wal l design procedures, however,are commonl y
adapted as part of the design of cantil ever wal l s and composite wal l systems.
Gravity wal l s are customaril y designed by one oftwo approaches: a seismic pressure-
based approach or a permanent displ acement-based approach. Al though the gravitywal l
design procedures are oriented toward prevention of sl iding fail ure, the
possibil ity ofoverturning due to bearing fail ure of the soil beneath the base of a
wal l must al so be consideredin design.11.8.1.1 Design Based on Seismic Pressures
Gravity wal l s have traditional l y been designed on the basis of seismic earth
pressures.The M-O method is most commonl y used al ong with an inertial force, using
the samepseudostatic accel eration appl ied to the active wedge, appl ied to the wal l
itsel f. Pseudostaticaccel erations are general l y considerabl y smal l er than
anticipated peak accel erations.Val ues between 0.05g and 0.15g, corresponding to
one-third to one-hal f of the peak groundsurface accel eration, are commonl y used
with factors of safety of 1.0 to 1.2 (Whitman,1990).Despite the considerabl e
simpl ification of their compl ex actual behavior, gravitywal l s designed by the
traditional approach have general l y performed quite wel l in earthquakes.The reason,
however, may have more to do with the conservatism commonl y usedin static wal l
design than with the accuracy of the M-O method. Design pressures thataccount for
backfil l ampl ification, such as those that can be obtained by the Steedman-Zeng
method, shoul d be considered for design of unusual l y tal l or unusual l y
displ acementsensitivewal l s.11 .8.1.2 Design Based on Al l owabl e Displ acementsGravity
wal l s are being designed on the basis of al l owabl e displ acements more andmore
frequentl y. This approach al l ows the designer to consider the consequences of
permanentdispl acement for an individual wal l when sel ecting an al l owabl e
displ acement fordesign. Design procedures based on the Richards-El ms and Whitman-
Liao procedures forestimation of permanent displ acement are avail abl e.The Richards-
El ms design procedure invol ves cal cul ation of the wal l weight thatwoul d be required
to ensure that permanent displ acements are l ess than or equal to someal l owabl e
val ue. The procedure can be summarized as fol l ows: Sec. I 1.8 Seismic Design
Considerations 495(11.45)(11.47)1. Sel ect an al l owabl e permanent displ acement, daB.
2. Cal cul ate the yiel d accel eration required to produce the al l owabl e permanent
displ acementfrom equation (11.41), rearranged in the formay= (0.087V;;'axai;,axf4
daB3. Cal cul ate PAEusing the M-O method with the yiel d accel eration from step 2 as
thepseudostatic accel eration. This represents the soil thrust that woul d be
expected tocause a maximum permanent displ acement equal to daB'4. Cal cul ate the
wal l weight required to l imit the permanent displ acement to the al l owabl epermanent
displ acement using equation (11.40) rearranged in the formW = PAEcos(8 + 8) -
PAEsin(8 + 8)tan <Ph (11.46)tan <Ph - a/gS. Appl y a factor of safety to the weight
of the wal l . Richards and El ms original l y suggesteda factor of safety of 1.5, but
subsequent research has shown that a wal l weightfactor of safety of 1.1 to 1.2
shoul d be sufficient to reduce the probabil ity of exceedingthe al l owabl e permanent
displ acement to 5% or l ess (Whitman and Liao, 1985).Using the Whitman-Liao
approach, gravity wal l s can be designed on the basis ofal l owabl e displ acements that
have defined probabil ities of exceedance. The sel ection of anacceptabl e probabil ity
of exceedance is a compl icated matter that may depend on the importanceof the wal l ,
the effects of fail ure, the cost of repair, and other technical and/or nontechnical
factors. The actual wal l design is accompl ished by using equation (11.43) tocompute
the yiel d accel eration as- 2amax 1 37Mvmax a y = -- n9.4 amaxdaBand then using the
M-O method to size the wal l to meet or exceed the computed yiel d accel eration.
Whitman and Liao suggest a conservative design (corresponding to a probabil ity of
exceedance of about 5%) can be obtained by assuming daB = 4dperm, where dperm is
cal cul atedfrom equation (11.42). A l ess conservative design (corresponding to about
10% probabil ityof exceedance) is obtained by taking daB = 2.5dperm.1 '1.8.2
Cantil ever Wal l sCantil ever wal l s are designed in much the same way as gravity
wal l s, except that bendingfail ure must al so be considered. Maximum bending moments
are usual l y cal cul ated usingthe M-O method to compute the maximum soil thrust,
which is taken to act at the heightgiven by equation (11.19). The maximum
overtvrning moment is used for structural designof wal l el ements (to prevent
fl exural fail ure of the wal l itsel f) and to determine the size ofthe wal l footing
required to prevent bearing fail ure of the supporting soil s.1 'I .8.3 Braced Wal l s
Because their l ateral displ acements are constrained by bracing el ements, braced
wal l s donot devel op minimum active (or maximum passive) earth pressures in the
vicinity of the496 Seismic Design of Retaining Wal l s Chap. 11bracing el ements. The
earth pressures that do devel op depend on the stiffness of the braceand the
rel ative stiffness of the wal l and soil . Anal ysis of actual soil -wal l -brace
interactionis quite compl icated, and simpl ified methods are general l y adopted for
design purposes.11.8.3.1 Non-Yiel ding Braced Wal l sWal l s that are braced
sufficientl y that they do not move at al l , such as braced wal l sfounded on rock, are
usual l y designed to resist the earth pressures predicted by Wood's(1973) anal ysis.
To avoid cracking or yiel ding of such wal l s, design pressures are usual l ybased on
the peak accel eration (Whitman, 1991). This approach, however, can l ead to veryhigh
design pressures and some data [e.g., Chang et al . (1990)], indicates that the
pressureson braced wal l s that can move sl ightl y are l ower. Dynamic finite-el ement
anal yses of stiff,embedded basement wal l s suggest that the M-O method with the peak
ground surface accel erationoutside the structure produces reasonabl e design earth
pressures (Whitman, 1990).Specifications for design of bridge abutments (e.g.,
AASHTO, 1991) recommend thatnonyiel ding abutments be designed to resist l ateral
thrust obtained from the M-O methodwith a pseudostatic horizontal accel eration 50%
greater than the effective peak accel eration.11.8.3.2 Fl exibl e Braced Wal l sThe
seismic response of fl exibl e braced wal l s such as anchored bul kheads and tieback
wal l s is particul arl y compl icated. Again, these compl exities require that
simpl ified methodsbe used for design. Because these types of wal l s are very
commonl y used in waterfrontareas, and because the great majority of earthquake-
induced retaining wal l fail ures occur inwaterfront areas, special attention to
their design is required. In the case of anchored bul kheads,the simpl ified design
procedures can be suppl emented by a useful empirical model ofdamage potential based
on observations of actual wal l performance.Anchored Bul l (heads. The design of
anchored
bul kheads in waterfront areasis strongl y infl uenced by l iquefaction hazards. If
widespread l iquefaction occurs, experienceindicates that bul khead fail ures are very
l ikel y. Consequentl y, steps shoul d be takenprior to construction to ensure that
such l iquefaction wil l not occur. Permanent seawardmovements of anchored bul kheads
in the absence of widespread l iquefaction, however, hasal so been observed.
Conventional design procedures seek to minimize this type of damageusing
pseudostatical l y determined design pressures.A recent design procedure uses the
free earth support method and Rowe's momentreduction method, with earthquake
effects represented by pseudostatic inertial forces. Abrief summary of the
procedure is presented bel ow; a detail ed description with a workedexampl e may be
found in Ebel ing and Morrison (1993).1. Design the anchored bul khead for static
l oading conditions.2. Sel ect pseudostatic accel erations ah and avo3. Compute the
active soil thrust on the back of the wal l using the M-O method. Theactive wedge is
assumed to originate at the bottom of the wal l .4. Compute the passive soil thrust
acting on the front of the wal l using the M-O method.The passive wedge is al so
assumed to originate at the bottom of the wal l .5. Compute the minimum required
depth of wal l penetration by summing momentsabout the wal l -tierod connection. Al l
water pressures (hydrostatic, hydrodynamic,and excess porewater, if present) must
be incl uded.Sec. 11.8 Seismic Design Considerations 4976. Compute the required
anchor resistance by summing the horizontal forces acting onthe wal l . Al l water
pressures (hydrostatic, hydrodynamic, and excess porewater, ifpresent) must be
incl uded. The computed anchor resistance is termed the free earthsupport anchor
resistance.7. Compute the distribution of bending moments over the height of the
wal l . Al l waterpressures (hydrostatic, hydrodynamic, and excess porewater, if
present) must beincl uded. The maximum bending moment is termed the free earth
support moment.8. Compute the design bending moment as the product of the free
earth support momentand Rowe's moment reduction factor (Rowe, 1952).9. Set the
design tierod force at a l evel 30% greater than the free earth support anchor
resistance.]l O. Determine the required size of the anchor bl ock to satisfy
horizontal force equil ibriumconsidering the active and passive pressures, as wel l
as al l water pressures, on bothsides of the bl ock. The effects of any water
pressures on the bottom and top surfacesof the anchor bl ock shoul d al so be
considered.11. Locate the anchor bl ock at a sufficient distance behind the wal l
that the active wedgebehind the wal l does not intersect the passive wedge in front
of the anchor bl ock.Since the active and passive fail ure surfaces are fl atter for
seismic l oading than forstatic l oading, seismic design may require a considerabl y
l onger tierod than staticdesign.12. Check the effects of redistribution of any
earthquake-induced excess porewater pressureafter earthquake shaking has ended.Case
histories of anchored bul khead performance (negl ecting cases in which widespread
l iquefaction was observed) suggest that anchored bul khead damage l evel s can be
predicted approximatel y with the aid of two dimensionl ess indices: the effective
anchorindex and the embedment participation index (Gazetas et al ., 1990). Referring
to Figure11.21a, the effective anchor index describes the rel ative magnitude of the
avail abl e anchorcapacity asEAI = ~H(11.48)where d is the horizontal distance
between the active wedge and the tierod-anchor connectionand H is the height of the
wal l . The critical active fail ure pl ane is taken to originate atthe effective point
of rotation of the wal l , which can be l ocated using soil -structure interaction
anal yses or estimated aswhereT> (1 + k; _ <I> - 200)H < D2 50° - (11.49)k'eabove
the water tabl ebel ow the water tabl e498 Seismic Design of Retaining Wal l s Chap. J J
The incl ination of the critical active fail ure pl ane can be approximated (Dennehy,
1985) as(11.50)(11.51)for 0.10 ~ k; ~ 0.50 and 25° ~ <I> ~ 35°. Beyond these
ranges, equation (11.17) can be usedto estimate the incl ination of the active
wedge. The embedment participation index isdefined asEPI = FpE(l +~)FAE f + Hwhere
FAE and FpEare the potential active and passive thrusts, respectivel y. For uniform
backfil l and foundation soil s,EPI",KKPEr2(1+r) (11.52)AEwhere r=f/l f+ H). Val ues
ofEAI and EPI have been computed for 75 anchored bul kheads forwhich degrees of
damage in earthquake had been categorized as indicated in Tabl e 11-5(Gazetas et
aI., 1990). Comparison of degrees of damage with EAI and EPI showed significant
trends in the characteristics of anchored bul kheads that performed wel l and those
that performedpoorl y. As il l ustrated in Figure 11.21b, anchored bul kheads with high
EAI and EPIval ues (zone I) general l y suffered l ittl e or no damage. Anchored
bul kheads with l ow EAI andEPI val ues (zone III) usual l y suffered severe damage.
Moderate damage was general l y associatedwith intermediate combinations ofEAI and
EPI (zone II). The chart of Figure 11.21b isa very useful tool for checking the
design of anchored bul kheads in waterfront areas.Tabl e 11-5 Qual itative and
Quantitative Descriptions of Reported Degrees of Damage toAnchored Bul kheads during
EarthquakesDegree ofDamage Description of Damageo No damageI Negl igibl e damage to
the wal l itsel f; noticeabl e damageto rel ated structures2 Noticeabl e damage to the
wal l 3 General shape of anchored sheetpil e preserved, butsignificantl y damaged4
Compl ete destructionSource: After Kitajima and Uwabe (1979).Permanent Horizontal
Displ acementat Top of Sheetpil e (em)<22-1010-3030-60>60Tieback Wal l s. Because their
use was original l y restricted to temporary supportof excavations, the seismic
performance of tieback wal l s has received rel ativel y l ittl e attention.As permanent
tieback wal l s have become more common, however, recognition of theneed to consider
seismic l oading in their design has increased. With respect to seismic performance,
tieback wal l s are simil ar to anchored bul kheads-the primary differences are that
tieback wal l s have mul tipl e anchors and that the anchors are general l y incl ined.
Al thoughSec. J J.8 Seismic Design ConsiderationsIe L -I499HEffective pointor
rotation(a)Figure 11.21 (a) Geometry and notation for eval uation of anchored
bul khead design;(b) correl ation between damage l evel s and dimensionl ess anchored
bul khead indices.After Gazetas et al . (1990). Empirical design method for
waterfront anchored sheetpil ewal l s, Design and Performance ofEarth Retaining
Structures. Reprinted by permissionof ASCE.formal ized design procedures are not
currentl y avail abl e, the resul ts of a l imited number ofexperimental and numerical
investigations provide insight into special considerations forseismic design of
tieback wal l s. Few observations of the seismic performance of tiebackwal l s have
been made, but those that are avail abl e general l y indicate good performance (Hoet
al ., 1990).Numerical anal yses indicate that the stiffness and spacing of the
anchors strongl yinfl uences the permanent displ acements of a tieback wal l (Sil l er
and Frawl ey, 1992). Ingeneral , tieback wal l s with stiff anchors wil l devel op
smal l er permanent displ acements thanwal l s with softer anchors. Tieback wal l s with
smal l er vertical spacing between anchors wil l experience smal l er and more uniform
permanent displ acements than wal l s with greateranchor spacing. Earthquake-induced
permanent displ acements al so appear sensitive tostatic design pressures-wal l s
designed for higher static pressures suffer smal l er permanentdispl acements during
earthquakes than wal l s designed for l ower static pressures. Tiebackwal l s al so tend
to devel op smal l er permanent when higher rather than l ower initial anchorprel oads
are used (Sil l er and Dol l y, 1992).Tieback wal l s can al so be infl uenced by phase
differences between the response of thesoil behind the wal l and the foundation soil
(Fragaszy et al ., 1987). Wal l el ements thatextend into the foundation soil s may be
subjected to very high bending moments at the baseof the wal l . Further, incl ined
anchors that extend bel ow the bottom of the excavation mayexperience very high
tensil e forces when the soil end moves one way and the wal l end theother.500
Seismic Design of Retaining Wal l s Chap. 112 ---o Degree of damageo 0ooo2340',\,IIo
o• • oo,,,,,,ooo 0ooooo0 00 O·o 0 o 0 0 000DO 0ooooo 0• o-0.5 ---1.5<i: ~xQ)"C .s0 s:
oc: : eel Q) >15 0.5Q): t: : Wo 0.5Embedment participation index (EPI)(b)Figure 11.21
(continued)11.8.4 Reinforced Soil Wal l sReinforced soil wal l s have become very
popul ar in recent years. Al though many differentmaterial s are used, reinforced soil
wal l s consist of a zone ofreinforced soil that retains unreinforcedsoil behind it.
During an earthquake, a reinforced soil wal l is subjected to adynamic soil thrust
at the back of the reinforced zone and to inertial forces within the reinforcedzone
in addition to static forces. The wal l must be designed to avoid external
instabil ity(sl iding or overturning fail ure of the reinforced zone) and internal
instabil ity (tensil eor pul l out fail ure of the reinforcement). Current design
procedures, such as those describedbel ow (Christopher et aI., 1990), use a
pseudostatic approach, but displ acement-baseddesign procedures are l ikel y to be
devel oped in the future.Sec. 1J .8 Seismic Design Considerations 501Reinforced soil
wal l s use different material s and provide support by different mechanismsthan do
conventional retaining wal l s. Their design requires careful consideration of
l oading, time, and environmental factors that may not be significant for
conventional retainingwal l s. The fol l owing sections are restricted to seismic
aspects of reinforced soil wal l design. Other aspects of design are discussed in
detail by Mitchel l and Vil l et (1987), Christopheret al . (1990), and Koerner (1994).
11.8.4.1 External Stabil ityFor eval uation of external stabil ity, a reinforced wal l
is treated much l ike a gravitywal l . As il l ustrated in Figure 11.22, the reinforced
zone is assumed to be acted on by its ownweight, W, and the static soil thrust,
PA . Earthquake l oading is represented pseudostatical l yby the dynamic soil
thrust, !!.PAE, and the inertial force on the reinforced zone, PIR . Theexternal
stabil ity of a particul ar wal l design can be eval uated by the fol l owing procedure:
1. Determine the peak horizontal ground surface accel eration, amax.2. Cal cul ate the
peak accel eration at the centroid of the reinforced zone from theequationa; =
( 1.45 - a~ax )amax3. Cal cul ate the dynamic soil thrust froma y(b) H2!!.PAE = 0.375
_C,-,-__gwhere i b) is the unit weight of the backfil l soil .4. Cal cul ate the
inertial force acting on the reinforced zone fromacy(r) HLPIR = gwhere iT) is the
unit weight of the reinforced zone.Reinforced zone(11.53)(11.54)(11.55)H(a)Backfil l
(b)Figure 11.22 (a) Geometry and notation for reinforced soil wal l s; (b) static and
pseudostatic forces acting on reinforced zone.502 Seismic Design of Retaining Wal l s
Chap. 115. Add PAE and 50% of PIR to the static forces acting on the reinforced
zone and checksl iding and overturning stabil ity (the reduced val ue of PIR is
al l owed to account for thefact that the maximum val ues of MAE and PIR are unl ikel y
to occur at the same time).For seismic design, factors of safety against sl iding
and overturning shoul d be greaterthan or equal to 75% of the minimum acceptabl e
factors of safety for static l oading.11.8.4.2 Internal Stabil ityInternal stabil ity
eval uation depends on the nature of the reinforcement since the critical internal
fail ure surface is different for inextensibl e and extensibl e reinforcement (Figure
11.23). Internal stabil ity for seismic conditions can be eval uated in the fol l owing
steps: 1. Determine the pseudostatic inertial force acting on the potential l y
unstabl e internal fail ure zone,(11.56)where WA is the weight of the fail ure mass
(the trapezoidal or triangul ar zones in Figurel 1.23a and b for inextensibl e or
extensibl e reinforcement, respectivel y).2. Distribute PIA to each reinforcement
l ayer in proportion to its resistant area (the areaof reinforcement that extends
beyond the potential internal fail ure surface). This processproduces a dynamic
component of tensil e force for each l ayer of reinforcement.3. Add the dynamic
components of tensil e force to the static components of tensil e forceto obtain the
total tensil e force for each l ayer of reinforcement.4. Check to see that the
al l owabl e tensil e strength of the reinforcement is at l east 75% ofthe total tensil e
force in each l ayer of reinforcement.5. Check to see that each l ayer of
reinforcement extends far enough beyond the potential internal fail ure surface to
avoid pul l out fail ure with a factor of safety not l ess than75% of the minimum
static factor of safety when the total tensil e force is appl ied.1~·3~1TH/2IH/2 Hj
(a) (b)Figure 11.23 Critical potential fail ure surfaces for eval uation of internal
seismicstabil ity of reinforced soil wal l s: (a) inextensibl e reinforcement; (b)
extensibl ereinforcement.Sec. 11.911.9 SUMMARYSummary 5031. A variety of different
systems are used to retain soil ; many innovative systems have beendevel oped and
used in the l ast 20 years. For seismic design purposes, most retaining systemsfal l
into three general categories: gravity wal l s, cantil ever wal l s, and braced wal l s.
Each of these types of wal l s resists the l ateral pressures of the retained soil
differentl y.2. Retaining wal l s can fail in many different ways. Gravity wal l s fail
by sl iding, overturning,or gross instabil ity. Cantil ever wal l s can fail in the same
ways as gravitywal l s, but can al so fail in bending. Braced wal l s can fail in the
same ways as cantil everwal l s, but al so by bracing el ement fail ure or rotation about
the brace connection.3. The seismic performance of a retaining wal l depends on the
total pressures (i.e., staticpl us dynamic pressures) that act on it during an
earthquake. Consequentl y, the l evel of dynamic pressures required to damage the wal l
depends on the l evel of static pressuresthat exist before the earthquake. Anal ysis
of the seismic behavior of a retainingwal l , therefore, requires an initial anal ysis
of the behavior under static conditions.4. The seismic response of retaining wal l s
is a compl ex exampl e of soil -structure interaction.Because of this compl exity,
design procedures are usual l y based on a numberof simpl ifying assumptions. Al though
the simpl ified procedures do not represent al l aspects of wal l -soil behavior
accuratel y, they have been shown to provide a reasonabl ebasis for design.5. The
soil pressure on a retaining wal l depends on whether the wal l is abl e to move
rel ativeto the soil . The dynamic pressures acting on a yiel ding wal l are usual l y
estimatedby the pseudostatic Mononobe-Okabe anal ysis. The dynamic pressures on
nonyiel ding wal l s are usual l y estimated by an el astic anal ysis.6. The presence of
water on either side of a retaining wal l strongl y infl uences the seismicbehavior of
the wal l . Water on the outboard side of the wal l can exert dynamic,in addition to
hydrostatic, pressures on the face of the wal l . Water within the backfil l can
infl uence the inertial forces acting on the wal l and can devel op hydrodynamic or
excess porewater pressures.7. Earthquake-induced displ acements of retaining wal l s
can be estimated by proceduresanal ogous to the Newmark sl iding bl ock anal ysis for
displ acements of sl opes. Refinementsto this approach al l ow consideration of factors
such as backfil l ampl ification,wal l til ting, kinematic constraints, and vertical
accel eration. Procedures for estimationof probabil ities of various displ acement
l evel s are al so avail abl e.8. Seismic design of retaining wal l s is general l y based
on seismic pressures or al l owabl edispl acements. In the former approach,
pseudostatic or pseudodynamic anal yses areused to estimate seismical l y induced wal l
pressures, and the wal l is designed to resistthose pressures without fail ing or
causing fail ure of the surrounding soil . The l atterapproach invol ves designing the
wal l such that its seismical l y induced permanent displ acementdoes not exceed a
predetermined al l owabl e displ acement.9. The majority of observed fail ures of
retaining wal l s during earthquakes have occurredin waterfront areas; many have
invol ved l iquefaction of the backfil l . Excessive permanentdispl acements of anchored
bul kheads in waterfront areas have often been associatedwith inadequate wal l
embedment depths and/or insufficient anchor capacity.504 Seismic Design of
Retaining Wal l sHOMEWORK PROBLEMSChap. 1111.1 Compute the static thrust and
overturning moment at the base of the concrete retaining wal l shown bel ow.412 ft
Figure PIl .Ic=o<iJ =38°y = 120 pef11.2 The wal l shown in Probl em 11.1 is subjected
to earthquake shaking with a peak horizontal accel eration of 0.28g. Assuming a
pseudostatic coefficient, kh = 0.3 amax/g, estimate (a) thetotal active thrust
acting on the wal l , (b) the height of the resul tant of the total active thrust, and
(c) the total overturning moment about the base of the wal l .11.3 Using the
Steedman-Zeng approach, pl ot the distributions of maximum and minimum l ateral
pressure if the wal l shown bel ow is subjected to a 0.15g harmonic base motion at
(a) a frequencyof 0.5 Hz, and (b) a frequency of 5.0 Hz.8mc=o<iJ= 36°15 = 24°P =2.0
Mg/m3Figure PIl .311.4 A 12 ft deep rigid basement wal l is backfil l ed with dense
gravel l y sand: c = 0, <l > = 45°, Y= 140pcv. Estimate the maximum dynamic thrust
acting on the wal l when subjected to a groundmotion with amax = 0.33g.11.5 Compute
the yiel d accel eration for the concrete gravity wal l shown bel ow.Gl ap.l 1 Homework
Probl ems 5058m1 m-11-I.. 3m ..IFigure Pl l .58c=ol i> =35°P = 1.9 Mg/m311.6 Using the
Richards-El ms procedure, estimate the permanent displ acement of the wal l ofProbl em
11.5 when subjected to (a) the Gil roy No.1 (rock) ground motion, and (b) the Gil roy
No.2(soil ) ground motion.11.7 Using the Whitman-Liao procedure, estimate the
permanent displ acement of the wal l of Probl em11.5 when subjected to (a) the Gil roy
No.1 (rock) ground motion, and (b) the Gil roy No.2(soil ) ground motion.U.S Estimate
the permanent displ acement of the wal l in Probl em 11.5 that woul d have a 10%
probabil ityof exceedance if subjected to the Gil roy No.2 (soil ) motion.U.9 A 15 ft
tal l gravity wal l is to retain a l evel sand backfil l (c = 0, l IJ = 35°, Y= 115
pcf). The wal l wil l be pl aced on a l ayer of dense sandy gravel (c = 0, l IJ = 44°, Y=
140 pcf). Determine the wal l weight that woul d be required to l imit the permanent
displ acement of the wal l to l /4-inch ifsubjected to the Gil roy No.2 (soil ) ground
motion.n.10 The anchored bul khead shown bel ow is l ocated in a seismical l y active
area. Negl ecting theeffect of vertical motions, estimate the minimum peak
accel eration l evel s that woul d be l ikel yto produce (a) moderate damage, and (b)
severe damage.13 m-- • • 10 m c=ol i> =35°P =1.78 Mg/m38mFigure Pl l .1012Soil
Improvement forMitigation of Seismic Hazards12.1 INTRODUCTIONSoil s have been
modified to improve their engineering properties for hundreds of years. Inthe past
75 years, however, improved knowl edge of soil behavior and geotechnical hazardshas
l ed to the devel opment and verification of many innovative soil improvement
techniques.Increased recognition of seismic hazards and improved understanding of
the factorsthat control them have l ed these techniques to be appl ied to the
mitigation of seismic hazardsin the past 30 years.In both
seismical l y active and inactive areas, soil improvement techniques are commonl y
used at sites where the existing soil conditions are expected to l ead to
unsatisfactoryperformance. Unsatisfactory performance can take many forms, but
usual l y invol ves unacceptabl yl arge soil movements. The movements may incl ude
horizontal or vertical (or both)components and may take pl ace during and/or after
earthquake shaking. In the absence ofearthquake shaking, unacceptabl e movements
usual l y resul t from insufficient soil strengthand/or stiffness. Consequentl y, most
soil improvement techniques were devel oped toincrease the strength and stiffness of
soil deposits. These techniques are described in detail in a number of useful
references (e.g., Wel sh, 1987; Van Impe, 1989; Hausmann, 1990;506Sec. 12.2
Densification Techniques 507Broms, 1991; Bel l , 1993; Mosel y, 1993); those which are
not commonl y used for mitigationof seismic hazards are not discussed in this
chapter.During earthquakes, other factors can contribute to unacceptabl e
performance. Inparticul ar,the buil dup of excess porewater pressure can l ead to very
l arge deformations. Consequentl y,commonl y used techniques for mitigation of seismic
hazards often invol vereducing the tendency of the soil to generate positive excess
porewater pressure duringearthquake shaking as wel l as increasing the strength and
stiffness of the soil .The topic of soil improvement is somewhat different from
topics presented in earl ierchapters. Advances in soil improvement technol ogy have
general l y resul ted from the initiativeand imagination of contractors. Research and
expl anatory "theories" have fol l owed,rather than l ed, impl ementation; for some
widel y used techniques, proven theories have yetto be devel oped. In such cases,
indirect or empirical evidence must be rel ied upon and thestudy of case histories
is particul arl y important. This chapter does not attempt to describe al l avail abl e
soil improvement techniques in detail ; instead, it presents an introduction to the
soil improvement techniques that are most commonl y used for mitigation of seismic
hazards.References to more compl ete descriptions of the techniques are presented.
Becausesoil improvement technol ogy changes rapidl y as new techniques are devel oped
and existingtechniques are tested by actual earthquakes, the rel evant geotechnical
engineering l iteratureshoul d be reviewed on a regul ar basis. Methods for
verification of the effectiveness of soil improvement techniques are al so described.
At present, a wide variety of soil improvement techniques are avail abl e for
mitigationof seismic hazards. The costs of these methods vary widel y, and the
conditions under whichthey can be used are infl uenced by the nature and proximity
of structures and constructedfacil ities. On the basis of the mechanisms by which
they improve the engineering propertiesof the soil , the most common of these can be
divided into four major categories: densificationtechniques, reinforcement
techniques, grouting/mixing techniques, and drainage techniques.However, not al l
soil improvement techniques fal l neatl y into a singl e category.12.2 DI: NSIFICATION
TECHNIQUESThe particl es that comprise a particul ar soil can be arranged in many
different ways. However,the strength and stiffness of the soil is higher when the
particl es are packed in a denseconfiguration than when they are packed l oosel y.
Al so, the tendency to generate positiveexcess porewater pressure due to cycl ic
l oading is l ower when the soil is dense than whenit is l oose. As a resul t,
densification is one of the most effective and commonl y used meansof improving soil
characteristics for mitigation of seismic hazards. At the same time, itshoul d be
recognized that the increased stiffness of a densified soil deposit wil l cause it
torespond differentl y to earthquake motion; displ acement ampl itudes are l ikel y to
decrease,but accel erations may be somewhat greater than they woul d have been had
the soil not beenimproved.Densification produces permanent vol ume changes that
often resul t in settl ement ofthe ground surface. Different densification techniques
produce different amounts of508 Soil Improvement for Mitigation of Seismic Hazards
Chap. J2settl ement; with some techniques additional soil is pl aced at or beneath
the ground surfaceduring the process of densification in order to minimize the
settl ement. Despite such efforts,some densification techniques are l imited to sites
without existing structures or facil itiesthat coul d be damaged by ground
settl ement.The most common approaches to densification incl ude vibro techniques,
dynamiccompaction, bl asting, and compaction grouting. Of these techniques, the
first three makeuse of the tendency of granul ar soil s to densify when subjected to
vibrations. As such, theireffectiveness is greatest for cohesionl ess soil s such as
cl ean sands and gravel s. Just as finestend to inhibit l iquefaction during
earthquakes, they tend to inhibit densification by vibration.12.2.1 Vibro
TechniquesVibro techniques use probes that are vibrated through a soil deposit in a
grid pattern to densifythe soil over the entire thickness of the deposit. Vibro
techniques can be divided intothose based on horizontal vibration (vibrofl otation)
and those based on vertical vibration(vibro rod systems). Vibro techniques are
among the most commonl y used techniques formitigation of seismic hazards.12.2.1.1
Vibrofl otationIn vibrofl otation, a torpedol ike probe (the vibrofl ot) suspended by a
crane is used todensify a soil deposit. Vibrofl ots, usual l y 12 to 18 in. (30 to 46
em) in diameter and about10 to 16 ft (3.0 to 4.9 m) l ong, contain weights mounted
eccentrical l y on a central shaftdriven by el ectric or hydraul ic power (Figure
12.1).The vibrofl ot is initial l y l owered to the bottom of the deposit by a
combination ofvibration and water or air jetting through ports in its pointed nose
cone. The vibrofl ot is thenincremental l y withdrawn in 2 to 3 ft (60 to 90 em)
interval s at an overal l rate of about 1ft/min (30 ern/min) whil e stil l vibrating.
Water may be jetted though ports in the upper partof the vibrofl ot to l oosen the
soil above the vibrofl ot temporaril y and aid in its withdrawal .The vibrations
produce a l ocal ized zone of temporary l iquefaction that causes the soil surrounding
the vibrofl ot to densify. A conical depression usual l y forms at the ground surface
above the probe. This depression can be fil l ed with granul ar material (such as
cl ean sand orgravel ) as the vibrofl ot is withdrawn. Al ternativel y, vibrofl ots with
bottom-feed systemscan introduce granul ar material through the tip of the
vibrofl ot. As the vibrofl ot is removed,it l eaves behind a densified col umn of soil .
When gravel or crushed stone is introduced intothe soil , the resul ting stone col umn
provides benefits of reinforcement and drainage in additionto densification. The
use of bottom-feed systems has increased rapidl y in recent years.Air del ivery
systems have become quite common and tend to be preferred over water del ivery
systems at congested sites and in environmental l y sensitive areas.Vibrofl otation is
most effective in cl ean granul ar soil s with fines contents l ess than20% and cl ay
contents bel ow 3%. In such soil s it typical l y produces high densities (rel ative
densities of about 100%) within 12 to 18 in. (30 to 46 cm) of the vibrofl ot and
l ower densitiesat greater radial distances. To densify an entire site,
vibrofl otation is performed in agrid pattern with a spacing that depends on the
soil conditions and the power of the vibrofl ot;spacings of 6 to l Oft (2 to 3 m) are
common. Vibrofl otation has been used successful l ySec. 12.2 Densification Techniques
5091.. ----- Vibrator section'=1<11----- Eccentric weightD~ Water hoses =1 '\~---
El ectric cabl e1f..'It----- El ectric motor(a) (b)Figure 12.1 (a) Schematic
il l ustration of a typical vibrofJot (after Bel l , 1993), and (b)vibrofJots
densifying l iquefiabl e soil s at a wastewater treatment facil ity in Cal ifornia(photo
courtesy of Hayward Baker).to densify soil s to depths of up to 115 ft (35 m). Case
histories of vibrofl otation have beenreported by Harder et al . (1984) and Dobson
(1987).12.2.1.2 Vibro RodVibro rod systems use a vibratory pil e driving hammer to
vibrate a l ong probe into thesoil . The probe is then withdrawn whil e stil l being
vibrated to densify the soil . To minimizecl ensification-induced settl ement,
additional soil may be introduced at the ground surface or510 Soil Improvement for
Mitigation of Seismic Hazards Chap. 12at depth. Several types of probes have been
used in vibrocompaction. In the Terraprobe system,a 30-in. (76-cm) open-ended steel
pipe is vibrated into the ground; the vibrations densifythe soil both inside and
outside the pipe. The Yibro- Wing consists of a central rod withdiametrical l y
opposed 31-in. (80-cm) "wings" spaced 19 in. (50 em) apart al ong the l engthof the
rod (Figure 12.2). The Franki Y-probe consists of three 19-in. (50-cm)-wide steel
pl ates wel ded to a central rod at 1200 angl es from each other. Horizontal cross-
ribs may bewel ded to the faces of the steel pl ates to facil itate densification. By
adjusting the frequencyof vibration, these probes can be "tuned" to the resonant
frequency of the soil -probe systemto increase vibration ampl itudes and densify the
soil more efficientl y (Massarch, 1991).Vibro rod systems are most effective in
soil s simil ar to those for which vibrofl otationis most effective. Because vibro
rods use vertical vibrations, their radius of infl uence is usual l ysmal l er than that
observed for vibrofl otation. As a resul t, the grid spacing for soil improvement by
vibro rods is general l y smal l er than for vibrofl otation. The effectiveness ofvibro
rods al so appears to vary with depth (Janes, 1973). Case histories ofvibro rod
systemshave been reported by Massarch (1991),
Neel y and Leroy (1991), and Senneset andNestvol d (1992).~;J// \g Vibratory hammer
1Vertical t vibrationRodCompactedfs;C: Jb=?J~- zoneFigure 12.2 Vibro-wing system.
Eachpair of wings is oriented at a 1200 angl e tothose l ocated immediatel y above and
bel ow. After Broms (1991).12.2.2 Dynamic CompactionDynamic compaction is performed
by repeatedl y dropping a heavy weight in a grid patternon the ground surface
(Figure 12.3). The weights, usual l y constructed of steel pl ates and/orreinforced
concrete, general l y range from 6 to 30 tons (53 to 267 kN), al though weights ofup
to 170 tons (1500 kN) have been used. Drop heights usual l y range from about 35 to
100Sec. 12.2 Densification Techniques 511Figure 12.3 Aerial view of siteundergoing
soil improvement by dynamiccompaction. The grid pattern on which thedynamic
compaction weight is dropped, andthe need for subsequent regrading andsurface
compaction, are evident from thisphotograph. (Photo courtesy of HaywardBaker).ft
(10 to 30 m), al though weights have been dropped from up to 130 ft (40 m). The
weightsare usual l y dropped three to eight times before moving to the next point on
the grid. Adetail ed description of dynamic compaction was prepared by Lukas (1986).
At a particul ar site, dynamic compaction is general l y performed in several stages,
orpasses. Empirical evidence suggests that the effective depth of infl uence (the
depth to whichsignificant improvement can be detected) increases with impact energy
and that the greatestdegree of improvement is usual l y observed at about hal f the
effective depth of infl uence(Mayne et al ., 1984). To avoid devel oping a shal l ow
zone of dense soil that coul d inhibit thetransmission of energy to greater depths,
the deepest soil is densified first with a series ofhigh-energy (heavy weight
ancl /or high drop height) drops on a widel y spaced grid. Afterthe craters produced
by the first pass have been fil l ed (preferabl y with wel l -graded granul arsoil ),
soil s at intermediate depth are then compacted using a greater number of drops from
a smal l er height at cl oser spacing (often hal f the spacing of the original grid).
Final l y, the512 Soil Improvement for Mitigation of Seismic Hazards Chap. 12near-
surface soil s are compacted by dropping rel ativel y l ight weights on a virtual l y
continuouspattern to smooth or "iron" the ground surface. Additional smoothing by
conventional surface grading and compaction equipment is usual l y required.The
kinetic energy of the weight at impact produces stress waves that travel throughthe
soil . The total energy del ivered to the soil is a function of the weight, drop
height, gridspacing, and number of drops per grid point. When the groundwater tabl e
is near the surface,pl acement of a gravel or sand bl anket may be required prior to
compaction. Al thoughdynamic compaction has been used successful l y for cohesive
soil s, its most common use formitigation of seismic hazards is for potential l y
l iquefiabl e soil s. At each grid point, a seriesof drops causes the porewater
pressure to increase so that the soil particl es can more easil ymove into a denser
configuration. Dissipation of the excess porewater pressure resul ts infurther
densification within a short period (1 to 2 days for sand and gravel s; 1 to 2 weeks
forsandy sil ts) after treatment.Dynamic compaction is general l y effective to depths
of 30 to 40 ft (9 to 12 m),al though extremel y high impact energies may produce
densification at greater depths.Because the process is rather intrusive-it can
produce considerabl e noise, dust, fl yingdebris, and vibration-it is rarel y used
near occupied or vibration-sensitive structures. Casehistories of dynamic
compaction ofpotential l y l iquefiabl e soil have been described by Russinand Al i
(1987), Kel l er et al . (1987), Koutsoftas and Kiefer (1990), Mitchel l and Wentz
(1991), and others.12.2.3 Bl astingLoose granul ar soil s have al so been compacted by
bl asting. Bl asting densification invol vesthe detonation of mul tipl e expl osive
charges vertical l y spaced 10 to 20 ft (3 to 6 m) apart indril l ed or jetted
borehol es. The borehol es are usual l y spaced between 15 to 50 ft (5 to 15 m)apart
and backfil l ed prior to detonation. To increase the efficiency of the densification
process,the charges at different el evations may be detonated at smal l time del ays.
Immediatel yafter detonation, the ground surface rises and gas and water are
expel l ed from fractures(Figure 12.4). The ground surface then settl es as the excess
gas and water pressure dissipates.Al though the efficiency of densification
decreases with each round of bl asting, two orthree rounds (with l ater rounds
detonated at l ocations between those of the earl ier rounds)are often used to
achieve the desired degree of densification.Bl asting is most effective in l oose
sands that contain l ess than 20% sil t and l ess than5% cl ay. Even smal l amounts of
cl ay, or smal l cl ay seams, can substantial l y reduce theeffectiveness of bl asting.
Bl asting can be effective in dry soil s, but the effects of capil l arytension and gas
bubbl es in partial l y saturated soil s virtual l y negates its effectiveness. As a
resul t, bl asting is most commonl y used to densify compl etel y saturated soil s. In
such soil sthe shock wave produced by the charges produces l ocal ized, temporary
l iquefaction whichal l ows the soil grains to move into a more dense configuration.
Al though bl asting is quite economical , its use is l imited by several practical
considerations.It produces strong vibrations that may damage nearby structures or
produce significantground movements. It requires the use of potential l y hazardous
expl osives forwhich strict regul ations on handl ing and storage usual l y appl y.
Final l y, its effectiveness isdifficul t to predict in advance. Case histories of the
use of bl asting to mitigate seismicSec. 12.2 Densification Techniques 513Figure
12.4 Ground surface shortl y after detonation of expl osives during bl ast
densification of l oose soil beneath an abutment prior to construction of Col dwater
Creekbridge near Mt. St. Hel ens in Washington state. (Photo by A. P. Kil ian; used
withpermission).hazards have been described by Kl ohn et al . (1981), Sol ymar and
Reed (1986), LaFosse andvon Rosenvinge (1992), and Hachey et al . (1994).12.2.4
Compaction GroutingSoft or weak soil s can be densified by injecting a very l ow
sl ump [general l y l ess than 1 in.(2.5 cm)] grout into the soil under high pressure,
a process known as compaction grouting.Because the grout is highl y viscous, it
forms an intact bul b or col umn that densifies the surroundingsoil by displ acement
(Figure 12.5). Compaction grouting may be performed at aseries of points in a grid
or al ong a l ine. Grout point spacings ranging from 3 to 15 ft ( 1 to4..6 m) have
been used. Because higher overburden pressures al l ow the use of higher grout
pressures, l arger spacings are general l y used when treating deeper soil s. At
shal l ow depths,compaction grouting may be used to l ift settl ed sl abs or structures;
indeed, remediation offoundation settl ement is probabl y the most common appl ication
of compaction grouting.Compaction grouting may be performed from the top down
(downstage grouting) orfrom the bottom up (upstage grouting). Upstage grouting is
l ess expensive and more commonl yused than downstage grouting. However, the
downstage procedure is preferred(Stil l ey, 1982; Bel l , 1993) for underpinning of
structures or for sites where l oose soil s514 Soil Improvement for Mitigation of
Seismic HazardsPumpChap. J2PressuregaugeCompactiongrout bul bFigure 12.5 Compaction
grouting. Lowsl umpgrout is pumped under high pressureto form a bul b that displ aces
and densifiesthe surrounding soil s. By raising the grouttube whil e pumping, a
col umn of grout canbe created in the soil . (After Hausmann,1990.)extend to the
ground surface. By working from the top down, pl acement of an upper groutbul b
reduces the possibil ity of subsequent grout escaping at the surface and grout
heave, andal so provides additional strength and confinement that al l ows the use of
higher groutingpressures at greater depths.Because it does not rel y on vibration,
compaction grouting can be used in al l soil types. It is most commonl y used in sands
and nonpl astic sil ts. Compaction grouting can beused to virtual l y any depth and can
easil y be used within a given range of depths. The sizeand shape of the grout bul b
or col umn is infl uenced by the stiffness and strength of the soil and al so by the
rate and pressure at which the grout is injected. An important feature of
compactiongrouting is that its greatest effects occur where the soil is softest and
weakest. Compactiongrout masses with diameters greater than 3 ft (1 m) are not
uncommon (Warner,1982). Compaction grouting has been used to depths of 100 ft (30
m). Case histories of compactiongrouting have been described by Sal l ey et al .
(1987), Warner (1982), Graf (1992),and Baez and Henry (1993).12.2.5 Areal Extent of
DensificationAn important consideration in the densification of soil s for
construction of individual structuresand foundations is the areal extent of soil
improvement required for satisfactory performanceduring earthquakes. The areal
extent shoul d be eval uated on a case-by-case basissince site-specific soil
conditions, performance requirements, and fail ure consequencesmust be addressed.The
required areal extent of improvement depends on the mechanism of fail ure thatthe
improvement is intended to el iminate. For potential stabil ity fail ures, the areal
extent ofimprovement wil l depend on the degree of improvement that can be achieved
and on theextent ofthe potential fail ure surface(s). By estimating the residual
strength ofthe soil afterimprovement, stabil ity anal yses can be used to estimate
the extent of improvement that wil l produce an acceptabl e l evel of stabil ity. To
minimize postearthquake settl ement
of a structureor foundation on l oose, saturated sand, densification is usual l y
performed within a zonedefined by a 30 to 45° l ine from the edge of the structure,
as il l ustrated in Figure 12.6. Avail abl eresearch and fiel d experience indicates
that this approach is l ikel y to produce a satisfactoryextent of improvement (Iai et
al ., 1988).Sec. 12.3 Reinforcement TechniquesStructure515Liquefiabl esoil Zone of
improvementFigure 12.6 Typical areal extent of improvement for densification of
potential l yl iquefiabl e soil beneath a structure.12.3 REINFORCEMENT TECHNIQUESIn
some cases it is possibl e to improve the strength and stiffness of an existing soil
depositby instal l ing discrete incl usions that reinforce the soil . These incl usions
may consist ofstructural material s, such as steel , concrete, or timber, and
geomaterial s such as densifiedgravel . Reinforcement of new engineered fil l s using
geosynthetic or metal l ic reinforcementis beyond the scope of this chapter.12.3.1
Stone Col umnsSoil deposits can be improved by the instal l ation of dense col umns of
gravel known as stonecol umns. Stone col umns may be used in both fine- and coarse-
grained soil s. In fine-grainedsoil s, stone col umns are usual l y used to increase
shear strength beneath structures andembankments by accel erating consol idation (by
al l owing radial drainage) and introducingcol umns of stronger material . For
mitigation of seismic hazards, they are commonl y usedfor improvement of l iquefiabl e
soil deposits.Stone col umns can be instal l ed in a variety of ways. As discussed
previousl y, stonecol umns may be constructed by introducing gravel during the
process of vibrofl otation(Brown, 1977). Several other methods of instal l ation are
al so avail abl e. In the Frankimethod, a steel casing initial l y cl osed at the bottom
by a gravel pl ug is driven to the desireddepth by an internal hammer (Figure 12.7).
At that depth, part of the pl ug is driven beyondthe bottom of the casing to form a
bul b of gravel . Additional gravel is then added and compactedas the casing is
withdrawn. The diameter of the resul ting stone col umn depends onthe stiffness and
compressibil ity of the surrounding soil ; in l oose sand, a 0.5- to 0.7-m (19to 28
in.) casing wil l typical l y produce a 0.8-m (31 in.) diameter col umn. Casings with
trapdoors at the bottom have al so been used to instal l stone col umns (Sol ymar and
Reed, 1986).The trap door al l ows the casing to be driven as a cl osed-end pil e but
al so al l ows gravel to bepl aced during withdrawal of the casing. The gravel can be
densified by pausing to redrivethe casing at various interval s during the
withdrawal process.Stone col umns combine at l east four different mechanisms for
improvement of l iquefiabl esoil deposits. First, they improve the deposit by virtue
of their own high density,strength, and stiffness-in this sense they reinforce the
soil deposit. Second, they providecl osel y spaced drainage boundaries that inhibit
the devel opment of high excess porewater516 Soil Improvement for Mitigation of
Seismic Hazards Chap. 12Figure 12.7 Methods of stone col umninstal l ation: (a) Franki
method; (b) by(b) vibrofl otation. (After Broms, 1991.)Internal drophammerHopper1(a)
..... : ;: .....Stonecompactedby pounding<, ,Nd\ ~..,.): .-.~~~,·: : ·.I~·;·: ~ : ~\JJ
.. : ._.''- -pressures (Section 12.6). Third, the processes by which they are
instal l ed densify the surroundingsoil by the combined effects of vibration and
displ acement. Final l y, the instal l ationprocess increases the l ateral stresses in
the soil surrounding the stone col umns. Thesemul tipl e benefits have made the use of
stone col umns very popul ar. Case histories of seismichazard mitigation by stone
col umns have been presented by Priebe (1991), Hayden andWel ch (1991), and Mitchel l
and Wentz, (1991).12.3.2 Compaction Pil esGranul ar soil s can be improved by the
instal l ation of compaction pil es. Compaction pil esare displ acement pil es, usual l y
prestressed concrete or timber, that are driven into a l oosesand or gravel deposit
in a grid pattern (Figure 12.8) and l eft there. Compaction pil esimprove the seismic
performance of a soil deposit by three different mechanisms. First, thefl exural
strength of the pil es themsel ves provides resistance to soil movement
(reinforcement).Second, the vibrations and displ acements produced by their
instal l ation cause densification.Final l y, the instal l ation process increases the
l ateral stresses in the soil surrounding the pil es.Compaction pil es general l y
densify the soil within a distance of 7 to 12 pil e diameters(Robinsky and Morrison,
1964; Kishida, 1967), and consequentl y, are usual l y instal l ed in agrid pattern.
Between compaction pil es, rel ative densities of up to 75 to 80% are usual l yachieved
(Sol ymar and Reed, 1986). Improvement can be obtained with reasonabl e economyto
depths of about 60 ft. Case histories describing the use of compaction pil es have
been presented by Lindqvist and Petaja (1981), Marcuson et al . (1991), Mitchel l and
Wentz(1991), and Kramer and Hol tz (1991).12.3.3 Dril l ed Incl usionsStructural
reinforcing el ements can al so be instal l ed in the ground by dril l ing or augering.
Dril l ed shafts, sometimes with very l arge diameters, have been used to stabil ize
manysl opes. Such shafts may be instal l ed cl osel y enough to form tangent or secant
pil e wal l s. Soil Sec. 12.4 Grouting and Mixing Techniques 517Figure 12.8 Compaction
pil es driven into the upstream embankment of Sardis Dam toreduce l iquefaction
hazards. The contractor drove the pil es to this l evel with a bargemounted
conventional hammer; the pil es were l ater driven bel ow the water surface witha
different hammer. (Photo by T. D. Stark; used with permission.)nail s, tiebacks,
micropil es, and root pil es have al so been used. The instal l ation of suchdril l ed
incl usions can be quite difficul t, however, in the l oose granul ar soil s that
contributeto many seismic hazards. Al though soil nail ed wal l s performed wel l in the
1989 Lama Prietaearthquake (Fel io et al ., 1990), there is currentl y no consensus on
their design for seismicl oading.12.4 GI1~OUTINGAND MIXING TECHNIQUESThe engineering
characteristics of many soil deposits can be improved by injecting or mixing
cementitious material s into the soil . These material s both strengthen the contacts
between soil grains and fil l the void space between the grains. Grouting techniques
invol vethe injection of such material s into the voids of the sail or into fractures
in the soil so that theparticl e structure of the majority of the soil remains
intact. Mixing techniques introducecementitious material s by physical l y mixing them
with the soil , compl etel y disturbing theparticl e structure of the soil . The mixing
can be accompl ished mechanical l y or hydraul ical l y.Grouting and mixing techniques
tend to be expensive but can often be accompl ishedwith minimal settl ement or
vibration. As a resul t, grouting and mixing techniques can oftenbe used in
situations where other soil improvement techniques cannot.51812.4.1 GroutingSoil
Improvement for Mitigation of Seismic Hazards Chap. J2The term grouting is used to
describe a variety of processes by which cementitious material is introduced into
the ground. Grouting techniques are often cl assified according to themethod by
which the grout is pl aced in the ground (Hausmann, 1990). In this chapter, however,
soil improvement techniques are cl assified according to the primary mechanisms by
which they produce improvement. As a resul t, compaction grouting is described with
otherdensification techniques in Section 12.3, and jet grouting is considered as a
mixing techniquein the fol l owing section. With this convention, there are two
primary types of groutingtechniques.12.4.1.1 Permeation GroutingPermeation grouting
invol ves the injection of l ow-viscosity l iquid grout into thevoids of the soil
without disturbing the soil structure (Figure 12.9). Particul ate grouts (i.e.,
aqueous suspensions of cement, fl y ash, bentonite, microfine cement, or some
combinationthereof) or chemical grouts (e.g., sil ica and l ignin gel s, or phenol ic
and acryl ic resins) maybe used.The suitabil ity of different types of grouts for
different soil conditions is moststrongl y infl uenced by the grain size of the soil .
Virtual l y any type of grout, even rel ativel yviscous cement grouts, can be used in
soil with l arge voids such as gravel s and coarse sands.Chemical grouts general l y
exhibit l ower viscosity than particul ate grouts (al though the viscosityof microfine
cements grouts may be as l ow as some chemical grouts) and can thereforebe used in
fine sands. The presence of fines can significantl y reduce the effectivenessof
permeation grouting.Grout pipes are typical l y instal l ed in a grid pattern at
spacings of 4 to 8 ft ( 1.2 to 2.4 m)(Hayden, 1994). The grout may be injected in
different ways. In stage grouting, a boring isadvanced a short distance before
grout is injected through the end of the dril l rod. After thegrout sets up, the
boring is advanced another short distance and grouted again. This processcontinues
until grout has been pl aced to the desired depth. In the tube-a-manchette approach,
a grout tube with injection ports every 12 to 24 in. (30 to 61 cm) al ong its l ength
is instal l edin a borehol e. Rubber sl eeves (manchettes) that serve as one-way val ves
cover the injectionports on the outer surface of the grout tube and internal packer
systems are used to control thedepths at which grout is injected.PumpPressuregauge
Grout penetrates~tt. sol l skel etonFigure 12.9 Permeation grouting. (AfterHausmann,
1990).Sec. 12.4 Grouting and Mixing Techniques 519Permeation grouting produces soil
improvement by two primary mechanisms. Firstthe grout tends to strengthen the
contacts between individual soil grains, thereby producinga soil skel eton that is
stronger and stiffer than that of the ungrouted
soil . Second, the grouttakes up space in the voids between soil particl es,
reducing the tendency for densification(or excess pore pressure generation) upon
cycl ic l oading. Soil s improved by permeationgrouting can have shear strengths of 50
to 300 psi (345 to 2070 kPa). Case histories in whichpermeation grouting was used
to mitigate seismic hazards were described by Zacher andGraf (1979), Graf (1992),
and Bruce (1992).12.4.1.2 Intrusion GroutingIn the process of intrusion grouting,
fl uid grout is injected under pressure to cause control l edfracturing of the soil
(Figure 12.10). Because the grout is not intended to fl ow throughthe smal l voids
between soil particl es, rel ativel y viscous (and strong) cement grouts can beused.
In theory, the first fractures shoul d be paral l el to the minor principal stress
pl anes, butobservations show that they usual l y fol l ow weak bedding pl anes. After
al l owing the initial l ypl aced grout to cure, repeated intrusion grouting fractures
the soil al ong different pl anes.Eventual l y, a three-dimensional network of
intersecting grout l enses can be formed. Somedensification of the soil may occur,
but the primary mechanism of improvement resul ts fromthe increased stiffness and
strength of the soil mass due to the hardened l enses of grout.Pump" Pressuregauge
Figure 12.10 Intrusion grouting process.(After Hausmann, 1990).12.4.2 Mixing
Local ized improvement of soil col umns can be achieved by in situ mixing of the soil
withcementitious material . Because the cementitious material is physical l y mixed
with the soil ,it need not have an extremel y l ow viscosity-strong, cement sl urries
are commonl y used.For mitigation of seismic hazards, this approach is most commonl y
accompl ished by soil mixing and jet grouting.12.4.2.1 Soil MixingThe term soil
mixing describes a specific technique in which cementitious material ismechanical l y
mixed into the soil using a hol l ow stem auger and paddl e arrangement (Figurel : U 1).
Soil mixing rigs may have singl e augers (0.5 to 4 m (1.6 to 13 ft) in diameter) or
520 Soil Improvement for Mitigation of Seismic Hazards Chap. 12Figure 12.11 Tripl e
auger soil mixing rig improving l iquefiabl e soil s at Jackson LakeDam in Wyoming.
Grout batch pl ant is at right (photo courtesy of SMW Seiko, Inc.)gangs of two to
eight augers (usual l y about 1 m (3.1 ft) in diameter). As the mixing augersare
advanced into the soil , grout is pumped through their stems and injected into the
soil attheir tips. The grout is thoroughl y mixed with the soil by the auger fl ights
and mixing paddl es.After the design depth has been reached, the augers are
withdrawn whil e the mixingprocess continues. The soil mixing process l eaves behind
a uniform (constant width) col umnof soil -cement. By overl apping the col umns before
the grout cures, wal l s and cel l ul arstructures can be constructed bel ow the ground
surface.Soil mixing can be used in virtual l y any type of inorganic soil . It has
been used todepths of over 20 m (66 ft) in the United States and up to 60 m (200
ft) in Japan. The strengthof the soil -cement mixture depends on the type of grout,
type of soil , and degree of mixing;strengths of 200 psi (1380 kPa) or more are
commonl y achieved. Case histories invol vingthe use of soil mixing for mitigation of
seismic hazards have been presented by Ryan andJasperse (1989), Babasaki et al .
(1991), and Taki and Yang (1991).12.4.2.2 Jet GroutingIn jet grouting, the soil is
mixed with cement grout injected horizontal l y under highpressure in a previousl y
dril l ed borehol e (Figure 12.12). The injection nozzl e is rotated toSec. 12.5
Drainage TechniquesPumpGroutjet521Figure 12.12 Jet grouting process. (After
Hausmann, 1990).al l ow the grout to be pl aced in al l directions. Air or air and
water may al so be injected to aidin the mixing process. Jet grouting begins at the
bottom of the borehol e and proceeds to thetop, l eaving behind a rel ativel y uniform
col umn of mixed soil -cement. By overl apping thecol umns before the grout cures,
wal l s and cel l ul ar structures can be constructed bel ow theground surface.The
diameter of a jet grouted col umn depends on the soil condition and the manner in
which the jet grouting is performed. Col umn diameters are general l y greater in
coarsegrainedsoil s than in fine-grained soil s. By varying the air, water, and grout
pressures andthe rates of rotation and l ifting of the grout tubes, a jet grouting
operator can control theeffective dimensions of the col umn. Diameters ranging from
0.4 to 0.5 m (16 to 20 in.) incl ayey sil t to 0.9 to 1.0 m (36 to 39 in.) in sandy
gravel can be expected using a singl e-jet(grout onl y) system (Bel l , 1993).
Diameters of 0.8 to 1.0 m (31 to 39 in.) in cl ayey sil t and2.0 to 2.4 m (6.5 to 7.9
ft) in sandy gravel can be expected with a tripl e-jet (air, water, andgrout)
system. Jet grouting can be performed in any type of inorganic soil to depths
l imitedonl y by the range of the dril l ing equipment. Case histories of the use of
jet grouting for mitigationof seismic hazards have been presented by Hayden (1994).
12.5 DRAINAGE TECHNIQUESUnacceptabl e movements of sl opes, embankments, retaining
structures, and foundationscan frequentl y be el iminated by l owering the groundwater
tabl e prior to earthquake shaking.A number of dewatering techniques have been
devel oped and proven useful in engineeringpractice. Procedures for the design of
dewatering systems are wel l establ ished andwidel y used (e.g., Cedergren, 1989;
Powers, 1992). These standard techniques may be usedto increase the stiffness and
strength of a soil deposit for mitigation of seismic as wel l asnonseismic hazards.
The buil dup of excess porewater pressure during earthquake shaking can be
suppressedusing drainage techniques, al though drainage al one is rarel y rel ied upon
for the mitigationof l iquefaction hazards. The instal l ation of stone col umns, for
exampl e, introduces522 Soil Improvement for Mitigation of Seismic Hazards Chap. J2
col umns offreel y draining gravel into a l iquefiabl e soil deposit (though mixing of
the gravel and the native soil during instal l ation may reduce the permeabil ity of
the stone col umn).Earthquake-induced excess pore pressures may be rapidl y
dissipated by horizontal fl ow ofporewater into the stone col umns. The rate of pore
pressure dissipation depends on thediameter and spacing of the stone col umns and on
the permeabil ity and compressibil ity ofthe surrounding soil . Seed and Booker (1976,
1977) devel oped procedures for sel ecting thesizes and spacings of gravel drains (or
stone col umns) for mitigation of l iquefaction hazards.The use of gravel drains for
suppression of excess porewater pressure requires careful attention to drain
permeabil ity and fil tration behavior of the drain-soil boundary. Eventhough
drainage techniques can mitigate l iquefaction hazards by suppressing excess
porewaterpressure buil dup, postearthquake settl ement may stil l occur. Case
histories of the useof drainage techniques for mitigation of seismic hazards have
been described by Ishihara etal . (1980), Aboshi et al . (1991), and Iai et al .
(1994).12.6 VERIFICATION OF SOIL IMPROVEMENTAl l attempts at soil improvement shoul d
be checked to confirm that the desired improvementhas taken pl ace. The most direct
way of verifying the effectiveness of a particul ar soil improvement technique is to
measure the soil characteristic that was considered deficientboth before and after
improvement. For exampl e, if the improvement was undertaken toincrease the strength
of the soil , measurement of the strength before and after improvementwoul d provide
the most direct verification of the effectiveness of the improvement process.
However, it is not al ways feasibl e to measure the deficient characteristic
directl y. In suchcases, verification is usual l y accompl ished using rel ated
characteristics that are more easil ymeasured.Verification may be based on the
resul ts of l aboratory or fiel d tests. Whil e l aboratorytests have historical l y been
commonl y used for verification of soil improvement, recentadvances in fiel d testing
techniques have provided additional means for verification. Fiel dtesting techniques
may be divided into in situ testing techniques and geophysical testingtechniques.
Common verification techniques were summarized by Ledbetter (1985).12.6.1
Laboratory Testing TechniquesLaboratory testing techniques have a number of
advantages over other methods for verificationof soil improvement, but they al so
suffer from drawbacks that can significantl y l imittheir useful ness for certain
types of soil improvement. The requirement of obtaining a sampl eof the improved
soil l eads directl y to many of the advantages of using l aboratory testingtechniques
and al so to many of the disadvantages. Obtaining a sampl e of improved soil al l ows
visual inspection of the effects of improvement. For many improvement techniques
(e.g., permeation grouting, soil mixing, etc.), the abil ity to inspect the treated
soil providesdirect and val uabl e evidence of the effectiveness of the treatment.
Laboratory tests al l owgreater control and more accurate measurement of stress,
strain, and environmental conditionsthan are possibl e in fiel d tests. In some cases
this fl exibil ity may al l ow more accuratecharacterization of the properties of the
improved soil .Sec. 12.6 Verification of Soil Improvement 523On the other hand,
l aboratory tests onl y provide verification at discrete points. Whensod improvement
is used to improve or el iminate l ocal ized zones or seams of weakness, verification
by methods that require discrete sampl ing may be ineffective. Laboratory tests may
al so be infl uenced by the inevitabl e effects of sampl e disturbance, a probl em that
is particul arl ysignificant in the improvement of l iquefiabl e soil s. The density
changes produced byeven thin-wal l ed sampl ers (Marcuson et aI., 1977; Seko and Tobe,
1977; Singh et al ., 1979)can l ead to considerabl e uncertainty in the eval uation of
improvement effectiveness.1 ~!.6.2 In Situ Testing TechniquesMany of the
l imitations of l aboratory testing based approaches to the verification of soil
improvement effectiveness may be overcome by the use of in situ tests. Indeed, the
use of insitu tests for verification of soil improvement effectiveness has
increased dramatical l y in thepast 15 to 20 years. Because many geotechnical seismic
hazards are eval uated using in situtest parameters, those parameters can provide
direct evidence of hazard mitigation. Indeed,soil improvement specifications may be
written to require that a certain parameter val ue(e.g., a minimum SPT resistance)
be achieved after improvement. Mitchel l (1986) and Wel sh(1986) described the use of
in situ tests for verification of soil improvement effectiveness.The SPT, CPT, PMT,
and DMT (Section 6.3.1.2) can al l be used for verification of soil improvement
effectiveness. The SPT and CPT tests are performed rel ativel y quickl y and
inexpensivel y compared to sampl ing and l aboratory testing. The CPT is particul arl y
useful because it provides a continuous record with depth. The PMT is more
expensive, but it al soal l ows measurement of l ateral stresses and direct measurement
of strength. For gravel l y soil s,the Becker hammer penetration test (Section
6.3.1.2) may be used for verification purposes.Interpretation of soil improvement
effectiveness from in situ test resul ts must be performedcareful l y. The penetration
resistance of granul ar soil s, for exampl e, is infl uenced notonl y by density and
overburden stress, but al so by l ateral stress. Soil improvement techniquesthat
resul t in increased l ateral stress may produce unconservative estimates of the
density of the improved soil if the postimprovement stress state is not careful l y
consideredin the interpretation of penetration test resul ts. Because time-dependent
changes in strength,stiffness, and penetration resistance are often observed after
densification (Mitchel l andSol ymar, 1984; Mitchel l , 1986), in situ tests performed
immediatel y after densification maynot refl ect the actual degree of improvement of
the soil . Verification testing is usual l y performedat l east 72 hours after
densification has taken pl ace. Many soil improvement techniquesare appl ied at a
grid of treatment points, and the degree of improvement usual l ydecreases with
distance from the treatment point. The rel ationship between the l ocations ofin situ
tests and the l ocations of treatment points shoul d be considered in the
interpretationof soil improvement effectiveness from in situ test resul ts. In situ
tests have l imited effectivenessfor verification of grouting effectiveness (Wel sh,
1986).1 ~!.6.3 Geophysical Testing TechniquesMany soil improvement techniques
increase the stiffness of the treated soil . The effectivenessof these techniques
can be verified using seismic geophysical techniques (Section 6.3.1.1). Inmost
cases it is desirabl e to perform seismic tests both before and after improvement.
524 Soil Improvement for Mitigation of Seismic Hazards Chap. J2Cross-hol e and
downhol e (incl uding seismic cone) tests are most commonl y used forverification of
soil improvement. These techniques can measure p- or s-wave vel ocitiesover
considerabl e distances, thereby providing spatial l y averaged stiffness
measurements.However, each requires at l east one borehol e. For sites where soil
improvement has beenperformed over a l arge area, seismic refl ection and seismic
refraction tests may be useful forverification purposes. SASW tests provide simil ar
information without the need for borehol es.At sites where stiffness changes
irregul arl y in two or three dimensions or sites thatcontain incl usions, the resul ts
of SASW tests may be very difficul t to interpret. Such testsmust al so be performed
when background noise (incl uding that produced by on-going soil improvement work)
wil l not adversel y affect their resul ts. Tests that measure average wavepropagation
vel ocities may not accuratel y refl ect the degree of improvement of thin, l oosezones
unl ess the distance over which vel ocities are averaged is quite smal l .12.7 OTHER
CONSIDERATIONSThe appl ication of soil improvement techniques to the mitigation of
seismic hazards is rel ativel ynew. The theoretical underpinnings of many soil
improvement techniques are poorl ydevel oped, and empirical observations of the
performance of improved soil in actual earthquakesare rare. Because of these
factors, it is particul arl y important to review the rel evantgeotechnical
engineering l iterature before attempting to mitigate seismic hazards by soil
improvement.The effectiveness of many soil improvement techniques can be difficul t
to predict inadvance for a particul ar site. Furthermore, the equipment, procedures,
experience, and skil l of the soil improvement contractor can strongl y infl uence soil
improvement effectiveness.For these reasons, it is frequentl y beneficial to
construct test sections before beginning productionwork or even before final
sel ection of a soil improvement technique. Test sectionsal l ow site- and procedure-
specific eval uation of soil improvement effectiveness at a moderatecost; their use
is advisabl e whenever possibl e.12.8 SUMMARY1. Unfavorabl e soil conditions can
frequentl y be improved using soil improvement techniques.A variety of soil
improvement techniques have been devel oped-some appl yto l ong-term, static l oading
conditions and others al so appl y to seismic l oading conditions.2. The cost of
different soil improvement techniques vary widel y. Costs are infl uencedby the
vol ume and extent of the soil to be treated, access to the site, site sensitivity
tovibration and permanent ground movement, and other factors.3. The presence of
existing structures, pipel ines, and other constructed facil ities canel iminate many
soil improvement techniques from consideration at a given site. Thetechniques that
can be used at such sites tend to be among the more expensive.4. Most soil
improvement techniques are intended to increase the strength and stiffnessof a soil
deposit. Increased strength and stiffness is general l y desirabl e for both staticand
seismic l oading conditions.Sec. 12.8 Summary 5255. Current soil improvement
techniques can be divided into four broad categories: densificationtechniques,
reinforcement techniques, grouting/mixing techniques, anddrainage techniques. Not
al l techniques fal l entirel y within a singl e category; forexampl e, stone col umns
can improve a soil deposit by densification, reinforcement,and drainage functions.
6. Several soil improvement techniques that are commonl y used to mitigate seismic
hazardsare intended to reduce the tendency of l oose, saturated granul ar soil s to
generateexcess porewater pressure during earthquake shaking. These techniques
typical l yinvol ve densification of the soil .7. Densification is probabl y the most
commonl y used soil improvement technique formitigation of seismic hazards. Most
densification techniques rel y on the tendency ofgranul ar soil s to densify when
subjected to vibration. Densification can produce substantial settl ement, al though
some procedures al l ow the introduction of new material to bal ance the vol ume change
caused by densification.8. Many densification techniques rel y on vibrations that
can be potential l y damaging tostructures, pipel ines, and other constructed
facil ities. Such vibrations may al so be tooobjectionabl e to peopl e who l ive or work
near sites that require improvement to al l owtheir use.9. Most vibratory techniques
produce a temporary, l ocal ized zone of l iquefaction inl oose, saturated sand.
Densification occurs as the sand particl es are rearranged duringreconsol idation.
The presence of fines, particul arl y pl astic fines, inhibits the devel opmentof high
pore pressures and the rearrangement of soil particl es. As a resul t, vibratory
techniques may have l imited effectiveness in soil s with significant fines contents.
10. Reinforcement techniques introduce discrete incl usions that stiffen and
strengthen asoil deposit. The high stiffness and strength of the incl usions al so
tend to reduce thestresses imposed on the weaker material between the incl usions.
11. Cementitious material s may be injected or mixed into a soil deposit. The
material simprove the soil by strengthening the contacts between individual grains
and fil l ingthe space between the grains.12. Grouting techniques invol ve the
injection of such material s into the voids of the soil orinto fractures in the soil
so that the particl e structure of the majority of the soil remainsintact. In
permeation grouting, very l ow viscosity grouts are injected into the voids ofthe
soil without disturbing the soil structure. In intrusion grouting, thicker and more
viscous grouts are injected under pressure to cause control l ed fracturing of the
soil .13. Mixing techniques introduce cementitious material s by physical l y mixing
them withthe soil , compl etel y disturbing the particl e structure of the soil . The
mixing can beaccompl ished mechanical l y (soil mixing) or hydraul ical l y Get
grouting). Both soil mixing and jet grouting l eave behind rel ativel y uniform col umns
of mixed soil cement.By overl apping the col umns, wal l s or cel l ul ar structures can be
constructedbel ow the ground surface.14. Drainage techniques minimize the buil dup of
porewater pressure during earthquakesby shortening the drainage paths in a soil
deposit. The instal l ation of drains general l yinvol ves some degree of densification
and the drains themsel ves may al so providesome reinforcement.526 Soil Improvement
for Mitigation of Seismic Hazards Chap. 1215. Verification of the effectiveness of
soil improvement is an important part of seismichazard mitigation. Direct or
indirect measurement of stiffness, strength, or densitycharacteristics
both before and after improvement can al l ow rel iabl e eval uation of soil improvement
effectiveness. These characteristics may be measured by l aboratory, insitu, or
geophysical tests. The rel ative advantages and l imitations of these types oftests,
discussed in detail in Chapter 6, appl y to their use in verification appl ications.A
Vibratory MotionA. 1 INTI~ODUCTIONMany different types of dynamic l oading can
induce vibratory motion in soil s and structures.To sol ve probl ems invol ving the
dynamic response of soil s and structures, it is necessaryto be abl e to describe
dynamic events. They can be described in different ways, andthe geotechnical
earthquake engineer must be famil iar with each. This appendix provides abrief
description of vibratory motion and introduces the nomencl ature and mathematical
forms by which it is usual l y described.A.2 TYPES OF VIBRATORY MOTIONVibratory
motion can be divided into two broad categories: periodic motion and nonperiodic
motion. Periodic motions are those which repeat themsel ves at regul ar interval s of
time.Mathematical l y, a motion, u(t), is periodic if there exists some period, Tf ,
for whichu(t + Tf ) = u(t) for al l t. The simpl est form of periodic motion is
simpl e harmonic motion inwhich displ acement varies sinusoidal l y with time.
Nonperiodic motions, which do notrepeat themsel ves at constant interval s, can
resul t from impul sive l oads (e.g., expl osions orfal l ing weights), or from l onger-
duration transient l oadings (e.g., earthquakes or traffic).Exampl es of periodic and
nonperiodic motions are shown in Figure A.I.527528Periodic motionVibratory Motion
Nonperiodic motionApp.Au (t)(a)u (t)(b)u (t)(e)u (t)(d)Figure A.I Periodic and
nonperiodic motion: (a) simpl e harmonic motion; (b) general periodicmotion; (c)
transient motion (response to impact l oading); (d) transient motion (earthquake
groundmotion).Some forms of periodic motion (e.g., Figure A.I b) may appear to be
much more compl exthan simpl e harmonic motion, but with the use of mathematical
techniques describedl ater in this appendix, they can be expressed as the sum of a
series of simpl e harmonicmotions. Even transient, nonperiodic motions such as those
of Figure A.l c and d can be representedas periodic motions by assuming that they
repeat themsel ves after some "quiet"zone during which no motion occurs (Figure
A.2). Using this technique, even a transientmotion can al so be expressed as a
periodic motion. This becomes a very powerful tool forthe dynamic anal ysis of
l inear systems, where the principl e of superposition al l ows theresponse to
transient l oading to be expressed as the sum of the responses to a series of simpl e
harmonic l oads.TransientmotionQuiet zoneu(t)Tf Tf~IFigure A.2 Representation of a
transient motion as a periodic motion using an artificial quiet zone. The motion
repeats itsel f indefinitel y at period ~,A.2.1 Simpl e Harmonic MotionSimpl e harmonic
motion can be characterized by sinusoidal motion at constant frequency.Its most
important features can be defined by three quantities: ampl itude, frequency, and
Sec. A.2 Types of Vibratory Motion 529phase. Simpl e harmonic motion can be
described in different ways, two of which wil l bepresented in the fol l owing
sections: using trigonometric notation or using compl ex notation.Both notations are
equival ent and both are commonl y used in geotechnical earthquakeengineering.A..2.2
Trigonometric Notation forSimpl e Harmonic MotionIn its simpl est form, simpl e
harmonic motion can be expressed in terms of a displ acement,u(t), using
trigonometric notation: for exampl e,u(t) = A sin (rot + <\» (A. I)where A
represents the displ acement ampl itude, ro the circul ar frequency, and <\> the phase
angl e. The time history of this simpl e harmonic displ acement is shown in Figure
A.3. Theampl itude, A, is occasional l y referred to as the singl e ampl itude to
distinguish it from the doubl eampl itude (which represents the peak-to-peak
displ acement) referred to in some of theol der geotechnical earthquake engineering
l iterature. The circul ar frequency describes therate of oscil l ation in terms of
radians per unit time, where 21tradians corresponds to one cycl eof motion. The
phase angl e describes the amount of time by which the peaks (and zero points)are
shifted from those of a pure sine function, as il l ustrated in Figure AA. The
displ acementwil l be zero when rot +<\> =0 or, consequentl y, when t= -<\>/ro. A
positive phase angl e indicatesthat the motion l eads the sine function; it l ags the
sine function if the phase angl e is negative.The concept of circul ar frequency is
more easil y understood by considering the motion of theu(t)TimeFigure A.3 Time
history of simpl e harmonic displ acement.Figure A.4 Infl uence of phase angl e on
position of sinusoid.530u(t)Vibratory Motion App.A\\\\Figure A.5 Rotating vector
representationof simpl e harmonic motion with zero phaseangl e.(A.3)rotating vector
of l ength A shown in Figure AS. If the vector rotates countercl ockwise aboutits
origin at an angul ar speed, ro, from its initial horizontal position, the
displ acement, u(t), isgiven by the vertical component of the vectoru(t) = A sinror
The vertical component increases to a maximum val ue at rot= rcl 2, then decreases
throughzero (at rot=rc) and reaches its maximum negative val ue at rot=3rc/2. It
continues back toits original position and then repeats the entire process.The time
required for the rotating vector to make one ful l revol ution is the timerequired
for one cycl e of the motion. This time is referred to as the period ofvibration, T,
and is rel ated to the circul ar frequency byT = angul ar distance for one revol ution
= 2rc (A2)angul ar speed coAnother common measure of the frequency of oscil l ation is
expressed in terms of the numberof cycl es that occur in a particul ar period of
time. Since the period of vibration representsthe time per cycl e, the number of
cycl es per unit time must be its reciprocal , that is,I rof = T = 2rcwhich is
usual l y expressed in cycl es per second or hertz (abbreviated Hz).Simpl e harmonic
motion can al so be described as the sum of a sine function and acosine function,
that is,u(t) = a cos rot + b sin rot (AA)As shown in Figure A6, the sum of the sine
and cosine functions is al so a sinusoid that oscil l atesat circul ar frequency, eo.
However, its ampl itude is not the simpl e sum of the ampl itudesof the sine and
cosine functions, and its peaks do not occur at the same times as those of thesine
or cosine functions. The rotating vector representation of this function is
il l ustrated inSec. A.2 Types of Vibratory Motion 53110 15 20Timea cos rot+ b sin
rot5o3,-------------------------,2\\-1 -- (-2 ---- ...a cos rot-
3L----------'-------~------~--------'ou(t)Figure A.6 Summation of sine and cosine
functions of the same frequency produces asinusoid of the same frequency. Ampl itude
and phase of the sinusoid depends on theampl itudes of the sine and cosine
functions.where AFigure A.7. Since cos e = sinCe + 90°), the rotating vector
ofl ength a must be 90° ahead ofthe vector of l ength b. The vertical components of
vectors a and b are a cos rotand b sin s»: respectivel y. As il l ustrated in Figure
A.7a, the total val ue of u(t) is given by u(t) =a cos rot+ b sin rot. The motion
can be expressed in a different form by considering the resul tant ofvectors a and
b, as in Figure A.7b. The l ength of the resul tant wil l be Ja2 + b2 and it wil l l ead
b by an angl e <\l = tan-\al b). Accordingl y, the vertical component of the resul tant
isu(t) = A sinerot + <\l ) (A.5)Ja2 + b2 is the ampl itude and <\l tan-1 (al b) is the
phase angl e of the motion.rotu u ~a2 + b2 = Atan? (al b)=<1>rot(a) (b)Figure A.7
Rotating vector representation of simpl e harmonic motion. Sum of vertical components
of sine and cosine components in (a) is equal to vertical component ofresul tant of
sine and cosine components in (b).A.2.2.1 Compl ex Notation forSimpl e Harmonic
MotionTrigonometric descriptions of simpl e harmonic motion use famil iar functions
that areeasy to visual ize. For many dynamic anal yses, however, the use of
trigonometric notation532 Vibratory Motion App.Al eads to very l ong and awkward
equations. These anal yses become much simpl er whenmotions are described using
compl ex notation (the word compl ex indicates that compl exvariabl es are used, not
that the notation is particul arl y compl icated). Compl ex notation canbe derived
directl y from trigonometric notation using Eul er's l aw: ei a = cos a + i sin a (A6)
where i is the imaginary number i = p.The quantity ei a is a compl ex number; it has
twoparts, a real part and an imaginary part, which can be written asRe(ei a ) = cos
aIm(ei a) sinrzEul er's l aw can be used to show thatsin a (A?)Substituting these
expressions into the general expression for harmonic motion equation(AA) givesu(t)
(A8)= a---ibeirot+a -+-ibe-irot2 2This form of the displ acement may be visual ized
as a pair of rotating vectors in an Arganddiagram. An Argand diagram represents a
compl ex number graphical l y as a vector withorthogonal real and imaginary
components. Al though usual l y drawn with the real axis orientedhorizontal l y, the
rotated Argand diagram of Figure A8a wil l hel p il l ustrate how thiscompl ex notation
describes simpl e harmonic motion. In the Argand diagram, the term eirotT(a)TimeReal
Vector sumatt= 0A(b)Figure A.S How counterrotating vectors of l ength A/2 produce
simpl e harmonicmotion. Note that the phase angl es are measured from the horizontal
axis in the directionof vector rotation.Sec. A.2 Types of Vibratory Motion 533is
represented by a vector of unit l ength rotating cl ockwise at an angul ar speed, co.
The terme-ioot =ei(-Ol )t therefore can be represented by a unit vector rotating
cl ockwise at angul arspeed, -co, which is equival ent to rotating countercl ockwise
at angul ar speed, co. Accordingl y,the first term in equation (A.8) can be
represented by a vector of real part, a/2, andimaginary part, -bl 2, rotating
cl ockwise at co, and the second term by another vector withthe same real part, but
an imaginary part, bl 2, rotating countercl ockwise at co. The l ength ofeach vector
is J (al 2)2 + (bI2)2 = ~Ja2 + b2. As shown in Figure A.8a, the sum of thevectors is
real (the imaginary parts al ways cancel each other). Figure A.8b shows how the
vector sum describes a simpl e harmonic motion of ampl itude A = Ja2 + b2 and
circul arfrequency co.A.2.3 Other Measures of MotionDispl acement is not the onl y
parameter that can be used to describe vibratory motion. In fact,other parameters
are often of greater interest. If the variation of displ acement with time isknown,
however, the other parameters of interest can be determined. Differentiating the
expressionfor simpl e harmonic displ acement produces expressions for vel ocity and
accel eration: (A.9c)(A.9a)(A.9b)accel erationvel ocityu(t) = Asin(cot + <1»
displ acement£l (t) = ~~ = coAcos (cot + <1»ii(t) = d2~ = _co2A sin (cot + <1» =
-co2udtNote that when the displ acement ampl itude is A, the vel ocity ampl itude is
coA,and the accel erationampl itude is co2A. Thus frequency and the displ acement,
vel ocity, and accel erationampl itudes of a harmonic motion are rel ated in such a way
that knowl edge of the frequencyand anyone ampl itude, or knowl edge of any two
ampl itudes, al l ows cal cul ation of al l otherquantities. This important and useful
property of harmonic motions al l ows the use of tripartitepl ots, in which a harmonic
motion can be compl etel y described in terms of frequencyand displ acement, vel ocity,
and accel eration ampl itudes by a singl e point. Tripartitepl ots, an exampl e of which
is shown in Figure A.9, are commonl y used to describe earthquakeground motions. It
is important to note that these rel ationships appl y onl y to harmonicmotions and
that the rel ationships between displ acement, vel ocity, and accel erationfor other
types of motion must be obtained by differentiation and/or integration.Examination
of equations (A.8) reveal s that in addition to having different ampl itudes,the
displ acement, vel ocity, and accel eration are out of phase with each other(Figure
A.I 0). The vel ocity can be seen to l ead the displ acement by nl 2 radians, or 90°,
andthe accel eration to l ead the vel ocity by the same amount. The rel ationships
between displ acement,vel ocity, and accel eration for harmonic motions, in both
trigonometric and compl exnotation, areu(t) = A sin cotu(t) = coAcoscot = coAsin(cot
+ n12)u(t) = Aeioot£l (t) icoAe iOl t(A. l Oa)(A.l Ob)534 Vibratory Motion
App.A,~;;c-...".: -~~ 100: '/',v.,"-... ; • • • <'-. ·1 80f ..··+ .• ..: ..: : ,.:
k·X: : ··.·'C) ·: : h : /; ·\· L: i : '- )~ ......• · ·: ,,'• • ,,'c··: ..·
;......• : .: ;...: ..~(: J• .• : : )j{ / : .: : i ·.• ..16040 [."..·+..··· ....,/....
V···.I,.·: ··...·,\'·.....+· ......·: ··/·.. ·: ...... ·...c/..--c...: ·.,j...• • ,., ...
"'.·..+ .......: ,.• : . +·.... ·.. : i: yr·.. ,.··: : ··.1·.·: : ·... : 'v: ·,··...... ·14020
10 10 "<D-!!! 8 8~"" 6 6'0a~4.1.04 .06 .08 .1 .2 .4 .6 .8 2 4 6 8 10 20Period (sec)
I I20 10 5.0 2.0 1.0 0.5 0.2 0.10 0.05Frequency (Hz)Figure A.9 Tripartite pl ot for
harmonic motion. Point at center describes harmonicmotion at a period of 0.65 sec
with displ acement ampl itude of 0.8 in., vel ocity ampl itudeof 8.0 inJsec, and
accel eration ampl itude of 0.20g. (After Richart, et al ., 1970.)Sec. A.2 Types of
Vibratory MotionU(t)535Li(t)D(t) -~Figure A.I0 Time histories of displ acement,
vel ocity, and accel eration. Note thataccel eration l eads vel ocity by one-quarter
cycl e and displ acement by one-hal f cycl e.aCt) = -ffi 2Asinffit = ffi2Asin(ffit+1t)
uCt) = i2ffi2Aeiwt = _ffi2Aeiwt (A.1Oc)The rel ationship between harmonic
displ acements, vel ocities, and accel erations can be visual l izedinterms of three
vectors rotating countercl ockwise at an angul ar speed co (Figure A.l l ).The
accel eration vector is 90° (or1t12 radians) ahead of the vel ocity vector and 180°
(or 1tradians)ahead of the displ acement vector.u,u,DVel ocity//IIIfAccel erationII-,\
\\\IIII//Displ acementFigure A.l l Rotating vector representationof displ acement,
vel ocity, and accel eration.Note how accel eration l eads vel ocity by 90°and
displ acement by 180°.536A.3 FOURIER SERIESVibratory Motion App.AWhil e studying heat
fl ow probl ems in the earl y nineteenth century, the French mathematician1. B. 1.
Fourier showed that any periodic function that meets certain conditions can be
expressed as the sum of a series of sinusoids of different ampl itude, frequency,
and phase.Since the conditions for existence of a Fourier series are nearl y al ways
met for functions thataccuratel y describe physical processes (Ramirez, 1985), it is
an extraordinaril y useful tool in many branches of science and engineering.
Geotechnical earthquake engineering is no exception. By breaking down a compl icated
l oading function such as that imposed by an earthquake ground motion into the sumof
a series of simpl e harmonic l oading functions, the principl e of superposition
al l ows avail abl esol utions for harmonic l oading to be used to compute the total
response (provided thatthe system is l inear), as il l ustrated schematical l y in
Figure A.12.~ + t Sol ution/\/\ 1\ 1\ 1\ 1\. for f---- rv v v.;: v v t - harmonict
l oadingt ~ + tAt 1- +t t: (a) (b) (c)-(d) (e)Figure A.12 Process by which Fourier
series representation of compl icated l oadingcan al l ow rel ativel y simpl e sol utions
for harmonic l oading to be used to produce the total response: (a) time history of
l oading; (b) representation of time history of l oading as sumof series of harmonic
l oads; (c) cal cul ation of response for each harmonic l oad; (d)representation of
response as sum of series of harmonic responses; (e) summation ofharmonic responses
to produce time history of response.A.3.1 Trigonometric FormSince a Fourier series
is simpl y a summation of simpl e harmonic functions, it can beexpressed using either
trigonometric notation or compl ex notation. The general trigonometricform of the
Fourier series for a function of period, Tf , isxU) = ao + L (ancos ffint + b; sin
ffint)n = 1where the Fourier coefficients are(A.Il )Sec. A.3 Fourier Series 537and
(On = 21tn/Tj- The term ao represents the average val ue of x(t) over the range t =
0 tot =: Tf ; its val ue is zero in many geotechnical earthquake engineering
appl ications. Note thatthe frequencies, (Om are not arbitrary; rather, they are
evenl y spaced at a constant frequencyincrement, Ll (O = 21t/Tj-Exampl eA.l The Fourier
coefficients are not difficul t to cal cul ate for simpl e functions. Consider the
squarewavefunction shown in Figure EA1. Over its period, Tfi the square wave is
described byT+A 0< t<---: !4': x(t) -ATt 3Tt -4<t-<4-+A3Tt """4 < is r,Since the
average val ue of x(t) is easil y seen to be zero, the coefficient ao = O. The val ue
of ajcan be computed as2 [ jTf/4 j3Tfl 4 jTf ] T A cos wl tdt-A cos wjtdt+A cos wl tdt
t a r.t« 3Tf l 4-2A-[s.m w-j-Tt-(s.m -3w-j-Ttsm. W-I-Tt) + (s.m wITt-. s3mWj-T-t)]
WITt 4 4 4 4Substituting wITf =21t yiel dsRepeating for al l n yiel dsaj = ~(1+2+l ) 4A
-1trAn1ta = -4A nn1t0b = 0 nn = 1,5,9, ...n = 3,7,11, ...n = even integersal l nx
(t)+ATf t-A - --Figure EA.! Square-wave function.538 Vibratory Motion App.Aso the
Fourier series iswhere WI = 21t1Tf" The sine terms are al l zero because the square
wave, l ike the cosine function,is an even/unction [i.e., one forwhich/(t) = fi-t)].
For an odd/unction Iftt) = ~-t)], the cosineterms are zero. For a function that is
neither even nor odd, the Fourier series wil l contain bothsine and cosine terms.The
Fourier series represents a function exactl y onl y for n =00. If the series is
truncatedat some finite val ue of n, the Fourier series onl y approximates the
function. For manyfunctions, however, the approximation can be quite good even when
n is rel ativel y smal l .This characteristic is often used to great advantage in
dynamic anal yses of soil s andstructures.From equations (A.5) and (A. I I), it is
apparent that the Fourier series can al so beexpressed asx(t) = co+
L,cnsin(ffint+<pn)n = I(A. 12)where Co = ao,cn = Ja~ + b~, and <Pn =tan-1(an/bn).
In this form, Cn and <Pn are theampl itude and phase, respectivel y, of the nth
harmonic. A pl ot of cn versus ffi nis known asa Fourier ampl itude spectrum; a pl ot
of <Pn versus ffingives a Fourier phase spectrum. Fourierampl itude spectra are very
useful in geotechnical earthquake engineering-as discussedin Chapter 3, they
effectivel y describe the frequency content of an earthquakemotion.<1>1n: 2<1>3-1t2
<1>5tt.2<1>7-n: 2C4 = 04AC5 = 5n: C6 = 04AC7 = 7n: Exampl eA.2The Fourier ampl itude and
phase spectra for the square wave of Exampl e A.I are easil y determined.The val ues
of c., and <l >n for the first eight terms of the series areCo = 04ACl = nC2 = 04A
3n: The spectra are pl otted in Figure EA.2.(Al 3)Sec. A.3 Fourier Series 539en4X <Pn
11: It/20>0>-It/2(a) (b)Figure EA.2 Fourier spectra for square wave of Exampl e A.I:
(a) Fourier ampl itudespectrum; (b) Fourier phase spectrum.Jl .3.2 Exponential Form
The Fourier series can al so be expressed in exponential form. Substituting
Equations A7into Al O for al l n gives() L (an- ibn iro t an + ibn -iro t)X t = a +
---e n + e n o 2 2n = 1Defining new Fourier coefficients,Co* = ao* an - ibnCn 2* an
+ ibnc.;2where the * indicates the compl ex nature of the coefficient, the Fourier
series can be rewrittenasx(t) = c~ + L (c: e iront + c: ne-iront)n = 1(AI4)Since OLn
=-(Om the l imits of summation can be changed to write the Fourier series in themore
compact form(AI5)n;;;: : _00The compl ex Fourier coefficients, c: , can be determined
directl y from x(t) as(AI6)540 Vibratory Motion
App.AExampl eA.3Compute the compl ex Fourier coefficients for the square wave of
Exampl e A. I.Sol ution Since the average val ue of the square wave is zero, c; = O.
For n = +1, equation(A. 16) givesc; f [X fT114eiW1tdt _ XJ3TI14 eiw1tdt + XJTj
eiW1tdt]f 0 Ij/4 3Ij/4-,--!.- [eiWITr14 _ (ei3WITj14 _ eiWITj14) + (eiWITj _
ei3WITjI4)]l WITfj£(2ei1t/2_ 2ei31t/2 + ei21t)l 21t'2X (2i + 2i + 0) = 2X1 1t 1tNote
that al though c~ = Co = ao, the definitions of c~ and c: n indicate that(~r+(~r=~=~
(~r +(~r =Ja~2+b~ =~(A. I?)(i.e., in exponential form, hal f of the ampl itude is
associated with positive frequencies and hal fwith negative frequencies). The phase
angl es at positive and negative frequencies are equal butof opposite sign;
consequentl y, the imaginary parts cancel each other [as they must if x(t) is areal
function]. The compl ex Fourier coefficients are sometimes (al though rarel y in
geotechnical earthquake engineering appl ications) used to pl ot two-sided spectra
which are rel ated to themore conventional one-sided spectra as shown in Figure
EA.3.Figure EA.3 Comparison of one- and two-sided Fourier spectra. The two-sided
Fourier ampl itude spectrum (a) is symmetrical with ampl itudes on each side of the
il l =0axis equal to hal f the ampl itude of the one-sided spectrum (b). (The
ampl itude at il l =0 isthe same for both.) The two-sided phase spectrum (c) is
antisymmetric, but phase atpositive frequencies is equal to phase of one-sided
spectrum (d).Sec. A.3 Fourier Series 541A.3.3 Discrete Fourier TransformIn many
geotechnical earthquake engineering appl ications, l oading or motion parametersare
described by a finite number of data points rather than by an anal ytical function.
In suchcases the Fourier coefficients are obtained by summation rather than
integration. For a variabl eX(tk), k =1,N, where tk=k St, the discrete Fourier
transform (DFT) is given byNX(ffin) = !'ItL x(tk)e-iffinl k (A18a)k = 1where ffin
=n!'l ffi =2rcn/N St. Using Eul er's l aw, the DFT can al so be written asNX(ffin)
= !'l tL [x(tk)cosffintk-ix(tk)sinffintk] (A18b)k = 1Note that the Fourier
coefficents of the DFT have units of the original variabl e mul tipl ied bytime.The
DFT can al so be inverted; that is, a set of data spaced at equal frequency
interval s,!'l ffi, can be expressed as a function of time, using the inverse
discrete Fourier transform(IDFT): orN!'l ffiL X(ffin) eiffi,hn = 1Nx(tk) = !'l ffiL
[X(ffin)COS ffintk+ iX(ffi n)sin ffintk]n = 1(A19a)(A.l 9b)Either of these
expressions can easil y be programmed on a personal computer; since n takeson N
different val ues, the summation operation wil l be performed N times. The time
required for computation of a DFT (or IDFT), therefore, is proportional to N2.
i'.3.4 Fast Fourier TransformThe DFT was devel oped l ong before computers were
avail abl e, and its use, for even modestval ues of N, was extremel y l abor intensive.
As earl y as 1805, the beginning of a more efficientapproach to the DFT was
described (Brigham, 1974). As digital computers weredevel oped in the 1960s, Cool ey
and Tukey (1965) devel oped a computational al gorithm forthe case where N is a power
of 2 that has become known as the fast Fourier transform (FFT).By performing
repeated operations on groups that start with a singl e number and increase insize
by a factor of2 at each ofjstages (whereN= 20, the time required to compl ete the
transformis proportional to N 10g2N. Consequentl y, the FFT is much more efficient
than theDFT. For exampl e, at N =2048, the FFT is more than 180 times faster than
the DFT. Theinverse fast Fourier transform (IFFT) operates with equal speed.542
A.3.S Power SpectrumVibratory Motion App.AThe Fourier ampl itude spectrum
il l ustrates how the strength of a quantity varies with frequency.This information
can al so be expressed in terms of power. The power of a signal ,x(t), that can be
expressed in the form of equation (A. II ) or (A.12), is defined asP( ) _ 1(2 b2) _
1 2ffin - 'l an + n - 'l c nNote that this definition of power can be appl ied to any
signal (it is not rel ated to mechanical -force times vel ocity-power). Power can be
pl otted as a function of frequency toobtain a power spectrum. The total power of
the signal is the same whether it is computedin the time domain or the frequency
domain: oo (r fl O"total power = L P(ffin ) = J" [x(t)] 2dt = ~ C~dffin = I 0 0Power
spectra are often used to describe earthquake-induced ground motions.BDynamics of
Discrete SystemsB.1 INT'RODUCTIONMany vibrating systems consist of discrete
el ements such as masses and springs, or can atl east be ideal ized as such. For most
practical probl ems of structural dynamics, the structureis ideal ized as a system of
rigid masses connected by massl ess springs. Even continuoussystems such as soil
deposits have been ideal ized as assembl ages of many discrete el ements,though that
approach is sel dom taken any more. Since the geotechnical earthquakeengineer often
provides input to the structural engineer, a firm understanding of thedynamic
response of discrete systems is required. Al so, many of the concepts and
terminol ogiesused in geotechnical earthquake engineering anal yses are anal ogous to
those of discretesystem dynamics and are more easil y introduced in that framework.
This appendix introduces the dynamics of discrete systems. It begins with very
simpl esystems, and adds compl icating factors such as damping, base motion, and
nonl inearity.Anal ytical and numerical sol utions in the time domain and frequency
domain are presented.Final l y, the response of mul tipl e-degree-of-freedom systems is
introduced. Whil e many ofthe basic concepts of structural dynamics are presented,
much more compl ete treatmentsmay be found in a number of structural dynamics texts
(e.g., Cl ough and Penzien, 1975;Paz, 1980; Berg, 1989; Chopra, 1995).543544B.2
VIBRATING SYSTEMSDynamics of Discrete Systems App. BVibrating systems can be
divided into two broad categories: rigid systems and compl iantsystems. A rigid
system is one in which no strains occur. Al l points within a rigid systemmove in
phase with each other, and the description of rigid-body motion is a rel ativel y
simpl ematter of kinematics. In compl iant systems, however, different points within
the systemmay move differentl y (and out of phase) from each other. A given physical
system maybehave very nearl y as a rigid system under certain conditions and as a
compl iant systemunder other conditions. Since neither soil s nor structures are
rigid, the dynamic response ofcompl iant systems is central to the study of soil and
structural dynamics and to earthquakeengineering.Compl iant systems can be
characterized by the distribution of their mass. Discretesystems are those whose
mass can be considered to be concentrated at a finite number ofl ocations, where the
mass of a continuous system is distributed throughout the system. Thenumber of
independent variabl es required to describe the position of al l the significant
masses of a system is the number of dynamic degrees offreedom of the system.
Systems ofinterest in earthquake engineering may have anywhere from 1 to an
infinite number ofdegrees of freedom. Figure B.l il l ustrates several commonl y
encountered systems withvarying numbers of degrees of freedom (DOF). Discrete
systems have a finite number ofdegrees of freedom; the number of degrees of freedom
of a continuous system is infinite.Certain types of anal yses ideal ize continuous
systems as discrete systems with l arge numbersof degrees of freedom, and other
types represent discrete systems with many degrees offreedom as continuous systems.
(a)m(b)mtm2ms7''7 7''7(c) (d) (e)Figure B.t Vibrating systems with various numbers
of degrees of freedom: (a) one DOF, vertical transl ation; (b) two DOF, vertical
transl ation and rocking; (c) three DOF, horizontal transl ation; (d)infinite DOF;
(e) infinite DOF.B.3 SINGLE-DEGREE·OF-FREEDOM SYSTEMSA discrete system whose
position can be described compl etel y by a singl e variabl e is knownas a singl e-
degree-of-freedom (SDOF) system. That singl e degree of freedom may represent
transl ational displ acement, as in the SDOF systems of Figure B.2a-e, or rotational
displ acement,as in the case of the pendul um of Figure B.2d.A typical SDOF system is
one in Whicha rigid mass, m, is connected in paral l el to aspring of stiffness, k,
and a dashpot of viscous damping coefficient, c, and subjected tosome external
l oad, Q(t), as shown in Figure B.3. The spring and dashpot are assumed to be
massl ess and the displ acement origin to coincide with the static equil ibrium
position.Sec. B.4 Equation of Motion for SDOF System 545(a) (b)m(c) (d)Figure B.2
Various SDOF systems. The degrees of freedom are (a) vertical transl ation, (b) and
(c) horizontal transl ation, and (d) rotation.'/OCt)Figure B.3 Damped SDOF system
subjected to external dynamic l oad, Q(t).8.4 EQUATION OF MOTION FOR SDOF SYSTEMMany
SDOF systems are acted upon by external l y appl ied l oads. In earthquake engineering,
dynamic l oading often resul ts from another source-movement of the supports of the
system.The dynamic response of a SDOF system such as that shown in Figure B.3 is
governedby an equation of motion. The equation of motion can be derived in a number
of ways; asimpl e, force equil ibrium approach wil l be used here.B: .4.1 Equation of
Motion: External LoadingWhen a dynamic l oad is appl ied to the mass of a SDOF system
(Figure R3), the tendencyfor motion is resisted by the inertia of the mass and by
forces that devel op in the dashpot andspring. Thus the external l oad, Q(t), acting
in the positive x-direction is opposed by threeforces (Figure B.4) that act in the
negative x-direction: the inertial force, fl , the viscousdamping force, [o. and the
el astic spring force, fs. The equation of motion can be expressedin terms of the
dynamic equil ibrium
of these forces: fl (t) + fvU) + fsU) = Q(t) (R!)'/OCt)Figure B.4 Dynamic forces
acting onmass from Figure B.3.546 Dynamics of Discrete Systems App. 8(B.2a)These
forces can al so be expressed in terms of the motion of the mass. Newton's second
l awstates that the inertial force acting on a mass is equal to its rate of change
of momentum,which for a system of constant mass producesfI(t) = !!: ...(mdU(t)) = m
d2u(t)= mii(t)dt dt drZFor a viscous dashpot, the damping force is proportional to
the vel ocity of the mass: duet) .f D(t) = C ----;[t = cu(t) (B.2b)and the force
provided by the spring is simpl y the product of its stiffness and the amount by
which it is displ acedfs(t) = ku(t) (B.2c)The behavior of these forces is
il l ustrated graphical l y in Figure B.5. The inertial force is proportional to the
accel eration and the constant of proportional ity is the mass. Simil arl y, theviscous
damping force and the el astic spring force are proportional to the vel ocity and
displ acementwith the damping and spring coefficients serving as the respective
constants ofproportional ity.Substituting equations (B.2) into equation (B. I), the
equation of motion for the SDOFsystem can be written asmii(t) + cit(t) + ku(t) =
Q(t) (B.3)This second-order differential equation is commonl y used to describe the
behavior of oscil l atingsystems ranging from the mechanical systems considered in
earthquake engineeringprobl ems to el ectrical circuits. The differential equation of
motion is l inear (i.e., al l of itsterms have constant coefficients). This l inearity
al l ows a cl osed-form anal ytical sol ution tobe readil y obtained and, importantl y, it
al l ows the principl e of superposition to be used.When any of the coefficients are
not constant, the behavior is not l inear and the sol utionbecomes considerabl y more
difficul t. Inmost cases, the response of nonl inear systems mustbe eval uated
numerical l y (Section B.7).tm(a)a(b) (c)kuFigure B.5 Variation of (a) inertial , (b)
viscous, and (c) el astic forces withaccel eration, vel ocity, and displ acement,
respectivel y.Sec. 8.5 Response of Linear SDOF Systems 547B.4.2 Equation of Motion:
Vibration of Supports(Base Shaking)For earthquake engineering probl ems, dynamic
l oading often resul ts from vibration of thesupports of a system rather than from
dynamic external l oads. To eval uate the response ofsuch systems, it is necessary to
devel op an equation of motion for l oading caused by baseshaking. Consider the
damped SDOF system shown in Figure B.6a. When subjected todynamic base shaking,
Ub(t), it wil l deform into a configuration that might l ook l ike thatshown in Figure
B.6b at a particul ar time, t. The total displ acement of the mass, ul t), can be
broken down as the sum of the base displ acement, Ub(t), and the displ acement of the
massrel ative to the base, u(t). The inertial force wil l depend on the total
accel eration of the mass,whil e the viscous damping and el astic spring forces wil l
depend on the rel ative vel ocity anddispl acement, respectivel y. Using the notation
shown in Figure B.6b, the equation ofmotion can be written asmii, + cit + ku = 0or
substituting iil t) =iib(t) + ii(t) and rearranging,mu = cit + ku = -rniu, (BA)In
other words, the response of the system to base shaking is equival ent to the
response thatthe system woul d have if its base was fixed and the mass was subjected
to an external l oadQ(t) = -mub(t). Thus any sol utions for the response of an SDOF
system subjected to external l oad can be used to eval uate the response of the system
to base shaking.1---1 Ut(t)___I 1--- u(t)r - -'---r-T---'I,--IIIIIIk2"mk2"r-D- I--_
--Ic(a) (b)Figure B.6 Damped SDOF system subjected to base shaking.B.5 RE~~PONSEOF
LINEAR SDOF SYSTEMSIn order to eval uate the dynamic response of a l inear SDOF
system, the differential equationof motion must be sol ved. There are several types
of conditions under which the dynamicresponse of SDOF systems are commonl y
cal cul ated. Forced vibration occurs when the mass548 Dynamics of Discrete Systems
App. Bis subjected to some external l oading, Q(t). The l oading may be periodic or
nonperiodic andit may correspond to an actual physical force appl ied to the mass or
to some known l evel ofbase shaking. Free vibration occurs in the absence of
external l oading or base shaking. It mayresul t from the rel ease of the mass from
some initial displ acement or may occur after sometransient forced vibration has
ended. The fol l owing sections wil l devel op sol utions to theequation of motion for
cases in which damping is and is not present, and for cases in whichexternal
l oading is and is not present. The resul ting four permutations of these conditions
are1. Undamped free vibrations: c= 0, Q(t) = 02. Damped free vibrations: c>O, Q(t)
= 03. Undamped forced vibrations: c=O, Q(t)"# 04. Damped forced vibrations: c>O,
Q(t)"# 0The sol ution of the equation of motion for each of these conditions wil l be
presented in turn.B.5.1 Undamped Free VibrationsA SDOF system undergoes free
vibration when it oscil l ates without being acted uponby any external l oads. When
damping is not present (c =0) the equation of motion (forundamped free vibration)
reduces tomii + ku = 0or after dividing both sides by the mass,.. k 0 u+-u = m(B.5)
(B.6)(B.9)(B.8)The sol ution to this simpl e differential equation can be found in
any el ementary text on differential equations asu = C1sinJft+Czcosgt (B.7)where the
val ues of the constants Cj and Czdepend on the initial conditions of the system.The
quantity ~ is very important-it represents the undamped natural circul ar frequency
of the system0)0 = ~Then the natural frequency, fa, and natural period ofvibration,
To, can be written as0)0 1 (kfa = 21t = 21t~mTo = 21t = 21t fii!: (B.IO)0)0 ~k
Substituting equation (B.8) into the sol ution for the equation of motion [equation
(B.7)] yiel ds(B.l l )Sec. 8.5 Response of Linear SDOF Systems 549(B. 12)which
indicates that an undamped system in free vibration wil l oscil l ate harmonical l y at
itsundamped natural frequency. Cj and Cb can be eval uated by assuming the initial
(t= 0) conditionsto be represented by an initial displ acement, uo, and initial
vel ocity, tio. ThenUo = C]sin(O) + C2cos(0) = C2Uo = OJoCjcos(0)-rooC2sin(0) =
rooCjTherefore, Cj = uo/roo and C2 = uo, so the compl ete sol ution to the undamped
free vibrationresponse of an SDOF system is given byUd . u = -sm root + uocos root
0J0The response of such a system is shown in Figure B.7.Referring back to equation
(A.S), the free vibration response can al so be expressed asu = A sin (root + <Ii)
where the ampl itude, A, and phase angl e, <Ii, are given by(B.l 3)A = ( . )2 2 UoUo +
rooth __ -1 uoroo 'I' tan -.-UoThe sol ution to the equation of motion of an
undamped system indicates that the responseof the system depends on its initial
displ acement and vel ocity. Note that the ampl ituderemains constant with time.
Because no energy is l ost in an undamped system, it wil l continueto oscil l ate
forever. Obviousl y, trul y undamped systems do not exist in the real worl d;however,
some systems can have such l ow damping that their response over short periodsof
time may approximate that of an undamped system.u__-----To = 21t-----_..rooFigure
B.7 Time history of displ acement for undamped free vibration with initial
displ acement Uoand initial vel ocity Uo .550 Dynamics of Discrete Systems App. B
Exampl e B.1The SDOF structure shown in Figure EB.1a consists of a 1O-kipweight
supported by a massl esscol umn. Appl ication of a 5-kip static horizontal force to
the weight produces a horizontal defl ection of 0.04 in. Compute (a) the natural
circul ar frequency, (b) the natural period of vibration,and (c) the time history of
response if the horizontal force was suddenl y removed.w= 10 kipsFigure EB.l a
Sol ution (a) The probl em statement indicates that the stiffness of the col umn isk =
5 kips = 125 kipsl in.0.04 in.The natural circul ar frequency is given by(125
kips/in.)(12 in./ft)(32.2 ft/sec 2)10 kips69.5 rad/sec(b) The natural period woul d
beTo = 2IT = 2IT rad = 0.09 secOl o 69.5 rad/sec(c) The horizontal force produced a
static defl ection of 0.04 in. Consequentl y, the initial conditions for free
vibration woul d beUo = 0.04 in. Uo = 0Thenu(t) = ~Sin Ol ot + Uo cosWot = (0.04 in.)
cos (69.5t)The response is pl otted in Figure EB.I b.
0.05 ,-----,---,-----,--,-----,------,----,--------,--..,-----,0.05 0.1 0.15 0.2
0.25 0.3 0.35 0.4 0.45Of---\---I------\,---+----\-------,r----\----+----'\---
r------t--i-0.05 '----_--'-__-'--_----'-__---'-__L.-_----'--__--'----_----'-__~_~o
Figure EB.l bSec. 8.5 Response of Linear SDOF Systems 551(RI9)8,,5.2 Damped Free
VibrationsIn real systems, energy may be l ost as a resul t of friction, heat
generation, air resistance, orother physical mechanisms. Hence the free vibration
response of a damped SDOF systemwil l diminish with time. For damped free
vibrations, the equation of motion is written asmil + cLi + ku = 0 (B.14)or,
dividing by m and substituting [from equation (B.17)] k = mco5, we haveii + 2 ~COou
+ co5u = 0 (B.l 5)2AjkmThe quantity 2j"k;, cal l ed the critical damping coefficient,
Co al l ows the damping ratio, ~,to be defined as the ratio of the damping
coefficient to the critical damping coefficient, thatis.,~ _ c _ c _ c ccoo (B.16)-
~ - 2j"k; - 2mcoo 2kWith this notation, the equation of motion can be expressed as
ii + 2~coou + co5u = 0 (B.l 7)The sol ution of this differential equation of motion
depends on the val ue of the dampingratio. When ~ < 100% (c < c.), the system is
said to be underdamped. When ~ = 100%(c = cc) the system is critical l y damped, and
when ~ > 100% (c > cJ the system is overdamped.Separate sol utions must be obtained
for each of the three cases, but structures ofinterest in earthquake engineering
are virtual l y al ways underdamped.For
the case in which damping is l ess than critical , the sol ution to the equation of
motion is of the formu = e-~(J)ot [ C) sin(cooJI - ~2 t) + C2 cos(COoJI - ~2 t) ]
(RI8)Note the exponential term by which the term in brackets is mul tipl ied. This
exponential term gets smal l er with time and eventual l y approaches zero, indicating
that the response ofan underdamped system in free vibration decays exponential l y
with time. The rate of decaydepends on the damping ratio-for smal l ~ the response
decays sl owl y and for l arger ~ theresponse decays more quickl y. Defining the damped
natural circul ar frequency of the systemas COd =COoJI - ~2 the sol ution can be
expressed asu = e-~(J)ot(c )sm. codt + C2COS codt)The natural frequency of a damped
system is al ways l ower than that of an undamped system,and it decreases with
increasing damping ratio.The coefficients C) and C2 can be determined from the
initial conditions in the samemanner as for the undamped case. The initial
displ acement and vel ocity areuo =e-~(J)o(O) [C]sin(O) + C2cos(0)] = C2Uo=e-~(J)o(O)
[COdC)COS COd(O) - codC2sincod(0)] - ~cooe-~(J)o(O)[ C) sincod(O) +C2COSCOd(0)]=
codC) - ~COOC2552 Dynamics of Discrete Systems App.8(B.20)(B.22)(B.21)Therefore, Cj
= (uo + S(Oouo)1(Od and C2 = UO, so the sol ution for damped free vibrationscan be
expressed asu = e-~(j)ot( Uo +~: ouo sin (Od t + Uo cos (Odt )The free vibration
response of an underdamped system is shown in Figure B.8. Note theexponential decay
of displ acement ampl itude with time. The ratio of the ampl itudes of anytwo
successive peaks wil l be~ = eXP( 21tS(OO)un + I (OdDefining the l ogarithmic
decrement as 0 =In (un/un + I); theno = 21tS(Oo =~(Od ~Rearranging al l ows the
damping ratio to be determined from the l ogarithmic decrementS = 0 (B.23)J41t2 + 02
For smal l val ues of 0,S"" 0/21t. Therefore, a simpl e way to estimate the damping
ratio of anSDOF system is to perform afree vibration test, in which the l ogarithmic
decrement is measuredwhen a system is displ aced by some initial displ acement, Uo,
and rel eased with initial vel ocity Uo =O.u(t) UoFigure B.8 Time history of damped
free vibration with initial displ acement Uo andinitial vel ocity it.Exampl eB.2The
structure shown in Figure EB.2a is rel eased from an initial displ acement of 1 em
with aninitial vel ocity of -5 em/sec. Compute (a) the damped natural frequency and
(b) the time historyof response of the mass.Sec. 8.5 Response of Linear SDOF
Systems 5531000 kg1000 20000 N/mN-msec~~~~~~~~/' Figure EB2.aSol ution (a) The
undamped natural frequency isfo = 0>0 = 1 [k I 20000 N/m = 071 Hz21t 21t~;;; = 21t
1000 kg .and the damping ratio is1; = _c_2jk;1000 N-m/sec = 0.1l 82,)(20000 N/m)
(l OOO kg)Thenfd = fo,)I-1;2 = (0.71 Hz),)I-(O.l l W = 0.70 Hz(b) The undamped and
damped natural circul ar frequencies wil l be 0>0 = 21tfo = 4.47rad/sec and O>d=
fi21tfd = 4.44 rad/sec, respectivel y. From equation (B.20), the displ acement
response isu = e~i;wo{ Uo +~: ouo sin O>dt + Uo cos O>dt )= exp [-(0.1l 8)(4.47)t] [-
0.05 + (0.1;.~~(4.47)(0.01) sine4.44t) + (I)COS(4.44t)]= e~0.527t [cos(4.44t) -
0.010 sin(4.44t)]which is pl otted in Figure EB.2b.4 5 5 7 8Time (sec)o 2 3Et~~: : ](5
Figure EB.2bEl i.5.3 Response of SDOF Systems to Harmonic LoadingA SDOF system is
said to undergo forced vibration when acted upon by some external dynamic force,
Q(t). Dynamic l oading may come from many different sources and may be554 Dynamics
of Discrete Systems App. Bperiodic or nonperiodic. For probl ems of soil and
structural dynamics, the response toharmonic l oading is very important. One form of
simpl e harmonic l oading Q(t) can beexpressed as Q(t) =Qo sin wt, where Qo is the
ampl itude of the harmonic l oad and W is thecircul ar frequency at which the l oad is
appl ied.8.5.3.1 Undamped Forced VibrationsThe equation of motion for an undamped
system subjected to such simpl e harmonicl oading ismil + ku = Qo sin Wt (B.24)The
general sol ution to this equation of motion is given by the sum of the
compl ementarysol ution (for the homogeneous case in which the right side of the
equation is zero)and the particul ar sol ution [which must satisfy the right side of
equation (B.24)].The homogeneous equation ismil + ku = 0so the compl ementary
sol ution is simpl y the sol ution to the undamped free vibration probl emuc(t) = C1sin
COot + C2cos coot (B.25)The portion of the response described by the compl ementary
sol ution is that which resul tsfrom the initial conditions of the system. It
consists of a simpl e harmonic oscil l ation at theundamped natural frequency of the
system.The particul ar sol ution describes the portion of the response caused by the
external l oading. This portion of the response can be assumed to be of the same form
and to be inphase with the harmonic l oading; thusup(t) = o; sin wt (B.26)where Va
is the ampl itude of the harmonic response. Substituting equation (B.26) into
equation(B.24) yiel ds- mw2 Va sin wt + kVo sin wt = Qo sin WtSubstituting kim = C06
and rearranging gives(B.27)Vo= Qo/k = Qo/k (B.28)I - w2I C06 1 _ ~2where ~ = wiCOo
is referred to as the tuning ratio. Now the general sol ution of the equationof
motion can be obtained by combining the compl ementary and particul ar sol utions: u(t)
= Uc(t) + up(t) = C1 si·n coot+ C2 cos coot+ 1Q_o/~k2 si.n c-ot (B.29)The general
sol ution must satisfy the initial conditions. From equation (B.29), the vel ocitycan
be written asu(t) = dudt= COoC1cos coot- COoC2 si·n coot+ C-O 1Q_o/~k2 cos c-ot
(B.30)Sec: . 8.5 Response of Linear SDOF Systems 555For a given initial
displ acement, uo, and initial vel ocity, l io,Uo = C1sin 000(0) + C2 cos 000(0) +
QO/k2 sin w(O) = C21-~andfrom whichC1= l io-W[(Qo/k)/(l _~2)] = l io Qo~000 000 k (1 _
~2)Now the general response can final l y be written as[l iO Qo~ u = - - 2 Js.m ooot +
Uo cos ooot + -Q-o/2k sm. o-ot000 k(l - ~ ) 1 - ~(B.31)(B.32)(B.33)(B.34)It is
interesting to consider the case in which the system is initial l y at rest in its
equil ibriumposition, (i.e., Uo = l io = 0). For this case the response is given byU
= -Qko--12 (ssim o- ot-I-A' s'm ooot)1-~(B.35)which indicates that the response has
two components. One component occurs in responseto the appl ied l oading and occurs
at the frequency of the appl ied l oading. The other is a freevibration effect
induced by the initial conditions; it occurs at the natural frequency of thesystem.
It is useful to real ize that the term QoIkin equation (B.35) represents the
displ acementof the mass that woul d occur if the l oad Qowas appl ied statical l y. The
term 1/(1 _ ~2)can then be thought of as a magnification factor that describes the
amount by which thestatic displ acement ampl itude is magnified by the harmonic l oad.
The magnification factorvaries with the tuning ratio, ~,as shown in Figure E.9.
Note that the displ acement ampl itudeis greater than the static displ acement for
l oading frequencies l ower than J2 000. At higherl oading frequencies, the
displ acement ampl itude is l ess than the static displ acement and canbecome very
smal l at high frequencies. However, the response of an undamped SDOF systembecomes
very l arge as wapproaches 000. When harmonic l oading is appl ied at the natural
frequency of an undamped SDOF system, the response goes to infinity indicating
resonance of the system. However, since trul y undamped systems do not exist, true
resonanceis never real l y achieved. The concept of the tuning ratio that rel ates the
frequency ofl oading to the natural frequency of the system is an important one, as
evidenced by its stronginfl uence on the response.556 Dynamics of Discrete Systems
App. BMF1.01.0 f3 = mWoFigure B.9 Variation of magnification factor with tuning
ratio for undamped SDOFsystem.Exampl eB.3From an initial stationary state, the
undamped SDOF system of Exampl e B.l is subjected to aharmonic base accel eration of
0.20g at a frequency of 2 Hz. Compute the response of thesystem.Sol ution Expressing
the base motion asUb(t) = (0.2)(32.2 ft/sec-) sin 41tt = 6.44 sin 41ttthe
equival ent external force woul d beQ(t) = - .!: !: : Ub(t) = - 10,000 Ib2 (6.44
ft/sec-) sin 41tt = -(2000 Ib) sin 41ttg 32.2 ft/secThe tuning ratio woul d beThen,
from equation (B.35),21t(2)69.50.181u(t) = -Qo-1- (ssiin w- t- ~.sm wot )k I _ ~2
5~~ ~ps/f 1 2 [sin 41tt-0.181 sin (69.5t)]1 IpS tI - (0.181)= 0.00138 sin 41tt -
0.00025 sin 69.5twhich is pl otted in Figure EB.3.Sec. 8.5 Response of Linear SDOF
Systems 5571.5 2 2.5 3 3.5Time (sec)0.5oC 0.001Q)EQ)al g. -0.001(5-0.002
'------'-------'------'-------'----'------'----~o0.002
,---------r---,--------,----,------,----,----, gFigure EB.38.5.3.2 Damped Forced
VibrationsThe most general case is that of a damped system subjected to forced
harmonic l oading.Each of the three prior cases can be considered as a subset of
this one since their equationsof motion can be obtained by setting various terms of
the equation of motion fordamped forced vibrations shown bel ow to zero. The
equation of motion for a damped SDOFsystem subjected to simpl e harmonic l oading of
the form Q(t) = Qo sin rot ismii + cit + ku = Qo sin rot (B.36)After dividing by m
and using the rel ationships ~ =cl 2mffio and ffi~ =kim, equation (B.36)can be
rewritten as.. 2t . 2 Qo.u+ .."ffiou + ffiOU = -m sm ffit(B.37)The compl ementary
sol ution represents the damped free vibration response, which wasexpressed for an
underdamped system by equation (B.19).uc(t) = e-~ffiot(C1 sin ffidt + C2 cos ffid
t)Since the response of a damped SDOF system is general l y out of phase with the
external l oading,a harmonic particul ar sol ution of the formUp(t) = C3 sin rot + C4
cos rotcan be assumed. The corresponding vel ocity and accel eration areup(t) =
C3rocos rot - C4ro sin rot(B.38a)(B.38b)iip(t) = -ro2C3 sin rot -r- ro2C4 cos rot
(B.38c)Substituting equations
(B.38) into the equation of motion [equation (B.37)] and groupingthe sinrot and
cosrot terms gives(C3ffi5 - C3ro2- 2~ffioC4ro) sin rot(B.39)558 Dynamics of
Discrete Systems App. BNow, at the instances where rot =0 + ni: (where n is any
positive integer), sintor =0 andcos rot =1. Thus the rel ationship(BAOa)must be
satisfied. Further, at rot = rrJ2 + nx, cos rot = 0 and sin rot = 1, which means
that(BAOb)(BAIa)(BAIb)(BA2)must al so be satisfied. Equations (BAO) represent two
simul taneous equations with the twounknowns C3 and C4. Sol ving for the unknowns
yiel dsC3= Qo 1- ~zk (1 _ ~z)z + (2~~)zC = Qo -2~~4 k (1- ~z)z + (2~~)zThe general
sol ution to the equation of motion for damped forced vibration can now beobtained
by combining the compl ementary and particul ar sol utions-~(J) t u (t) = e 0 (Cj sin
O)dt + Cz cos O)dt)+ Qko / [( 1 - ~z) sin rot - 2~~ cos ffit](1 - ~z) + (2~~)zwhere
the constants Cj and Cz depend on the initial conditions. There are several
importantcharacteristics of this sol ution. Note that the compl ementary sol ution
(which represents theeffects of the initial conditions) decays with time. The
compl ementary sol ution thereforedescribes a transient response caused by the
requirement of satisfying the initial conditions.After the transient response dies
out, onl y the steady-state response described by the particul arsol ution remains.
The steady-state response occurs at the frequency of the appl iedharmonic l oading
but is out of phase with the l oading.Exampl e B.4The SDOF system shown in Figure
EB.4a is at rest when the sinusoidal l oad is appl ied. Determinethe transient,
steady state, and total motion of the system./ 1001b/in.Figure EB.4aQ (t) =(100 Ib)
sin tt tSol ution From equation (B.42), the total response is given bySec. 8.5
Response of Linear SDOF Systemsu (t) = e-~(j'o'(C1 sin ffidt + C2 cos ffidt)+ Qo 21
[(1 - ~2) sin wt - 2S~ cos Wt]k (1 - ~2) + (2SW559For zero initial displ acement,u(t
=Q) = 0= e-~(j)o(O) [C1 sin ffiAO) + C2 cos ffiAO)]+ Qo 21 [( I _ ~2) sin 00 (0) -
2S~ cos 00(0)]k(I_~2) +(2SW= C + Qo -2s~2 k (1 _ ~2)2 + (2s~)2orC = Qo 2sP2 k(I-
~2)2+(2SWFor zero initial vel ocity,Ii = 0= ffid-~(j)o(O) [CI cos ffid(O) - C2 sin
ffid(O)] - Sffio e -~(j)ot [C1 sin ffiAO) - C2 cos ffiAO)]+ Qko 200 2[(1- ~2)COS
00(0) + 2S~ sin 00(0)](1- ~2) + (2S~)C S C Qo 00(1 - ~2)= ffid 1- ffio 2+ k (I _
~2)2 + (2S~)2orC _ Qo 00 p2 _ 1I - k ffid (1 _ ~2)2 + (2sfWThen the transient
motion is given byuc(t) = Qko 21 e-~(j)ot [ il l (~2 + 2S2-1) sin ffidt + 2S~ cos
ffidtJ(I - ~2) + (2S~)2 ffidand the steady-state motion byup(t) = Qko 2 / 2 [(1-
~2) sin wt - 2S~ cos wt](l -~) +(2S~)The total motion is the sum of the transient
and steady-state motions. For the system shown inFigure EB.l a,{k [kg (100 l b/in.)
(12 in./ft) (32.2 ft/sec/sec) _ 622 d/ffio = ~m = ~W = 1000l b - . ra secS = _c_ = ~
= (3 Ib-sec/in.)(12 in./ft)(32.2 ft/sec/sec) = 0.0932mffio 2Wffio 2(1000 Ib)(6.22
rad/sec)ffid = ffio~ = JI - (0.092) 2 = 6.19 rad/sec~ = 00 = 1t rad/sec = 0.505ffio
6.22 rad/sec560 Dynamics of Discrete SystemsSubstituting these val ues into the
sol utions gives the response shown in Figure EB.4bApp.B2§.CQl E 0 Ql oel l C.(fJ (5 -1
Transient-20 2 4Time (sec)Figure EB.4b6 8 10The steady-state response coul d al so be
described byu = A sin (COt + <1»where(B.43)A =Qo 1k J(l - P2)2 + (2~P)2<I> = tan-l (-
~)1- p2The steady-state response can be visual ized with the aid of rotating
vectors, both for theresponse and for the forces induced in the system, as shown in
Figure B.1O. Note that thespring, dashpot, and inertial forces act opposite to the
displ acement, vel ocity and accel erationvectors, and that the displ acement l ags the
appl ied l oading vector by the negative phaseu, it, il AkAFigure RIO Rotating vector
representation of response and forces in vibrating SDOFsystem.Sec. 8.5 Response of
Linear SDOF Systems 561(b)-0: 7070.40 ~_ (j)S=0.01 0.20 : - 0)01.0 2.01t -----
----------------<I> 1t: 21.0 2.0(a)o~= (j)L ------L ---'--- 0)0o3Figure B.l l
Variation of (a) magnification factor, and (b) phase angl e with damping ratio and
tuning ratio.angl e, <\>. For harmonic l oading the phase angl e varies with both
damping ratio and tuningratio, as shown in Figure B.l l a.The infl uence of the tuning
ratio can be il l ustrated by the use of the magnification factor,again defined as
the ratio of the ampl itude to the static displ acement: M=~ = 1 (B.44)Qo/k J(l - ~2)2
+ (2~~)2The variation of the magnification factor with tuning ratio and damping
ratio is shown inFigure B.l l b. The damping ratio infl uences the peak magnification
factor and al so the variationof magnification factor with frequency. The
magnification factor curves broaden withincreasing damping ratio. Note that the
magnification is unbounded (resonance) onl y for~ =0 and ~ =1. For nonzero damping,
there is some maximum magnification, Mmax,M max = (B.45)2~~which occurs when the
tuning ratio ~ = J1 - 2~2 . The shape of the magnification curve isobviousl y
control l ed by the damping ratio. Al though a system with l ow damping may produce
l arge magnification at a tuning ratio near 1, it wil l exhibit significant
magnificationover a smal l er range of frequencies than a system with higher damping.
a: .5.4 Response of SDOF Systems to Periodic LoadingThe sol utions for the response
of a SDOF system to harmonic l oading devel oped in the precedingsection can be used
to devel op sol utions for the more general case of periodic l oading.As shown in
Appendix A, periodic l oading can be approximated by a Fourier series(i.e., as the
sum of a series of harmonic l oads). The response of a SDOF system to the periodic
l oading, using the principl e of superposition, is simpl y the sum of the responses
to eachterm in the l oading series. The required cal cul ations can be performed using
trigonometricor exponential notation.562 Dynamics of Discrete Systems App. 88.5.4.1
Trigonometric NotationFrom equation (A. 11) a periodic l oad, Q(t), can be expressed
by the Fourier seriesQ(t) = ao + L an cos (On t + b; sin (On tn = 1where the
Fourier coefficients areao = ; fT! Q(t) dtf 02 fT! an T Q(t) cos (On t dtf 0b; = i
fT! Q(t) sin (Ont dtf 0and (On = 2rrn/Tj- Using the steady-state portion of
equation (B.42), the response to each sineterm in the Fourier series isu.; sin(t) =
bkn 21 [(I-~~)sinrot-2~~ncosrot], (l - ~~) + (2~~n)2where ~n =(OnTi2rr. In the same
way, the steady-state response to each cosine term can beshown to be= akn 21[( 1 -
~~) cos rot + 2~~ n sin rot](1 - ~~) + (2~~n)2Since the steady-state response to
the constant l oad term is the static displ acement, Uo =ao/k,the total steady-state
response is given byu(t) = Uo + L un, sin(t) + Un,cas(t)n= 1= ~(a + L 2 212
{ [an2~~n + bn(l - ~~)] sin (On to n = I (l - ~n) + (2~13n)+ [an(l - ~~) - bn2~13n]
cos (Ont })(B.46)2.5.4.2 Exponential NotationPeriodic l oading can al so be described
by the Fourier series in exponential form.Using equation (A. 15), a periodic l oad
can be expressed asn =_00Sec. 8.5 Response of Linear SDOF Systems 563The compl ex
Fourier coefficients, q~, can be determined directl y from Q(t) asq~ = i rfQ(t)e-
iOl "l dt fJoThe response of a SDOF system l oaded by the nth harmonic woul d be
governed by theequation of motionmiin(t) + (cun(t) + kun(t» = q~eiOl "tThe response
of the system can be rel ated to the l oading by(B.47)(B.48)where H(ron) is a
transfer function [i.e., a function that rel ates one parameter (in this case,the
displ acement of the oscil l ator) to another (the external l oad)]. Substituting
equation(B.48) into the equation of motion givesorI (B.49)k(- P~ + 2iPn~ + 1)H(ron)
= = ----,,--------- mro~ + icas; + kSince A" = a + ib = Aei8, where the modul us, A
= Ja z + bZ, and the argument, e =tan-1(bl a), the transfer function can al so be
written as11k (. -12~Pn)H(ron) = exp I tan -z-- J(l - P~)z + (2~Pn)z Pn - 1Note the
cl ose rel ationship between the modul us of the transfer function and the
magnificationfactor of equation (B.44). Because the transfer function can be used
for any frequencyin the series, the principl e of superposition gives the total
response as(B.50)n =_00Many different transfer functions can be devel oped. For
exampl e, a transfer functionrel ating the accel eration of the SDOF system to the
external l oad coul d have been devel opedjust as easil y. The advantages of the
transfer function approach l ie in its simpl icity and in theease with which it
al l ows computation of the response to compl icated l oading patterns.The transfer
function may be viewed as afil ter that acts upon some input signal to producean
output signal . In the case just considered, the input signal was the time history
ofl oading, Q(t), and the output was the displ acement, u(t). If the input signal has
Fourierampl itude and phase spectra, Fi(ron) and I\>;(ron), the Fourier ampl itude
spectra of the outputsignal wil l be given byFo(ron) = H(ron)Fi(ron)<Po(ron) =
H(ron)<Pi(ron)(B.51a)(B.51b)564 Dynamics of Discrete Systems App.BThus the
procedure for Fourier anal ysis of SDOF system response can be summarized in the
fol l owing steps: 1. Obtain the Fourier series for the appl ied l oading (or base
motion). In doing so, thel oading (or base motion) is expressed as a function of
frequency rather than a functionof time.2. Mul tipl y the Fourier series coefficients
by the appropriate val ue of the transfer functionat each frequency, (On- This wil l
produce the Fourier series of the output motion.3. Express the output motion in the
time domain by obtaining the inverse Fourier transformof the output motion.It is
precisel y this approach that forms the backbone of several of the most commonl yused
methods for anal ysis of ground response and soil -structure interaction. These
methodsare presented in Chapter 7.8.5.5
Response of SDOF Systems to Genera' LoadingNot al l l oading is harmonic or even
periodic. To determine the response of SDOF systemsto general l oading conditions, a
more general sol ution of the equation of motion is required.8.5.5.1 Response to
Step LoadingConsider a damped SDOF system subjected to a step l oad of intensity,
Qo, which isappl ied instantaneousl y at t = 0 and removed instantaneousl y at t = t]
as shown in FigureB.12. For t: : ; t], the compl ementary sol ution to the equation of
motion for this system [equation(B.19)],~~ro tuc(t) = e 0 [C] sin (Odt + Cz cos
(Odt]describes the transient response of the system. The equation of motion for the
steady-statecondition is given bymiip + cUp + kup = QoSince the appl ied l oad does
not vary with time, the steady-state response wil l be a constantdispl acement,o(t)0
0 1----------,oFigure B.12 Time history of step l oading.Sec. 8.5 Response of Linear
SDOFSystems 565The general sol ution to the step l oading probl em for t : s; tl can
then be written asu(t) = ~o + e-~°l ol (C 1 sin (Odt + C2 cos (Odt) (B.52)with free
vibration occurring at t > t] (when no extemal l oad is appl ied). The constants are
determined by the initial conditions, Uo and uo' At t =0,Q ~~ ~Uo = / + e [C] sin
(Od(O) + C2 cos (Od(O)] = T + C2Uo = e-~roo(o) [(OdC] cos (Od(O) - (OdC2 sin
(Od(O)]-S(Ooe-~roo(o) [C 1 sin (Od(O) + C2 cos (Od(O)] = (OdC] - S(OOC2from whichso
thatQo -~root [uo + S(Oo(uo - Qol k) . (Qo) ]u(t) = T + e (Od sm (Odt + Uo - T cos
(Odtdescribes the response of the system up to the beginning of free vibration at t
= tI'(B.53)8.5.5.2 Dirac Pul seA particul ar type of step l oading can be described
using a Dirac del ta function. ADirac del ta function is one whose val ue is zero at
al l val ues of u except one at which it goesto infinity in such a way that the area
under the function is unity. Mathematical l y, the Diracdel ta function satisfies the
conditions{o for x"i= ao(x) = 00 for x = af= o(x)dx = I-c-oo(B.54a)(B.54b)(B.55)
Define a Dirac pul se as a constant force Qoappl ied over a duration t] that
approaches zeroas shown in Figure B.l 3. From impul se-momentum principl es, Qot] =muo
(t]). As t 1approaches zero, the effect of the Dirac pul se is to cause an initial
vel ocity Uo =Qotl im, withno initial displ acement. Thus the steady-state response
occurs onl y over an infinitesimal period of time, and the system is immediatel y set
into free vibration. From equation (Bol O),the response to a Dirac pul se disturbance
at t = 0 is given by-~ro u(t) = e 0 t(-Qo-t j.sm (Odt )m(Od566 Dynamics of Discrete
Systems App. BFigure B.13 Dirac pul se l oading.(B.56)8.5.5.3 Duhamel Integral A
general l oading function such as, that shown in Figure B.14 can be thought of as a
train of l oad pul ses, each of infinitesmal duration. Looking at one of these
pul ses, the pul seof duration di occurring at t ='T. (Figure B.14), the response it
causes at a l ater time, t = t,fol l ows from equation (B.55): d (- ) -~roo(i-t)
Qo('T.)d'T.. (- )u t = e SIl l ood t - 'T.mOOdThe response induced by the entire
train of l oad pul ses can be obtained by summing theresponses of al l of the
individual pul ses up to the time t = t, that is,(B.57)nu(l : ) = _1_" Q('T.Jsin
ood(t-'T.;)d'T.moodL...;= Iwhere n is the total number of pul ses up to t =t. As dt
approaches zero, the summationbecomes an integral with which the total response can
be cal cul ated as- 1 fl (B.58) -~ro (i-t) - u(t) = - Q('T.)e 0 sin ood(t- 'T.) dt
mOOd 0This equation describing the response of a l inear system is known as
Duhamel 's integral . Itis usual l y very difficul t to sol ve anal ytical l y, but can be
integrated numerical l y by a varietyof procedures. Its use, however, is constrained
to l inear systems.Q(t)1..--------·1 1---d1: Figure B.14 Pul se of duration d1:
occurring at I = 1: .Sec. B.68.6 DAMPINGDamping 567Energy is dissipated in soil s and
structures by several mechanisms, incl uding friction, heatgeneration, and pl astic
yiel ding. Fot specific soil s and structures, however, the operativemechanisms are
not understood sufficientl y to al l ow them to be expl icitl y model ed. As aresul t, the
effects of the various energy l oss mechanisms are usual l y l umped together and
represented by some convenient damping mechanism.8.6.1 Viscous DampingThe most
commonl y used mechanism for representing energy dissipation is viscous damping.When
a viscous damped SDOF system such as that shown in Figure B.3 is subjected toa
harmonic displ acementu(t) = uosincotthe net force exerted on the mass by the spring
and dashpot isF(t) = ku(t) + cu(t) = kuosinil l t + cil l uocosil l tEval uating these
functions from time to to time to + 21t/co yiel ds the force-displ acementval ues for
one cycl e of a hysteresis l oop. When the viscous damping coefficient, c, is zero,
the force and displ acement are in phase and proportional to each other, impl ying a
l inearel astic stress-strain rel ationship. For nonzero damping, however, the
hysteresis l oop iseil l iptical ,as shown in Figure B.l S. Note that when the
displ acement is zero, the spring forceis zero and the net force comes entirel y from
the dashpot. Simil arl y, when the vel ocity iszero (at cot = 1t/2 + n1t), the dashpot
force vanishes and the net force consists entirel y ofthe spring force. The aspect
ratio of the hysteresis l oop decreases with increasing damping;the l oop becomes a
circl e when c = kl co.F/u/Figure B.IS Stress-strain behaviorimpl ied by viscous
damping. Hysteresisl oop is el l iptical .568 Dynamics of Discrete Systems App. B
Obviousl y, the shape of the hysteresis l oop depends on the viscous damping
coefficientand therefore on the damping ratio. Hence we shoul d be abl e to determine
the dampingratio from a known hysteresis l oop. The energy dissipated in one cycl e
of oscil l ation isgiven by the area inside the hysteresis l oop and can be obtained
fromto + 2rr/iii dWD =r F d~dt = ncffiu6 (B.59)oAt maximum displ acement, the
vel ocity is zero and the strain energy stored in the system isgiven by1 2 Ws =
"2kuo (B.60)Equations (B.59) and (B.60) show that c = WDI (nO)u6) and k = 2Wsl u6.
Substitutingthese into equation (B.16) with (j) = 0)0 gives an expressionJ'- WD ..,
= 4nWsthat is commonl y used for graphical determination of the damping ratio from a
measuredhysteresis l oop. Referring to Figure B.16, the damping ratio is taken as
the ratio of the areaof the hysteresis l oop to the area of the shaded triangl e, al l
divided by 4n. This graphical eval uation of the damping ratio is commonl y used in
the interpretation of many of the l aboratorytests discussed in Chapter 6.FuFigure
R16 Graphical eval uation ofdamping ratio from measured hysteresisl oop. The damping
ratio is proportional tothe ratio of the shaded area to the area ofthe hysteresis
l oop.The damping characteristics of a l inear system can al so be eval uated from its
frequencyresponse characteristics. Setting the magnification factor expression
[equation(B.44)] equal to Mmaxl J2, the hal f-power tuning ratios, shown in Figure
B.17, can beapproximated as~l "'" 1_~_~2~2 "'" 1 + ~ - ~2Sec. 8.6 Damping 569Figure
B.17 Hal f-power tuning ratios foreval uation of damping ratio frommagnification
curve.Therefore, the damping ratio is given by hal f the difference between the
hal f-power tuningratios(B.61)or, when the response is expressed in terms of
frequency, where (01 =PI (00 and ~ =P2 (00,j: . = (02 - (01~.~ (B.62)(02 + (01Thus
the damping ratio of a system can be measured by exciting the system at different
frequenciesand determining the ampl itude of the magnification factor at each
frequency.EI.6.2 Other Measures of Energy DissipationIn addition to the damping
ratio, ~, a number of other parameters have been used to describeenergy dissipation
characteristics. Seismol ogists, for exampl e, often work with the qual ityfactorIn
vibration anal ysis, the l oss factorQ12~(B.63)and specific damping capacity11 = 2~
'If = 21t~(B.64)(B.65)are often used (Goodman, 1988).It is important to remember
that the damping ratio, and any of these other parameters,are simpl y parameters
used to describe the effects of phenomena that are often poorl yunderstood. They
al l ow the effects of energy dissipation to be represented in a mathematical l y
convenient manner. For most soil s and structures, however, energy is dissipated
hysteretical l y(i.e., by yiel ding or pl astic straining of the material ). In such
cases the behavioris more accuratel y characterized by eval uating the nonl inear
response of the system.5708.6.3 Compl ex StiffnessDynamics of Discrete Systems App.
BA viscousl y damped system can be represented convenientl y in a different but
equival entway for a cl ass of techniques known as compl ex response anal ysis.
Consider a dampedSDOF system subjected to simpl e harmonic l oading of ampl itude
Qoand l oading frequencyffi. The l oading can be represented byQ(t) = QoeiWl Assuming
that u(t) = UoeiWl , the steady-state sol ution to the equation of motionmil + cit +
ku = QoeiWl is(B.66)(B.67)u(t) = Qo eiWt (B.68)k - mffi2 + icffiNow consider the
SDOF system of Figure B.18, which has no dashpot but which hasa spring of compl ex
stiffness k* = k j + ik2• The equation of motion for this system ismil + k*u =
QoeiWl Again assuming that u(t) =UoeiW, the steady-state sol ution can be expressed
asQou(t) = e'Mk* -2 - (f) m(B.69)(B.70)Comparing equations (B.68) and (B.70), it is
apparent thatk* = k + icffi (B.71)By the appropriate choice of k*, the displ acement
ampl itude of equation (B. 70) can be madeequal to that of equation (B.68),
(al though a smal l phase difference between the two sol utionswil l remain). To
accompl ish this, the compl ex stiffness is represented as(B.72)where ~ ~ 1. For the
usual smal l damping ratios considered in earthquake engineering probl ems,the ~2
terms can be negl ected so that k* '"k(l + 2i~). Using this
expression for k*, theerror in phase angl e between the responses given by
equations (B.68) and (B.70) is ~e '"2~/(l + ~). As a resul t, a viscousl y damped
system can be represented as an undamped systemwith compl ex stiffness. The use
ofthis approach, however, is restricted to cases ofharmonicmotion. For probl ems in
which l oading is characterized as periodic (and therefore asthe sum as a series of
harmonic l oads), the use of compl ex stiffness greatl y simpl ifies cal cul ationof the
response of damped systems./Figure B.18 SDOF system with spring of/ compl ex
stiffness.Sec. 8.7 Response Spectra 571(B.74)(B.73b)(B.73a)For smal l damping
ratios, the compl ex stiffness then consists of real and imaginarypartsRe(k*) = k
Im(k*) = 2k~Consequentl y, the damping ratio can be expressed as_ Im(k*)~ - 2Re(k*)
which is useful to remember in the interpretation of quantities such as compl ex
impedancefunctions, which are usual l y expressed in terms of their real and
imaginary parts.B.7 RESPONSE SPECTRAFor earthquake-resistant design, the entire
time history of response may not be required.Instead, earthquake-resistant design
may be based on the maximum (absol ute) val ue of theresponse of a structure to a
particul ar base motion. Obviousl y, the response wil l depend onthe mass, stiffness,
and damping characteristics of the structure and on the characteristics ofthe base
motion.The response spectrum describes the maximum response of a singl e-degree-of-
freedom(SDOF) system to a particul ar input motion as a function of the natural
frequency (ornatural period) and damping ratio of the SDOF system (Figure B.l 9).
The response may beexpressed in terms of accel eration, vel ocity, or displ acement.
The maximum val ues of eachof these parameters depend onl y on the natural frequency
and damping ratio of the SDOF......' ./////////Natural periodof vibrationFigure
B.19 Response specrum. Spectral accel erations are the maximum accel eration
ampl itudes of SDOF systems in response to the same input motion. The response
systemis obtained by pl otting the spectral accel erations against the periods of
vibrations of theSDOF systems.572 Dynamics of Discrete Systems App. Bsystem (for a
particul ar input motion). The maximum val ues of accel eration, vel ocity, and
displ acement are referred to as the spectral accel eration (Sa), spectral vel ocity
(S,), andspectral displ acement (Sd), respectivel y. Note that a SDOF system of zero
natural period(infinite natural frequency) woul d be rigid, and its spectral
accel eration woul d be equal tothe peak ground accel eration.Appl ication of the
Duhamel integral to a l inear el astic SDOF system produces expressionsfor the
accel eration, vel ocity, and displ acement time histories that are proportional (bya
factor of ro), except for a phase shift. Because the phase shift does not
significantl y infl uencethe maximum response val ues, the spectral accel eration,
vel ocity, and displ acementcan be approximatel y rel ated to each other by the
fol l owing simpl e expressions: Sd = l ui max (B.75a)(B.75b)(B.75c)where u and roo are
the displ acement and natural frequency of the SDOF system, PSV is thepseudospectral
vel ocity, and PSA is the pseudospectral accel eration. Al though the PSVand PSA are
not the true maximum val ues of vel ocity and accel eration, they are usual l y very
cl ose to the maxima for recorded strong ground motions. In practice, the
pseudospectral val uesare general l y assumed to be equal to the spectral val ues.B.8
RESPONSE OF NONLINEAR SDOF SYSTEMSTO GENERAL LOADINGNumerical integration of the
Duhamel integral is very useful for cal cul ation of the responseof l inear systems to
general l oading. Many systems for which the seismic response is to becal cul ated,
however, exhibit nonl inear behavior. In such systems the mass is usual l y constant,
but the damping coefficient and/or the stiffness may vary with time, defl ection, or
vel ocity. It wil l be useful to devel op methods for anal ysis of the response of
nonl inear systems,recognizing that they wil l be appropriate for l inear systems as
wel l when damping andstiffness val ues are hel d constant.The most common approach to
nonl inear anal ysis is the direct integration of incremental equations ofmotion that
govern the response of the system over smal l time increments.The response is
cal cul ated for each time increment after adjusting the stiffness anddamping at the
beginning of the increment. By using the conditions at the end of one timeincrement
as the initial conditions for the next time increment, the nonl inear system is
approximated as an incremental l y changing l inear system.B.8.1 Incremental Equation
of MotionConsider the SDOF system shown in Figure B.20, which has a nonl inear
spring and dashpot(i.e., the spring force is not proportional to displ acement and
the dashpot force is not proportional to vel ocity). Dynamic equil ibrium at time 't
requires that(B.76)Sec. 8.8 Response of Nonl inear SDOF Systems to General Loading
573/I} Ms (1: )IIIIIU, u't+L'n: U0(1)m ~Figure B.20 SDOF system with nonl inear
damping and spring forces.and thatf/(1: + Ll 1: ) + fD(1: + Ll 1: ) + fs(1: + Ll 1: ) =
Q(1: + Ll 1: )at time 1: + Ll 1: . Defining(B.77)Ll f/(1: ) = f/(1: + Ll 1: ) - f/(1: )
Ll fD(1: ) = fD(1: + Ll 1: ) - fD(1: )Ll fs(t) fs(t + Ll t) - fs(t)Ll Q(t) Q(t + Ll t) - Q(t)
and subtracting equation (B.76) from equation (B.77), the incremental equation of
motionfor the time interval from t to r + Ll t is(B.78)or expressing the incremental
forces in terms of incremental displ acements, vel ocities, andaccel erations, asm
Ll ii(t) + c(t) Ll u(t) + k(t) Ll U(t) = Ll Q(t) (B.79)By integrating this incremental
equation of motion in a series of smal l time steps, theresponse of the nonl inear
system can be obtained. It shoul d be noted that this approach canbe used to
cal cul ate the response of l inear el astic, nonl inear el astic, or nonl inear inel astic
material s with stress-strain behaviors shown in Figure B.2l . The third of these is
particul arl yimportant because it al l ows representation of the hysteretic damping
displ ayed bycycl ical l y l oaded soil s.u u u(a) (b) (c)Figure B.21 Stress-strain
behavior of (a) l inear el astic, (b) nonl inear el astic, and (c)nonl inear inel astic
material s under cycl ic l oading conditions.5748.8.2 Numerical IntegrationDynamics of
Discrete Systems App. BThere are many ways to numerical l y integrate the incremental
equation of motion. One ofthe simpl est and most easil y coded of these is the l inear
accel eration method. It is based onthe assumption that the accel eration varies
l inearl y within each time increment. If the accel erationin the time increment
varies l inearl y, the vel ocity and displ acement wil l vary quadratical l yand
cubical l y, respectivel y, as shown in Figure B.22. Therefore, expressions forthe
incremental vel ocity and displ acement can be written in terms of the incremental
accel eration,that is,L1.u('t) = ii('t)M + L1.ii('t)~tL1.u('t) = u('t)M +
L1.ii('t)L1.t + L1.ii('t)~2(B.80)(B.8!)Rearranging, the incremental accel eration
and vel ocity can be expressed in terms of theincremental displ acementL1.ii('t)
(B.82a)L1.u('t) = ~tL1.U('t) - 3u('t) - ~tii('t)Substituting equations (B.82) into
the incremental equation of motion givesm [L1.~2L1.U('t) - : /l ('t) - 3ii('t)] +
c('t) [~/U('t) - 3u('t) - ~t ii('t)]+ k('t) L1.u('t) = L1.Q('t)Linear ~(B.82b)
(B.83)Figure B.22 Variation of accel eration,vel ocity, and displ acement over time
increment.Sec. 8.9 Mul tipl e-Degree-of-Freedom Systems 575which can be sol ved for
the unknown incremental displ acementAuCt) = AQ('t) + m [(6/At)u('t) + 3ii('t)] +
c('t)[3u('t) + (AtI2)ii('t)] (B.84)k('t) + (6/At2 )m + (3/At)c('t)Equation (B.84)
shows that if the displ acement, vel ocity, and accel eration at time 't areknown, the
incremental displ acement during the succeeding time increment A't based on the
l oading and the stiffness and damping during that time increment can be cal cul ated.
Fromthis incremental displ acement, the incremental vel ocity and accel eration, and
from these thedispl acement, vel ocity, and accel eration at the end of the time
increment, can be determined.The conditions at the end of the time increment are
then taken as the initial conditionsfor the next time increment and are used to
cal cul ate the appropriate stiffness anddamping val ues for the next time increment.
To prevent the accumul ation of errors resul tingfrom the assumptions of the l inear
accel eration method, the accel eration at the beginning ofeach time step shoul d be
cal cul ated by subtracting the damping and spring forces from thetotal external l oad
and dividing the resul t by the mass. This wil l ensure that total equil ibriumis
satisfied at each step of the anal ysis.For num~rical stabil ity, it is necessary
that the time steps be rel ativel y smal l , typical l yl ess than about 55% of the
undamped natural period of the system. These smal l time stepscan l ead to
considerabl e computational effort, particul arl y when such direct integrationmethods
are appl ied to mul tipl e-degree-of-freedom systems. A number of other numerical
integration techniques are avail abl e; Berg (1989) describes the appl ication of
several tostructural dynamics probl ems.8.9 MUILTIPLE-DEGREE-OF-FREEDOM SYSTEMSIn
most physical systems, the motion of the significant masses cannot be described by
a singl evariabl e; such systems must be treated as mul tipl e-degree-of-freedom (MDOF)
systems.With the exception of onl y the simpl est cases, the types of buil dings,
bridges, and otherstructures that are of interest in earthquake engineering have
mul tipl e degrees of freedom.Some structures can be ideal ized with onl y a few
degrees of freedom; others may requirehundreds or even thousands.In many respects,
the response of MDOF systems is simil ar to the response of SDOFsystems, and
procedures for anal ysis are anal ogous to those described previousl y for SDOF
systems.
Al though the additional degrees of freedom compl icate the al gebra, the procedures
are conceptual l y quite simil ar. In fact, a very useful approach to the response of
l inearMDOF systems al l ows their response to be computed as the sum of the responses
of a seriesof SDOF systems.1: 1.9.1 Equations of MotionIn eval uating the response of
an MDOF system, dynamic equil ibrium of al l masses must beensured simul taneousl y.
Consider the ideal ized two-story structure shown in Figure B.23.The structure has
two degrees of freedom: horizontal transl ation of the upper mass and horizontal
transl ation of the l ower mass. For each mass the external l y appl ied l oad must be
bal ancedby the inertial , damping, and el astic forces that resist motion: 576
Dynamics of Discrete Systems App. BIII + IDl + l SI = qj (t)112+ID2+l s2 = q2(t)or,
in matrix form,Figure B.23 Two-degree-of-freedomsystem. Displ acements of masses 1
and 2from equil ibrium positions are Uj and U2,respectivel y.(B.85a)(B.85b)(B.86)If
the structure exhibits l inear behavior, the principl e of superposition is val id.
Thenthe forces that resist motion at each l evel can be expressed in terms of
coefficients by whichthe motion parameter at al l l evel s are mul tipl ied. For
exampl e, the el astic force resistingmotion at l evel 1 can be expressed asl SI =
kIl uj +k12U2where the stiffness coefficients kijrepresent the force induced at
l evel i due to a unit displ acementat l evel } (with the displ acements at al l l evel s
except) hel d equal to zero). Inmatrix formorfs = kuin which k is the stiffness
matrix of the structure.Simil arl y, a damping matrix arid a mass matrix can be
devel oped in which the el ementscij(or mi) represent the damping (or inertial )
forces resisting motion at l evel i due toa unit vel ocity (or accel eration) of
l evel }. Dynamic equil ibrium of the MDOF system canthen be described by a set of
simul taneous equations of motion, which can be expressed inmatrix form asmti + cu +
ku = q(t) (B.87)MDOF systems al so respond to base motions. The equation of motion
for the case ofbase shaking is easil y devel oped fol l owing the same procedure
appl ied to the SDOF case inSection B.4.2. The resul ting equation of motion isrnii +
cu + ku = -ml iibCt) (B.88)Sec. B.9 Mul tipl e-Degree-of-Freedom Systems 571Equation
(B.95) indicates that the response of an N-story structure to base motion is equal
to the response to equival ent external l oads, where qi=-miub(t) (i =1,N) is the
l oad appl iedto the ith fl oor.8.9.2 Undamped Free VibrationsFor undamped free
vibrations, al l terms of the damping matrix are zero, so the equations ofmotion
reduce tomii + ku = 0 (B.89)Assuming that the response of each mass (degree of
freedom) is harmonic, we haveutr) = Usin (tor + 9) (B.90)where U is a vector
containing the displ acement ampl itudes and 6 is a vector containing thephase angl es
at each l evel of the structure (or for each degree of freedom). Differentiating
equation (B.90) twice givesii(t) = -m2Usin(mt + 9) = -m2u(t) (B.9l )Substituting the
expressions for displ acement [equation (B.90)] and accel eration [equation(B.91)]
into the equation of motion [equation (B.89)] yiel ds-mm2Usin(mt+9) + kUsin(mt+ 9) =
0or(B.92)which is a set ofl inear al gebraic equations with unknown U. A nontrivial
sol ution (one thatgives val ues other than U = 0) can be obtained onl y ifdet (k -
m2m) = Ik - m2ml = 0 (B.93)Equation (B.93) is the frequency equation (or
characteristic equation) of the system, whichfor a system of N degrees of freedom,
wil l give a pol ynomial of Nth degree in m2• The Nroots ofthe frequency equations
{mr, m~, m~, ... , m~} representthefrequencies at whichthe undamped system can
oscil l ate in the absence of external forces. These frequencies arecal l ed the
natural circul ar frequencies of the system.Exampl eB.5Compute the natural
frequencies of the three-story structure shown in Figure EB.5.10,000 kg2000 kN/m
12,000 kg2500 kN/m15,000 kg3000 kN/m~~~~{}~7 Figure EB.5578 Dynamics of Discrete
SystemsSol ution The mass matrix for this simpl e structure isApp. B[10 0 o~m = 1000
0 12 0o 0 15kgThe stiffness matrix can be determined by appl ying a unit
displ acement to each fl oor (with zerodispl acement at the other fl oors) and
eval uating the resul ting forces. By this procedure the stiffnessmatrix isThen[5.5
-2.5 0 l k = 1,000,000 -2.5 4.5 -2.0o -2.0 2.0N/mOl 2a=--1000k - 002m =
1,000,000[5.5: 02.~oa 4.5-: ·: 2a -~.o J where-2.0 2.0 - 15aSetting the determinant [k
- 002ml = 0 gives the frequency equation1800a3-1905a2+459.5a-15 = 0The roots of the
frequency equation are al = 0.0386, a2 = 0.3000, and a3 =0.7197. Consequentl y,j
(O0l 1~) = j 30308..06) jOl 1) j 6.21) jfl ) 10.99) =} Ol 2 = 17.62 ~: ~ =} fz = 2.76 Hz
Ol ~ 719.7 003 26.83 f3 4.27Each natural frequency is associated with a mode of
vibration of the system. At thenatural frequencies, the ampl itude of the
displ acement vector, U, is indeterminate [scal ingthe displ acements up or down by a
constant factor wil l stil l satisfy equation (B.92)]. Thevector U does describe the
shape of the vibrating system, which is different at each natural frequency. This
shape is often made dimensionl ess by dividing the el ements of U by one(often the
first, sometimes the l argest) el ement. The resul ting vector describes the mode
shape; the mode shape for the nth mode of vibration woul d be« I>~ = [<PIn <P2n ...
<PNn] = Ul rUIn U2n '" 1] (B.94)NnAl l mode shapes satisfy the rel ationship, Ik -
co~ml « I>n = 0 for: n = [1, N]. The val ues of thevector « I>n at each natural
frequency describe the mode shape of the corresponding mode ofvibration. Thus a
system of N degrees of freedom wil l have N natural frequencies correspondingto N
modes of vibration. Each mode of vibration occurs at a particul ar natural frequency
and causes the structure to deform with a particul ar mode shape. The mode
corresponding to the l owest natural frequency is cal l ed the first mode
orfundamental mode,the second l owest natural frequency is cal l ed the second mode,
and so on. The mode shapescan be shown to be orthogonal , that is, for m *" n
« I>~m« l >n 0« I>~k« l >n = 0Sec. 8.9 Mul tipl e-Degree-of-Freedom Systems 579Exampl eB.6
Compute the mode shapes for the structure shown in Exampl e B.5.Sol ution
Substituting the fundamental frequency into equation (B.99) yiel ds[5.114 -2.5 ° ]
1UIl ) 1°) 1,000,000 -2.5 4.037 -2.0 U21 = °° -2.0 1.421 U31 °Normal izing by the top
fl oor displ acement, U310 the mode shape+, = ~" m: l = !;: )must satisfy [k - mym] <I>
1 = O. Then, using the known val ue of <\>31 =1, <\>11 and <\>21 can bedetermined.
The process can be repeated to yiel d the mode shapes of al l three modes of
vibration: !0.347 )<I>! = 0.7111.000 !- 1.250 )<1>2 = -1.2501.000 !6.479 )<1>3 =
-4.3981.000The mode shapes are shown graphical l y in Figure EB.6.1.000 1.000 1.000
0.7110.347Mode 1'1 = a.99HzMode 2'2 = 2.76HzFigure EB.6Mode 3 '3 = 4.27HzEl .9.3
Mode Superposition MethodFor l inear structures with certain types of damping, the
response in each mode of vibrationcan be determined independentl y of the response
in the other modes. The independentmodal responses can then be combined to
determine the total response. This is the basis ofthe mode superposition method.
Recal l ing that the mode shape vector, <l >m describes onl y the shape of the nth
mode, thedispl acements can be expressed as the product of the mode shape and the
modal ampl itude, Yn: 580 Dynamics of Discrete Systems App. B(B.9S)Then, by
substituting equation (B.9S) into equation (B.87) and premul tipl ying each term by
cP~ , the equation of motion can be written for the nth mode of vibration asMnYn+
CnYn+ KnYn = Qn(t) (B.96)where M; = cP~mcPn' Cn = cP~ccPm K; = cP~kcPm and Qn(t) =
cP~q(t). This equation ofmotion is based on the assumption that the damping matrix
is orthogonal (i.e., thatcP~ ccPn = 0 for m =I- n). Rayl eigh damping, in which the
damping matrix can be broken intoa component proportional to the mass matrix and a
component proportional to the stiffnessmatrix, satisfies the orthogonal ity
requirement. Other procedures are described in standardstructural dynamics texts.
Al ternativel y, the equation of motion can be written as.. 2): . 2 Qn(t)Yn +
': >nWnYn+ WnYn = -MnFor the case of base shaking, the equation of motion can be
expressed as(B.97)(B.98)Nwhere i; = L mj<pjn'j = 1By this process, the system of N
simul taneous equations (the original equationsof motion) is transformed to a system
of N independent equations. Each of these independentequations can be sol ved for
Yn(t) using the SDOF procedures described earl ier inthis appendix. Then the total
displ acement is obtained by superposition of the modal contributions: (B.99)Once the
displ acements are known, they can be used to compute forces, stresses, and other
parameters of interest. The displ acements can al so be used to compute a set of
equival entl ateral forces, f(t), which woul d produce the displ acements u(t) if they
were appl ied as staticl oads: (B.100)Internal forces can be computed by static
anal ysis of the structure subjected to the equival entl ateral forces. These internal
forces can be used for design of the various el ements of thestructure.Exampl eB.7
Compute the response of the structure shown in Exampl e BA to the Gil roy NO.2 E-W
earthquakemotion using the mode superposition method. Assume 5% damping for al l
modes.Sol ution The Gil roy No.2 earthquake motion, il l ustrated in Figure 3.1, was
recorded onthe surface of a thick deposit of stiff soil in the 1989 Loma Prieta
earthquake. Use of themode superposition method requires eval uation of the modal
equations of motion. For the firstmode,Sec. 8.9 Mul tipl e-Degree-of-Freedom Systems
581M] = cl >fmcl >! = (1000 kg) {0.347 0.711 l .000} [I~o I~ ~l J~: ~~~) = 22,270 kgo
l ;jl 1.000L 1 = m,<pl I + m2<P21 + m3<P31= (10,000)(0.347) + (12,000)(0.71 I) +
(15,000)(
1.000)= 27,002 kgso the equation of motion [equation (B.98)] isY, + 0.62IY, +
38.56Yl = -1.212iib(t)Repeating this process for the second and third modes gives
Y2+ 1.732Y2 + 300.0Y2 = 0.253iib(t)Y3+2.683Y3+719.8Y3 = -0.04Iiib(t)8.9.4 Response
Spectrum Anal ysisThe mode superposition method produces the entire time history of
the response of the structure.For design purposes, however, the entire time history
may riot be needed; the maximumresponse val ues may be sufficient. Because each mode
of vibration can be treated as an independentSDOF system, maximum val ues of modal
response can be obtained from the responsespectrum. The modal maxima can then be
combined to estimate the maximum total response.8.9.4.1 Cal cul ation of Modal
Response MaximaLet Sdm Svm and San denote the spectral displ acement, vel ocity, and
accel eration associatedwith the nth mode of vibration, respectivel y (these val ues
woud be obtained from theresponse spectrum at a period, T; =2rrJwn) . Then the
maximum modal displ acement is given by(B. 101)Using equation (B.95), the maximum
displ acement of the jth fl oor woul d be2 Ln . LnTn(Ujn)max = M Sdn<!>jn = 4 2M San<!
>jnn 1t nThe maximum val ue of the equival ent l ateral force at the jth fl oor isLn
(fjn)max = M mAjnSann(B. 102)(B.103)Maximum val ues of the internal forces can be
computed by static anal ysis of the structuresubjected to the maximum equival ent
l ateral forces.8.9.4.2 Combination of Modal Response MaximaSection B.9.4.1 showed
how the response spectrum can be used to predict maximumval ues of various modal
response parameters. The mode superposition method showed that582 Dynamics of
Discrete Systems App. B(B.104)time histories of modal response can be combined by
simpl e superposition to obtain thetotal time history of response. However,
combination of modal response maxima to obtainthe maximum total response is not as
straightforward.The exact val ue of the maximum total response cannot be obtained
directl y from themodal maxima because the modal maxima occur at different times.
Direct superposition ofthe modal maxima, which impl ies that the maxima do occur
simul taneousl y, produces anupper bound to the maximum total response; for any
response parameter r(t),Nrmax: S; L (rn)maxn = 1This upper bound val ue is usual l y
too conservative and is rarel y used for design. Instead,modal combination
procedures based on random vibration theory are used. The simpl est ofthese is the
root-sum-square val ueNrmax = L (rn)~axn = 1(B. l OS)The root -sum-square procedure
provides a good estimate of maximum total response whenthe natural periods are wel l
separated (by a factor of about 1.5 or more for 5% damping).Procedures that account
for correl ation between modes are avail abl e (Newmark and Rosenbl ueth,1971; Chopra,
1995) for cases of cl osel y spaced modes.8.9.5 DiscussionThe mode superposition
method and response spectrum anal ysis procedures both rel y onrepresentation of a
MDOF system by a set of SDOF systems. The characteristics of the setof SDOF systems
are such that those corresponding to the l ower natural frequencies contributemore
to the total response than those corresponding to the higher natural frequencies.
For practical purposes, the response of a MDOF system can be computed with
reasonabl eaccuracy by considering onl y the l ower modes that contribute
significantl y to the total response of the structure. For some structures, onl y a
smal l number of modes may need tobe considered. Al l of the anal yses described in
this section appl y to l inear structures. Proceduresfor anal ysis of nonl inear MDOF
structures are avail abl e but are wel l beyond thescope of this appendix.CProbabil ity
ConceptsC.1 INTJrODUCTIONGeotechnical earthquake engineering probl ems are fraught
with uncertainty. At a particul arsite, earthquake-induced l oading depends on the
size and l ocation of the earthquake-noneof which can be predicted with certainty.
Because of the inherent variabil ity of soil s and theinevitabl e l imits on
expl oration of subsurface conditions, the resistance of the soil to thatl oading is
not known with certainty. When both l oading and resistance are uncertain, the
resul ting effects are uncertain as wel l . A number of geotechnical earthquake
engineeringanal yses attempt to quantify the uncertainty in the various input
parameters for a particul arprobl em, and compute the resul ting uncertainty in the
output.In this appendix we provide a brief introduction to some basic concepts of
probabil ityand describe several probabil ity distributions that are used in the body
of the book. Moredetail ed information on these topics can be found in texts such as
Benjamin and Cornel l (1970), Ang and Tang (1975a,b, 1984), and Harr (1987).C.2
SAMIPLE SPACES AND EVENTSProbabil ity theory deal s with the resul ts, or outcomes, of
processes that are usual l ydescribed in a general sense as experiments. The set of
al l possibl e outcomes of an experimentis cal l ed the sampl e space, and each outcome
of an experiment is cal l ed a sampl e583584 Probabil ity Concepts App.Cpoint. The
sampl e space therefore consists of al l possibl e sampl e points. The sampl e spacemay
be continuous, in which case the number of sampl e points is infinite, or it may be
discrete,as when the number of sampl e points are finite and countabl e.An event is a
subset of a sampl e space, and therefore represents a set of sampl e points.A singl e
event consists of a singl e sampl e point, and a compound event consists of more than
one sampl e point. If Q represents a sampl e space and A represents an event, the
compl ementaryevent, A, is the set of al l sampl e points in Q that are not inA. The
interrel ationshipsamong sets can be convenientl y il l ustrated by means of a Venn
diagram (Figure C,l ). InFigure C, I the sampl e space is represented by the
rectangl e Q and the event A by the circl e.Thus A is a subset of Q. The
compl ementary event A corresponds to the part of the rectangl ethat l ies outside the
circl e. Because no sampl e points are in both A and A, the intersectionof A and A is
the nul l set, <I> (i.e., A n A =<1». Simil arl y, al l sampl e points are in eitherA or
A, so the union of A and A is Q (i.e., A u A = Q). Two events, A and B, are said to
be mutual l y excl usive if they share no common sampl e points (i.e., A n B =<1».o Q
Figure C.I Venn diagram il l ustratingevent A in sampl e space Q.Exampl e C.IConsider
the Venn diagram for the three events, A, B, and C, shown in Figure EC.I.Q8Figure
EC.IA n B = regions 1 and 4B u C = regions 1,2,3,4,5, and 6A n B n C = region 1C.3
AXIOMS OF PROBABILITYA n jj = regions 3 and 7(A u B) n C = regions 1,2, and 311 n
jj n C = region 8A probabil ity measure, P, can be assigned to each sampl e point or
set of sampl e points in asampl e space. The probabil ity of an event A is denoted by
the symbol P[A]. The entire theoryof probabil ity is based on the fol l owing three
fundamental axioms.Sec. C.4 Probabil ities of Events 585Axiom 1. The probabil ity of
an event is represented by a number greater than or equal to zero but l ess than or
equal to I: 0: : ; P [A] : : ; IAxiom 2. The probabil ity of an event equal to the entire
sampl e space Q is I: P [Q] = I(C.l a)(C.l b)Axiom 3. The probabil ity of an event
representing the union of two mutual l y excl usiveevents is equal to the sum of the
probabil ities of the events: P [A u B] = P [A] + P [B] (C.l c)These axioms can be
used to devel op the rul es and theorems that comprise the mathematical theory of
probabil ity.C.4 PROIIABILITIES OF EVENTSProbabil ities are often thought of in terms
of rel ative frequencies of occurrence. If the existenceof a water content greater
than the optimum water content in a compacted fil l is consideredto be an event, the
probabil ity of that event can be estimated by determining therel ative frequency of
water content measurements that exceed the water content. If the total number of
water content measurements is smal l , the rel ative frequency may onl y approximatethe
actual probabil ity, but as the number of measurements becomes l arge, the rel ative
frequency wil l approach the actual probabil ity. This frequentist point of view is
not veryhel l pful ,however, for situations in which an experiment cannot be repeated.
In such cases,probabil ities can be viewed as rel ative l ikel ihoods (or degrees of
bel ief), as in the probabil itythat a newl y discovered faul t is capabl e of producng
maximum earthquake magnitudesof7.0 or 7.5. The l atter interpretation l ends itsel f
to the subjective eval uation of probabil ity.Regardl ess of how probabil ities are
interpreted, the axioms of probabil ity al l ow statementsto be made about the
probabil ities of occurrence of singl e or mul tipl e events. Thesecan be visual ized
with the hel p of Venn diagrams drawn such that the area of the rectangl e
representing the sampl e space Q is I and the areas of al l events within the sampl e
space areequal to their probabil ities. Consider the nonexcl usive events A and B in
Figure C.2. Theevent AnB (which means that both A and B occur) is represented by
the shaded region inFigure C.2a; P[A n B] is given by the area of the shaded
region. The event A u B (whichmeans that either A orB occurs) is represented by the
shaded region in Figure C.2b; P[A uB]is given by the area of that shaded region, or
P [A u B] = P [A] + P [B] - P [A n B] (C.2)In many instances, the probabil ity of
one event depends on the occurrence of anotherevent. The conditional probabil ity of
event A given the occurrence of event B is denotedP [A IB] and is defined (for P[B
> 0]) byP [Al B] = P [A n B]P [B](C.3)586 Probabil ity Concepts App. C(a)Q(b)QFigure
C.2 Venn diagrams for events A and B in sampl e space Q. (a) The setA n B isgiven by
the shaded area; if the area ofQ is I, P[A n B] is equal to the area that isshaded.
(b) The setA u B is given by the shaded area; P[A u B] is equal to the shadedarea.
The
conditional probabil ity is easil y visual ized with the Venn diagram (Figure C2a) as
theratio of the area of A n B to the area of B. Event A is statistical l y
independent of event B ifthe occurrence of B does not affect the probabil ity of
occurrence of A; that is,P[AIB] = P[A] (CA)Rearranging equation (C2), the
probabil ity that both A and B occur, is given byP[AnB] = P[AIB]P[B] (CS)which if A
and B are statistical l y independent becomesP [A n B] = P [A] P [B] (C6)This is
known as the mul tipl ication rul e and can be extended to the mul tipl e, mutual l y
independentevents A,B, C, ..., N byP [A n B n C n ... n N] = P [A] P [B] P [C] ...P
[N] (C7)The mul tipl ication rul e states that the probabil ity of joint occurrence of
statistical l y independentevents is equal to the product of their individual
probabil ities.Exampl e C.2Consider the rol l ing of a singl e fair die as an
experiment. Then the resul ting sampl e space,Q={ I, 2, 3, 4, 5, 6), is the set of
al l possibl e outcomes of the experiment. Let the fol l owing threeevents be defined
asA = {I} (a singl e rol l produces a 1)B = {I, 3, 5} (a singl e rol l produces an odd
number)C = {4, 5, 6} (a singl e rol l produces a number greater than 3)Define the
sets A n B, A u B, and B u C, and compute their probabil ities.Sol ution The set A n
B incl udes al l outcomes that are in both A and B (i.e., A n B = ( 1}). Theset A u B
incl udes al l outcomes that are in either A or B (i.e., A u B = ( 1, 3, 5}). The set
B u Cincl udes al l outcomes that are in either B or C (i.e., B u C = ( 1,3,4,5,6}).
The probabil ities ofeach set can be computed asP[AnB] = P[A[B]P[B] = mm = ~P [A u
B] = P [A] + P [B] - P [A n B] = ~ + ~ - ~ = ~P[BuC] = P[B] +P[C] -P[BnC] = ~+~-~ =
~Sec: . C.4 Probabil ities of Events 587Exampl eC.3One hundred fiel d compaction tests
were performed in the earl y stages of construction of anearth dam. The resul ts of
the tests are presented in terms of the numbers that satisfied specificationsfor
minimum rel ative compaction and for compaction water content in the tabl e bel ow.
Rel ative CompactionWater ContentAcceptabl eNot acceptabl eAcceptabl e806Not Acceptabl e
104Assume that the contractor's performance in the future wil l be the same as in
the first 100 testsand that the fil l material does not change. Estimate the
probabil ity that the rel ative compactionspecification wil l be satisfied in the next
test if the water content specification is satisfied. Estimatethat probabil ity for
the case in which the water content specification is not satisfied.Sol ution Define
two events, Rand W, such thatR = rel ative compaction specification satisfiedW =
water content specification satisfiedFrom the tabl e the probabil ity that both the
rel ative compaction and water content specificationsare satisfied can be estimated
as P [W n R] = 80/100. Then the probabil ity that the rel ativecompaction
specification wil l be satisfied in the next test if the water content specification
is satisfied is the conditional probabil ity P[R I W], which can be computed as
P[RIW] = P[WnR] = 80/100 80 0889P[W] 80/100+101100 = 90 = .The probabil ity that the
rel ative compaction specification is~atisfied given that the water content
specification is not satisfied can be estimated as P [R IW], orP [W n R] 6/100 6
P[RIW] = P[W] = 6/100+4/100 = 10 = 0.600For a set of events, Bl > B2, • • • , BN, which
are mutual l y excl usive (B, n Bj = </» for al l i: : l -: j but col l ectivel y exhaustive
(B] u B2 u ... U BN =il ), l ike that shown in the Venndiagram of Figure C.3, the
probabil ity of another event A can be expressed asP [A] = P [A n Btl + P [A n B2 ]
+ ... + P [A n BN] (e.8)Using equation (e.S) for each term on the right side of
equation (C.8) yiel dsP [A] = P [AIBtl P [Btl + P [AIB2 ] P [B2 ] + ... + P [AIBN] P
[BN ]N= LP [Al B;] P [Bi ]i= ](e.9)which is known as the total probabil ity theorem.
The total probabil ity theorem forms thebackbone of the probabil ity cal cul ations
required for probabil istic seismic hazard anal yses(Chapter 4).588 Probabil ity
Concepts App. CB1Figure C.3 Intersection of eventA withmutual l y excl usive but
col l ectivel yexhaustive events B;Exampl e C.4A structural engineer has determined
that a structure wil l col l apse in an earthquake that producesa peak accel eration of
0.3g. The probabil ity that a given earthquake on faul t A, B, or Cwoul d be strong
enough to cause the structure to col l apse are 0.5, 0.2, and 0.1, respectivel y. The
probabil ities that that such earthquakes wil l occur on faul ts A, B, and C during
the l ife of thebuil ding are 0.01,0.05, and 0.08, respectivel y. What is the
probabil ity that the structure wil l col l apsein an earthquake?Sol ution Define the
fol l owing events asA = the structure col l apses in an earthquakeD, = an earthquake
capabl e of col l apsing the structure occurs on faul t ADz = an earthquake capabl e of
col l apsing the structure occurs on faul t BD3 = an earthquake capabl e of col l apsing
the structure occurs on faul t CThen the probabil ity that the structure col l apses in
an earthquake is given byP [A] = P [AIDd P [Dd + P [AIDz]P [Dz] + P [AID3 ] P [D3]
(0.5) (0.01) + (0.2) (0.05) + (0.1) (0.10)= 0.025c.s RANDOM VARIABLESAl l fiel ds of
science and engineering attempt to describe various quantities or phenomenawith
numerical val ues. In most cases, the precise numerical val ue cannot be predicted in
advance of some process, or experiment, of interest. In such cases, a particul ar
quantity orphenomenon is described by a random variabl e. The random variabl e is
used to describe anevent in a sampl e space in quantitative terms.A continuous
random variabl e can take on any val ue within one or more interval s.Because a
continuous random variabl e can take on any of an infinite number of val ues, the
probabil ity of it taking on any specific val ue is 1/00 =O. The probabil ity
distribution of acontinuous random variabl e can al so be described by its
probabil ity density function orPDF,fx<x), which must satisfy the conditionsfx(x) ~
0 for al l xr~fx(X)dXb f fx(x) dxaSec. C.b Expected Val ues and Standard Deviations
589According to these conditions, the area under the PDF between two val ues a and b
representsthe probabil ity that the random variabl e wil l have a val ue in the
interval bounded bya and b. The probabil ity distribution of a random variabl e can
al so be described by its cumul ativedistribution function (CDF), which is given by
Fx(x) = P[X~x] = f=fx(X)dX (C.l O)Therefore, the probabil ity that a random variabl e,
X, fal l s between tWQ val ues a and b isP[a~X~b] = Fx(b) -Fx(a) (C. II)(C.l 2)
Obviousl y, the PDF and CDF are cl osel y rel ated-one can be obtained from the other
byintegration or differentiation. The PDF and CDF of a typical probabil ity
distribution areshown in Figure CA.From the total probabil ity theorem and the
definition of the PDF, the probabil ity ofthe: random variabl e Yhaving some val ue y
given that the random variabl e Xis between twovaIues, a and b, can be expressed as
P[Y=y] = P[Y=Yl a~X~x2]P[Xl ~X~X2]b =f P[Y =yl a~X~x2]fx(X)dxaa x1.0 - - - - - - - - -
- - -~----1P[X<a] 1--....,-----,(a xarea = P[X<a](a) (b)Figure C.4 (a) POP for a
random variabl e, X. The probabil ity that X < a is given by thearea under the POP to
the l eft of a. (b) COP for the same random variabl e. Theprobabil ity that X < a is
given by the val ue of the COP at X=a.C.6 EXPE: : CTED VALUES AND STANDARD DEVIATIONS
The uncertainty of a random variabl e can often be characterized with reasonabl e
accuracyby a few statistical parameters. The mean, or expected val ue, of a
continuous random variabl e,X, is given byf= xfx(x) dx (C.l 3)The mean is a very
useful measure of the central tendency of the random variabl e. By itsel f,however,
it does not adequatel y describe the shape of the PDF. The dispersion of the random
590 Probabil ity Concepts App. Cvariabl e about the mean is al so very important. This
dispersion is usual l y characterized bythe varianceor the standard deviationa = 9 x
Ajox(C.14)(C.IS)Both of these parameters refl ect how widel y the random variabl e is
dispersed about themean. Because its units are the same as those of the random
variabl e, the standard deviationis more commonl y used than the variance. This
characteristic al so al l ows the dispersion toexpressed in dimensionl ess form by the
coefficient of variationCOy = axx x (C.l 6)The mean and standard deviation (or mean
and coefficient of variation) go far towarddescribing the uncertainty in a random
variabl e. Many simpl e probabil ity distributions,incl uding those most commonl y used
in geotechnical earthquake engineering, are compl etel ydescribed by these two
parameters. Other distributions may require additional parameters to characterize
their symmetry, l imits, and/or other characteristics.C.7 COMMON PROBABILITY
DISTRIBUTIONSThe resul ts of statistical experiments often exhibit the same general
type of behavior. As aresul t, the random variabl es associated with those
experiments can be described by essential l ythe same PDF. Many probabil ity density
functions exist, but onl y a few are requiredfor the geotechnical earthquake
engineering anal yses described in this book.C. 7.1 Uniform DistributionThe simpl est
probabil ity distribution is one in which al l possibl e val ues of the random variabl e
are equal l y l ikel y. Such a random variabl e is described by a uniform distribution.
ThePDF for a continuous random variabl e, X, that is uniforml y distributed between
two val uesa and b isfx(x) =I;~a for x $ afor a < x $ bfor x> b(C.l 7)The PDF and CDF
for a uniform distribution are il l ustrated in Figure C.S.Sec. C.7 Common
Probabil ity Distributions 5911/(b- 61) - - - ..,-------,a(a)b1.0 ------------~--~a
b(b)(C.IS)Figure C.S Uniform distribution: (a) probabil ity density function; (b)
cumul ativedistribution function.C"7.2 Normal DistributionThe most commonl y used
probabil ity distribution in statistics
is the normal distribution (orGaussian distributions. Its PDF, which pl ots as the
famil iar bel l -shaped curve of Figurec..6a, describes sets of data produced by a
wide variety of physical processes. The normal distribution is compl etel y defined by
two parameters: the mean and standard deviation.Mathematical l y, the PDF of a
normal l y distributed random variabl e X with mean x andstandard deviation o, is
given byfx(x) = ~()x exp [-~(x;xXJ]The PDF and CDF for a normal distribution are
il l ustrated in Figure C.6. Exampl es of normal pdf's for random variabl es with
different means and standard deviations are shown inFigure C.7.Integration of the
PDF of the normal distribution does not produce a simpl e expressionfor the CDF, so
val ues of the normal CDF are usual l y expressed in tabul ar form. The1.0 - - - - - -
- - - - - - - ""-~-...,0.5(a) (b)Figure C.6 Normal distribution: (a) probabil ity
density function; (b) cumul ativedistribution function.592 Probabil ity Concepts
App.C(a) (b)Figure C.? Normal distributions for (a) two random variabl es, Xl and
Xl > withdifferent means but the same standard deviation, and (b) two random
variabl es, X3 andX4, with the same mean but different standard deviations.normal
CDF is most efficientl y expressed in terms of the standard normal variabl e, Z,which
can be computed for any random variabl e, X, using the transformationX-xZ= -ax(C.19)
Whenever X has a val ue, x, the corresponding val ue of Z is z = (x - X )l ax- Thus,
the meanval ue of Z is z =0 and the standard deviation is az=1. Tabul ated val ues of
the standard normal CDF are presented in Tabl e C-l .Exampl e C.SGiven a normal l y
distributed random variabl e, X, with x =270 and ax=40, compute the probabil itythat
(a) X < 300, (b) X > 350, and (c) 200 < X < 240.Sol ution (a) For X = 300,Z = X-x =
300-270 = 0.75ax 40ThenP [X < 300] = P [Z < 0.75] = F, (0.75)(b) For X = 350,1 - F,
(-0.75) 1 - 0.2266 = 0.7734Z = X - x = 350 - 270 = 2.0ax 40ThenP[X>350] = P[Z>2.0]
(c) For X = 200,1 - F; (2.0) = F, (-2.0) = 0.0228Z = X-x = 200-270 = -175ax 40 .For
X= 240,Z = X -x = 240-270 = -0.75ax 40Sec. C.7 Common Probabil ity Distributions 593
ThenP [200<X < 240] = P [-1.75 < Z < -0.75] = F, (-0.75) - F, (-1.75)= 0.2266 -
0.0401 = 0.1865TABLE C-1 Val ues of the CDF ofthe standard normal distribution,
Fz(z) = 1 - Fz(-z)z C).OO 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09-3.4 0.0003
0.0003 0.0003 0.0003 0.0003 0.0003 0.0003 0.0003 0.0003 0.0002-3.3 0.0005 0.0005
0.0005 0.0004 0.0004 0.0004 0.0004 0.0004 0.0004 0.0003-3.2 0.0007 0.0007 0.0006
0.0006 0.0006 0.0006 0.0005 0.0005 0.0005 0.0005-3.1 0.0010 0.0009 0.0009 0.0009
0.0008 0.0008 0.0008 0.0008 0.0007 0.0007-3.0 0.0013 0.0013 0.0013 0.0012 0.0012
0.0011 0.0011 0.0011 0.0010 0.0010-2.9 0.0019 0.0018 0.0017 0.0017 0.0016 0.0016
0.0015 0.0015 0.0014 0.0014-2.8 0.0026 0.0025 0.0024 0.0023 0.0023 0.0022 0.0021
0.0021 0.0020 0.0019-2.7 0.0035 0.0034 0.0033 0.0032 0.0031 0.0030 0.0029 0.0028
0.0027 0.0026-2.6 0.0047 0.0045 0.0044 0.0043 0.0041 0.0040 0.0039 0.0038 0.0037
0.0036-2.5 0.0062 0.0060 0.0059 0.0057 0.0055 0.0054 0.0052 0.0051 0.0049 0.0048-
2.4 0..0082 0.0080 0.0078 0.0075 0.0073 0.0071 0.0069 0.0068 0.0066 0.0064-2.3
00107 0.0104 0.0102 0.0099 0.0096 0.0094 0.0091 0.0089 0.0087 0.0084-2.2 0..0139
0.0136 0.0132 0.0129 0.0125 0.0122 0.0119 0.0116 0.0113 0.0110-2.1 00179 0.0174
0.0170 0.0166 0.0162 0.0158 0.0154 0.0150 0.0146 0.0143-2.0 0.0228 0.0222 0.0217
0.0212 0.0207 0.0202 0.0197 0.0192 0.0188 0.0183-1.9 0.0287 0.0281 0.0274 0.0268
0.0262 0.0256 0.0250 0.0244 0.0239 0.0233-1.8 0.0359 0.0352 0.0344 0.0336 0.0329
0.0322 0.0314 0.0304 0.0301 0.0294-1.7 0.0446 0.0436 0.0427 0.0418 0.0409 0.0401
0.0392 0.0384 0.0375 0.0367-1.6 0.0548 0.0537 0.0526 0.0516 0.0505 0.0495 0.0485
0.0475 0.0465 0.0455-1.5 0.0668 0.0655 0.0643 0.0630 0.0618 0.0606 0.0594 0.0582
0.0571 0.0559-1.4 0.0808 0.0793 0.0778 0.0764 0.0749 0.0735 0.0722 0.0708 0.0694
0.0681-1.3 0.0968 0.0951 0.0934 0.0918 0.0901 0.0885 0.0859 0.0853 0.0838 0.0823-
1.2 0.1151 0.1131 0.1112 0.1093 0.1075 0.1056 0.1038 0.1020 0.1003 0.0985-1.1
0.1357 0.1335 0.1314 0.1292 0.1271 0.1251 0.1230 0.1210 0.1190 0.1170-1.0 0.1587
0.1562 0.1539 0.1515 0.1492 0.1469 0.1446 0.1423 0.1401 0.1379-0.9 0.1841 0.1814
0.1788 0.1762 0.1736 0.1711 0.1685 0.1660 0.1635 0.1611-0.8 0.2119 0.2090 0.2061
0.2033 0.2005 0.1977 0.1949 0.1922 0.1894 0.1867-0.7 0.2420 0.2389 0.2358 0.2327
0.2296 0.2266 0.2236 0.2206 0.2177 0.2148-0.6 0.2743 0.2709 0.2676 0.2643 0.2611
0.2578 0.2546 0.2514 0.2483 0.2451-0.5 0.3085 0.3050 0.3015 0.2981 0.2946 0.2912
0.2877 0.2843 0.2810 0.2776-0.4 0.3446 0.3409 0.3372 0.3336 0.3300 0.3264 0.3228
0.3192 0.3156 0.3121-0.3 0.3821 0.3783 0.3745 0.3707 0.3669 0.3632 0.3594 0.3557
0.3520 0.3483-0.2 0.4207 0.4168 0.4129 0.4090 0.4052 0.4013 0.3974 0.3936 0.3897
0.3859-0.1 0.4602 0.4562 0.4522 0.4483 0.4443 0.4404 0.4365 0.4325 0.4286 0.4247-
0.0 0.5000 0.4960 0.4920 0.4880 0.4840 0.4801 0.4761 0.4721 0.4681 0.4641594
Probabil ity Concepts App. C(C.20)C.7.3 Lognormal DistributionSome probl ems,
particul arl y those invol ving ground motion parameters (Chapter 3), areformul ated in
terms of the l ogarithm of a parameter rather than the parameter itsel f. IfX isa
random variabl e, then Y =InX is al so a random variabl e. If Y is normal l y
distributed, thenX is l ognormal l y distributed. In other words, a random variabl e is
l ognormal l y distributedif its l ogarithm is normal l y distributed. The PDF of a
l ognormal l y distributed random variabl eX is given byfx(x) = I exp [_!( l nx -l iiX)2J
x J2ic(j l nx 2 o l nxThe shape of the l ognormal distribution is shown in Figure e.S.
Note that the PDF is notsymmetric, and that it assigns zero probabil ity to negative
val ues of the random variabl e.These characteristics can be very useful for some
random variabl es [the normal distribution,for exampl e, assigns nonzero
probabil ities for val ues ranging from -00 to +00; when appl iedto a random variabl e
such as soil density, it can assign some (hopeful l y smal l ) probabil itythat the soil
wil l have a negative density].Val ues of the CDF of the l ognormal distribution are
usual l y obtained from Tabl e C-I,using the modified transformationz = InX -l iiX(jInx
(C.21)fx(x)InX(a)X(b)Figure C.S Two views of the l ognormal distribution. (a)
Because the l ogarithm of al ognormal l y distributed random variabl e, X, is normal l y
distributed, the probabil itydensity function of InX is a bel l -shaped curve. (b) The
probabil ity density function of Xitsel f has no negative val ues and is not
symmetric.Exampl e C.6A random variabl e, X, is l ognormal l y distributed with Inx = 5
and O"l nx= 1.2. Compute (a) theprobabil ity that X < 100, and (b) the val ue of X
that has a 10% probabil ity of being exceeded.Sol ution (a) For X =100,Z = InX -
Il l .X In 100 - 5 = -0.33O"l nx 1.2Sec: . C.7 Common Probabil ity Distributions 595From
Tabl e C-I,P[X<100] =P[Z<-0.33] = Fz(-0.33) =0.3707(b) From Tabl e C-I, the val ue of
Z that woul d have a 10% probabil ity of exceedance is 1.282[i.e., Fz(1.282) = 0.90].
Then, rearranging equation (C.21) yiel dsInX = ZO"Inx + l nx = (1.282) (1.2) + 5 =
6.54sox = e6.54 = 691