Near-Fault Seismic Site Response

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/266496826
Near-Fault Seismic Site Response
Article · January 2000
CITATIONS READS
70 718
1 author:
Adrian Rodriguez-Marek
Virginia Polytechnic Institute and State University
102 PUBLICATIONS 1,625 CITATIONS
SEE PROFILE
Some of the authors of this publication are also working on these related projects:
Resilient and Sustainable Buildings: Performance-Based Engineering View project
Bayesian Methods for Performance Based Engineering View project
All content following this page was uploaded by Adrian Rodriguez-Marek on 12 March 2015.
The user has requested enhancement of the downloaded file.

Near-Fault Seismic Site Response
by
Adrian Rodriguez-Marek
B.S. (Washington State University) 1994

M.S. (Washington State University) 1996
A dissertation submitted in partial satisfaction of the
requirements for the degree of
Doctor of Philosophy
in
Engineering-Civil Engineering
in the
GRADUATE DIVISION
of the
UNIVERSITY OF CALIFORNIA, BERKELEY
Committee in charge:
Professor Jonathan D. Bray, Chair

Professor Juan M. Pestana
Professor Alexandre J. Chorin
Fall 2000
The dissertation of Adrian Rodriguez-Marek is approved:
________________________________________________
Chair Date
________________________________________________
Date
________________________________________________
Date
University of California, Berkeley
Fall 2000
Abstract
NEAR-FAULT SEISMIC SITE RESPONSE
by
Adrian Rodriguez Marek
Doctor of Philosophy in Engineering-Civil Engineering
University of California, Berkeley
Professor Jonathan D. Bray, Chair
The understanding of ground motions in the near-fault region is important for
seismic risk assessment in a number of populated areas that overlie fault zones.
Understanding the effect of local site conditions is important for a complete
characterization of near-fault ground motions. A geotechnically-based site classification
scheme that includes soil depth is proposed for the evaluation of site effects. This
classification scheme is used to evaluate site response for the 1989 Loma Prieta and 1994
Northridge earthquakes. Regression analyses resulted in estimates of ground motion with
lower uncertainties than those made with the simpler classification schemes typically
used in attenuation relationships. These results emphasize the need to better define the
baseline site category used in attenuation relationships. Site amplification factors for
each site category are proposed. These factors, however, are based on data recorded at
distances generally greater than 10 km from the fault and are not directly applicable to
pulse-like near-fault forward-directivity ground motions. Thus, a separate database of
near-fault forward-directivity ground motions is studied. Each record in the database is
characterized by a pulse period and its peak ground velocity (PGV). Regression analyses
1
on these parameters indicate that site response alters the characteristics of forward-
directivity pulses.
The large scatter of the near-fault ground motion data precludes definite
conclusions regarding site response based exclusively on empirical analyses. Numerical
site response studies are performed to develop a better understanding of near-fault site
response. A nonlinear site response methodology using a bounding surface plasticity
model developed by Borja and Amies (1994) is implemented in the finite element
program GeoFEAP. Site response analyses using pulse-like simplified velocity time-
histories as input motions are performed on generalized site profiles representing average
site conditions for different site categories. Site effects influence the characteristics of
input velocity-pulses by amplifying PGV and elongating the period of the input pulses.
The extent of PGV amplification and pulse period elongation is a function of site
condition, profile depth, and input pulse characteristics. The numerical results presented
agree with the observations made in the empirical analysis of near-fault ground motions.
____________________________
Jonathan D. Bray, Thesis Advisor
____________________________
Date
2
For Tina
i
TABLE OF CONTENTS
ABSTRACT ................................................................................................................... 1
TABLE OF CONTENTS ............................................................................................... ii
LIST OF TABLES ........................................................................................................ vi
LIST OF FIGURES ....................................................................................................... ix
ACKNOWLEDGMENTS........................................................................................... xix
CHAPTER 1 - INTRODUCTION .............................................................................. 1

1.1 PROBLEM ..........................................................................................................1
1.2 SIGNIFICANCE .................................................................................................3
1.3 OBJECTIVE........................................................................................................4
1.4 OUTLINE............................................................................................................5
CHAPTER 2 - LITERATURE REVIEW.................................................................. 7

2.1 INTRODUCTION...............................................................................................7
2.2 SITE RESPONSE ...............................................................................................8
2.2.1 Site Response Studies and Code Development............................................8
2.2.2 Cyclic Soil Models and Site Response Methodologies..............................12
2.3 SITE EFFECTS IN ATTENUATION RELATIONSHIPS...............................18
2.4 SUMMARY ......................................................................................................24
CHAPTER 3 - CHARACTERIZATION OF SITE RESPONSE GENERAL SITE

CATEGORIES ........................................................................................................... 30
3.1 INTRODUCTION.............................................................................................30
3.2 METHODOLOGY............................................................................................32
3.3 SITE CLASSIFICATION .................................................................................34
3.3.1 Classification Scheme ...............................................................................34
3.3.2 Site Classification......................................................................................37
3.4 GROUND MOTION DATA.............................................................................38
3.5 STATISTICAL ANALYSIS .............................................................................39
3.5.1 General......................................................................................................39
ii
3.5.2 Northridge Earthquake .............................................................................41
3.5.3 Loma Prieta Earthquake ...........................................................................44
3.5.4 Results .......................................................................................................45
3.6 EVALUATION OF RESULTS.........................................................................46
3.6.1 General......................................................................................................46
3.6.2 Comparison With a "Soil vs. Rock" Classification System .......................46
3.6.3 Comparison With a Code-Based Site Classification System.....................48
3.6.4 Subdivision of Site C .................................................................................50
3.6.5 Subdivision of Site D .................................................................................51
3.6.6 Effect of Depth to Basement Rock .............................................................51
3.6.7 Amplification Factors................................................................................52
3.6.8 Recommended Factors ..............................................................................54
3.7 SUMMARY ......................................................................................................56
3.7.1 Findings.....................................................................................................56
3.7.2 Recommendations......................................................................................58
CHAPTER 4 - EMPIRICAL CHARACTERIZATION OF NEAR-FAULT

GROUND MOTIONS.............................................................................................. 103
4.1 INTRODUCTION...........................................................................................103
4.2 FREQUENCY DOMAIN REPRESENTATION OF NEAR-FAULT
GROUND MOTIONS.....................................................................................106
4.3 TIME DOMAIN REPRESENTATION OF NEAR-FAULT GROUND
MOTIONS.......................................................................................................108
4.3.1 General....................................................................................................108
4.3.2 Parameterization of velocity pulses ........................................................110
4.3.3 Statistical Evaluation of Equivalent Pulse parameters in the Fault
Normal Direction ................................................................................................117
4.3.4 Characteristics of the fault parallel component of motion .....................128
4.3.5 Development of bi-directional simplified pulses for use as baseline
input motions .......................................................................................................130
4.4 SUMMARY AND FINDINGS .......................................................................134
CHAPTER 5 - SITE RESPONSE ANALYSIS METHODOLOGY ................... 199

5.1 INTRODUCTION...........................................................................................199
iii
5.2 CONSTITUTIVE MODEL .............................................................................203
5.2.1 Selection of a constitutive model.............................................................203
5.2.2 Mathematical development .....................................................................206
5.3 FINITE ELEMENT IMPLEMENTATION ....................................................216
5.3.1 General....................................................................................................216
5.3.2 Local integration of the constitutive equations .......................................219
5.3.3 Radiation Boundary Conditions..............................................................224
5.3.4 Model Parameters ...................................................................................228
5.3.5 General comments...................................................................................230
5.4 VALIDATION ................................................................................................231
5.4.1 General....................................................................................................232
5.4.2 Lotung Array ...........................................................................................233
5.4.3 Chiba Downhole Array ...........................................................................239
5.4.4 Analysis of Shaking Table Tests ..............................................................241
5.4.5 General Comments..................................................................................246
CHAPTER 6 ....... - SITE RESPONSE ANALYSIS OF NEAR-FAULT GROUND

MOTIONS ................................................................................................................ 284
6.1 INTRODUCTION...........................................................................................284
6.2 GENERALIZED SITE PROFILES.................................................................285
6.2.1 General....................................................................................................285
6.3 COMPARISON OF RESULTS FOR RECORDED AND SIMPLIFIED
NEAR-FAULT MOTIONS.............................................................................292
6.3.1 General....................................................................................................292
6.4 SITE RESPONSE TO SIMPLIFIED PULSES ...............................................297
6.4.1 General....................................................................................................297
6.4.2 Stiff Soil Profiles (Sites C and D)............................................................300
6.4.3 Soft clay profile .......................................................................................307
6.4.4 General Observations .............................................................................310
6.5 FINDINGS ......................................................................................................314
CHAPTER 7 - CONCLUSION............................................................................... 370

7.1 SUMMARY ....................................................................................................370
iv
7.2 FINDINGS ......................................................................................................372
7.2.1 General....................................................................................................372
7.2.2 Characterization of site response............................................................372
7.2.3 Empirical study of near-fault, forward-directivity ground motions........373
7.2.4 Site response analyses to near-fault ground motions..............................375
7.3 RECOMENDATIONS FOR FURTHER RESEARCH ..................................378
REFERENCES ......................................................................................................... 381
APPENDIX A - LIST OF GROUND MOTION SITES WITH CORRESPONDING

SITE CLASSIFICATION .......................................................................................... 398
APPENDIX B - SITE VISITS TO SELECTED GROUND MOTION SITES......... 419
APPENDIX C - EQUATIONS TO OBTAIN COMBINED SPECTRAL

ACCELERATION RATIOS FOR THE NORTHRIDGE AND LOMA PRIETA
EARTHQUAKES ...................................................................................................... 441
v
LIST OF TABLES
Table 3.1 Geotechnical Site Categories (after Bray and Rodríguez-Marek 1997)......60
Table 3.2 Sites located on the footwall (FW) and hanging-wall (HW) in the
Northridge Earthquake (adapted from Abrahamson and Somerville 1996)......61
Table 3.3 Regression coefficients and Standard Error for spectral acceleration
values at 5% damping for (a) Northridge Earthquake (b) Loma Prieta
Earthquake.........................................................................................................62
Table 3.4 Standard deviations for the Northridge Earthquake compared with
standard deviations from Somerville and Abrahamson (Somerville,
personal comm.). Values of the standard deviation of the sample standard
deviation are given in parenthesis. ....................................................................64
Table 3.5 Subdivision of sites classified according to the presented classification
system by means of the 1997 UBC shear wave velocity-based
classification system..........................................................................................65
Table 3.6 Comparison of standard errors at selected periods for an analysis based
on the classification system presented herein and an analysis based on the
1997 UBC average shear wave velocity-based classification system. ..............66
Table 3.7 Spectral acceleration amplification factors with respect to Site B and
standard deviations for corresponding soil type. (a) Geometric mean of the
Loma Prieta and Northridge earthquakes. (b) Variance weighted mean of
the Loma Prieta and Northridge earthquakes. ...................................................67
Table 3.8 Spectral acceleration amplification factors with respect to Site C and
standard deviations for corresponding soil type. (a) Geometric mean of the
Loma Prieta and Northridge earthquakes. (b) Variance weighted mean of
the Loma Prieta and Northridge earthquakes. ...................................................69
Table 3.9 (a) Short-period (Fa) and mid-period (Fv) spectral amplification
factors from the 1997 Uniform Building Code. (b) Average spectral
amplification periods over the short-period range (0.1 s – 0.5 s) and the
mid-period range (0.4 s – 2.0 s), denoted by Fa and Fv, respectively................71
Table 4.1. Modification to ground motion parameters to account for directivity
effects. .............................................................................................................138
Table 4.2. Near-source factors from the 1997 Uniform Building Code (UBC).........139
Table 4.3. Earthquakes included in the study of near-fault ground motions. Fault
parameters are obtained from Somerville et al. (1997). ..................................140
Table 4.4. Parameters used to define the simplified sine-pulse ground motions. ......141
Table 4.5. Stations included in the analysis of near-fault ground motions. ...............142
Table 4.6. Pulse parameter for stations included in this study. ..................................144
vi
Table 4.7. Ground motion recordings satisfying geometric requirements for
forward-directivity conditions not included in this study. Listed are only
stations for earthquakes with Mw ≥ 6.5 and R < 20 km. ..................................146
Table 4.8. Values of Tv-p/Tv for multiple-record events. ............................................147
Table 4.9. Number of half-cycle pulses (N) by event for the recordings considered
in this study. Value in parenthesis is the number of half-cycle pulses that
corresponds to a cut-off value of 33% of the PGV (as opposed to 50% used
to define N)......................................................................................................148
Table 4.10. Parameters from the regression analyses for the period of the pulse of
maximum amplitude, Tv, and the period corresponding to the maximum
pseudo-velocity response spectral value, Tv-p (Equation 4.8)..........................149
Table 4.11. Inter-event error term from the random effects model for the
attenuation relationship for pulse period.........................................................150
Table 4.12. Comparison of pulse periods between rock and soil stations for
recordings in the Gilroy area in the 1989 Loma Prieta earthquake. ................151
Table 4.13. Parameters from the regression analyses for peak ground velocity
(Equation 4.12)................................................................................................152
Table 4.14. Inter-event error term from the random effects model for the
attenuation relationship for peak ground velocity. ..........................................153
Table 4.15. Ratios of pulse period in the fault parallel to pulse period in the fault
normal direction. .............................................................................................154
Table 4.16. Ground motions included in the determination of typical pulse
shapes. .............................................................................................................155
Table 5.1. Parameters needed for the model implementation. ...................................251
Table 5.2. Model and numerical parameters used for analysis of Lotung array.
The shear-wave velocity profile is given in Figure 5.10. ................................252
Table 5.3. An example of the rate of convergence of the Residual Norm in the
finite element implementation.........................................................................253
Table 5.4. Model and numerical parameters used in the analysis of the Chiba
downhole array. The shear-wave velocity profile is given in Figure 5.16. ....254
Table 5.5. Model and numerical parameters used in the analysis of the Shaking
Table tests. Shear-velocity profile is given in Figure 5.23.............................255
Table 5.6 Peak response values for the Shaking Table Runs.....................................256
Table 6.1. Number of profiles for each Site Type (based on UBC classification
scheme, Table A3) in the database of shear wave velocity profiles of Silva
(personal comm.).............................................................................................319
Table 6.2. Model parameters used for the soils considered in the Stiff Soil
profiles. The shear-wave velocity profile is given in Figure 6.1....................320
vii
Table 6.3. Model parameters used for the soils considered in the Soft Clay
profiles. The shear-wave velocity and density profile is given in Figure
6.4....................................................................................................................321
Table 6.4. Input motions and results of site response analyses on recorded near-
fault ground motions. ......................................................................................322
Table 6.5. Values of parameters used in the parametric study of site response to
simplified pulse motions. ................................................................................323
viii
LIST OF FIGURES
Figure 2.1. PGA amplification in soft soils (from Idriss 1990). ..................................25
Figure 2.2. Spectral shapes for different site conditins (from Seed and Idriss
1983)..................................................................................................................26
Figure 2.3. Site Factors used in some current ground motion attenuation
relationships for estimating spectral acceleration at 5% damping. ...................27
Figure 2.4. Site factors for varying levels of PGA from the Abrahamson and Silva
(1997) spectral acceleration (5% damping) attenuation relationship. ...............28
Figure 3.1. Relationship between structural damage intensity and soil depth in the
Caracas earthquake of 1967 (From Seed and Alonso 1974). ............................73
Figure 3.2a. Shear wave velocity versus depth for a generic stiff clay deposit.
Shear wave velocity of underlying bedrock is 1220 m/s.c ................................74
Figure 3.2b. Spectral accelerations for the stiff soil deposit shown in Figure 3.2a,
with PGA = 0.3 g for a Mw = 8.0 earthquake. ..................................................75
Figure 3.2c. Spectral acceleration amplification ratio for the stiff soil deposit
shown in Figure 3.2a with PGA = 0.3 g for a Mw = 8.0 earthquake. The
predominant period of the site is indicated by a circle......................................75
Figure 3.3a. Shear wave velocity profiles for generic sites. Shear wave velocity
of underlying bedrock is 1220 m/s. ...................................................................76
Figure 3.3b. Spectral acceleration amplification ratio for the soil profiles shown
in Figure 3b, with PGA = 0.3 g for a Mw = 8.0 earthquake. ............................76
Figure 3.4a. Distribution of data by site type for the Northridge Earthquake. .............77
Figure 3.4b. Distribution of data by site type for the Loma Prieta Earthquake............77
Figure 3.5. Number of recordings as a function of period. ...........................................78
Figure 3.6. Regression coefficients for the Northridge Earthquake. ............................79
Figure 3.7. Regression coefficients for the Loma Prieta Earthquake. The
coefficient "c" is equal to 1.0 for all periods. ....................................................80
Figure 3.8. Comparison of response spectra before smoothing and after
smoothing regression coefficients. Corresponds to the Northridge
Earthquake at R = 10 km...................................................................................81
Figure 3.9. Response spectra for the Northridge Earthquake. Thick lines
represent median values, thin lines represent ± one standard deviation............82
Figure 3.10. Response spectra for the Loma Prieta Earthquake. Thick lines
represent median values, thin lines represent ± one standard deviation............83
Figure 3.11. Median spectral values vs. distance for the Northridge Earthquake........84
Figure 3.12. Median spectral values vs. distance for the Loma Prieta Earthquake......85
ix
Figure 3.13. Comparison of results with an earthquake specific attenuation
relationship by Somerville and Abrahamson (1998). Response spectra at
5% damping for the Northridge Earthquake at R = 20 km. ..............................86
Figure 3.14. Residuals with respect to regression analysis for Site C. All sites
shown are classified as C sites in the classification system proposed in this
study, but are differentiated with respect to their corresponding UBC
classification based in the average shear wave in the upper 30 m. ...................87
Figure 3.15a. Residuals for Site C, Northridge Earthquake. Table gives mean of
residuals for each subgroup...............................................................................88
Figure 3.15b. Residuals for Site C, Loma Prieta Earthquake. Table gives mean of
Figure 3.16a. Residuals for Site D, Northridge Earthquake. Table gives mean of
Figure 3.16b. Residuals for Site D, Loma Prieta Earthquake. Table gives mean of
Figure 3.17. Residuals for D sites within the Los Angeles Basin plotted as a
function of depth to basement rock. ..................................................................92
Figure 3.18a. Amplification factors with respect to Site B for the Northridge
Earthquake.........................................................................................................93
Figure 3.18b. Amplification factors with respect to Site C for the Northridge
Earthquake.........................................................................................................94
Figure 3.18c. Amplification factors with respect to Site B for the Loma Prieta
Earthquake.........................................................................................................95
Figure 3.18d. Amplification factors with respect to Site C for the Loma Prieta
Earthquake.........................................................................................................96
Figure 3.19a. Amplification factors with respect to Site B. Geometric mean of
the Northridge and Loma Prieta Earthquakes. ..................................................97
Figure 3.19b. Amplification factors with respect to Site C. Geometric mean of
the Northridge and Loma Prieta Earthquakes. ..................................................98
Figure 3.20a. Amplification factors with respect to Site B. Weighted mean of the
Northridge and Loma Prieta Earthquakes. ........................................................99
Figure 3.20b. Amplification factors with respect to Site C. Weighted mean of the
Northridge and Loma Prieta Earthquakes. ......................................................100
Figure 3.21. Earthquake weighting scheme used for calculating spectral
amplification factors. Shown here is an average of the weights used for all
periods. Weights are inversely proportional to the sample variance..............101
x
Figure 3.22. Short-period (Fa) and intermediate-period (Fv) spectral amplification
factors. Dotted lines are code values (UBC 1997), and continuous lines
are values obtained from this study.................................................................102
Figure 4.1 Schematic diagram of rupture directivity effects for a vertical strike-
slip fault. .........................................................................................................156
Figure 4.2. Definition of parameters used in defining rupture directivity
conditions (adapted from Somerville et al. 1997). ..........................................157
Figure 4.3. Predictions from the Somerville et al. (1997) relationship for varying
directivity conditions. .....................................................................................158
Figure 4.4. Simplified pulses used by other researchers. ...........................................159
Figure 4.5. Recorded and simplification of the dominant pulses for selected fault-
normal velocity time-histories. Thick lines correspond to motions
representing a simplification of the dominant pulses......................................160
Figure 4.6. Fault normal (FN) and fault parallel (FP) velocity time-histories and
horizontal velocity-trace plots for two near-fault recordings. Both
recordings have significant fault normal velocities, but the Meloland
Overpass recording from the Imperial Valley earthquake has much lower
fault parallel velocities than the West Pico Canyon Road record from the
Northridge earthquake.....................................................................................161
Figure 4.7. Parameters needed to define the fault parallel and fault normal
components of simplified velocity pulses. Subscripts N and P indicate
fault normal and fault parallel motions, respectively. .....................................162
Figure 4.8. Comparison of two different measures of pulse period: weighted
average period and period of maximum pulse (Tv). .......................................163
Figure 4.9. Determination of pulse period (Tv) for cases in which the pulse is
preceded by a small drift in the velocity time-history. ..................................164
Figure 4.10. Determination of period corresponding to the peak pseudo-velocity
response spectral value. The pseudo-velocity response spectra shown is
for the Erzincan record of the Erzincan, Turkey, earthquake..........................165
Figure 4.11. Comparison of the period of the maximum pseudo-velocity response
spectral value (Tv-p) with the period of the maximum pulse, Tv...................166
Figure 4.12. Normalized power spectral densities of velocity time histories for
selected ground motions. The two ground motions on the left have lower
Tv-p than Tv, while those on the right have roughly coinciding values of
Tv-p and Tv.....................................................................................................167
Figure 4.13. Velocity time-histories for the 1966 Parkfield earthquake. The
Temblor record corresponds to the fault normal direction. The
seismograph at the Cholame 02 station recorded only acceleration in one
direction that corresponds to 15° degrees from the fault normal direction.
Dashed lines correspond to 50% and 33% of the PGV...................................168
xi
Figure 4.14. Fault normal velocity time-histories for the Pacoima Dam record in
the 1971 San Fernando earthquake. Dashed lines correspond to 50% and
33% of the PGV. .............................................................................................169
Figure 4.15a. Fault normal velocity time-histories for the 1979 Imperial Valley
earthquake. Dashed lines correspond to 50% and 33% of the PGV. .............170
Figure 4.15b. Fault normal velocity time-histories for the 1979 Imperial Valley
Figure 4.16. Fault normal velocity time-histories for the 1984 Morgan Hill
Figure 4.17. Fault normal velocity time-histories for the 1987 Superstition Hills
Figure 4.18. Fault normal velocity time-histories for the 1989 Loma Prieta
Figure 4.19. Fault normal velocity time-histories for the Erzincan record in the
1992 Erzincan, Turkey, earthquake. Dashed lines correspond to 50% and
33% of the PGV. .............................................................................................175
Figure 4.20. Fault normal velocity time-histories for the Lucerne record in the
1992 Landers earthquake. Dashed lines correspond to 50% and 33% of
the PGV. ..........................................................................................................176
Figure 4.21a. Fault normal velocity time-histories for rock records from the 1994
Northridge earthquake. Dashed lines correspond to 50% and 33% of the
PGV.................................................................................................................177
Figure 4.21b. Fault normal velocity time-histories for soil records from the 1994
Northridge earthquake. Dashed lines correspond to 50% and 33% of the
PGV.................................................................................................................178
Figure 4.22. Fault normal velocity time-histories for records from the 1995 Kobe,
Japan, earthquake. Dashed lines correspond to 50% and 33% of the PGV. ..179
Figure 4.23. Fault normal velocity time-histories for records from the 1999
Kocaeli, Turkey, earthquake. Dashed lines correspond to 50% and 33% of
the PGV. ..........................................................................................................180
Figure 4.24. Attenuation relationship for pulse period (Tv). .....................................181
Figure 4.25. Comparison of results form regression analysis with relationships
proposed by other researchers. The definition of pulse period of
Somerville (1998) is similar to that used in this study (Tv). On the other
hand, the pulse period of Alavi and Krawinkler (2000) is the period of the
maximum pseudo-velocity response spectral value (Tv-p).............................182
Figure 4.26. Attenuation relationship for pulse period (Tv) showing one standard
deviation band. ................................................................................................183
xii
Figure 4.27a. Velocity time-histories and velocity-trace plots for sites in the
Gilroy area recorded in the 1989 Loma Prieta earthquake. .............................184
Figure 4.27b. Pseudo-velocity response spectra for the sites in Figure 2.27a.
Observe the significant difference in frequency content at long periods
between the soil and the rock sites. .................................................................185
Figure 4.28. Distribution of near-fault sites considered in this study. .......................186
Figure 4.29. Attenuation relationship for PGV in the near-fault region. ...................187
Figure 4.29b. Comparison of results from regression analysis for PGV with
relationships proposed by other researchers....................................................188
Figure 4.30. Dependence of ratio of PGV from soil to rock on magnitude and
distance............................................................................................................189
Figure 4.31. Relationship between the ratio of fault parallel to fault normal peak
ground velocity (PGVP/N) to fault normal peak ground velocity (PGV).
Regression line and equation are for Rock sites..............................................190
Figure 4.32. Selected motions with one dominant half-cycle pulse (N = 1). .............191
Figure 4.33. Selected motions with a dominant full cycle of motion (N = 2)............193
Figure 4.34. Simplification of the dominant pulses in the recordings in Figure
4.33..................................................................................................................195
Figure 4.35a. Simplified sine-pulse representation of near-fault ground motions.
The fault parallel PGV is set to 50% of the fault normal PGV. ......................197
Figure 5.1. Schematic representation of bounding surface plasticity model..............257
Figure 5.2. Fraction of critical damping versus frequency for Rayleigh damping.
is the target fraction of critical damping. ........................................................258
Figure 5.3. Stress-strain loops for two different loading paths ..................................259
Figure 5.4. Schematic representation of site response problem. ................................260
Figure 5.5. Influence of hardening parameter h on modulus reduction and
damping curves. R/Gmax = .02, m = 1, Ho = 0 (adapted from Borja and
Amies 1994). ...................................................................................................261
Figure 5.6. Influence of hardening parameter m on modulus reduction and
damping curves. R/ Gmax = .02, h/ Gmax = 1, Ho = 0 (adapted from
Borja and Amies 1994). ..................................................................................262
Figure 5.7. Representative stress-strain loops at different amplitudes. .....................263
Figure 5.8. Finite element response when subjected to cyclic seismic loading
R/Gmax = .02, m = 1, h/Gmax = 1, Ho = 001. The loading was the KJMA
station in the Kobe earthquake. .......................................................................264
Figure 5.9. Influence of the slope of the rebound modulus in the model by
Pestana and Lok...............................................................................................265
xiii
Figure 5.10. Shear wave velocity (Vs) profile for Lotung Site (Borja et al. 1999,
EPRI 1993). .....................................................................................................266
Figure 5.11. Modulus degradation and damping curves for the Lotung site..............267
Figure 5.12. Comparison of recorded and computed ground motions at the
surface for the Lotung array. ...........................................................................268
Figure 5.13. Recorded and computed acceleration time histories in the Lotung
array. Thick lines are recorded accelerations. The input motion is at Elev.
47. ...................................................................................................................269
Figure 5.14. Recorded and computed velocity time histories in the Lotung array.
Thick lines are recorded velocities. The input motion is at Elev. 47. ............270
Figure 5.15. Comparison of uni-directional and bi-directional analyses for the
Lotung array. ...................................................................................................271
Figure 5.16. Shear wave velocity (Vs) profile for the Chiba site (Katayama et al.
1990)................................................................................................................272
Figure 5.17. Modulus degradation and damping curves for soils in the Chiba site.
Thin lines correspond the curves predicted by the model. Model
parameters are matched to curves from the indicated references....................273
Figure 5.18. Recorded and computed acceleration time histories in the Chiba
array. Thick lines are recorded accelerations. The input motion is at Elev.
40. The accelerogram at 5 m depth did not trigger for this event. .................274
Figure 5.19. Recorded and computed velocity time histories in the Chiba array.
Input motion is at Elev. 40. .............................................................................275
Figure 5.20. Acceleration response spectra for the analyses of the North-South
component of motion in the Chiba downhole array. .......................................276
Figure 5.21. Strains predicted by SHAKE91 (Idriss and Sun 1992) and the
GeoFEAP analysis for the North-South component of the Chiba downhole
array. ..............................................................................................................277
Figure 5.22. Shear-wave velocity profile of the model clay soil used in the
Shaking Table Test 2.53, including location of accelerograms.......................278
Figure 5.23. Modulus degradation and damping curves for model soil used in
shaking table test. ............................................................................................279
Figure 5.24. Velocity response spectral for recorded and calculated motions in
Test 2.16. .........................................................................................................280
Test 2.53.. ........................................................................................................281
Test 2.55. .........................................................................................................282
xiv
Test 2.58. .........................................................................................................283
Figure 6.1. Selected shear wave velocity profile for the Stiff Soil profile.
Included in the graph are the median and ± one standard deviation for a
database of 343 profiles (Silva, personal comm.) classified as Site D by the
Uniform Building Code (e.g. 180 m/s < Vs ≤ 360 m/s)..................................324
Figure 6.2. Shear modulus degradation and equivalent viscous damping curves
for the soils used in the Stiff Soil profile. .......................................................325
Figure 6.3. Dynamic undrained shear strength for the Stiff Soil profile. ...................326
Figure 6.4. Generic Soft Clay shear wave velocity and density profiles. The
profile represents typical Bay Mud sites from the San Francisco Bay
region...............................................................................................................327
Figure 6.5. Dynamic shear strength profile for the generic Soft Clay site. ................328
for the Soft Clay soil. The Large-Strain model curve is given for the
average strength value of the clay. The curves shift to the right for an
increase in strength, and to the left for a decrease in strength.........................329
for the Pleistocene clay used in the generic Soft Clay profile. The upper-
and lower-bound model curve corresponds to the upper bound and lower
bounds, respectively, for the strength parameter R. The upper bound curve
is used when soil yielding is not of concern....................................................330
Figure 6.8. Selected shear wave velocity profile for the Very Stiff Soil profile
(Site C). Included in the graph are the median and ± one standard
deviation for a database including 227 profiles (Silva, personal comm.)
classified as Site C by the Uniform Building Code (e.g. 360m/s < Vs ≤ 760
m/s)..................................................................................................................331
Figure 6.9. Power spectral densities, PSD (normalized by the maximum value of
the PSD) for the ground motions used in the site response analyses. PSD
are given for the full recorded ground motion, for the time interval of the
ground motion containing the velocity pulse, and for the simplified
representation of the velocity pulse.................................................................332
Figure 6.10. Site response analyses results. Pacoima Dam record for the 1971
San Fernando Earthquake................................................................................333
Figure 6.11. Site response analyses results. Gilroy Gavilan College record for the
1989 Loma Prieta Earthquake. ........................................................................334
Figure 6.12. Site response analyses results. Pacoima Dam downstream record for
the 1994 Northridge Earthquake. ....................................................................335
Figure 6.13. Site response analyses results. KJMA record for the 1995 Kobe
Earthquake.......................................................................................................336
xv
Figure 6.14. Site response analyses results. Input (rock) and output (soil) particle
velocity-trace plots. .........................................................................................337
Figure 6.15. Fault-normal acceleration and velocity time-histories for the
Pacoima Dam record of the 1971 San Fernando earthquake. The dominant
velocity pulse is highlighted to show that the maximum accelerations do
not coincide with the fault-normal velocity pulse. ..........................................338
Figure 6.16. Calculated strains using the recorded velocity pulse and the
simplified velocity pulse as input motions......................................................339
Figure 6.17. Response of a linear-elastic soil column subject to the simplified
velocity pulses. The natural period of the soil deposit is 1.0 s.......................340
Figure 6.18. Amplification of input motions by a linear-elastic profile (ω is the
frequency of the input motion, Vs is the shear wave velocity of the elastic
soil, and H is the soil depth (from Kramer 1996)............................................341
Figure 6.19. Results from the site response analyses for the 60 m deep Stiff Soil
profile. Particle velocity-trace plots for the input (rock) and output (soil)
motions. Input motion set is Set 6. PGVp/n of input motion is 0.5. .............342
motions. Input motion set is Set 7. PGVp/n of input motion is 0.5. .............343
motions. Input motion set is Set 8. PGVp/n of input motion is 0.5. ............344.
Figure 6.22. Results of site response analyses for the 60 m deep Stiff Soil site for
three different input motion sets. Results are plotted with fault normal
input pulse period in the abscissa....................................................................345
Figure 6.23. Site response analyses results for the 30 m deep Stiff Soil profile.
The input motion is Set 6, with a PGVp/n ratio of 0.5. ..................................346
Figure 6.24. Comparison of site response studies for two profiles with equal
depth to bedrock (30 m) and varying profile stiffness. Input motion is Set
6 with a PGVp/n ratio of 0.5. ..........................................................................347
Figure 6.25. Comparison of site response analyses results for the Stiff Soil
profile. The depth to bedrock is varied from 30 m to 200 m. The input
motion is Set 6 with a PGVp/n ratio of 0.5. ....................................................348
Figure 6.26. Ratio of pulse period in soil to pulse period in rock for site response
analyses on Stiff Soil profiles with varying depth to bedrock. The input
motion is Set 6 with a PGVp/n ratio of 0.5. ....................................................349
Figure 6.27. Comparison of site response analyses for soils with different shear
modulus reduction and damping curves. Results for input motion Set 6
with a PGVp/n ratio of 0.5. .............................................................................350
xvi
Figure 6.28. Comparison of calculated ratio of soil pulse period to rock pulse
period. Results for site response analyses for soils with different shear
modulus reduction and damping curves. Input motion is Set 6, with
PGVp/n ratios of 0.5........................................................................................351
Figure 6.29. Comparison of site response analyses results to input motions with
varying levels of fault parallel velocity. Results for a 45 m deep Stiff Soil
for PGVp/n ratios of 0, 0.25 and 0.75. Input fault-normal PGV is 75 cm/s...352
Figure 6.30. Comparison of site response analyses to input motions with different
fault-parallel velocities. Results for the 45 m deep Stiff Soil profile, Input
motion Set 6. ...................................................................................................353
Figure 6.31. Comparison of site response analyses results for the different input
motion sets. The analyses are for the Stiff Soil profiles with varying
depth. The input PGVp/n ratio is 0.25 for Sets 7 and 8, and 0.5 for Set 6.....354
Figure 6.32. Comparison of strain levels for runs with (a) equal PGV, and (b)
equal PGA. Stiff Soil profile with a depth of 45 m. Input motion is Set 6
with PGVp/n = 0.5. .........................................................................................355
Figure 6.33. Calculated strains for various Stiff Soil profiles. Input motion is Set
6 with PGVp/n ratios of 0.5, and input PGV of 160 cm/s...............................356
Figure 6.34. Results of site response analyses. Particle velocity-trace plots of
input and output motions. Soft Clay site, input motion is Set 6 with
PGVp/n ratios of 0.5........................................................................................357
PGVp/n ratios of 0.25......................................................................................358
PGVp/n ratios of 0.25......................................................................................359
Figure 6.37. Calculated stresses and strains for the Soft Clay profile. Input
motion is Set 6 with PGVp/n = 0.5 .................................................................360
Figure 6.38. Calculated profiles of displacement, velocity, and acceleration for
the Soft Clay site. Input motion is Set 6 with PGVp/n of 0.5 and input
PGV of 160 cm/s. ............................................................................................361
Figure 6.39. Comparison of results of site response analyses for two profiles with
equal depth to bedrock but different soil profiles. Ratios of PGV in soil to
PGV in rock. Input motion is Set 6 with PGVp/n of 0.5................................362
Figure 6.40. Comparison of site response analyses for the Soft Clay profile and
the Stiff Soil profile with equal depth (60 m). Results shown are for all
input motion sets with PGVp/n values ranging from 0.25 to 0.5....................363
xvii
Figure 6.41. Comparison of results of site response analyses for two profiles with
equal depth to bedrock but different soil profiles. Ratios of pulse period in
soil to pulse period in rock. Input motion is Set 6 with PGVp/n of 0.5. ........364
Figure 6.42. Effect of site response on the PGVp/n ratio. Results for three
different profiles and for input motion sets 7 and 8. PGV of input motion
is 75 cm/s.........................................................................................................365
Figure 6.43. Calculated amplification of peak ground velocity. Results for site
response analyses for all the Stiff Clay profiles and all the simplified
velocity poulse input motion sets. Large symbols are the predicted PGV
amplification for the Stiff Soil site with a depth of 45 m using the
indicated records as input motions. .................................................................366
Figure 6.44. Relationship between PGV in soil and PGV in rock. Results for the
45 m deep Stiff Soil deposit. Input motion is Set 6 with PGVp/n ratio of
0.5. Observe the dependence of PGV soil to rock ratio on pulse period
(Tvp) and intensity of motion (PGVrock).......................................................367
Figure 6.45. Calculated ratios of pulse period in the soil over pulse period in rock.
Results from site response analyses on Stiff Clay profiles using all
simplified velocity pulse input motions sets and PGVp/n ratios ranging
from 0.25 to 0.75. ..........................................................................................368
Figure 6.46. Calculated ratio of pulse period in soil to pulse period in rock for the
45 m deep Stiff Soil profile. Results are shown for all the input motion
sets with input PGV lower than 300 m/s. Results for the indicated ground
motions are included for comparison with the results of the simplified
velocity pulses. ................................................................................................369
xviii
ACKNOWLEDGMENTS
This work would not have been possible without the help of a large number of
individuals to whom I feel deeply indebted. First, I would like to thank my advisor,
Professor Jonathan D. Bray, for his advice, support, and friendship. He has contributed to
this work by providing innumerable ideas and exciting intellectual challenges. Most of
all, I greatly appreciate his mentorship and his support throughout my doctoral studies.
I would also like to thank Professor Juan Manuel Pestana for reviewing this
dissertation, and especially for his teachings and insights on soil behavior; and Professor
Alexandre J. Chorin for reviewing this dissertation and allowing me to feed from his
great knowledge of numerical methods. My years at U.C. Berkeley have been the greatest
intellectual experience of my life. My deepest thanks go to the faculty in the Civil
Engineering Department for fostering the environment that made this intellectual
experience possible. In particular, I would like to mention Professor Nicholas Sitar for
his support during my first year in the doctoral program.
The discussions during meetings with Dr. Norman A. Abrahamson of Pacific Gas
and Electric Company planted the seeds for most of the research presented in this work.
In addition, I would also like to extend my thanks to Professor Robert L. Taylor for his
invaluable comments regarding time integration schemes and the finite element method;
Professor Govindjee for his guidance on constitutive model implementation; Dr. Walter
Silva for his assistance in providing the ground motion database used in this study; Dr.
François E. Heuze of the Lawrence Livermore National Laboratory for his assistance in
sharing their geotechnical database; Dr. Paul Somerville for providing event-specific
xix
attenuation relationships for the Northridge and Loma Prieta earthquake and for providing
useful comments on the near-fault studies in this work; Prof. Raymond B. Seed for his
insights into site response; Dr. Mladen Vucetic, Dr. Sands Figuers, and Dr. David Rogers
for providing essential geotechnical data for ground motion sites; Dr. Tomoya Iwashita
for his help in obtaining data from the Kobe region; Dr. Jonathan Stewart for the
information on the Oakland BART Station site and numerous sites in Southern
California, for the ground motion data of forward-directivity parameters, and for the
many instructive discussions; Dr. Susan Chang for her willingness to share knowledge
and information on ground motion site classification; Dr. Ellen Rathje for providing the
Kocaeli records; Dr. Philip Meymand, Dr. Thomas Lok and Dr. Michael Riemer for their
help in using the U.C. Berkeley Shaking Table test data; Mr. Ignacio Romero for
providing a linear-elastic brick element and helping me in the finite element
implementation; Mrs. Laurie Gaskins and Professor Steve Glaser for the data from
downhole arrays; Mrs. Catherine O'Sullivan for reviewing this manuscript; Dr.
Christopher Hunt for his jokes and his help using Matlab; and Dr. Laurent Luccioni for
his willingness to come to my help anytime it was needed.
I would also like to extend my thanks to my classmates and friends for their
friendship, support, and for all the great soil talks we had over the years. Each of them
has contributed to making the years in Berkeley not only instructive, but also fun. Even if
I were to try to list them all, I would invariably fail to mention all of them; their names
might be unlisted here, but their friendship will always remain alive.
Financial support was provided by the Pacific Gas and Electric Company, the
David and Lucile Packard Foundation, and the Pacific Earthquake Engineering Center.
xx
My thanks go to my fellow researchers in PEER for their instructive comments at
research meetings.
Finally, I would like to acknowledge the love and support of my family. I would
not be here if it were not for my parents' support. I would not be the same if it were not
for my brothers' friendship. I would not have finished if it were not for my wife's love.
My greatest thanks go to her.
xxi
CHAPTER 1
INTRODUCTION
1.1 PROBLEM
A number of major metropolitan centers lie in close proximity to major faults.
For example, in the United States, the San Andreas fault system crosses densely
populated areas of the metropolitan Los Angeles and Bay Area regions, while other active
faults lie beneath cities in the Pacific Northwest. The proximity of these fault systems to
large population centers motivates research seeking a better understanding of ground
motions close to faults.
The estimation of ground motions in zones that are close to the causative faults of
medium to large magnitude earthquakes should account for the special characteristics of
near-fault ground motions. Of particular importance in the near-fault region are the
effects of forward-directivity. Forward-directivity conditions produce ground motions
characterized by a strong pulse or series of pulses of long period motions. In recent years,
many efforts have been devoted to the characterization of forward-directivity motions
(e.g., Somerville et al. 1997, Krawinkler and Alavi 1998, Sasani and Bertero 2000).
These studies have highlighted the important contribution of near-fault ground motions to
the seismic risk of structures. Building on this understanding, more research is needed to
account for the potential effects of local soil conditions on these motions.
1
The seismic site response problem has been studied extensively in recent years,
both by empirical analysis of the existing database of ground motions (e.g., Borcherdt
1994, Crouse and McGuire 1996) and by means of analytical site response tools (e.g.,
Seed et al. 1991, Dobry et al. 1994). The empirical characterization of site response is
constrained by the availability of data. Until the 1994 Northridge, 1995 Kobe, 1999
Kocaeli, and especially the 1999 Chi-Chi earthquakes, very few recordings at close
distances to the causative fault were available for study. Consequently, most of the
empirical study of site response has been concentrated on ground motions recorded at
more than 15 km from the zone of energy release. The same unavailability of near-fault
data also caused the direction of seismic site response studies to be driven towards
analyzing site response to "normal" ground motions, that is, ground motions recorded at
intermediate to long distances from the zones of energy release.
Near-fault ground motions, and in particular forward-directivity motions, have
specific characteristics that differentiate these motions from those recorded at
intermediate to long distances from the zone of energy release. These particular
characteristics raise the question of whether the current understanding of site response
applies directly to near-fault ground motions. This dissertation aims at helping to resolve
this question and as a consequence, provide a better understanding of the effect of local
site conditions on near-fault ground motions. In addition, a new site classification
scheme is introduced for the evaluation of site response at intermediate distances from the
zone of energy release. This classification scheme is used to evaluate data from the 1989
Loma Prieta and 1994 Northridge earthquakes.
2
1.2 SIGNIFICANCE
The relevance and importance of correctly characterizing seismic site response
and in particular local soil condition effects cannot be over-stated. Local soil conditions
have played a major role in the damage and loss of life in earthquakes such as the 1985
Mexico City, 1989 Loma Prieta, 1994 Northridge, and 1995 Kobe earthquakes.
Moreover, large population centers across the globe can potentially be struck by an
earthquake originating on nearby faults. A better understanding of the mechanisms that
control near-fault ground motions and the interaction of near-fault ground motions with
soil deposits is imperative to evaluate better seismic risk in these regions. This
understanding, in turn, will permit the development of adequate code-based regulations
that ensure the safety of engineered structures.
The assessment of seismic demand on a structure within the framework of
performance-based design requires not only an estimation of the median expected levels
of ground motion intensity, but also the standard error associated with this estimated
median. A complete assessment of site conditions, if sufficient data are available, should
reduce the uncertainty associated with the predictions of ground motions. However, a
fine balance is necessary when dealing with sparse data sets, such as the ground motion
database. Site classification schemes that are too elaborate would result in a data set that
is too small to yield statistically meaningful results. On the other hand, a site
classification scheme that is too simplistic might not capture trends in ground motions
that could significantly reduce the uncertainty associated with ground motion prediction.
The use of adequate classification schemes should result in better estimates of ground
3
motion for engineering design. A reduction in uncertainty associated with better site
classification schemes could also lead to better risk assessment methodologies useful for
the private sector (e.g., insurance companies), as well as the public sector (e.g., advanced
emergency planning).
A better understanding of site amplification and, in particular, site amplification in
the near-fault region, is useful beyond its application in the estimation of seismic demand
on structures. The recent earthquakes in Kocaeli, Turkey, and Chi-Chi, Taiwan, have
increased the available database of strong motions recordings by an order of magnitude,
particularly in the near-fault region. Undoubtedly, the larger database will result in a
push for new attenuation relationships that can directly account for near-fault effects. A
proper understanding of the influence of local site conditions on ground motions will be
of paramount importance in the evaluation of the current database of strong motions
recordings. This dissertation will contribute toward this goal by providing both a
qualitative and a quantitative evaluation of seismic site response to near-fault ground
motions.
1.3 OBJECTIVE
The objectives of this dissertation are to address the site response problem in
general, and the site response to near-fault ground motions in particular. These objectives
can be summarized as follows:
4
• Development and evaluation of a site classification scheme that accounts for both
profile depth and material stiffness. This classification scheme will be used to
evaluate the effect of profile depth on site response.
• Evaluation of the empirical database of near-fault ground motions, with emphasis on
the development simplified pulses that can be used for seismic site response analyses,
and on the evaluation of differences in ground motions recorded on rock from those
recorded on soil.
• Evaluation of the implementation of a dynamic constitutive model into a finite
element computer code for the analysis of site response to near-fault ground motions.
• Quantification of trends in site response to near-fault ground motions by means of
finite element analysis of the dynamic response of different site profiles.
1.4 OUTLINE
This dissertation is organized into seven chapters. The first chapter outlines the
problem of near-fault seismic site response and presents the motivating factors that led to
this dissertation. Chapter Two presents a brief review of the existing literature on site
response with particular emphasis on current methodologies used to solve the site
response problem.
Chapter Three consists of an exhaustive analysis of site response for ground
motions recorded during two important recent earthquakes, the 1989 Loma Prieta and
1999 earthquakes. This chapter presents and evaluates a site classification scheme that is
used to evaluate the effect of profile depth on reducing the uncertainties associated with
5
ground motion prediction. Site amplification factors based on the classification scheme
are developed from ground motions recorded during the 1989 Loma Prieta and 1994
Northridge earthquakes. It is found that the database does not include many ground
motion recordings in the near-fault region and is applicable only to sites at intermediate
distances from the zone of energy release.
Chapter Four presents an empirical analysis of the database of motions that are
located in the near-fault and that experienced forward-directivity effects. The study
concentrates on developing a parameterization of this type of motions. Differences
between motions recorded in rock and in soil are analyzed and the observed trends are
discussed.
Chapter Five contains the description of a finite element implementation of a
constitutive model used to solve the site response problem. The constitutive model used
is first presented in detail. The finite element implementation is then validated through a
series of case studies, with an emphasis on its application to near-fault ground motions.
Chapter Six presents a series of site response analyses using the soil model and its
finite element implementation presented in Chapter Five. The analyses are performed for
recorded near-fault motions, as well as for simplified motions developed from the
empirical analysis in Chapter Four. The trends in seismic site response noted in Chapter
Four are used to validate the results obtained in this chapter.
Chapter Seven presents a summary of the results and the findings resulting from
this work. It includes recommendations for future research.
6
CHAPTER 2
LITERATURE REVIEW
2.1 INTRODUCTION
The present study involves contributions from a variety of disciplines. The
empirical study of site response in Chapters Three and Four behooves an up-to-date
assessment of many of the issues relating to site response and an understanding of the
state of the art in attenuation relationships. A significant amount of research on near-fault
ground motions has been published in the seismological literature, and many of the issues
raised in the analysis of near-fault site response (Chapters Four and Six) are intrinsically
tied to this research. Additionally, the presentation of the finite element implementation
of the soil model utilized to analyze the site response problem feeds from a great amount
of work in finite element technology and the development of constitutive models of soil
response to dynamic loading. A comprehensive review of all these areas is beyond the
scope of this research project. Attempts to cover such a variety of topics in a
comprehensive manner would undoubtedly fail to cover adequately all of the important
research performed in each of these individual areas of study. In this chapter, only a brief
synopsis of recent work in seismic site response analysis and the treatment of site
amplification by current ground motion attenuation relationships are presented. The
objective is to provide the reader with a general knowledge of the foundations upon
which this work is grounded.
7
2.2 SITE RESPONSE
2.2.1 Site Response Studies and Code Development
The influence of local site conditions on ground motions has been observed since
the early days of earthquake engineering. Observations from as early as the 1800s exist in
the literature indicating the effects of local geology on ground motions (EPRI 1993).
Traditionally, site amplification had been studied by seismologists as part of the larger
problem of seismic wave propagation (e.g., Sezawa and Kanai 1932, Kanai 1950,
Thompson 1950, Haskell 1960). Seismologists have traditionally treated soil as a linear
material and rarely considered soil nonlinearity in their assessment of site conditions
(Finn 1991). The pioneering work of Seed and Idriss (1969) brought attention to the
nonlinear behavior of soils during seismic shaking. This work stemmed form
observations of the earthquakes in Niigata and Alaska in 1964, and the 1967 Caracas
earthquake. Since then, site response has become an integral part of geotechnical
earthquake engineering.
Following the pioneering work of the late Professor H. B. Seed and his
colleagues, site response has been studied in a large number of earthquakes since the
1960s. The 1985 Michoacan Earthquake is particularly significant due to the large levels
of spectral amplification recorded in the soft lakebed sediments in Mexico City. Prior to
the 1985 Michoacan earthquake, soft soils were thought to de-amplify motions at peak
ground accelerations (PGAs) larger than 0.1 to 0.2 g (e.g., Seed and Idriss 1983), while
motions at stiff soils were thought to be largely unaffected by the ground motion
intensity. However, the response of the high plasticity Mexico City clays was more linear
8
than expected for the observed levels of ground motion. This resulted in a reassessment
of the amplification of PGAs in soft ground, as reflected in Figure 2.1 (Idriss 1990). The
Mexico City earthquake also brought attention to the need for a better understanding of
the dynamic properties of soft clays (Finn 1991).
The development of design codes has followed the advancements in
understanding of site response. Seed et al. (1976) and Mohraz (1976) developed average
spectral shapes for various soil conditions based on a statistical study of more than 100
records from 21 earthquakes (Figure 2.2). While the spectral shapes in Figure 2.2 are
relatively independent of site condition at short periods, they are largely site dependent at
periods longer than 0.5 s. This led to the adoption of simplified response spectra shapes
by the Applied Technology Council (1978) and later, with some variations, by the
Uniform Building Code (1988). The use of spectral shapes without amplification factors
for peak acceleration reflected the observations by Seed et al. (1976) that accelerations in
soils and rocks were approximately equal.
The 1989 Loma Prieta earthquake provided a large database of strong ground
motion recordings in a variety of site conditions. This database was analyzed by a
number of authors (e.g., Silva and Stark 1992, Borcherdt 1994, Dickenson 1994, Chang
and Bray 1995, Boatwright and Seekings 1997, among others). Borcherdt (1994)
introduced a classification system based uniquely on the average shear wave velocity of
the upper 30 m of the profile at the site, Vs . This classification system was used to
develop spectral amplification factors with respect to a baseline condition. The
amplification factors were obtained by performing a statistical regression analysis on the
9
Loma Prieta data using a continuous function of Vs . These factors are applicable mainly
to low intensity motions as most of the recordings of the 1989 earthquake were located at
significant distances (i.e. > 20 km.) from the causative fault.
The Loma Prieta earthquake, following in the wake of the devastating 1985
Michoacan earthquake, served as a catalyst for the update of building codes. This effort
was initiated in a workshop convened by the National Center for Earthquake Engineering
Research (now MCEER: Multidisciplinary Center for Earthquake Engineering Research)
in 1991 (Whitman 1992). At a later workshop in Los Angeles in 1992 (Martin 1994), a
consensus was reached that resulted in the adoption of the 1994 National Earthquake
Hazard Reduction Program Provisions (NEHRP 1994), which later were adopted by the
1997 UBC, and are unaltered in the 2000 International Building Code (Dobry et al. 2000).
The 1994 NEHRP provisions were based on the shear wave velocity based site
classification system by Borcherdt (1994). The low intensity amplification factors for the
Loma Prieta earthquake developed by Borcherdt (1994) were extrapolated to larger
intensities using linear and nonlinear site response analyses (Seed et al. 1991, Dobry et al.
1994). The design spectra in the 1994 NEHRP Provisions differ from the previous codes
in that two separate site-dependent coefficients are used to represent the spectral response
over the short period and long period spectral region. This step is taken as recognition of
the potential differences in amplification between the short and the long period regions of
the spectra. An extensive review of the development of code provisions up to the 1997
NEHRP code is presented in Dobry et al. (2000). Further discussion of these is also
included in Chapter Three.
10
Crouse and McGuire (1996) used an extended database to validate the site factors
adopted in the 1994 NEHRP provisions. They performed a regression analysis on sites C
and D in the UBC classification scheme (See Table A3 in Chapter Three) using a
weighting scheme proposed by Campbell (1990), which gave the same total weight to
recordings from each earthquake within each magnitude and distance interval. The
regression lines for sites C and D where then scaled to fit data of sites B and E using a
least squares fit. The scaling factor was assumed to be independent of ground motion
intensity. The authors concluded that the extended database supports the factors
implemented in the NEHRP provisions. However, the large degree of nonlinearity of
short period amplification factors in the NEHRP Provisions was not observed in the
larger database.
The use of the average shear wave velocity over the upper 30 m of a profile was
reviewed by a number of authors (Day 1996, Anderson et al. 1996, Darragh and Idriss
1998). Darragh and Idriss (1998) performed equivalent-linear analysis on the Gilroy
Array in Northern California with records from the Loma Prieta earthquake. The authors
concluded that the use of Vs is justified for input levels near 0.45 g. Anderson et al.
(1996) performed wave propagation analyses on horizontal elastic layers with depths
extending to 5 km and concluded that the upper 30 m have a large influence on
amplifications of peak amplitudes and root mean square acceleration. However, the
ground motions at the surface also depend on the attenuation structure of deeper deposits.
The site classification scheme adopted in the 1994 NEHRP Provisions does not
directly include soil depth as a classification parameter, as did earlier codes with the
11
direct use of site period for site classification (UBC 1976). Studies using generalized soil
profiles with varying depth to bedrock have indicated the important effect of soil depth in
amplification factors (EPRI 1993, Silva 1998). Chang and Bray (1995) and Chang et al.
(1997) observed that the ground motions at deep stiff soil sites were larger than those
predicted by the corresponding Vs -based site class.
2.2.2 Cyclic Soil Models and Site Response Methodologies
Ever since the nonlinear behavior of soil was recognized as an important factor in
site response, the development of cyclic soil models used to represent the dynamic
response of soil has attracted a great deal of attention from engineers. Fundamental to the
development of cyclic soil models was the improvement in the understanding of cyclic
response of soil through laboratory experimentation (e.g., Seed and Idriss 1970, Hardin
and Drnevich 1972, Seed et al. 1984, Sun et al. 1988). In later years, the improvement of
local strain measurement techniques has increased the understanding of the small-strain
behavior of soils. A review of recent work in dynamic testing is presented in Bray et al.
(1999). For a compilation of testing data on small strain soil properties see Lanzo and
Vucetic (1999). More recently, data from downhole arrays has also been used to
characterize the nonlinear stress-strain behavior of soils (Zeghal et al. 1995). The use of
downhole array data has the advantage of avoiding sample disturbance problems, but
introduces additional problems such as definition of boundary conditions.
The site response problem is generally solved either by equivalent-linear analysis
(Seed and Idriss 1969), or by a fully nonlinear approach. The site response code used
most frequently in engineering practice is the equivalent-linear program SHAKE

12
(Schnabel et al. 1972, Idriss and Sun 1992). Details on the equivalent-linear method are
given in Chapter 5. In the following paragraph, some of the site response codes and
cyclic soil models in current use are summarized.
Lok (1999) classifies nonlinear models into three separate classes: Mechanical, in
which soil behavior is represented by single mechanical elements, such as springs,
dashpots, and sliders, placed in series or in parallel; empirical, which use empirically-
derived functions to fit the nonlinear stress-strain behavior of soils; and plasticity models,
based on the framework of plasticity theory.
Mechanical models are commonly of the type described by Iwan (1967). Iwan-
type models consist of elastoplastic elements placed either in series or in parallel. Each
element consists of a linear spring and a Coulomb slider. Any stress-strain curve can be
described within a specified degree of accuracy by increasing the number of elastoplastic
elements. The series model, more apt to describe strain in terms of stress, was used by
Joyner and Chen (1975) and the parallel model, apt for strain-driven problems, was used
by Taylor and Larkin (1978).
Empirical models that are commonly used include the Ramberg-Osgood
(Ramberg and Osgood 1943), the Davidenkov, and the hyperbolic model (Kondner 1963).
These models are typically used to describe first loading and the backbone curve. The
hysteresis loops observed in cyclic tests of soil can be easily constructed from the
backbone curve by the use of the Masing rules. These rules are stated as (Pyke 1979): (1)
The shear modulus on each loading reversal assumes a value equal to the initial tangent
modulus for the initial loading curve; and (2) the shape of the unloading or reloading
13
curves is the same as that of the initial loading curve. Pyke (1979) discusses the
extension of monotonic models into cyclic models and suggests adding two more
Masing-type rules to adequately represent cyclic soil behavior: (3) The unloading and
reloading curves should follow the initial loading curve if the previous maximum shear
strain is exceeded; and (4) if the current loading or unloading curve intersects the curve
described by a previous loading or unloading curve, the stress-strain relationship follows
that previous curve (Newmark and Rosenblueth 1971, Rosenblueth and Herrera 1964).
Alternatively, stress-strain curves compatible with the Masing rules are obtained using
plasticity based models or appropriate interpolation rules (Pyke 1979).
Empirical models are used in a number of implementations of site response
analyses. CHARSOIL (Streeter et al. 1974) uses the Ramberg-Osgood model and solves
the site response problem using the method of characteristics. Martin and Seed (1982)
used both the Ramberg-Osgood and the Davidenkov model to perform one-dimensional
ground response analyses. A hyperbolic model is used by Lee and Finn (1978, 1991) in
the program DESRA. A variation of the constitutive model by Lee and Finn (Matasovic
and Vucetic 1993) is used in the program D-MOD. Pyke (1992) uses a hyperbolic model
and the Cundall-Pyke hypothesis (Pyke 1979) in the finite difference code TESS.
Plasticity-based models provide the most flexibility in representing details of soil
behavior, including yielding, pore pressure generation, and soil response to multi-
directional loading paths. However, plasticity-based models are not always amenable to a
robust numerical implementation. Since the inception of critical state soil mechanics
(Schofield and Wroth 1968), critical state models have been used successfully to model
14
monotonic soil behavior. Plasticity based models have also been incorporated into site
response procedures (e.g., Scott 1985, Finn 1988, Li et al. 1998, Borja et al. 1999).
Typically, plasticity models are implemented in finite element or finite difference codes.
The basic elements of a plasticity formulation are (Prévost 1977): (1) the yield
condition, specifying states of stress for which plastic flow occurs; (2) a flow rule,
indicating the direction of the stress increment for a given plastic strain increment; and
(3) a hardening rule, specifying the evolution of the loading surface. In addition, the
conditions for loading and unloading from the yield surface must be specified (Simo and
Hughes 1998), as well as definition of the soil response before the state of stress reaches
the yield condition (typically elastic). In cyclic models, condition (3) must ensure that
plastic strains accumulate immediately after stress reversal, as observed in laboratory
cyclic tests. Prévost (1977) proposed the use of a combination of isotropic and kinematic
plastic hardening rules to represent soil response. Additionally, Prévost (1977) uses a
field of shear moduli (originally proposed by Mróz 1967) defined in stress space by a
collection of nested yield surfaces that define regions of constant shear moduli. The
multiple yield surface model was initially proposed in connection to metal plasticity by
Mróz (1967). The innermost surface in the collection of nested surfaces can be
represented as a single point in order to model the plastic strains observed in soils at the
onset of loading. On the other hand, the outermost surface plays the role of a failure
surface, outside which states of stress are not possible. The model by Prévost (1977)
permits the modeling of essential elements of cyclic soil behavior, such as small strain
nonlinearity and hysteretic dissipation of energy. It also automatically satisfies the
extended Masing rules (Pyke 1979). The model by Prévost is applied in the site response
15
program DYNA1D (Prévost 1989). Other multiple yield surface models were proposed
by Mróz, Norris, and Zienkiewicz (1978) and Hirai (1987). Mróz et al. (1981) presented
an improvement of the multi-surface model by introducing an infinite number of nested
loading surfaces. In this way, the plastic moduli were made continuous in the whole
domain enclosed by the outer surface.
Small strain nonlinearity can also be described through what is known as
Bounding Surface Plasticity. The bounding surface plasticity concept was independently
proposed by Dafalias and Popov (1975, 1976) and Krieg (1975). Bounding surface
plasticity rests on the idea that any stress point inside a bounding surface has a unique
'image' point on the surface, defined by means of a specific rule. The instantaneous value
of the plastic modulus depends on the distance between the stress point and its 'image'
(Dafalias and Herrmann 1982). Bounding surface plasticity models have been used by a
number of authors, including Dafalias and Herrmann (1982), Dafalias (1986), Bardet
(1989), Mróz et al. (1979), Borja and Amies (1994), Li et al. (1998), and others. These
models vary in the definition of the elastic region or in the type of bounding surface used.
Bounding surface models incorporated into site response analysis include SPECTRA
(Borja et al. 1999) and SUMDES (Wang et al. 1990).
Other models used to describe cyclic soil behavior are based on the concept of
hypoplasticity and endochronic theory. Hypoplasticity provides general expressions for
the constitutive tensors satisfying all necessary requirements (Zienkiewicz et al. 1999).
Darve and Labanieh (1982) suggested that the constitutive tensor can be interpolated from
constitutive tensors defined along specified loading paths. Other hypoplasticity models
16
include the model by Dafalias (1986). This model, mentioned earlier along with the
bounding surface plasticity models, presents extensions to the bounding surface plasticity
model within the framework of hypoplasticity. Endochronic models describe nonlinearity
by a sequence of events leading to successive states of the material (Finn 1982).
Endochronic theory was developed by Valanis, and applied to liquefaction problems by
Bazant and Krizek (1976) and Zienkiewicz et al. (1978).
Several of the plasticity models listed in the above paragraphs are able to account
for pore pressure generation leading to liquefaction in loose soil deposits. The
liquefaction problem, however, requires the use of models able to reproduce contractive
behavior of soils, and in some cases, performed coupled mechanic-pore pressure
dissipation analysis. For a review of such models, see Zienkiewicz et al. (1999).
The ability of constitutive models in representing cyclic soil behavior can be
evaluated by comparing model predictions to laboratory data. Typically, model
predictions of shear modulus reduction and damping curves can be compared to
laboratory curves. In many cases, model parameters that are used to match shear moduli
curves fail to adequately match damping curves. Lok (1999) presents an enhancement of
a hysteretic model introduced by Whittle and Kavvadas (1994, Pestana and Whittle 1999)
within a generalized plasticity model. This model permits a simultaneous match of shear
modulus reduction and material damping curves. Some details of the model are given in
Chapter Five.
17
2.3 SITE EFFECTS IN ATTENUATION RELATIONSHIPS
Attenuation relationships are used widely to predict ground motion levels for risk
assessment and seismic design. The use of attenuation relationships permits a more
flexible assessment of seismic hazard in design as opposed to the fixed levels of 2% and
10% probability of occurrence in fifty years traditionally used in the codes. Moreover,
within the context of probabilistic seismic hazard assessment, the estimate of the level of
uncertainty of a ground motion is as important as the prediction of its median values.
Attenuation relationships vary mainly from the database used in their development, the
differing definitions of distance, and the assumptions made in the statistical analysis. A
complete review of current attenuation relationships is presented by Abrahamson and
Shedlock (1997). Most attenuation relationships include a 'site factor' that accounts for
site amplification effects. The following discussion focuses specifically on the way in
which site effects are incorporated into some of the most widely used attenuation models.
A comparison of all the predicted amplification factors is given in Figure 2.3.
Abrahamson and Silva (1997)
The relationship by Abrahamson and Silva (1997) is developed for shallow crustal
earthquakes in active tectonic regions. The site factor (f5) is included as an additive term
to the natural logarithm of the corresponding spectral value. The amplification term is a
function of the expected PGA for rock (PGArock). The form of the equation was also used
by Youngs (1993), but the parameter c5 is included by Abrahamson and Silva (1997):
f5 (PGArock) = a10 + a11 ln (PGArock + c5), (2.1)

18
where a10, a11, and c5 are parameters of the regression analysis. Amplification factors as
a function of period and PGA in the rock are given in Figure 2.4. The site factor f5
decreases with increasing PGA in the rock for spectral periods lower than 1.0 s, however,
it increases with intensity for spectral periods larger than 1.0 s. This is consistent with the
expected shift in site period from nonlinear soil effects (Chapter Three). However, the
rock PGA at which attenuation of accelerations are observed is lower than that suggested
by other researchers (Seed et al. 1991, Dickenson 1994, Chang et al. 1997). Note that
although rock and shallow (< 20m) soils are grouped in a single category, it is likely that
a site with 20 m of soil will have a motion different to that of a rock site (Abrahamson
and Silva 1997). Moreover, Abrahamson and Silva (1997) indicate that the site factor
could also have magnitude dependence.
Boore, Joyner and Fumal
Joyner and Boore (1981, 1982) proposed an attenuation relationship for shallow
crustal earthquakes in active tectonic regions. Further reviews were published by Boore,
Joyner and Fumal (1993, 1994a and b) and are summarized in Boore et al. (1997). These
reviews will be denoted by BJF 1994a and BJF 1994b.
In BJF 1994a the site factor was changed from a constant for each site condition
to a continuous function of average shear wave velocity over the upper 30 m ( Vs ). The
site factor is determined as an additive term to the natural log of the ground motion
parameter, and is given by:
19
SITE FACTOR = bv ln ( Vs /Va), (2.2)
where both bv and Va are regression parameters. The same functional form for
amplification factors was used by Borcherdt (1994). The amplification factors from 'soil'
to 'rock' are given in Figure 2.3. 'Soil' is defined by Vs = 310 m/s and 'Rock' is defined by
Vs = 620 m/s (Boore et al. 1997). The site effect was first obtained as an average site
amplification for USGS Site B (soil with 360 m/s < Vs ≤ 750 m/s) and USGS Site C (180
m/s < Vs < 360 m/s) relative to Site A ( Vs > 750) while accounting for distance and
magnitude dependence. (BJF 1993). Residuals with respect to USGS Site A for
recordings made at sites with shear wave velocity measurements were used in BJF 1994a
to develop parameters for the Vs relationship (Equation 2.2). Boore, Joyner and Fumal
(1997) indicate that a better parameter than Vs is the average shear-wave velocity over a
quarter of the wavelength of the motion for each spectral period. Similarly, they suggest
that a 'depth of the attenuating layer' should be included in the relationships to explain
systematic differences in motion depending on site condition. Site factors are obtained
from stations located at long distances from the fault and are assumed to be independent
of intensity (i.e., the BJF relationship does not account for soil nonlinearity). For short
distances, the data are controlled by soil sites in the Imperial Valley, thus motions for
rock at short distances are obtained from the same factors obtained for long distances.
Therefore, results for rock at short distances from the fault are not considered reliable.
20
Toro, Abrahamson and Schneider (1997)
The attenuation relationship by Toro, Abrahamson, and Schneider (1997) was
developed for shallow crustal earthquakes in stable continental regions. The soil factors
used were developed by Silva et al. (in EPRI 1993) and are shown in Figure 2.3. The
authors recommend that when calculating uncertainty in soil motions, the uncertainty for
rock motions should be used as a conservative estimate of soil-site uncertainty. This is
counter-intuitive if one thinks of the uncertainty of site amplification being applied on top
of the uncertainty of rock motion predictions. However, Abrahamson and Sykora (1993)
show from dense arrays that uncertainty in soil is lower than rock. Toro et al. (1997)
propose that either soil is a homogenizing factor, or uncertainty in bedrock motions is
lower than in outcrop motions. Moreover, the apparent paradox might occur because
uncertainties in soil and rock are mutually dependent. The decrease in uncertainty for
soils is also evident for high intensity motions. Toro et al. (1997) observed from
equivalent-linear analyses a decrease in uncertainty for high intensity motions. This
decrease apparently offsets the increased uncertainty associated with high-strain dynamic
properties of soils.
Youngs et al. (1997)
The relationship of Youngs et al. (1997) was originally developed for subduction
zone earthquakes. The authors initially divided the database into Deep Stiff Soil, Shallow
Stiff Soil, and Rock. The results, however, are only presented for Rock and Deep Stiff
Soils. In an initial analysis, the ratio of soil to rock PGA increased as PGA increased,
which contradicts intuitive soil behavior. To correct for this apparent error, the
21
relationship was constrained so that soil and rock PGA are equal for small distances to
compensate for sparse soil data at small distances.
The attenuation relationship is calculated separately for soil and rock. The site
effect can be reduced to a series of additive terms to the constant term, the magnitude
scaling term (both linear and cubic), and the distance scaling term in the rock attenuation
relationship. The stiff soil to intermediate rock factors shown in Figure 2.3 are in general
greater than factors derived from other attenuation relationships (note that the Atkinson
and Boore (1995) and the EPRI (1993) site factors are for stiff soil relative to hard rock).
Campbell (1997)
The Campbell (1997) ground motion attenuation relationship was developed for
shallow earthquakes in active tectonic regions. The database is divided into Hard Rock,
Soft Rock and Firm Soil categories. The baseline attenuation relationship is developed
for Firm Soil, with factors for Hard Rock and Soft Rock. The relationship includes a
depth-to-basement term that defines a depth to crystalline basement. The attenuation
relationship is developed for PGA and normalized spectral periods. The depth-to-
basement term affects only the normalized spectra.
Site factors are given as functions of distance in the attenuation relationship for
PGA. For normalized spectral periods, the site factor is added as a parameter that varies
with period but not with distance. When the depth-to-basement term is ignored, the site
amplification with respect to hard rock is independent of period, that is, the long period
site amplification is a direct function of depth-to-basement. The ratio of ground motions
22
in the Firm Soil to the Soft Rock categories is included in Figure 2.3 for a depth-to-
basement of 1 km.
Sadigh et al. (1997)
The attenuation model by Sadigh et al. (1997) is developed for shallow crustal
earthquakes in active tectonic regions. The database is divided into rock (bedrock is at
least 1 m from surface) and deep soil (a minimum of 20 m of firm soil). Soft soils are
excluded from the database. Sadigh et al. (1997) indicate that a significant number of
'rock' sites have Vs < 750 m/s. The relationship is regressed independently for PGA and
for normalized spectra. Two different sets of coefficients are given, resulting in both a
PGA and a magnitude dependence in the site amplification factors.
There is a large degree of nonlinearity for PGA, in fact, PGA in soil is lower than
PGA in rock for PGA values in rock greater than about 0.2 g. The same degree of
nonlinearity was inferred by Abrahamson and Silva (1997). In general, amplification
factors are close to those obtained by Abrahamson and Silva (1997).
Spudich et al. (1997): Extensional regimes
The relationship of Spudich et al. (1997) for extensional regimes is a modification
of the attenuation relationship of Boore et. al. (1997). Site factors are constant for each
period. Relatively low amplification levels at long periods are predicted compared to the
other relationships (Figure 2.3).
23
Summary
The treatment of site response in the ground motion attenuation relationships
discussed herein is summarized in Figure 2.3. The site amplification factors predicted by
the different models have relatively large variability; however, all relationships predict
the same trends with varying period, namely, larger site factors for longer periods. The
relationships that include a dependence on intensity of motion (e.g. Abrahamson and
Silva 1997, Sadigh et al. 1997, Campbell 1997, EPRI 1993, Youngs et al. 1997) predict
lower amplifications for more intense ground motions. The variability in the estimates of
site response illustrated in Figure 2.3 highlights the need for further study of seismic site
response.
2.4 SUMMARY
A brief review of the relevant work in the areas of site response, constitutive
modeling of cyclic soil behavior, and treatment of soil response in attenuation
relationships was presented. Additionally, this dissertation feeds from advancements in
finite element technology (e.g., Zienkiewicz and Taylor 1989, Taylor 1998),
computational mechanics (e.g., Simo and Hughes 1998), and seismology (e.g. Somerville
et al. 1997). Research of particular relevance to this work in these areas is presented
when appropriate in the later Chapters of this dissertation.
24
Figure 2.1. PGA amplification in soft soils (from Idriss 1990).
25
Figure 2.2. Spectral shapes for different site conditins (from Seed and Idriss 1983).
26
2.5

2
Atkinson and Boore (1995)
Boore et al. (1997)

Site Factor
1.5
Campbell (1997)
Sadigh et al. (1997)

1
Spudich et al. (1997)
Youngs et al. (1997)

0.5
EPRI (1993)
0
0.01 0.1 1 10
Period (s)
a) Baseline PGA = 0.1 g
2.5
2
Atkinson and Boore (1995)
1.5
Site Factor
Boore et al. (1997)

Campbell (1997)
1 Sadigh et al. (1997)

Spudich et al. (1997)
0.5 EPRI (1993)
0 b) Baseline PGA = 0.4 g

0.01 0.1 1 10
Period (s)
b) Baseline PGA = 0.4 g
Figure 2.3. Site Factors used in some current ground motion attenuation relationships for
estimating spectral acceleration at 5% damping. Factors are from "Rock" to "Soil". In
Campbell (1997), the factor is from "Firm Soil" to "Soft Rock" and the Depth to
Basement term equal to 1 km. In Boore et al. (1997), "Soil" is defined as V s = 310 m/s
and "Rock" as V s = 620 m/s. The EPRI (1993) curves are given for a profile with an
average depth of 76 m. When necessary, the magnitude used to get the PGA at the
baseline site condition was set to 7.0. Amplification factors for Atkinson and Boore
(1995) and EPRI (1993) are with respect to harder Eastern North America rock.
27
2
1.8
1.6
1.4
Factor, f 5
1.2
PGA = 0.1 g
PGA = 0.2 g
1
PGA = 0.3 g
0.8
PGA = 0.4 g
0.6
0.4
0.2
0
0.01 0.1 1 10
Period (s)
Figure 2.4. Site factors for varying levels of PGA from the Abrahamson and Silva (1997)
spectral acceleration (5% damping) attenuation relationship.
28
BLANK
29
CHAPTER THREE
CHARACTERIZATION OF SITE RESPONSE
GENERAL SITE CATEGORIES
3.1 INTRODUCTION
The effect of local site conditions on the amplification of ground motions has long
been recognized (e.g., Seed and Idriss 1982). Recent earthquakes, such as the 1985
Mexico City, 1989 Loma Prieta, 1994 Northridge, and 1995 Kobe earthquakes have
resulted in significant damage associated with amplification effects due to local geologic
conditions (e.g., Seed et al 1987, Chang et al. 1996). While potentially other factors lead
to damage (such as topographic and basin effects, liquefaction, ground failure, or
structural deficiencies), these events emphasize the need to characterize the potential
effect of local soil deposits on the amplification of ground motions.
Extensive studies of seismic site response have been performed over the last thirty
years. Recently, Borcherdt (1994) developed intensity-dependent, short and long period
amplification factors based on the average shear wave velocity measured over the upper
30 m of a site. Concurrently, Seed et al. (1991) developed a geotechnical site
classification system based on shear wave velocity, depth to bedrock, and general
geotechnical descriptions of the soil deposits at a site. Seed et al. (1991) then developed
intensity-dependent site amplification factors to modify the baseline "rock" peak ground
acceleration (PGA) to account for site effects. With this site PGA value and a site-
30
dependent normalized acceleration response spectra, a site-dependent design spectra can
be developed. Work by these researchers along with work by Dobry (Dobry et al. 1994)
has been incorporated into the 1997 Uniform Building Code (UBC) based primarily on the
site classification system and amplification factors developed by Borcherdt (1994). A
shear wave velocity based classification system, however, has two important limitations:
(a) it requires a relatively extensive field investigation, and (b) it overlooks the potential
importance of depth to bedrock as a factor in site response. Recent work completed at
the University of California at Berkeley based on results from the Northridge and Loma
Prieta earthquakes reflects the importance of introducing a measure of depth in a site
classification system (Chang and Bray 1995, Chang et al. 1997). Moreover, the Borcherdt
(1994) site amplification factors are based primarily on observations from the 1989 Loma
Prieta Earthquake, which shows significant nonlinear site response effects; whereas,
observations from the 1994 Northridge Earthquake indicate that site amplification factors
should not decrease as rapidly with increasing ground motion intensity. Hence, the
current code site factors may be unconservative, and this should be re-evaluated using the
extensive Northridge ground motion database.
A probabilistic seismic hazard assessment requires not only an estimation of the
median expected levels of ground motion intensity, but also the standard error associated
with such a median estimation. Current ground motion attenuation relationships provide
this information (e.g., Abrahamson and Silva 1997, Campbell 1997, Sadigh et al. 1997,
Boore et al. 1997). However, most current attenuation relationships have a simplified
classification scheme for site conditions in which all sites are divided into two or three
broad classifications, e.g., rock/shallow soils, deep stiff soils, and soft soils. A notable
31
exception is the attenuation relationship by Boore et al. (1997). In this relationship, the
factor that accounts for site response (site factor) is a continuous function of the average
shear wave velocity measured over the upper 30 m of a site. However, Boore et al.
(1997) ignore the effects of ground motion intensity on the site factor, which contradicts
measured observations of nonlinear site response (e.g., Trifunac and Todorovska 1996).
Conversely, studies involving a more elaborate site classification scheme encompassing
stiffness, depth, and intensity of motion, currently lack an appropriate estimate of the
statistical uncertainty involved (e.g., Seed et al. 1991).
The significant quantity of ground motion data recorded in the 1994 Northridge
and 1989 Loma Prieta earthquakes provides an opportunity to assess and to improve
empirically based predictions of seismic site response. The objective of this work is to
develop site amplification factors that are both intensity-dependent and frequency-
dependent. The site amplification factors will be estimated based on a new proposed site
classification system that includes soil stiffness and soil depth as key parameters. The
uncertainty levels resulting from the proposed classification system will be compared with
those resulting from a simplified "rock vs. soil" classification system and the more
elaborate code-based system which uses average shear wave velocity measured over the
upper 30 m of a site.
3.2 METHODOLOGY
The following three steps constitute the methodology used in the development of
the proposed empirically based site-dependent amplification factors:
32
(1) A site classification scheme was developed with the objective of encompassing the
factors that have the greatest influence on seismic site response. The proposed
scheme utilizes only general geological and geotechnical information, including
depth to bedrock or to a significant impedance contrast. More elaborate
measurements, such as average shear wave velocity ( Vs ), are utilized only as a
guideline and are not essential to the classification system. The classification
scheme will be described in detail in the next section.
(2) Two major recent earthquakes, the Loma Prieta Earthquake of October 17, 1989,
and the Northridge Earthquake of January 17, 1994, were considered in this study.
The strong motion sites that recorded these earthquakes were classified according
to the site classification scheme developed in this study. Distance-dependent
attenuation relationships for 5% damped elastic acceleration response spectra were
developed for each earthquake and for each site condition. For simplicity,
hereinafter, any reference to response spectral values will imply linear elastic
acceleration response spectra at 5% damping.
(3) These attenuation relationships were utilized to develop site-dependent
amplification factors with respect to the baseline site condition, Site Class B,
"California Rock." The site-dependent amplification factors are a function of both
spectral period and intensity of motion. Amplification factors estimated for the
Northridge and Loma Prieta earthquakes were combined to develop
recommendations that can be generalized to other events.
33
3.3 SITE CLASSIFICATION
3.3.1 Classification Scheme
The amplification of ground motions at a nearly level site is significantly affected
by the natural period of the site (Tn = 4H/Vs; where Tn = natural period, H = soil depth,
and Vs = shear wave velocity; i.e., both dynamic stiffness and depth are important). Other
important seismic site response factors are the impedance ratio between surficial and
underlying deposits, the material damping of the surficial deposits, and how these seismic
site response characteristics vary as a function of the intensity of the ground motion, as
well as other factors. To account partially for these factors, a site classification system
should include a measure of the dynamic stiffness of the site and a measure of the depth of
the deposit. Although earlier codes made use of natural period as a means to classify site
conditions (e.g., 1976 UBC), recent codes such as the 1997 UBC disregard the depth of
the soil deposit and use mean shear wave velocity over the upper 30 m as the primary
parameter for site classification.
Both analytical studies and observation of previous earthquakes indicate that depth
is indeed an important parameter affecting seismic site response. Figure 3.1 shows a
measure of building damage as a function of site depth in the Caracas Earthquake of 1967.
Damage is concentrated in buildings whose natural period matches the natural period of
the soil deposit (Seed and Alonso 1974). To illustrate the effect of soil profile depth on
surface ground motions, a one-dimensional wave propagation analysis was performed
using the equivalent-linear program SHAKE91 (Idriss and Sun 1992). A synthetic motion
for an earthquake of moment magnitude 8.0 (Mw = 8.0) on the San Andreas Fault in the
34
San Francisco Bay was used as an input outcropping rock motion for a soil profile with
varying thickness. The input rock motion was modified to match the Abrahamson and
Silva (1997) attenuation relationship for an earthquake of moment magnitude 7.5 (Mw =
7.5) at a distance of 30 km. The soil profile represents a generic stiff clay site. The upper
30 m of the profile was kept constant, while the depth of the profile was varied between
30 m and 150 m (Figure 3.2a). Figure 3.2b shows the resulting surface linear elastic
acceleration response spectra, and Figure 3.2c shows the corresponding spectral
amplification factors. Observe that an increase in depth shifts the fundamental period,
where amplification is most significant, toward higher values. This results in significantly
different surface motions as a function of the depth to bedrock. An increase in depth also
results in a longer travel path for the waves through the soil deposit. This accentuates the
effect of soil material damping, resulting in greater attenuation of high frequency motion.
However, the significantly higher response at longer periods for deep soil deposits is an
important expected result that should be accommodated in a seismic site response
evaluation.
The same input motion was applied to the four profiles illustrated in Figure 3.3a.
The depth to bedrock for the four profiles is kept constant at 30 m. The four different
profiles correspond to a dense sand, a stiff clay, a loose sand, and a soft clay profile. The
shear modulus reduction curves proposed by Iwasaki et al. (1976) were used for the dense
and loose sand, along with the damping curves for sand proposed by Seed and Idriss
(1970). The Vucetic and Dobry (1991) shear modulus reduction and damping curves for
clays with PI = 30 were used for the stiff clays, whereas for the soft clays the shear
modulus reduction and damping curves for Holocene Bay Mud proposed by Sun et al.
35
(1988) were used. Figure 3.3b illustrates the resulting spectral amplification factors.
Observe that the effect of different average dynamic shear wave velocities over the upper
30 m is similar to the effect of changing the depth to bedrock, as observed in Figure 3.2;
that is, the peak spectral amplification factor shifts toward higher periods. Hence, case
records and analytical studies support a site classification scheme that captures both the
important influences of soil stiffness and soil depth on seismic site response and resulting
damage.
Seismic site response is also a function of the intensity of motion due to the
nonlinear stress-strain response of soils. The effect of nonlinearity is largely a function of
soil type (e.g., Vucetic and Dobry 1991). Factors such as cementation and geologic age
may also affect the nonlinear behavior of soils. The effect of soil nonlinearity is two-fold:
(a) the site period shifts toward longer values, as illustrated in the previous example, and
(b) material damping levels in the soils at a site increase. The increased damping levels
result in lower spectral amplifications for all periods. The effect of damping, however, is
more pronounced for high frequency motion. Hence, PGA is more significantly affected
by soil damping. The consequences of the shift toward longer site periods depend on the
soil type and the input motion. For some sites, the site period may be shifted toward
periods containing high-energy input motion, resulting in large spectral amplification
factors with an associated increase in PGA. Conversely, the site period may be shifted to
periods where the energy of the input motion is low, resulting in large spectral
amplification at long periods associated with a decrease of amplification for short periods.
This may result in lower levels of PGA, and possibly even in attenuation of PGA.
36
The site classification system proposed herein is an attempt to encompass the
factors affecting seismic site response while minimizing the amount of data required for
site characterization. The site classification system is based on two primary parameters
and two secondary ones. The primary parameters are:
(1) Type of deposit, i.e., hard rock, competent rock, weathered rock, stiff soil, soft
soil, and potentially liquefiable sand. These general divisions introduce a
measure of stiffness (i.e., average shear wave velocity) to the classification
system. However, a generic description of a site is sufficient for classification,
without the need for measuring shear wave velocity over the upper 30 m.
(2) Depth to bedrock (defined by Vs ≥ 760 m/s) or to a significant impedance
contrast between surficial soil deposits and material with Vs ≈ 760 m/s. .
The secondary parameters are depositional age and soil type. The former divides
soil sites into Holocene or Pleistocene groups, the latter into primarily cohesive or
cohesionless soils. These subdivisions are introduced to capture the anticipated different
nonlinear responses of these soils. Table 3.1 summarizes the site classification scheme.
3.3.2 Site Classification
The list of sites with the corresponding site classification based on the proposed
classification system is given in Appendix A (Tables A-1 and A-2). The sites are also
classified according to the 1997 UBC and the Seed et al. (1991) systems (Tables A-3 and
A-4 in Appendix A). The references used for the classification of each site are also
included in Appendix A. Due to the lack of consistent information for all the sites, the
37
subdivision of Site D into a very deep site sub-category (D3) was omitted. Additionally,
sufficient information was not available to categorize sites by a precise depth to bedrock
parameter, so that a regression analysis could not be performed using this parameter as a
continuous variable.
The references listed in Appendix A were complemented with site visits for some
of the sites where information was incomplete. A list of the visited sites is given in
Appendix B. Note that an important source of information, particularly for sites belonging
to the University of Southern California, was the database of Vucetic and Doroudian
(1995). The shear wave velocity values presented in this database have recently been
challenged (e.g., Wills 1998, Boore and Brown 1998). In light of these observations, the
shear wave velocities for these sites were used, whenever possible, only as a secondary
reference. For those sites where the only data available was those in the Vucetic-
Doroudian database, these shear wave velocity data were used after incorporating the
comments made by Boore and Brown (1998).
3.4 GROUND MOTION DATA
Ground motion data from two recent earthquakes, the 1989 Loma Prieta
Earthquake and the 1994 Northridge Earthquake, were used in this study. The ground
motion recordings were obtained from a database provided by Dr. Walter Silva from
Pacific Engineering and Analysis (personal comm. 1998). The database consists of
computed elastic spectral acceleration values at 5% damping.
38
The ground motion database provided by Dr. Walter Silva was complemented with
four additional motions for the Northridge Earthquake and eleven motions for the Loma
Prieta Earthquake. The baseline corrected motions were obtained from the Internet from
sites supported by the institutions in charge of the instruments (see Appendix A). A total
of 149 and 70 recorded "free-field" ground motions were used from the 1994 Northridge
and 1989 Loma Prieta earthquakes, respectively.
The ground motion recordings used in the study are listed in Appendix A. The
number of recordings is a function of spectral period, because of the acceptable filtering
parameters used in the processing of the data. The response spectral values are only used
if the frequency is greater than 1.25 times the high-pass-corner frequency and less than
1/1.25 times the low pass-corner frequency (Abrahamson and Silva 1997). The
distribution of recorded motions with distance as a function of site type for spectral
periods between 0.055 seconds and 1.0 seconds is given in Figure 3.4 for each earthquake.
The number of recordings as a function of period is given in Figure 3.5 for each
earthquake.
3.5 STATISTICAL ANALYSIS
3.5.1 General
The ground motion sites were divided into the major categories indicated in the
site classification scheme (Table 3.1). A regression analysis was performed to develop
event and site specific attenuation relationships for acceleration response spectral values
39
(5% damping) at selected periods. A basic form of an attenuation relationship was
selected for this study, that is,
ln[Sa] = a + b ln(R + c) + ε (3.1)
where ln[Sa] is the natural logarithm of the spectral acceleration at a specified
period, T, R is the closest distance to the rupture zone (Sadigh et al. 1993), ε is an error
term, and a, b, and c are regression coefficients. Equation (3.1) is used for various
spectral periods, thus the regression coefficients a, b, and c, and the error term ε are
functions of period. This functional form was previously used by Idriss (1994) and
Somerville (personal comm.). Note that this functional form does not capture the effects
of directivity (Somerville et al. 1997) and the hanging-wall effect (Abrahamson and
Somerville 1996). Forward directivity effects result in larger response spectral values at
periods larger than one second for near-fault sites and for the component of motion
perpendicular to the fault orientation. Similarly, the hanging wall effect increases the
motion of sites located directly above the rupture zone. Directivity and hanging-wall
effects introduce a systematic bias for the affected near-fault sites. This bias, however,
generally affects all sites regardless of their type, thus the effect on amplification factors
should not be significant. A list of sites located over the hanging-wall and the footwall for
the Northridge Earthquake is provided in Table 3.2.
The regression coefficients were estimated by means of a maximum likelihood
estimate (Benjamin and Cornell 1970). The error term ε in Equation (3.1) is assumed to
40
be normally distributed with mean zero and standard deviation σ. With this assumption,
maximum likelihood estimates are equivalent to ordinary least squares estimates (Benjamin
and Cornell 1970). The natural logarithm of the likelihood function is thus given by:
    S ai,    
2
   ln  est  
    S   
  ai , meas 
 
N
1 
∑ ln  σ exp − 0.5 (3.2)
i =1  σ  
    
     
 
where σ is the standard deviation, ln(Sai)meas is the natural logarithm of the measured
spectral acceleration and ln(Sai)est is the mean value given by Equation (3.1). Similar to
Equation (3.1), Equation (3.2) is applied at each spectral period. The values of the
coefficients a, b, and c from Equation (3.1) and the standard deviation σ result from the
minimization of the negative of the natural logarithm of the likelihood function (Equation
3.2). The regression analysis was performed using the software JMP (SAS Institute, Inc.
1995). This analysis was performed separately for each earthquake and for selected
periods. In the following sections, the details of the analyses for each event are described.
3.1.2 Northridge Earthquake
The data distribution by site type is shown in Figure 3.4a. Initially, a separation
between sites C1, C2 and C3 (weathered/soft rock, shallow stiff soil, and intermediate
depth stiff soil, respectively) was assumed, but no significant differences were observed in
the resulting attenuation relationships. Consequently, the subdivision of Site C was
ignored in the preliminary analysis. Similarly, differences for deep soil sites based on age
41
and soil type (i.e., Holocene or Pleistocene and primarily cohesive or cohesionless) were
also not considered in the preliminary analysis.
The response of potentially liquefiable sand deposits (Site F) is mainly a function of
whether or not liquefaction is triggered or partially triggered (i.e., significant pore pressure
generation develops) at the site. Triggering of liquefaction is a function of the intensity
and duration of ground motion, the relative density of the soil, the permeability of the soil,
the fines content of the soil, as well as other factors. If liquefaction is triggered or nearly
triggered, ground motion is a function of a number of parameters, including rate of excess
pore pressure generation, dissipation of pore pressure, reduction of effective stress, shear
modulus degradation, duration of motion, as well as other factors. The analysis of these
sites is beyond the scope of this project; thus, ground motion sites that are classified as
Site F will be excluded from the analysis.
Most of the ground motion sites are concentrated between 20 and 70 km of the
zone of energy release (Figure 3.4a). Accordingly, the resulting attenuation relationships
are judged to be appropriate for sites located within this distance range from an active
fault. Of all the sites located closer than 20 km from the rupture plane, most sites are C
and D sites, and only one Site B (California rock) is located within 20 km.
Equation (3.1) is defined for all distance values only if the coefficient c is non-
negative. Accordingly, this coefficient was assumed to be non-negative for all periods.
Moreover, initial analyses yielded a large standard deviation for the coefficient c, implying
that changes in this coefficient did not result in an increase of the overall standard
deviation in Equation (3.1). The coefficient c was held constant across site conditions to
42
avoid the coupling of uncertainty in the coefficient c with uncertainty in the amplification
factors. This is consistent with a number of previous studies (e.g., Somerville, personal
comm.; Abrahamson and Silva 1997).
Preliminary results yielded spectral accelerations at long periods larger at rock sites
(Site B) than at soil sites (Sites C and D) for distances greater than 70 km, a result that
contradicts both previous analyses (Abrahamson and Silva 1997, Borcherdt 1994) and
theoretical considerations (Dobry et al. 1997). This result is thought to be primarily a
consequence of the poor sampling for Site B across all distances. Data for Site B are
concentrated within a distance range of 20 to 40 km (See Figure 3.4a), thus there is
limited data to constrain adequately all of the coefficients in Equation (3.1). The approach
taken was to assume that the coefficient a is equal for both B and C sites.
In summary, the regression analysis for the Northridge Earthquake proceeded in three
steps:
(1) The value of coefficient c was determined using the whole data set (Sites B, C
and D). The values of the coefficient c obtained in this manner for different
spectral periods were fitted to a piece-wise linear function.
(2) Using the values obtained in step 1 for the coefficient c, the coefficient a for
site types B and C was obtained using the data for these two site types.
(3) Finally, the remaining coefficients were determined using the data set for each
site type separately.
43
3.1.3 Loma Prieta Earthquake
The distribution of data by site type was given in Figure 3.4b. Due to the limited
quantity of data, the distinction between C1, C2, and C3 sites (i.e., weathered/soft rock,
shallow stiff soil, and intermediate depth soil, respectively) and the difference in age or soil
type (i.e., Holocene or Pleistocene and primarily cohesive or cohesionless sites) were not
considered in the preliminary stage of the analysis. Similar to the Northridge Earthquake,
potentially liquefiable sand and peat deposits (Site F) were excluded from the analysis.
Separate regression analyses were performed for sites B, C, D, and E. Most of the data
are concentrated within a distance range of 10 km to 90 km from the zone of energy
release. Moreover, there are only two sites located within 10 km of the fault rupture
plane, implying that the resulting attenuation relationships are poorly constrained for close
distances to the zone of energy release. Consequently, the results presented in this report
will not reflect the localized effects of near-fault ground motions on seismic site response.
Numerical simulations will be required to provide insight into the near-fault seismic
response of soil sites, and this is the objective of another ongoing research project by the
authors.
Only three rock sites (Site B) are located within 20 km of the zone of energy
release. This poor sampling implies that the attenuation relationship is not well
constrained for short distances. This is especially important since Site B is taken as the
baseline site for developing amplification factors. The low number of soft clay sites (7
sites) and the poor distribution with distance (see Figure 3.4) results in a poorly
constrained attenuation relationship for this site class. Previous studies (e.g., Seed et al.
44
1991) complemented this lack of empirical data with numerical simulations. Since the
objective of this work was the development of empirically based amplification factors, soft
clay sites were excluded from the analysis.
As previously indicated for the analysis of the Northridge data, the coefficient c in
Equation (3.1) was constrained to be non-negative. Unconstrained regression analysis
yielded a negative value of c for all periods. Consequently, the coefficient c was set to 1
for all periods. As a result of the better sampling in the Loma Prieta data set, there was no
need to constrain the parameter a in the analysis.
3.1.4 Results
The coefficients a, b, and c found for each earthquake were smoothed by a
convolution with a triangular function with a window-width of three. The convolution
was repeated until no further improvement was obtained. The smoothed coefficients are
illustrated in Figures 3.6 and 3.7 and listed in Table 3.3. A comparison between the
resulting smoothed and non-smoothed spectra is shown in Figure 3.8 for the Northridge
earthquake. The resulting attenuation relationships are illustrated in Figures 3.9 and 3.10
for selected distances. Spectral acceleration values as a function distance for selected
periods are shown in Figures 3.11 and 3.12.
45
3.6 EVALUATION OF RESULTS
3.6.1 General
The attenuation relationships obtained using the classification system introduced in
this work are compared with results from a simplified "Rock vs. Soil" classification
system, as well as with a more elaborate code-based classification system (1997 UBC).
Residuals for sites C and D were evaluated to judge whether a further subdivision is
justified.
3.6.2 Comparison With a "Soil vs. Rock" Classification System
Most current attenuation relationships use a broad and general site classification,
dividing sites in either rock/shallow soil or deep stiff soil, in addition to deep soft clay sites
(e.g., Abrahamson and Silva 1997). This classification is also often applied in design
practice (Abrahamson, personal comm.). Results from this study, however, show that this
classification is an oversimplification, and further division into additional categories is
warranted.
As a basis for comparison, the earthquake specific attenuation model developed by
Somerville and Abrahamson (Somerville, personal comm.) will be compared with the
model developed in this study. The Somerville and Abrahamson model will be denoted as
S&A. This model divides sites into rock/shallow soil (rock) and deep stiff soils (soil).
Deep soft clay sites are excluded. Figure 3.13 shows a comparison of the results at a
distance of 20 km. Note that the spectra for soil sites in S&A generally match the spectra
for Site D (deep stiff soils). However, the spectra for rock sites in S&A generally match
46
the spectra for Site C (shallow and intermediate depth soils and weathered/soft rock).
This result reflects the fact that for the joint database of rock and shallow soil sites, 83%
of the sites are shallow soil or weathered rock sites, and only 17% of these sites actually
belong to the Site B classification (competent rock sites). Note that the spectrum for Site
B falls significantly below that for Site C (20% to 40% lower, 30% on average). Similar
results were obtained independently by Idriss and Silva (personal comm. 1999). These
authors performed an analysis of strong motions at competent rock and weathered rock
sites and concluded that response spectra at weathered rock sites is 20% lower on average
than response spectra at competent rock sites.
Attenuation relationships are commonly used in engineering practice to predict ground
motions at baseline "outcropping rock" sites. These baseline motions are often later
modified for local soil conditions. The results presented herein point to the need of
redefining the baseline "outcropping rock" condition into a more restrictive category.
Many "rock" attenuation relationships used in practice actually correspond to a mix site
condition that is dominated by shallow stiff soil and soft/weathered rock records due to
the preponderance of these sites in the ground motion database. Hence, most "rock"
attenuation relationships reflect the amplified response of soft/weathered rock and shallow
stiff soil sites. The acceleration response spectrum at 5% damping for the true baseline
rock condition (California rock) is consistently overestimated by that developed using
currently available "rock attenuation relationships. Thus, updated "rock" attenuation
relationships that use a more restrictive baseline rock category are required.
47
A significant difference in response spectra was observed between the proposed
site categories (see Figures 3.9 and 3.10). Again, Site B (California rock) data plot
significantly below that for Site C (weathered rock/shallow stiff soil), which illustrates that
a further subdivision from the 'rock' vs. 'soil' classification is warranted. More significant,
however, is the reduction of uncertainty that results from the proposed classification
system. Table 3.4 compares the standard deviations from the S&A relationships with
those from the relationships proposed in this report. The decrease in the standard
deviation for Site B compared with S&A rock sites is between 30% and 40%. A similar
reduction is observed for soil sites (S&A Soil vs. Site D). Standard deviations for Site C,
however, remain high and are only marginally lower than standard deviations for rock in
the S&A model. A reduction in the uncertainty bands for sites B and D reflects the more
selective grouping criteria applied in this study.
3.6.3 Comparison With a Code-Based Site Classification System
The data set for both earthquakes was also divided according to the 1997 UBC
(i.e., using the average shear wave velocity measured over the upper 30 m of the site).
The UBC classification system is presented in Table 3.4 in Appendix A. Differences in the
classification of ground motion sites using both systems are shown in Table 3.5. For
simplicity, sites classified according to the system presented in this work (Table 3.1) will
simply be denoted by Site X, while sites classified according to the UBC system will be
denoted UBC X. Note that in the Northridge database, there is a significant number of C
sites that correspond to either UBC B or UBC D sites. The former are weathered rock
sites lying on top of harder, intact rock (such as Lake Hughes #9), and the latter are either
48
shallow soil or weathered rock sites with depth to bedrock ranging from 25 to 60 m.
There are also five D sites in Northridge and one in Loma Prieta that classify as UBC C
sites. These sites correspond to stiff clay or sand deposits with shear wave velocities only
slightly larger than the boundary values determined by the UBC classification system (such
as Sepulveda VA Hospital). This overlap results in different attenuation relationships
depending on the classification system. Because shear wave velocity measurements were
not taken at all ground motion stations used in this study, the classification of sites
according to the scheme presented in this work probably is more accurate than the
classification of sites according to their average shear wave velocity value. The same
finding carries over to the results of the regression analyses.
Table 3.6 compares the standard deviations at selected periods resulting from the
regression analysis using both classification systems. For the Loma Prieta Earthquake,
standard deviations for both classification systems are comparable. This is expected
because there is little overlap between classification systems for the Loma Prieta database
(Table 3.5). For the Northridge Earthquake, standard deviations vary slightly from one
classification system to the other. Standard deviations for Site B are slightly lower for the
proposed classification system. For Sites C and D, standard deviations are equal for a
period of 0.3 seconds, but vary slightly at a period of one second. With the exception of
Site D at a period of one second, the differences of the standard deviations resulting from
both classification systems are within the ranges of the estimates. Given that the spectral
amplification factors change significantly with depth at a period close to one second
(Figure 3.2), the exclusion of sites shallower than 60 m from Site D in the proposed
classification system result in a reduction of the scatter in the data.

49
3.6.4 Subdivision of Site C
For the Northridge Earthquake, the standard deviations of Site C at long periods
are larger than those of Sites B and D. This observation motivated a closer examination
of the results for Site C. Figure 3.14 shows the residuals for Site C for the Northridge
Earthquake, along with the mean value of the residuals. Each site is identified by its
corresponding UBC classification. Note that well-defined trends are observed for periods
larger than 0.1 seconds. UBC D sites plot significantly above the median while UBC C
sites plot below the median, illustrating that a further subdivision for Site C according to
shear wave velocity may be warranted. Similarly, these results demonstrate that whereas
for deep stiff soil sites and rock sites the additional expense of a shear wave velocity
characterization may not be justified, for intermediate depth soil sites characterization
using average shear wave velocity may reduce the uncertainty in the prediction of ground
motions.
A subdivision of Site C as indicated in Table 3.1 was also studied. Residuals for
Site C are plotted in Figure 3.15a for the Northridge Earthquake. Observe that no specific
trends for sites C1, C2, and C3 are observed, as opposed to the trend observed when C
sites were divided according to an average shear wave velocity-based classification
system. This observation implies that for shallow and intermediate depth soils, the
average shear wave velocity may be the discriminating additional factor.
50
3.6.5 Subdivision of Site D
A further subdivision for deep soil sites (Site D) according to age and soil type is
also studied. As shown in the classification system (Table 3.1), Site D is subdivided as
either Holocene or Pleistocene, or as primarily clayey or sandy. Figure 3.16a shows the
residuals for sites D for the Northridge Earthquake. Mean residuals consistently greater
than zero are observed for clay sites at all periods. These mean residuals are considered
important, however, not overly significant when compared with the standard deviation for
the entire distribution of around 0.4. This trend is magnified when only Pleistocene sites
(D2) are considered. However, since the number of such sites is low, further studies are
needed to confirm this trend. No apparent trend based solely on the age of the deposit is
observed. The same trends are observed in the Loma Prieta Earthquake (Figure 3.16b),
but the small number of sites precludes any definite finding in this regard.
In general, it appears that greater amplification can occur at clay sites, especially if
Pleistocene, and this is consistent with the concept that higher plasticity soils have higher
threshold strains and hence exhibit less shear modulus reduction and less material damping
at intermediate levels of ground motion. However, until additional ground motion and site
classification data are obtained, the limited number of sites and records, and the level of
scatter associated with Site D, precludes further subdivision at this time.
3.6.6 Effect of Depth to Basement Rock
In an effort to assess the ability of a depth to basement rock term to capture
seismic site effects, sites within the Los Angeles basin were investigated. The depth to
51
basement bedrock was obtained from a map by Conrey (1967), and is defined as the depth
to the top of Pliocene bedrock. All of the selected sites are classified as Site D, with the
exception of the USC 54 site (LA Centinela), which is a C3 site. Residuals for periods of
one and two seconds as a function of depth to bedrock are plotted in Figure 3.17 for sites
located in the Los Angeles basin. No trend is observed for sites shallower than 2000 m,
but residuals are higher than zero for sites deeper than 2000 m. Positive residuals may be
due to basin effects rather than to local site amplification.
3.6.7 Amplification Factors
The attenuation relationships developed in this work are event-specific relations
that cannot be generalized to other events. To extend the applicability of the results
presented in this work, amplification factors with respect to a baseline site condition were
obtained. The baseline site condition was taken to be rock (Site B). However, since
current "rock" attenuation relationships reflect mostly site condition C, factors with
respect to Site C will also be presented. The amplification factors are obtained by dividing
Equation (3.1) for the two site conditions. The resulting relationship can be written as
ln (FC / B ) = a C / B + bC / B ln( R + c ) (3.3)
where FC/B is the spectral acceleration amplification factor for Site C with respect
to Site B, R is the closest distance to the rupture plane, and aC/B and bC/B are coefficients
defined by:
52
aC/B = a(Site C) - a(Site B) (3.4a)
bC/B = b(Site C) - b(Site B) (3.4b)
The regression coefficient a was assumed to be independent of site conditions and
therefore no site-dependent ratio is necessary for this coefficient. These two coefficients
were smoothed for all periods. The smoothed coefficients are listed in Appendix C. The
resulting amplification factors are shown in Figure 3.18 for each earthquake. For the
Loma Prieta Earthquake, a reduction in spectral amplification factors for increasing levels
of base rock motion is observed for periods shorter than one second. This trend is
consistent with nonlinear soil behavior. At periods greater than one second, spectral
amplification values do not necessarily decrease with increasing levels of base rock
motion, as soil response nonlinearity would also tend to increase the response at larger
periods as the site softened. Other issues may have affected the data in this period range,
such as basin effects and surface waves. In addition, rather than a reflection of soil
response, these observations may be a result of the significant scatter of the data at long
periods. Moreover, for high values of PGA, the attenuation relationships are not well
constrained due to the lack of near-fault data for the Loma Prieta Earthquake.
Amplification factors from the Northridge Earthquake do not show the same
degree of nonlinearity, as do the results from Loma Prieta. Because the current UBC is
based mainly on observational data from the Loma Prieta Earthquake (e.g., Borcherdt
53
1994), amplification factors presented in the UBC may lead to underestimation of high
amplitude ground motions.
3.6.8 Recommended Factors
The spectral amplification factors from each earthquake were combined to develop
a set of recommended amplification factors. The factors were combined at equal PGA
values. Note that because event-specific attenuation relationships are used for each
earthquake, the relationship between PGA and distance is not unique for both earthquakes.
Two different weighting schemes were utilized. One weighting scheme gives equal weight
to each earthquake, while the other gives a weight inversely proportional to the variance
of the sample mean. The equations and coefficients used to determine the amplification
factors are given in Appendix C. The resulting amplification factors are shown in Figures
3.19 and 3.20, and are given in Tables 3.7 and 3.8. The standard deviations for each site
condition were averaged using the same weighting schemes, and are also presented in
Tables 3.7 and 3.8.
For long periods (T > 1.0 s) the difference in amplification factors between
earthquakes is significantly smaller than the difference in amplification factors between site
type. For shorter periods, however, differences between earthquakes are comparable to
differences due to site type.
Amplification factors with respect to Site B (Figures 3.19a and 3.20a) show a
significant degree of nonlinearity. On the other hand, spectral amplification factors from
Site D to Site C are nearly linear, mainly because of the linearity observed in the
Northridge data (Figure 3.18b). This effect is increased when weighting factors inversely
54
proportional to sample variance are applied (Figure 3.20b) as a result of the larger number
of Site C and Site D data points in the Northridge Earthquake (see Figure 3.21 for the
weights for each earthquake).
A comparison of Figures 3.20a and 3.20b illustrates the dramatic difference in
spectral amplification factors that results from taking either rock (Site B) or weathered-
soft rock/shallow stiff soil (Site C) as the baseline site condition. Current practice takes an
intermediate site condition as reference. The large differences in behavior between these
site conditions illustrated in this work serve to highlight the need to define a unique
baseline site condition.
For the sake of comparison with current code provisions, the spectral amplification
factors were averaged over a range of periods to obtain short-period and mid-period
amplification factors. The period range for the short-period amplification factor (Fa) is 0.1
to 0.5 seconds, and the period range for the mid-period amplification factor (Fv) is 0.4 to
2.0 seconds (Borcherdt 1994). The mean factors were averaged from a double
logarithmic plot of amplification factors versus period. The values of the code factors
(UBC 1997) and the factors obtained in this work are given in Table 3.9, and are
presented graphically in Figure 3.22.
The short-period amplification factors (Fa) obtained in this work are larger than
the code values. This is due in large part to the larger levels of motion observed in the
Northridge earthquake, which was not included in the studies that led to the adoption of
the 1997 UBC factors. Additionally, the site classification scheme adopted for the 1997
UBC differs from that proposed in this study, so that some sites are classified differently
55
(see Section 3.6.2 and Table 3.5). In addition, the intermediate-period amplification
factors from the proposed site classification show less nonlinearity than the code factors.
This is consistent with "soil" to "rock" amplification factors obtained by Abrahamson and
Silva (1997) using a much broader database. Overall, the differences in the amplification
factors are generally less than 25%, which is acceptable for design. However, as
additional data become available, these two site classification systems warrant further
comparisons.
3.7 SUMMARY
3.7.1 Findings
The strong ground motion data from the Loma Prieta and Northridge earthquakes
were analyzed and used to evaluate a proposed new site classification scheme developed
to account for site effects in probabilistic seismic hazard assessments. The proposed
classification system is based on a general geotechnical characterization of the site that
includes soil depth and stiffness.
The proposed classification scheme results in a significant reduction in standard
error when compared with a simpler "rock vs. soil" attenuation relationship approach.
Additionally, the generic "outcropping rock" category used in many current attenuation
relationships groups competent rock and weathered soft rock/shallow stiff soils. The
results shown herein indicate a significant difference in the seismic responses of these two
site classes, witch competent rock spectral ordinates being about 30% lower than the
corresponding spectral ordinates for weathered soft rock/shallow stiff soil sites for the
56
same magnitude and distance. Moreover, "rock" attenuation relationships used in practice
are dominated by the weathered soft rock/shallow stiff soil site category due to the
preponderance of these sites in the database. The results presented herein point to the
need of redefining the baseline "outcropping rock" condition into a more restrictive site
category.
The standard errors resulting from the proposed classification system are
comparable to the standard errors obtained using the 1997 UBC shear wave velocity-
based classification system. However, for profiles within the same UBC classification (i.e.
equal average shear wave velocity), the inclusion of soil depth leads to a significant
reduction of uncertainty. Conversely, for sites with a shallow soil depth (i.e. < 60 m),
refinement of the estimate of profile stiffness in the form of average shear wave velocity
reduces uncertainty levels. For deep stiff soil sites, measurement of the shear wave
velocity over the upper 30 m of the site does not significantly reduce uncertainty in the
prediction of seismic site response. These observations indicate that a classification
scheme including both soil stiffness and depth leads to better estimates of ground motion.
The relative merits of subdividing deep stiff clay sites (Site D) by soil type (e.g.,
either sand or clay) and depositional age (e.g., either Holocene or Pleistocene) was also
evaluated. Results indicate a trend of higher response for clay sites, especially if
Pleistocene. This is consistent with the concept that higher plasticity soils have higher
threshold strains and hence, exhibit less shear modulus reduction and less material
damping at intermediate levels of ground motion.
57
Spectral amplification factors with respect to competent bedrock (PEER 1998)
and Site C (i.e., "rock" classification for most attenuation relationships, e.g., Abrahamson
and Silva 1997) are presented. The intensity-dependent, period-dependent spectral
acceleration amplification factors can be obtained using the formulas given in Appendix C.
Overall, the spectral amplification factors do not vary significantly from the factors
presented in the 1997 UBC; however, the difference is sufficient to warrant further
reevaluation of the code factors. Additionally, the spectral amplification factors presented
herein can be used in probabilistic seismic hazard assessments, because, unlike the code
site factors, the proposed sit amplification factors include quantification of the underlying
uncertainty in the site dependent ground motion estimate. However, caution should be
exercised when using these factors because they are obtained from a data set containing
only two earthquake. Hence, intra-event scatter could be assessed for these two
earthquakes, but inter-event scatter could not be evaluated satisfactorily. Moreover, due
to the scarcity of the data, results are not well defined for near-fault conditions. This issue
will be addressed in detail in Chapter Four.
3.7.2 Recommendations
Based on the results of this analysis of the Loma Prieta and Northridge earthquake
ground motion databases, it is judged that the site-dependent, period-dependent
amplification factors given in Tables 3.7 and 3.8 and in equation form in Appendix C, can
be used in general probability seismic hazard assessments. However, caution should be
exercised when using these factors, because they are obtained from a data set containing
only two earthquakes. Hence, intra-event scatter could be assessed for these two
58
earthquakes, but inter-event scatter could not be evaluated satisfactorily based on only
two earthquake events. Moreover, due to the scarcity of the data, results are not well
defined at high acceleration levels. Amplification factors are given with respect to site
condition B (rock) and Site C (i.e., "rock" classification for most attenuation relationships,
e.g., Abrahamson and Silva 1997). Current attenuation relationships (i.e., Abrahamson
and Silva 1997) and probabilistic maps use an intermediate baseline site condition of B-C.
Due to the relative scarcity of data of B sites relative to C sites, their “rock” sites are more
closely analogous to Site C. This should be taken into consideration when applying the
recommended amplification factors to current attenuation relationship and probabilistic
map values.
The results of this study strongly support the development of an attenuation
relationship based on the proposed site classification scheme. With this new relationship,
spectral acceleration values for a site could be estimated directly without the use of
amplification factors. A better estimate of the uncertainty involved in ground motion
estimation could be made with this direct approach, rather than the approach applied
herein that required ratios of spectral ordinates.
59
Table 3.1 Geotechnical Site Categories (after Bray and Rodríguez-Marek 1997)
Site Categories
A Hard Rock
B Rock
C Weathered/Soft Rock or Shallow Stiff Soil
D Deep Stiff Soil
E Soft Clay
F Special, e.g., Liquefiable Sand
Site Description Site Comments

Period
A Hard Rock ≤ 0.1 s Hard, strong, intact rock (Vs ≥ 1500 m/s)
B Rock ≤ 0.2 s Most "unweathered" California rock cases
(Vs ≥ 760 m/s or < 6 m. of soil)
C-1 Weathered/Soft Rock ≤ 0.4 s Vs ≈ 360 m/s increasing to > 700 m/s,
weathered zone > 6 m and < 30 m
-2 Shallow Stiff Soil ≤ 0.5 s Soil depth > 6 m and < 30 m
-3 Intermediate Depth Stiff ≤ 0.8 s Soil depth > 30 m and < 60 m
Soil
D-1 Deep Stiff Holocene Soil, ≤ 1.4 s Soil depth > 60 m and < 200 m. Sand has
either S (Sand) or C low fines content (< 15%) or non-plastic
(Clay) fines (PI < 5). Clay has high fines content
(> 15%) and plastic fines (PI > 5).
-2 Deep Stiff Pleistocene ≤ 1.4 s Soil depth > 60 m and < 200 m. See D1
Soil, S (Sand) or C for S or C sub-categorization.
(Clay)
-3 Very Deep Stiff Soil ≤2s Soil depth > 200 m
E-1 Medium Depth Soft Clay ≤ 0.7 s Thickness of soft clay layer 3 m to 12 m
-2 Deep Soft Clay Layer ≤ 1.4 s Thickness of soft clay layer > 12 m
F Special, e.g., Potentially ≈1s Holocene loose sand with high water table
Liquefiable Sand or Peat (zw ≤ 6 m) or organic peat.
60
Table 3.2 Sites located on the footwall (FW) and hanging-wall (HW) in the Northridge
Earthquake (adapted from Abrahamson and Somerville 1996).
Organization Station # Location Classification

CDMG 24047 FW B
CDMG 24207 FW B
CDMG 24279 FW C3
CDMG 24469 FW B
CDMG 24514 FW D1C
CDMG 24575 FW C2
CDMG 24607 FW C1
USC 90057 FW D1S
USGS 127 FW C1
CDMG 24396 HW C1
DWP 75 HW D1S
DWP 77 HW C2
USC 90003 HW D1C
USC 90049 HW C2
USC 90053 HW C3
USC 90055 HW C2
USGS 637 HW D2C
USGS 655 HW F
USGS 5080 HW B
USGS 5081 HW C2
USGS 5108 HW C1
61
Table 3.3a Regression coefficients and Standard Error for spectral acceleration values at
5% damping for the Northridge Earthquake
B Sites C Sites D Sites

T a b c σ a b c σ a b c σ
PGA 2.3718 -1.2753 6.3883 0.3209 2.3718 -1.1538 6.3883 0.4686 2.6916 -1.2161 6.3883 0.3559
0.055 3.5192 -1.4829 10.2486 0.4343 3.5192 -1.3869 10.2486 0.4661 3.5126 -1.3703 10.2486 0.3560
0.06 3.7423 -1.5138 11.8103 0.4343 3.7423 -1.4266 11.8103 0.4655 3.7970 -1.4257 11.8103 0.3654
0.07 4.3982 -1.6291 14.5768 0.4310 4.3982 -1.5480 14.5768 0.4636 4.4475 -1.5472 14.5768 0.3705
0.08 4.8097 -1.7006 16.9734 0.4180 4.8097 -1.6152 16.9734 0.4619 4.9774 -1.6422 16.9734 0.3754
0.09 4.9993 -1.7175 18.0000 0.3935 4.9993 -1.6366 18.0000 0.4617 5.2637 -1.6826 18.0000 0.3779
0.1 4.9768 -1.6855 18.0000 0.3615 4.9768 -1.6089 18.0000 0.4642 5.3000 -1.6679 18.0000 0.3774
0.11 4.9365 -1.6614 18.0000 0.3457 4.9365 -1.5844 18.0000 0.4667 5.2529 -1.6439 18.0000 0.3766
0.12 4.8748 -1.6330 18.0000 0.3322 4.8748 -1.5530 18.0000 0.4703 5.1563 -1.6072 18.0000 0.3759
0.13 4.7753 -1.5991 18.0000 0.3226 4.7753 -1.5140 18.0000 0.4750 5.0044 -1.5586 18.0000 0.3758
0.14 4.6161 -1.5564 17.3303 0.3179 4.6161 -1.4646 17.3303 0.4808 4.7947 -1.4991 17.3303 0.3766
0.15 4.3937 -1.5041 16.0757 0.3182 4.3937 -1.4037 16.0757 0.4877 4.5454 -1.4330 16.0757 0.3786
0.16 4.1376 -1.4471 14.9021 0.3232 4.1376 -1.3364 14.9021 0.4952 4.2958 -1.3685 14.9021 0.3820
0.17 3.8807 -1.3907 13.7997 0.3315 3.8807 -1.2694 13.7997 0.5030 4.0778 -1.3133 13.7997 0.3865
0.18 3.7555 -1.3635 12.7603 0.3368 3.7555 -1.2373 12.7603 0.5069 3.9820 -1.2900 12.7603 0.3893
0.19 3.6370 -1.3378 11.7771 0.3418 3.6370 -1.2069 11.7771 0.5105 3.8913 -1.2680 11.7771 0.3918
0.2 3.4048 -1.2891 10.8444 0.3531 3.4048 -1.1508 10.8444 0.5174 3.7044 -1.2249 10.8444 0.3974
0.24 2.9146 -1.1904 7.5290 0.3759 2.9146 -1.0449 7.5290 0.5285 3.2196 -1.1160 7.5290 0.4071
0.28 2.6754 -1.1429 5.8000 0.3872 2.6754 -0.9965 5.8000 0.5330 2.9725 -1.0610 5.8000 0.4106
0.3 2.5178 -1.1149 4.9000 0.3983 2.5178 -0.9682 4.9000 0.5372 2.8087 -1.0250 4.9000 0.4129
0.34 2.4645 -1.1197 4.4254 0.4176 2.4645 -0.9768 4.4254 0.5463 2.7212 -1.0067 4.4254 0.4145
0.36 2.4594 -1.1242 4.3606 0.4242 2.4594 -0.9870 4.3606 0.5515 2.6916 -0.9999 4.3606 0.4142
0.4 2.4375 -1.1239 4.2415 0.4276 2.4375 -0.9935 4.2415 0.5570 2.6466 -0.9915 4.2415 0.4133
0.44 2.4279 -1.1279 4.1337 0.4277 2.4279 -1.0049 4.1337 0.5627 2.6269 -0.9946 4.1337 0.4119
0.5 2.4692 -1.1545 3.9890 0.4198 2.4692 -1.0526 3.9890 0.5739 2.7651 -1.0629 3.9890 0.4066
0.55 2.4447 -1.1582 3.8812 0.4140 2.4447 -1.0682 3.8812 0.5792 2.8613 -1.1091 3.8812 0.4023
0.6 2.3687 -1.1540 3.7828 0.4090 2.3687 -1.0710 3.7828 0.5843 2.9263 -1.1469 3.7828 0.3968
0.667 2.2699 -1.1513 3.6630 0.4060 2.2699 -1.0675 3.6630 0.5892 2.9650 -1.1752 3.6630 0.3901
0.7 2.1804 -1.1550 3.6084 0.4059 2.1804 -1.0660 3.6084 0.5937 2.9956 -1.1995 3.6084 0.3826
0.75 2.1276 -1.1664 3.5303 0.4090 2.1276 -1.0746 3.5303 0.5977 3.0096 -1.2199 3.5303 0.3750
0.8 2.1239 -1.1848 3.4573 0.4151 2.1239 -1.0966 3.4573 0.6009 2.9754 -1.2294 3.4573 0.3680
0.85 2.1516 -1.2064 3.3887 0.4235 2.1516 -1.1267 3.3887 0.6030 2.8866 -1.2261 3.3887 0.3621
0.9 2.1703 -1.2244 3.4413 0.4332 2.1703 -1.1539 3.4413 0.6041 2.7784 -1.2185 3.4413 0.3579
0.95 2.1451 -1.2353 3.2629 0.4435 2.1451 -1.1701 3.2629 0.6041 2.6965 -1.2187 3.2629 0.3556
1.0 2.0734 -1.2443 3.2048 0.4538 2.0734 -1.1775 3.2048 0.6033 2.6601 -1.2333 3.2048 0.3551
1.1 1.9888 -1.2635 3.0970 0.4637 1.9888 -1.1873 3.0970 0.6017 2.6461 -1.2583 3.0970 0.3563
1.2 1.9252 -1.2983 2.9986 0.4726 1.9252 -1.2071 2.9986 0.5995 2.6099 -1.2804 2.9986 0.3587
1.3 1.8811 -1.3390 2.9080 0.4799 1.8811 -1.2317 2.9080 0.5962 2.5295 -1.2884 2.9080 0.3618
1.4 1.8327 -1.3706 2.8242 0.4799 1.8327 -1.2510 2.8242 0.5909 2.4272 -1.2846 2.8242 0.3649
1.5 1.7582 -1.3853 2.7461 0.4799 1.7582 -1.2588 2.7461 0.5850 2.3331 -1.2785 2.7461 0.3664
1.7 1.5420 -1.3800 2.6045 0.4799 1.5420 -1.2565 2.6045 0.5800 2.1862 -1.2817 2.6045 0.3811
2.0 1.3896 -1.3970 2.4206 0.4799 1.3896 -1.2933 2.4206 0.5700 2.0500 -1.3154 2.4206 0.4130
2.2 1.2440 -1.3983 2.3128 0.4799 1.2440 -1.3004 2.3128 0.5600 1.8906 -1.3182 2.3128 0.4244
2.6 0.9829 -1.3739 2.1238 0.4799 0.9829 -1.2719 2.1238 0.5400 1.6293 -1.2941 2.1238 0.4145
3.0 0.6859 -1.3338 2.0000 0.4799 0.6859 -1.2207 2.0000 0.5200 1.3413 -1.2536 2.0000 0.3877
62
Table 3.3b Regression coefficients and Standard Error for spectral acceleration values at
5% damping for the Loma Prieta Earthquake

T A b c σ a b c σ a b c σ
PGA 0.7219 -0.7954 1.0000 0.4713 0.8212 -0.7502 1.0000 0.3111 0.5716 -0.6032 1.0000 0.3896
0.055 1.6308 -0.9794 1.0000 0.4566 1.4230 -0.8769 1.0000 0.3708 1.3201 -0.7767 1.0000 0.4334
0.06 1.8207 -1.0119 1.0000 0.4561 1.4804 -0.8841 1.0000 0.3747 1.2568 -0.7489 1.0000 0.4338
0.07 1.9001 -1.0181 1.0000 0.4554 1.4819 -0.8734 1.0000 0.3798 1.2413 -0.7315 1.0000 0.4340
0.08 2.0559 -1.0383 1.0000 0.4538 1.5348 -0.8701 1.0000 0.3886 1.3041 -0.7271 1.0000 0.4331
0.09 2.1619 -1.0489 1.0000 0.4518 1.5875 -0.8642 1.0000 0.3973 1.4037 -0.7300 1.0000 0.4303
0.1 2.2305 -1.0551 1.0000 0.4500 1.6419 -0.8595 1.0000 0.4027 1.5122 -0.7400 1.0000 0.4269
0.11 2.2946 -1.0607 1.0000 0.4481 1.7031 -0.8527 1.0000 0.4074 1.6341 -0.7524 1.0000 0.4220
0.12 2.3215 -1.0625 1.0000 0.4472 1.7361 -0.8492 1.0000 0.4091 1.6890 -0.7575 1.0000 0.4187
0.13 2.3462 -1.0642 1.0000 0.4464 1.7665 -0.8461 1.0000 0.4108 1.7395 -0.7622 1.0000 0.4157
0.14 2.3659 -1.0613 1.0000 0.4451 1.8339 -0.8450 1.0000 0.4124 1.7916 -0.7621 1.0000 0.4084
0.15 2.3410 -1.0484 1.0000 0.4448 1.9079 -0.8523 1.0000 0.4120 1.7724 -0.7458 1.0000 0.4004
0.16 2.2804 -1.0268 1.0000 0.4460 1.9696 -0.8621 1.0000 0.4095 1.7156 -0.7191 1.0000 0.3924
0.17 2.2370 -1.0125 1.0000 0.4476 1.9792 -0.8631 1.0000 0.4071 1.6926 -0.7066 1.0000 0.3888
0.18 2.1960 -0.9991 1.0000 0.4491 1.9882 -0.8640 1.0000 0.4049 1.6710 -0.6949 1.0000 0.3853
0.19 2.0939 -0.9675 1.0000 0.4545 1.9513 -0.8531 1.0000 0.3989 1.6405 -0.6754 1.0000 0.3797
0.2 1.9861 -0.9352 1.0000 0.4626 1.8633 -0.8291 1.0000 0.3923 1.5961 -0.6551 1.0000 0.3761
0.24 1.8523 -0.8933 1.0000 0.4797 1.6772 -0.7749 1.0000 0.3837 1.5154 -0.6295 1.0000 0.3754
0.28 1.8136 -0.8775 1.0000 0.5001 1.5268 -0.7272 1.0000 0.3796 1.5140 -0.6328 1.0000 0.3796
0.3 1.8860 -0.8959 1.0000 0.5149 1.4800 -0.7104 1.0000 0.3812 1.5933 -0.6598 1.0000 0.3859
0.34 1.9996 -0.9271 1.0000 0.5292 1.4613 -0.7033 1.0000 0.3868 1.7028 -0.6968 1.0000 0.3950
0.36 2.0373 -0.9373 1.0000 0.5358 1.4510 -0.7011 1.0000 0.3916 1.7453 -0.7128 1.0000 0.4012
0.4 2.0412 -0.9393 1.0000 0.5524 1.3972 -0.6925 1.0000 0.4092 1.7829 -0.7364 1.0000 0.4219
0.44 1.8966 -0.9057 1.0000 0.5600 1.3081 -0.6785 1.0000 0.4251 1.6708 -0.7148 1.0000 0.4396
0.5 1.5766 -0.8357 1.0000 0.5658 1.0905 -0.6402 1.0000 0.4486 1.3791 -0.6481 1.0000 0.4659
0.55 1.3683 -0.7909 1.0000 0.5678 0.9405 -0.6134 1.0000 0.4616 1.1859 -0.6031 1.0000 0.4808
0.6 1.2193 -0.7593 1.0000 0.5685 0.8299 -0.5944 1.0000 0.4699 1.0459 -0.5707 1.0000 0.4906
0.667 1.0380 -0.7209 1.0000 0.5694 0.6953 -0.5713 1.0000 0.4799 0.8757 -0.5314 1.0000 0.5025
0.7 0.9158 -0.6959 1.0000 0.5700 0.5954 -0.5543 1.0000 0.4867 0.7392 -0.4998 1.0000 0.5112
0.75 0.7412 -0.6602 1.0000 0.5708 0.4527 -0.5302 1.0000 0.4965 0.5444 -0.4547 1.0000 0.5235
0.8 0.6212 -0.6371 1.0000 0.5719 0.3418 -0.5106 1.0000 0.5038 0.3623 -0.4116 1.0000 0.5335
0.85 0.5083 -0.6155 1.0000 0.5728 0.2376 -0.4923 1.0000 0.5106 0.1913 -0.3712 1.0000 0.5428
0.9 0.2964 -0.5761 1.0000 0.5760 0.0693 -0.4630 1.0000 0.5215 -0.1385 -0.2932 1.0000 0.5598
0.95 0.0614 -0.5335 1.0000 0.5803 -0.0415 -0.4494 1.0000 0.5296 -0.3583 -0.2461 1.0000 0.5739
1.0 -0.1915 -0.4913 1.0000 0.5854 -0.0967 -0.4555 1.0000 0.5354 -0.4193 -0.2456 1.0000 0.5852
1.1 -0.4301 -0.4563 1.0000 0.5904 -0.1041 -0.4806 1.0000 0.5401 -0.3485 -0.2856 1.0000 0.5936
1.2 -0.6336 -0.4304 1.0000 0.5941 -0.0738 -0.5215 1.0000 0.5450 -0.2165 -0.3463 1.0000 0.5996
1.3 -0.8156 -0.4103 1.0000 0.5953 -0.0320 -0.5691 1.0000 0.5511 -0.0920 -0.4091 1.0000 0.6035
1.4 -1.0118 -0.3912 1.0000 0.5931 -0.0357 -0.6071 1.0000 0.5593 -0.0276 -0.4612 1.0000 0.6063
1.5 -1.2503 -0.3703 1.0000 0.5874 -0.1493 -0.6191 1.0000 0.5697 -0.0722 -0.4911 1.0000 0.6087
1.7 -1.5259 -0.3501 1.0000 0.5785 -0.3975 -0.6017 1.0000 0.5819 -0.2535 -0.4925 1.0000 0.6117
2.0 -1.7950 -0.3397 1.0000 0.5674 -0.7453 -0.5663 1.0000 0.5950 -0.5395 -0.4756 1.0000 0.6160
2.2 -1.9108 -0.3426 1.0000 0.5611 -0.9419 -0.5467 1.0000 0.6018 -0.7079 -0.4662 1.0000 0.6187
2.6 -2.0796 -0.3504 1.0000 0.5508 -1.2418 -0.5188 1.0000 0.6119 -0.9767 -0.4513 1.0000 0.6233
3.0 -2.1924 -0.3596 1.0000 0.5428 -1.4567 -0.5011 1.0000 0.6189 -1.1824 -0.4400 1.0000 0.6268
3.4 -2.3459 -0.3686 1.0000 0.5302 -1.7104 -0.4873 1.0000 0.6263 -1.5020 -0.4122 1.0000 0.6310
4.0 -2.4736 -0.3683 1.0000 0.5170 -1.8745 -0.4834 1.0000 0.6284 -1.7876 -0.3769 1.0000 0.6319
63
Table 3.4 Standard deviations for the Northridge Earthquake compared with standard
deviations from Somerville and Abrahamson (Somerville, personal comm.). Values of the
standard deviation of the sample standard deviation are given in parenthesis.
Period This Study This Study This Study Somerville & Somerville &
Site B Site C Site D Abrahamson: Abrahamson:
Rock Soil
PGA .32 (.07) .47 (.04) .36 (.03) .53 .48
0.3 .40 (.08) .54 (.05) .41 (.04) .60 .51
1 .45 (.11) .60 (.05) .36 (.03) .62 .48
2 .48 (.12) .57 (.05) .41 (.04) .57 .60
64
Table 3.5 Subdivision of sites classified according to the presented classification system
by means of the 1997 UBC shear wave velocity-based classification system.
NORTHRIDGE
Site Vs based Number of

Classification Classification sites
(from this work)
B UBC B 11
UBC C 0
C UBC B 9
UBC C 41
UBC D 20
D UBC C 5
UBC D 54
LOMA PRIETA
Site Classification Vs based Number

(from this work) Classification of sites
B UBC B 13
UBC C 5
C UBC B 1
UBC C 21
UBC D 4
D UBC C 1
UBC D 18
65
Table 3.6 Comparison of standard errors at selected periods for an analysis based on the
classification system presented herein and an analysis based on the 1997 UBC average
shear wave velocity-based classification system. Values in parenthesis are standard
deviations of the estimate of the standard error.
Northridge Loma Prieta

T = 0.3 s T = 1.0 s T = 0.3 s T = 1.0 s
Site This UBC This UBC This UBC This UBC
Study Study Study Study
B .40(.08) .46(.07) .45(.11) .52(.09) .51(.10) .52(.10) .58(.11) .61(.11)
C .54(.05) .54(.06) .60(.05) .54(.06) .38(.05) .36(.05) .53(.08) .52(.07)
D .41(.04) .42(.03) .36(.03) .41(.03) .39(.07) .39(.06) .59(.11) .64(.10)
66
Table 3.7a Spectral acceleration amplification factors with respect to Site B and standard
deviations for corresponding soil type. Geometric mean of the Loma Prieta and
Northridge earthquakes.
Site C Site D
T PGA = PGA = PGA = PGA = σ PGA = PGA = PGA = PGA = σ
0.1 g 0.2 g 0.3 g 0.4 g 0.1 g 0.2 g 0.3 g 0.4 g
PGA 1.43 1.35 1.31 1.28 0.39 1.75 1.58 1.49 1.43 0.37
0.055 1.30 1.20 1.14 1.11 0.42 1.57 1.37 1.27 1.20 0.39
0.06 1.30 1.20 1.14 1.11 0.42 1.57 1.37 1.27 1.20 0.40
0.07 1.30 1.20 1.14 1.11 0.42 1.57 1.37 1.27 1.20 0.40
0.08 1.30 1.20 1.14 1.10 0.43 1.57 1.37 1.27 1.20 0.40
0.09 1.31 1.20 1.14 1.10 0.43 1.58 1.38 1.28 1.21 0.40
0.1 1.32 1.21 1.15 1.11 0.43 1.59 1.39 1.29 1.21 0.40
0.11 1.33 1.22 1.16 1.11 0.44 1.61 1.40 1.29 1.22 0.40
0.12 1.35 1.23 1.16 1.12 0.44 1.62 1.41 1.30 1.23 0.40
0.13 1.36 1.24 1.17 1.13 0.44 1.63 1.42 1.31 1.24 0.40
0.14 1.38 1.25 1.19 1.14 0.45 1.65 1.44 1.33 1.26 0.39
0.15 1.39 1.27 1.20 1.15 0.45 1.68 1.46 1.35 1.27 0.39
0.16 1.41 1.28 1.21 1.16 0.45 1.70 1.48 1.37 1.29 0.39
0.17 1.42 1.29 1.22 1.17 0.46 1.72 1.50 1.38 1.30 0.39
0.18 1.43 1.29 1.22 1.17 0.46 1.73 1.50 1.39 1.31 0.39
0.19 1.44 1.30 1.23 1.18 0.45 1.74 1.52 1.40 1.32 0.39
0.2 1.45 1.31 1.23 1.18 0.45 1.76 1.54 1.42 1.34 0.39
0.24 1.46 1.31 1.23 1.17 0.46 1.79 1.56 1.44 1.36 0.39
0.28 1.46 1.30 1.22 1.16 0.46 1.80 1.57 1.46 1.38 0.40
0.3 1.46 1.30 1.21 1.15 0.46 1.81 1.58 1.47 1.39 0.40
0.34 1.44 1.29 1.20 1.14 0.47 1.83 1.60 1.49 1.40 0.40
0.36 1.44 1.28 1.20 1.14 0.47 1.83 1.61 1.50 1.42 0.41
0.4 1.42 1.27 1.19 1.13 0.48 1.83 1.62 1.51 1.43 0.42
0.44 1.41 1.26 1.18 1.13 0.49 1.84 1.63 1.52 1.45 0.43
0.5 1.38 1.25 1.18 1.13 0.51 1.85 1.66 1.55 1.48 0.44
0.55 1.36 1.24 1.17 1.13 0.52 1.85 1.67 1.57 1.50 0.44
0.6 1.35 1.24 1.17 1.13 0.53 1.86 1.68 1.59 1.52 0.44
0.667 1.34 1.23 1.17 1.13 0.53 1.87 1.70 1.60 1.54 0.45
0.7 1.33 1.23 1.17 1.13 0.54 1.88 1.71 1.62 1.56 0.45
0.75 1.32 1.23 1.18 1.14 0.55 1.89 1.73 1.64 1.58 0.45
0.8 1.32 1.23 1.18 1.14 0.55 1.91 1.75 1.67 1.61 0.45
0.85 1.31 1.23 1.19 1.15 0.56 1.92 1.77 1.69 1.63 0.45
0.9 1.31 1.24 1.20 1.18 0.56 1.95 1.81 1.73 1.67 0.46
0.95 1.31 1.26 1.22 1.20 0.57 1.98 1.85 1.78 1.72 0.46
1.0 1.31 1.27 1.25 1.23 0.57 2.02 1.89 1.83 1.78 0.47
1.1 1.31 1.29 1.27 1.26 0.57 2.05 1.94 1.88 1.84 0.47
1.2 1.31 1.30 1.29 1.29 0.57 2.09 1.99 1.94 1.90 0.48
1.3 1.32 1.32 1.32 1.32 0.57 2.12 2.04 2.00 1.96 0.48
1.4 1.32 1.33 1.34 1.34 0.58 2.15 2.09 2.05 2.02 0.49
1.5 1.32 1.35 1.36 1.36 0.58 2.18 2.13 2.10 2.08 0.49
1.7 1.33 1.36 1.37 1.38 0.58 2.22 2.18 2.16 2.14 0.50
2.0 1.33 1.37 1.38 1.39 0.58 2.25 2.22 2.20 2.18 0.51
2.2 1.33 1.37 1.39 1.40 0.58 2.26 2.24 2.22 2.20 0.52
2.6 1.33 1.37 1.39 1.40 0.58 2.27 2.25 2.23 2.22 0.52
3.0 1.33 1.37 1.39 1.40 0.57 2.27 2.25 2.24 2.23 0.51
67
Table 3.7b Spectral acceleration amplification factors with respect to Site B and standard
deviations for corresponding soil type. Variance weighted geometric mean of the Loma
Prieta and Northridge earthquakes.
Site C Site D
0.1 g 0.2 g 0.3 g 0.4 g 0.1 g 0.2 g 0.3 g 0.4 g
PGA 1.45 1.37 1.33 1.30 0.37 1.74 1.60 1.53 1.48 0.36
0.055 1.29 1.19 1.13 1.09 0.42 1.56 1.38 1.28 1.22 0.37
0.06 1.29 1.19 1.13 1.09 0.42 1.56 1.38 1.28 1.22 0.38
0.07 1.29 1.19 1.13 1.10 0.43 1.56 1.38 1.28 1.22 0.38
0.08 1.30 1.19 1.14 1.10 0.43 1.57 1.38 1.29 1.23 0.38
0.09 1.31 1.20 1.15 1.11 0.44 1.57 1.39 1.30 1.24 0.39
0.1 1.33 1.22 1.17 1.13 0.44 1.58 1.41 1.32 1.26 0.39
0.11 1.35 1.24 1.18 1.14 0.44 1.59 1.42 1.34 1.28 0.38
0.12 1.36 1.25 1.19 1.16 0.45 1.60 1.44 1.35 1.29 0.38
0.13 1.38 1.27 1.21 1.17 0.45 1.61 1.45 1.36 1.31 0.38
0.14 1.40 1.28 1.22 1.18 0.45 1.63 1.47 1.38 1.32 0.38
0.15 1.41 1.30 1.23 1.19 0.46 1.66 1.49 1.40 1.34 0.38
0.16 1.43 1.31 1.24 1.20 0.46 1.68 1.50 1.41 1.35 0.38
0.17 1.44 1.31 1.25 1.20 0.46 1.70 1.52 1.42 1.36 0.39
0.18 1.45 1.32 1.25 1.20 0.46 1.72 1.53 1.43 1.36 0.39
0.19 1.45 1.32 1.25 1.21 0.46 1.73 1.54 1.44 1.37 0.39
0.2 1.46 1.33 1.25 1.21 0.45 1.75 1.56 1.45 1.39 0.39
0.24 1.47 1.32 1.24 1.19 0.45 1.79 1.58 1.47 1.40 0.40
0.28 1.47 1.32 1.23 1.18 0.45 1.80 1.60 1.49 1.41 0.40
0.3 1.46 1.31 1.23 1.17 0.45 1.81 1.61 1.50 1.42 0.41
0.34 1.45 1.30 1.21 1.16 0.46 1.83 1.63 1.52 1.44 0.41
0.36 1.44 1.29 1.21 1.15 0.46 1.84 1.64 1.53 1.45 0.41
0.4 1.43 1.28 1.20 1.15 0.48 1.84 1.65 1.55 1.48 0.42
0.44 1.42 1.28 1.20 1.15 0.49 1.85 1.67 1.57 1.50 0.42
0.5 1.39 1.26 1.20 1.15 0.51 1.86 1.70 1.61 1.55 0.42
0.55 1.38 1.26 1.20 1.15 0.53 1.87 1.72 1.63 1.57 0.41
0.6 1.36 1.25 1.19 1.15 0.53 1.87 1.73 1.65 1.60 0.41
0.667 1.35 1.25 1.19 1.15 0.54 1.88 1.75 1.68 1.63 0.40
0.7 1.34 1.25 1.19 1.15 0.55 1.89 1.77 1.70 1.65 0.40
0.75 1.33 1.24 1.19 1.16 0.56 1.91 1.79 1.72 1.68 0.39
0.8 1.33 1.24 1.20 1.16 0.56 1.92 1.81 1.75 1.70 0.38
0.85 1.32 1.24 1.20 1.17 0.57 1.94 1.83 1.77 1.72 0.38
0.9 1.32 1.25 1.22 1.19 0.57 1.97 1.86 1.80 1.76 0.37
0.95 1.32 1.26 1.23 1.21 0.58 2.00 1.90 1.84 1.80 0.37
1.0 1.32 1.28 1.25 1.24 0.58 2.04 1.94 1.89 1.85 0.37
1.1 1.32 1.29 1.28 1.26 0.58 2.08 1.99 1.94 1.90 0.37
1.2 1.32 1.31 1.30 1.29 0.58 2.12 2.03 1.99 1.95 0.37
1.3 1.33 1.32 1.32 1.32 0.58 2.15 2.08 2.04 2.01 0.37
1.4 1.33 1.34 1.34 1.34 0.58 2.19 2.13 2.09 2.06 0.38
1.5 1.33 1.35 1.36 1.36 0.58 2.23 2.17 2.14 2.11 0.38
1.7 1.34 1.36 1.37 1.38 0.58 2.28 2.23 2.19 2.17 0.39
2.0 1.35 1.37 1.38 1.39 0.58 2.31 2.27 2.24 2.21 0.43
2.2 1.35 1.37 1.39 1.39 0.57 2.32 2.28 2.25 2.23 0.44
2.6 1.35 1.37 1.39 1.40 0.56 2.33 2.29 2.26 2.24 0.43
3.0 1.34 1.37 1.39 1.41 0.54 2.32 2.29 2.27 2.25 0.41
68
Table 3.8a Spectral acceleration amplification factors with respect to Site C and standard
deviations for corresponding soil type. Geometric mean of the Loma Prieta and
Northridge earthquakes.
Site B Site D
0.1 g 0.2 g 0.3 g 0.4 g 0.1 g 0.2 g 0.3 g 0.4 g
0.68 0.72 0.74 0.76 0.40 1.24 1.18 1.15 1.12 0.37
0.055 0.74 0.81 0.85 0.88 0.45 1.23 1.16 1.13 1.10 0.39
0.06 0.74 0.81 0.85 0.88 0.45 1.23 1.16 1.13 1.10 0.40
0.07 0.74 0.81 0.85 0.88 0.44 1.23 1.16 1.13 1.10 0.40
0.08 0.74 0.81 0.85 0.88 0.44 1.23 1.16 1.13 1.10 0.40
0.09 0.73 0.80 0.85 0.88 0.42 1.23 1.17 1.13 1.11 0.40
0.1 0.72 0.80 0.84 0.88 0.41 1.22 1.17 1.13 1.11 0.40
0.11 0.72 0.79 0.84 0.87 0.40 1.22 1.17 1.13 1.11 0.40
0.12 0.71 0.78 0.83 0.86 0.39 1.22 1.17 1.13 1.11 0.40
0.13 0.70 0.78 0.82 0.86 0.38 1.22 1.17 1.13 1.11 0.40
0.14 0.69 0.77 0.81 0.85 0.38 1.22 1.17 1.13 1.11 0.39
0.15 0.68 0.76 0.80 0.84 0.38 1.22 1.17 1.14 1.11 0.39
0.16 0.68 0.75 0.80 0.83 0.38 1.22 1.17 1.14 1.12 0.39
0.17 0.67 0.74 0.79 0.82 0.39 1.22 1.17 1.14 1.12 0.39
0.18 0.66 0.74 0.79 0.82 0.39 1.22 1.18 1.15 1.13 0.39
0.19 0.66 0.74 0.78 0.82 0.40 1.23 1.18 1.15 1.13 0.39
0.2 0.65 0.73 0.78 0.81 0.41 1.23 1.18 1.16 1.14 0.39
0.24 0.65 0.73 0.78 0.82 0.43 1.24 1.20 1.18 1.17 0.39
0.28 0.65 0.73 0.78 0.82 0.44 1.24 1.21 1.20 1.19 0.40
0.3 0.65 0.73 0.78 0.83 0.46 1.25 1.22 1.21 1.20 0.40
0.34 0.65 0.74 0.79 0.83 0.47 1.27 1.25 1.24 1.23 0.40
0.36 0.66 0.74 0.80 0.84 0.48 1.28 1.26 1.25 1.24 0.41
0.4 0.67 0.75 0.80 0.84 0.49 1.29 1.28 1.27 1.26 0.42
0.44 0.67 0.75 0.81 0.84 0.49 1.31 1.29 1.28 1.28 0.43
0.5 0.69 0.77 0.81 0.85 0.49 1.34 1.33 1.32 1.31 0.44
0.55 0.70 0.77 0.82 0.85 0.49 1.36 1.34 1.33 1.33 0.44
0.6 0.71 0.78 0.82 0.85 0.49 1.38 1.36 1.35 1.34 0.44
0.667 0.71 0.78 0.82 0.85 0.49 1.39 1.37 1.36 1.36 0.45
0.7 0.72 0.78 0.82 0.85 0.49 1.41 1.39 1.38 1.37 0.45
0.75 0.73 0.79 0.82 0.85 0.49 1.43 1.41 1.39 1.39 0.45
0.8 0.73 0.79 0.82 0.85 0.49 1.45 1.42 1.41 1.40 0.45
0.85 0.74 0.79 0.82 0.84 0.50 1.47 1.44 1.42 1.41 0.45
0.9 0.74 0.78 0.81 0.83 0.50 1.50 1.46 1.44 1.42 0.46
0.95 0.75 0.78 0.80 0.81 0.51 1.53 1.48 1.45 1.43 0.46
1.0 0.75 0.77 0.79 0.80 0.52 1.55 1.50 1.47 1.45 0.47
1.1 0.75 0.77 0.77 0.78 0.53 1.58 1.52 1.49 1.46 0.47
1.2 0.75 0.76 0.76 0.76 0.53 1.61 1.54 1.51 1.48 0.48
1.3 0.76 0.75 0.75 0.75 0.54 1.63 1.56 1.52 1.50 0.48
1.4 0.76 0.75 0.74 0.74 0.54 1.65 1.58 1.54 1.52 0.49
1.5 0.76 0.74 0.73 0.73 0.53 1.67 1.60 1.56 1.53 0.49
1.7 0.75 0.73 0.72 0.72 0.53 1.70 1.63 1.59 1.56 0.50
2.0 0.75 0.73 0.72 0.71 0.52 1.72 1.65 1.61 1.58 0.51
2.2 0.75 0.73 0.72 0.71 0.52 1.72 1.65 1.61 1.59 0.52
2.6 0.75 0.73 0.72 0.71 0.52 1.73 1.66 1.62 1.59 0.52
3.0 0.76 0.73 0.72 0.71 0.51 1.73 1.66 1.63 1.60 0.51
69
Table 3.8b Spectral acceleration amplification factors with respect to Site C and standard
deviations for corresponding soil type. Variance weighted geometric mean of the Loma
Prieta and Northridge earthquakes.
Site B Site D
0.1 g 0.2 g 0.3 g 0.4 g 0.1 g 0.2 g 0.3 g 0.4 g
PGA 0.67 0.71 0.73 0.75 0.36 1.17 1.15 1.14 1.14 0.36
0.055 0.74 0.81 0.86 0.89 0.45 1.14 1.12 1.10 1.09 0.37
0.06 0.74 0.81 0.86 0.89 0.45 1.14 1.12 1.10 1.09 0.38
0.07 0.74 0.81 0.86 0.89 0.44 1.14 1.12 1.10 1.09 0.38
0.08 0.74 0.81 0.85 0.89 0.44 1.14 1.12 1.11 1.10 0.38
0.09 0.73 0.80 0.85 0.88 0.42 1.14 1.12 1.11 1.10 0.39
0.1 0.72 0.79 0.83 0.86 0.40 1.14 1.12 1.11 1.10 0.39
0.11 0.71 0.78 0.82 0.85 0.39 1.13 1.12 1.11 1.10 0.38
0.12 0.70 0.77 0.81 0.84 0.37 1.13 1.12 1.11 1.10 0.38
0.13 0.70 0.76 0.80 0.83 0.36 1.13 1.12 1.11 1.10 0.38
0.14 0.69 0.75 0.79 0.82 0.36 1.13 1.12 1.11 1.10 0.38
0.15 0.68 0.74 0.79 0.81 0.36 1.13 1.12 1.11 1.10 0.38
0.16 0.67 0.74 0.78 0.81 0.36 1.14 1.12 1.11 1.10 0.38
0.17 0.66 0.73 0.77 0.81 0.37 1.14 1.13 1.12 1.11 0.39
0.18 0.66 0.73 0.77 0.81 0.38 1.15 1.13 1.12 1.11 0.39
0.19 0.66 0.73 0.77 0.80 0.38 1.15 1.14 1.13 1.12 0.39
0.2 0.65 0.72 0.77 0.80 0.39 1.16 1.15 1.14 1.13 0.39
0.24 0.65 0.72 0.77 0.81 0.42 1.18 1.17 1.16 1.16 0.40
0.28 0.65 0.72 0.78 0.81 0.43 1.19 1.18 1.18 1.18 0.40
0.3 0.65 0.73 0.78 0.82 0.44 1.20 1.20 1.19 1.19 0.41
0.34 0.65 0.73 0.79 0.83 0.46 1.23 1.22 1.22 1.22 0.41
0.36 0.66 0.74 0.79 0.83 0.47 1.24 1.24 1.24 1.24 0.41
0.4 0.66 0.74 0.79 0.83 0.48 1.25 1.26 1.26 1.26 0.42
0.44 0.67 0.75 0.80 0.84 0.48 1.27 1.28 1.28 1.28 0.42
0.5 0.69 0.76 0.80 0.84 0.47 1.30 1.31 1.32 1.33 0.42
0.55 0.69 0.76 0.81 0.84 0.47 1.32 1.34 1.34 1.35 0.41
0.6 0.70 0.77 0.81 0.84 0.46 1.34 1.36 1.37 1.38 0.41
0.667 0.71 0.77 0.81 0.84 0.46 1.36 1.38 1.39 1.40 0.40
0.7 0.72 0.78 0.81 0.84 0.46 1.37 1.40 1.41 1.42 0.40
0.75 0.72 0.78 0.81 0.84 0.46 1.39 1.42 1.43 1.45 0.39
0.8 0.73 0.78 0.81 0.84 0.47 1.41 1.44 1.45 1.47 0.38
0.85 0.73 0.78 0.81 0.83 0.48 1.43 1.45 1.47 1.48 0.38
0.9 0.74 0.78 0.80 0.82 0.49 1.45 1.48 1.49 1.50 0.37
0.95 0.74 0.77 0.79 0.81 0.50 1.48 1.50 1.51 1.52 0.37
1.0 0.74 0.77 0.78 0.80 0.51 1.50 1.52 1.53 1.54 0.37
1.1 0.75 0.76 0.77 0.78 0.51 1.53 1.54 1.55 1.56 0.37
1.2 0.75 0.76 0.76 0.77 0.52 1.55 1.56 1.57 1.58 0.37
1.3 0.75 0.75 0.75 0.75 0.53 1.58 1.59 1.59 1.60 0.37
1.4 0.75 0.74 0.74 0.74 0.53 1.60 1.61 1.61 1.62 0.38
1.5 0.75 0.74 0.73 0.73 0.53 1.63 1.63 1.63 1.64 0.38
1.7 0.74 0.73 0.72 0.72 0.52 1.67 1.66 1.66 1.67 0.39
2.0 0.74 0.73 0.72 0.72 0.52 1.70 1.69 1.69 1.69 0.43
2.2 0.74 0.73 0.72 0.71 0.52 1.71 1.70 1.70 1.70 0.44
2.6 0.74 0.73 0.72 0.71 0.51 1.71 1.71 1.71 1.71 0.43
3.0 0.75 0.73 0.72 0.71 0.51 1.72 1.71 1.70 1.70 0.41
70
Table 3.9a Short-period (Fa) and mid-period (Fv) spectral amplification factors from the
1997 Uniform Building Code.
Fa PGA = .08 g PGA = .15 g PGA = .2 g PGA = .3 g PGA = .4 g

B 1.0 1.0 1.0 1.0 1.0
C 1.2 1.2 1.2 1.1 1.0
D 1.6 1.5 1.4 1.2 1.1
Fv PGA = .08 g PGA = .15 g PGA = .2 g PGA = .3 g PGA = .4 g

B 1.0 1.0 1.0 1.0 1.0
C 1.7 1.7 1.6 1.5 1.4
D 2.4 2.1 2.0 1.8 1.6
Table 3.9b Average spectral amplification periods over the short-period range (0.1 s – 0.5
s) and the mid-period range (0.4 s – 2.0 s), denoted by Fa and Fv, respectively.
Fa PGA = .08 g PGA = .15 g PGA = .2 g PGA = .3 g PGA = .4 g

B 1.0 1.0 1.0 1.0 1.0
C 1.5 1.3 1.3 1.2 1.2
D 1.8 1.6 1.6 1.5 1.4
Fv PGA = .08 g PGA = .15 g PGA = .2 g PGA = .3 g PGA = .4 g

B 1.0 1.0 1.0 1.0 1.0
C 1.4 1.3 1.3 1.3 1.2
D 2.1 2.0 1.9 1.8 1.8
71
BLANK
72
Figure 3.1. Relationship between structural damage intensity and soil depth in the
Caracas earthquake of 1967 (From Seed and Alonso 1974).
73
Shear Wave Velocity (m/s)
0 500 1000 1500 2000 2500
0
25
50
Depth (m)
75
100
125
150
Figure 3.2a. Shear wave velocity versus depth for a generic stiff clay deposit. Shear
wave velocity of underlying bedrock is 1220 m/s.c
1.8
1.6
Spectral Acceleration (g)
5% damping
1.4
1.2 Depth to Bedrock
30 m
1
60 m
0.8
150 m
0.6 Input
0.4
0.2
0
0 01 0.1 1 10
Period (s)
Figure 3.2b. Spectral accelerations for the stiff soil deposit shown in Figure 3.2a, with
PGA = 0.3 g for a Mw = 8.0 earthquake.
74
5
Spectral Amplification Ratio

4
Depth to Bedrock
3
30 m
2 60 m
150 m
1
0
0.01 0.1 Period (s) 1 10
Figure 3.2c. Spectral acceleration amplification ratio for the stiff soil deposit shown in
Figure 3.2a with PGA = 0.3 g for a Mw = 8.0 earthquake. The predominant period of the
site is indicated by a circle.
75
Shear Wave Velocity (m /s)
0 100 200 300 400 500 600 700 800 900
0
10
Depth (m)
15
20
25
30
Stiff Clay\ Loose Sand Dense Sand Soft Clay
Figure 3.3a. Shear wave velocity profiles for generic sites. Shear wave velocity of
underlying bedrock is 1220 m/s.
5
4.5
Ratio of Response Spectra
4
3.5
Stiff Clay
3
Soft Clay
2.5
Loose Sand
2 Dense Sand
1.5
1
0.5
0
0.01 0.1 1 10
Period (s)
Figure 3.3b. Spectral acceleration amplification ratio for the soil profiles shown in Figure
3b, with PGA = 0.3 g for a Mw = 8.0 earthquake.
76
B sites B 11
C1 28
C2 27
C sites
C3 15
D1C 11
D1S 29
D sites 58.9 D2C 10
D2S 9
F 9
1 10 100
Distance (km)
Figure 3.4a. Distribution of data by site type for the Northridge Earthquake.
B sites B 18
C1 11
C2 11
C sites
C3 4
D1C 10
D1S 2
D2C 3
D sites
D2S 3
E 7
E sites
F 1
1 10 100
Distance (km)
Figure 3.4b. Distribution of data by site type for the Loma Prieta Earthquake.
77
80
70
60
Number of Recordings
B - Northridge
50 C - Northridge
D - Northridge
40 C - Loma Prieta
B - Loma Prieta
30 D - Loma Prieta
20
10
0
0 1 2 3 4 5
Period (s)
Figure 3.5. Number of recordings as a function of period.
78
a -0.8 b
6
5 B and C
D -1.2
4
b
3
a
-1.6
2 B
C
1 D
0 -2
0.01 0.1 Period (s) 1 10 0.01 0.1 1 10
Period (s)
20 c 0.8 σ
16
0.6
12
0.4
c
σ
8
B
4 0.2 C
D
0 0.0
0.01 0.1 1 10 0.01 0.1 1 10
Period (s) Period (s)
Figure 3.6. Regression coefficients for the Northridge Earthquake.
79
3.0 a
2.0 B
C
1.0 D
a 0.0
-1.0
-2.0
-3.0
0.01 0.1 Period (s) 1 10
b
0
-0.2
-0.4
b -0.6
B
-0.8
C
D
-1
-1.2
0.01 0.1 Period (s) 1 10
σ
0.7
0.6
0.5
0.4
σ B
0.3 C
0.2 D
0.1
0
0.01 0.1 1 10
Period (s)
Figure 3.7. Regression coefficients for the Loma Prieta Earthquake. The coefficient "c"
is equal to 1.0 for all periods.
80
1.2
5% damping
1
Spectral Acceleration (g's)

0.8
0.6 Site B
Site D
0.4 Site C
0.2
0
0.01 0.1 1 10
Period (s)
Figure 3.8. Comparison of response spectra before smoothing and after smoothing
regression coefficients. Corresponds to the Northridge Earthquake at R = 10 km.
81
1.4
15 km
1.2
5% damping
Spectral Acceleration (g's) 1
0.8 Site B
Site C
0.6 Site D
0.4
0.2
0
0.01 0.1 Period (s) 1 3 10
30 km
0.8
5% damping
0.6
Site B
0.4 Site C
Site D
0.2
0
0.01 0.1 Period (s) 1 3 10
0.6
50 km
5% damping
0.4
Site B
Site C
Site D
0.2
0
0.01 0.1 Period (s) 1 3 10
Figure 3.9. Response spectra for the Northridge Earthquake. Thick lines represent
median values, thin lines represent ± one standard deviation.
82
15 km
1.4
1.2
5% damping
1
0.8 Site B
Site C
0.6 Site D
0.4
0.2
0
0.01 0.1 Period (s) 1 4 10
30 km
0.8
5% damping
0.6
Site B
0.4 Site C
Site D
0.2
0
0.01 0.1 Period (s) 1 4 10
0.8
50 km
5% damping
0.6
Site B
0.4 Site C
Site D
0.2
0
0.01 0.1 Period (s) 1 4 10
Figure 3.10. Response spectra for the Loma Prieta Earthquake. Thick lines represent
median values, thin lines represent ± one standard deviation.
83
PGA
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
T = 0.3 s
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
T = 1.0 s
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
Figure 3.11. Median spectral values vs. distance for the Northridge Earthquake.
84
PGA
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
T = 0.3 s
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
T = 1.0 s
10
1
Site B
Site C
Site D
0.1
0.01
1 10 100
Distance (km)
Figure 3.12. Median spectral values vs. distance for the Loma Prieta Earthquake.
85
0.8
Site B (11 sites)

5% damping
0.6 Site C (70 sites)
Site D (59 sites)

S&A: Rock
S&A: Soil
0.4
0.2
0
0.01 0.1 Period (s) 1 10
Figure 3.13. Comparison of results with an earthquake specific attenuation relationship

by Somerville and Abrahamson (1998). Response spectra at 5% damping for the
Northridge Earthquake at R = 20 km.
86
PGA T = 0.3 s
1.5 1.5
1 1
Residual (in Ln scale)

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
Distance (km) Distance (km)
UBC D UBC C UBC B UBC D UBC C UBC B
T = 1.0 s T = 2.0 s
1.5 1.5
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
UBC D UBC C UBC B UBC D UBC C UBC B
PGA T=0.1 T=0.3 T=1 T=2

UBC B -0.08 -0.08 -0.14 -0.34 -0.44
UBC C -0.08 -0.07 -0.09 0.08 0.10
UBC D 0.14 0.13 0.19 0.19 0.18
Figure 3.14. Residuals with respect to regression analysis for Site C. All sites shown are
classified as C sites in the classification system proposed in this study, but are
differentiated with respect to their corresponding UBC classification based in the average
shear wave in the upper 30 m.
87
PGA T = 0.3 s
1.5 1.5
1 1

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
C1 C2 C3 C1 C2 C3
T = 1.0 s T = 2.0 s
1.5 1.5
1 Residual (in Ln scale) 1

0.5 0.5
0
0
-0.5
-0.5
-1
-1
-1.5
-1.5
1 10 100
1 10 100
C1 C2 C3
C1 C2 C3
PGA T=0.3 T=1 T=3

C1 0.08 0.08 0.10 0.01
C2 -0.08 -0.10 0.01 0.05
C3 -0.10 -0.06 0.05 0.16
Figure 3.15a. Residuals for Site C, Northridge Earthquake. Table gives mean of
residuals for each subgroup.
88
PGA T = 0.3 s
1.5 1.5
1 1

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
C1 C2 C3 C1 C2 C3
T = 1.0 s T = 2.0 s
1.5 1.5
1
1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
C1 C2 C3 C1 C2 C3
PGA T=0.3 T=1 T=3

C1 -0.04 0.07 0.05 -0.04
C2 0.07 0.06 -0.01 0.01
C3 -0.08 -0.29 0.10 0.16
Figure 3.15b. Residuals for Site C, Loma Prieta Earthquake. Table gives mean of
89
1.5
PGA 1.5
T = 0.3 s
1 1

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
D1C D1S D2C D2S D1C D1S D2C D2S
1.5
T = 1.0 s 1.5
T = 2.0 s
1 1
0.5 Residual (in Ln scale) 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
1 10 100 1 10 100
PGA T=0.3 T=1 T=3

D1C 0.03 0.14 0.13 0.15
D1S 0.00 -0.02 0.01 -0.07
D2C 0.18 0.13 0.18 0.04
D2S -0.26 -0.29 -0.13 0.07
D1 0.01 0.02 0.04 -0.01
D2 -0.03 -0.07 0.04 0.05
DC 0.10 0.14 0.16 0.10
DS -0.06 -0.09 -0.02 -0.04
Figure 3.16a. Residuals for Site D, Northridge Earthquake. Table gives mean of
90
1.5
PGA 1.5
T = 0.3 s
1 1

0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
10 Distance (km) 100 10 100
Distance (km)
1.5
T = 1.0 s T = 2.0 s
1.5
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
10 100 10 Distance (km) 100
Distance (km)
PGA T=0.3 T=1 T=3

D1C 0.00 -0.02 0.21 0.24
D1S 0.15 0.23 0.04 -0.10
D2C 0.20 0.19 0.18 -0.32
D2S -0.46 -0.30 -0.78 -0.78
D1 0.03 0.02 0.18 0.19
D2 -0.13 -0.05 -0.30 -0.55
DC 0.05 0.03 0.20 0.11
DS -0.21 -0.09 -0.45 -0.51
Figure 3.16b. Residuals for Site D, Loma Prieta Earthquake. Table gives mean of
91
T = 1.0 s
1
0.5
-0.5
-1
0 500 1000 1500 2000 2500 3000
Depth to Basement Rock (m)
T = 2.0 s
1
0.5
-0.5
-1
0 500 1000 1500 2000 2500 3000
Depth to Basement Rock (m)
Figure 3.17. Residuals for D sites within the Los Angeles Basin plotted as a function of
depth to basement rock.
92
C sites
2.40
2.20
2.00
1.80 PGA = 0.1 g

Factor
PGA = 0.2 g
1.60 PGA = 0.3 g
PGA = 0.4 g
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
D sites
2.40
2.20
2.00
1.80
PGA = 0.1 g
Factor
1.60 PGA = 0.2 g

PGA = 0.3 g
1.40 PGA = 0.4 g
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
Figure 3.18a. Amplification factors with respect to Site B for the Northridge Earthquake.
93
B sites
1.90
1.70
1.50
PGA = 0.1 g
Factor
1.30
PGA = 0.2 g
1.10 PGA = 0.3 g

PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
D sites
1.90
1.70
1.50
PGA = 0.1 g
Factor
1.30
PGA = 0.2 g
1.10 PGA = 0.3 g
PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
Figure 3.18b. Amplification factors with respect to Site C for the Northridge Earthquake.
94
C sites
2.40
2.20
2.00
PGA = 0.1 g
1.80
PGA = 0.2 g
Factor
PGA = 0.3 g
1.60
PGA = 0.4 g
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
D sites
2.40
2.20
2.00
PGA = 0.1 g
1.80
PGA = 0.2 g
Factor
PGA = 0.3 g
1.60
PGA = 0.4 g
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
Figure 3.18c. Amplification factors with respect to Site B for the Loma Prieta
Earthquake.
95
B sites
1.90
1.70
1.50 PGA = 0.1 g

PGA = 0.2 g
Factor
1.30 PGA = 0.3 g

PGA = 0.4 g
1.10
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
D sites
1.90
1.70
1.50
PGA = 0.1 g
Factor
1.30 PGA = 0.2 g

PGA = 0.3 g
1.10 PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
Figure 3.18d. Amplification factors with respect to Site C for the Loma Prieta
Earthquake.
96
C sites
2.40
2.20
2.00
1.80
PGA = 0.1 g
Factor
PGA = 0.2 g
1.60
PGA = 0.3 g
PGA = 0.4 g
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
D sites
2.40
2.20
2.00
1.80
PGA = 0.1 g
Factor
PGA = 0.2 g
1.60
PGA = 0.3 g
1.40 PGA = 0.4 g
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
Figure 3.19a. Amplification factors with respect to Site B. Geometric mean of the
Northridge and Loma Prieta Earthquakes.
97
B sites
1.90
1.70
PGA = 0.1 g
1.50 PGA = 0.2 g
PGA = 0.3 g
Factor
1.30 PGA = 0.4 g
1.10
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
D sites
1.90
1.70
1.50
PGA = 0.1 g
PGA = 0.2 g
Factor
1.30
PGA = 0.3 g
1.10 PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
Figure 3.19b. Amplification factors with respect to Site C. Geometric mean of the
98
C sites
2.40
2.20
2.00 PGA = 0.1 g

PGA = 0.2 g
1.80 PGA = 0.3 g
PGA = 0.4 g
Factor
1.60
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
D sites
2.60
2.40
2.20
2.00
PGA = 0.1 g
1.80
Factor
PGA = 0.2 g
1.60 PGA = 0.3 g

PGA = 0.4 g
1.40
1.20
1.00
0.80
0.01 0.1 Period (s) 1 10
Figure 3.20a. Amplification factors with respect to Site B. Weighted mean of the
99
B sites
1.90
1.70
1.50
PGA = 0.1 g
Factor
1.30
PGA = 0.2 g
PGA = 0.3 g
1.10
PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
D sites
1.90
1.70
1.50
PGA = 0.1 g
1.30
Factor
PGA = 0.2 g
PGA = 0.3 g
1.10 PGA = 0.4 g
0.90
0.70
0.50
0.01 0.1 Period (s) 1 10
Figure 3.20b. Amplification factors with respect to Site C. Weighted mean of the
100
D/B ratios C/B Ratios
Loma
Prieta Loma
39% Prieta
Northridge 46% Northridge
61% 54%
D/C ratios
Loma Prieta
26%
Northridge
74%
Figure 3.21. Earthquake weighting scheme used for calculating spectral amplification
factors. Shown here is an average of the weights used for all periods. Weights are
inversely proportional to the sample variance.
101
Fa
2.50
2.30
2.10
Amplification Factor, Fa
1.90 UBC C
UBC D
1.70 C (This Study)
D (This Study)
1.50
1.30
1.10
0.90
0 0.1 0.2 0.3 0.4 0.5
PGA (g)
Fv
2.50
2.30
2.10
Amplification Factor, Fv
1.90
UBC C
UBC D
1.70
C (This Study)
D (This Study)
1.50
1.30
1.10
0.90
0 0.1 0.2 0.3 0.4 0.5
PGA (g)
Figure 3.22. Short-period (Fa) and intermediate-period (Fv) spectral amplification factors.
Dotted lines are code values (UBC 1997), and continuous lines are values obtained from
this study.
102
CHAPTER 4
EMPIRICAL CHARACTERIZATION OF NEAR-FAULT
GROUND MOTIONS
4.1 INTRODUCTION
The estimation of ground motions in close proximity to the ruptured fault for
medium to large magnitude earthquakes must account for the special characteristics of
near-fault ground motions. The near-fault zone is typically assumed to be restricted to
within 10 to 15 km of the causative fault. Of particular importance in the near-fault are
the effects of forward-directivity. Forward-directivity conditions produce ground motions
characterized by a strong pulse or series of pulses best observed in ground velocity-time
histories. The importance of forward-directivity records has prompted a number of
studies directed at addressing the prediction of this type of motions and their effect on
structural response (e.g. Somerville 1998, Krawinkler and Alavi 1998, Sasani and Bertero
2000). These studies have highlighted the importance of characterizing near-fault
records. However, more research is needed to account for the potential effects of local
soil conditions on these motions.
The effects of rupture directivity are generated because the velocity of fault
rupture is only slightly lower than the shear-wave propagation velocity. As the rupture
front propagates from the hypocenter, a shear wave front is formed by the accumulation
of the shear waves traveling ahead of the rupture front. When a site is located at one end
103
of the fault and rupture initiates at the other end of the fault and travels towards the site,
the arrival of the wave front is seen as a large pulse of motion occurring at the beginning
of the record (Somerville et al. 1997). This condition is known as forward-directivity,
and is illustrated in Figure 4.1. The radiation pattern of the shear dislocation on the fault
causes this large pulse of motion to be oriented in a direction perpendicular to the fault
plane (Somerville et al. 1997). The pulse is typically a long-period pulse that is best
observed in the velocity or displacement-time history. However, if a site is located at one
end of the fault and rupture propagates away from the site, the opposite effect is observed
and the motion is characterized by longer duration and lower amplitude ground motions.
This condition is termed backward-directivity.
Pulse-like motions can also be generated by a concentration of high slip in the
region of the fault near the recording site (Abrahamson, personal communication). Slip-
induced pulses, herein termed “fling-step”, have different characteristics than forward-
directivity pulses and are modeled by different terms in seismological fault modeling.
The fling-step normally generates one sided velocity pulses. It is also observed as a
discrete step in displacement-time histories that occurs parallel to strike of the fault with
strike-slip earthquakes, and in the dip direction for dip-slip events. Given their different
characteristics, it is desirable to treat the pulse motions originated from forward-
directivity and fling-step slip effects separately. In strike-slip events, forward-directivity
pulses can be identified by positive fault normal to fault parallel spectral ratios at long
periods. Whereas forward-directivity records have larger fault normal motions at long
periods, slip-induced pulses tend to have equal energy in both directions. In dip-slip
faults, permanent displacement can also occur in the fault normal direction. For these
104
motions, the fling-step should be removed before the effects of a forward-directivity pulse
can be evaluated.
Forward-directivity conditions can be present both for strike-slip and dip-slip
events. In strike-slip events, forward-directivity conditions are typically largest in sites
near the end of the fault when the rupture front is moving towards the site. In dip-slip
events, forward-directivity conditions occur in sites located in the up-dip projection of the
fault.
Currently, there is a lack of data with regards to the effects of site response on the
characteristic of pulse-type motions. This in itself constitutes an important motivating
factor for pursuing research on site effects for near-fault ground motions. Further
understanding of near-fault site effects is important for two reasons. First, pulse-type
motions have been identified as critical in the design of structures in the near-fault region
(e.g. Krawinkler and Alavi 1998, Sasani and Bertero 2000). Preliminary analysis of
elastic and inelastic multiple degree of freedom systems indicated that the amplitude and
period of the pulse are parameters that control the demand on the structure. Site effects
have the potential to significantly alter these parameters. The second important
motivating factor lies in the evaluation of existing near-fault records. Recent events, such
as the 1999 Kocaeli, Turkey, and Chi-Chi, Taiwan earthquakes, have increased
significantly the available data for large magnitude earthquakes in the near-fault. As
those data become available, improved empirical characterization of near-fault ground
motions is possible. A complete understanding of the effects of site conditions on these
105
motions will greatly aid in the development of predictive relationships for near-fault
events.
This chapter presents an evaluation of the currently available recordings of near-
fault, forward-directivity motions. Section 4.2 presents a review of the current
attenuation relationships used to predict near-fault effects in the frequency domain.
Section 4.3 studies the time domain representation of forward-directivity, near-fault
motions. Section 4.4 presents the development of base-line pulse-motions to be used as
input for site response analyses in Chapter 6. All the analyses in the remaining parts of
this work concentrate on pulses generated by forward-directivity effects.
4.2 FREQUENCY DOMAIN REPRESENTATION OF NEAR-
FAULT GROUND MOTIONS
In current practice, rupture directivity effects are generally taken into account by
modifications to the elastic acceleration response spectrum (at 5% damping). A detailed
model for the amplitude and duration effects of rupture directivity is presented by
Somerville et al. (1997). This model is widely used in conjunction with attenuation
relationships for the estimation of ground motions in the near-fault region. Besides from
its application on ground motion estimation, the model is important in its development of
a parameterization of the geometric conditions that lead to forward and backward
directivity conditions.
The ground motion parameters that are modified to account for the effects of
directivity are the average horizontal response spectra, the ratios of fault normal to fault
106
parallel response spectra, and the duration of the ground motion. The model parameters
used to define the geometric conditions for rupture directivity are illustrated in Figure 4.2.
The spatial variation of directivity effects depends on two parameters. First, the angle
between the direction of rupture propagation (θ for strike-slip faults, and φ for dip-slip
faults) and the direction of waves travelling from the fault to the site. Second, the
fraction of the fault rupture surface (X for strike-slip faults and Y for dip-slip faults) that
lies between the hypocenter and the site. The smaller the angle, the larger the directivity
conditions that are experienced at a site. Similarly, if a larger fraction of the fault lies
between the site and the hypocenter, the effects of directivity are larger. Somerville et al.
(1997) chose to model rupture directivity effects using the functions Xcosθ and Ycosφ for
strike-slip and dip-slip fault, respectively. The equations developed by Somerville et al.
(1997) are presented in Table 4.1 and their effect on spectral amplification ratios are
illustrated in Figure 4.3. The parameters that identify forward-directivity conditions in
the Somerville model can also be used to identify sites with a potential for experiencing
forward-directivity conditions. This is discussed further in Section 4.3.
The 1997 UBC accounts for near-fault effects by means of near-fault factors Na
and Nv applied to the low period (acceleration) and intermediate period (velocity) parts of
the acceleration response spectrum, respectively. The near source factors are specified
for distances less than 15 km and for three different fault types (Table 4.2). The near-
source factors in the UBC are compatible with the average of the fault normal and fault
parallel component in the Somerville et al. (1997) model. However, the code does not
specifically address the larger ground motions in the fault normal component (Somerville
1998).
107
4.3 TIME DOMAIN REPRESENTATION OF NEAR-FAULT
GROUND MOTIONS
4.3.1 General
Ground motion recordings at sites subject to forward-directivity are characterized
by the arrival of most of the seismic energy in a single large pulse of motion at the
beginning of the record (Somerville 1997). As described earlier, this large pulse of
motion is oriented in the direction perpendicular to the strike of the fault. Although some
design codes characterize near-fault ground motions by means of an amplified
acceleration response spectrum, there is a growing recognition that a time history
representation is better able to capture the effects of near-fault ground motions on
structures (e.g. Somerville 1998, Alavi and Krawinkler, 2000, Sasani and Bertero, 2000).
A time-domain representation is desired because the frequency domain characterization
of ground motion (i.e. through a response spectrum) implies a stochastic process having a
relatively uniform distribution of energy throughout the duration of the motion. When
the energy is concentrated in a single pulse of motion, the resonance phenomenon that the
response spectrum was conceived to represent has insufficient time to build up
(Somerville 1998). Somerville (1998) states:
… the inadequacy of the response spectrum as a sole design

criterion becomes readily apparent when time histories are selected
to represent a design response spectrum. It is well known that a
suite of different ground motion time histories which all match the
same response spectrum produce variations in the response of a
structure subjected to non-linear time history analysis. However,
when the input time history is a near-fault pulse this effect is
accentuated to the point where small modifications of a near-fault
108
time history that have no significant effect on the response
spectrum can have a major effect on the response of a structure
when subjected to non-linear time history analysis. This
demonstrates that the current standard of practice does not provide
a reliable basis for providing near-fault ground motions time
histories that are specified solely on the basis of a design response
spectrum.
Studies by structural engineers, such as Krawinkler and Alavi (1998) and Sasani
and Bertero (2000), have shown that simplified representations of the velocity pulse are
capable of capturing the salient response features of structures subjected to near-fault
ground motions. The model by Krawinkler represents the observed pulses in a velocity
time history by means of the zero crossings and peak ground velocities of the dominant
pulses in a ground motion. The peaks can be joined by straight lines or can be fitted by
means of a spline to obtain a better fit (Somerville 1998). The period of the equivalent
pulse in Krawinkler's model is identified by a clear and global peak in the velocity
response spectrum of the ground motion. The equivalent pulse amplitude is obtained by
minimizing the differences between the maximum story ductility demand from the near-
fault record and the corresponding demand obtained from the equivalent pulse for a
certain range of ductility. Results indicated that in almost all cases, the equivalent pulse
velocity lies within 20% of the peak ground velocity of the record (Alavi and Krawinkler
2000). The simplified pulses developed by Krawinkler and Alavi (1998) and Sasani and
Bertero (2000) are shown in Figure 4.4.
These previously mentioned studies point to the importance of studying near-fault
motions in the time domain. In particular, structural response is sensitive to the long
period pulses that are best observed in a velocity time history. In the following sections,
109
the available empirical database of near-fault ground motions is evaluated with two main
objectives. First, to obtain a characterization of forward-directivity motions that includes
effects of site conditions, and second, to develop simplified velocity pulses that will be
used as input motions in site response studies. The results of the site response studies
will be validated with the trends observed in the data.
4.3.2 Parameterization of velocity pulses
Velocity time histories of near-fault ground motions can be satisfactorily
approximated by means of simplified pulse shapes. For the sake of simplicity, sine
functions will be used to represent the observed pulses in velocity time histories. Figure
4.5 illustrates the fit of sine pulses to selected near-fault recordings. The advantage of
using sine functions is that velocity pulses can be characterized by a reduced number of
parameters. This, in turn, will facilitate a systematic study of the effects of site response
on the characteristics of velocity pulses. Section 6.3 presents a detailed site response
analysis using the motions in Figure 4.5 as input motions. The site response analyses are
done using both the recorded and the simplified pulses as input motions. Results are
remarkably similar. This serves as further validation for the use of simplified pulses for
the study of site response in the near-fault region.
The larger fault-normal component of motion is considered critical and is
commonly used for structural response studies (Alavi and Krawinkler 2000). Ground
motion in the fault parallel direction, however, can also be significant and can induce
large strains in an affected soil column. Consequently, both directions of motion will be
considered in developing simplified velocity time histories. This is not to say that the
110
fault normal component is not the dominant component in near-fault recordings, but that
significant fault-parallel velocities are sometimes observed to occur in phase with fault-
normal pulses. Figure 4.6 illustrates two different recordings that are both oriented in the
fault normal direction, but have significantly different fault parallel components of
motion. The use of both components of motion will allow the evaluation of the effect of
fault parallel motions in site response. Figure 4.6 includes horizontal velocity-trace plots.
The horizontal velocity-trace plot is a plot of the fault parallel versus the fault normal
components of motion. It represents the magnitude and the orientation of the particle's
velocity during the seismic event. Horizontal velocity-trace plots illustrate well the
orientation of velocity pulses.
The simplified sine-pulse representations of velocity time histories are fully
defined by the number of equivalent half-cycles of motion, the period of each half-cycle,
and their corresponding amplitudes. In general, simplified velocity pulses can be defined
by their peak amplitude or peak ground velocity (PGV), approximate period of the
primary pulse, and the number of half-cycles of pulse motions. The time lag between the
initiation of the fault normal and fault parallel component must also be defined in
addition to the equivalent sine-pulse representation of the fault parallel component.
Figure 4.7 illustrates the parameters needed for a full characterization of the simplified
time histories. In the following section, the methodology to obtain the equivalent sine
pulse parameters is described.
111
Equivalent Pulse Representation of Ground Motions
Near-fault, forward-directivity records were selected from the ground motion
database described in Chapter Three (Silva, personal comm.). This ground motion
database was complemented with records from the 1999 Kocaeli, Turkey. Processing of
these recent records is described in Rathje et al. (2000). The parameters describing
directivity conditions for each station were compiled by Stewart (personal
communication). Forward-directivity conditions were determined by means of the
predicted ratio of fault normal to fault parallel spectral acceleration at a period of three
seconds. This ratio, termed RDI by Stewart (personal comm.) is predicted by an updated
version of the Somerville et al. (1997) model (Table 4.1). For strike-slip faults it is given
by (Stewart, personal comm.):
exp(− 0.605+ 2.506 X cos φ ) TR (R )TM (M w ) if X cosθ < 0.4

RDI =  (4.1a)
exp(0.395) TR (R )TM (M w ) otherwise
and for dip-slip faults by:
RDI = exp(− 0.327 + 0.559Y cos φ ) TR (R )TM (M w )End (4.1b)
where X, Y, θ, and φ are defined in Figure 4.2; R is closest distance to the fault plane
(Sadigh et al. 1993); Mw is moment magnitude; End is 0 if the station is on the ends of the
112
fault and 1 otherwise (Figure 4.2); and TR and TM are taper functions given by (Stewart,
personal comm.):
1 if R < 30 km
 R - 30
TR ( R) = 1 - if 30 km ≤ R < 60 km (4.2a)
 30
0 otherwise
1 if M w > 6.5
 M - 30
TM ( M w ) = 1 - w if 6.0 km < M w ≤ 6.5 (4.2b)
 30
0 otherwise
Geometric conditions for forward-directivity exist for an RDI larger than 1.0. It is
important to note, however, that even when the geometric conditions for forward-
directivity are satisfied, the effects of forward-directivity may not exist. This could
happen if a station is at the end of a fault and rupture occurs towards the station (i.e.
forward-directivity conditions), but slip is concentrated near the end of the fault where the
station is located. Under these conditions, the forward-directivity pulse may not be
generated. Note, however, that forward-directivity conditions occur also from up-dip
rupture propagation. Hence, in a strike slip event where the hypocenter is at depth,
forward-directivity conditions might arise, especially if dip-slip movement along the fault
occurs also.
Earthquakes with moment magnitude equal or larger than 6.1 were considered.
Table 4.3 lists the earthquakes included in the study, along with the fault strike used to
determine fault normal orientation. Stations at a distance (closest distance to the fault
113
plane) less than or equal to 20 km for earthquakes with Mw ≥ 6.5 and 15 km for lower
magnitude events, and an RDI larger than 1.0 were considered. In addition, recordings
not possessing at least some features of forward-directivity characteristics were excluded
from the analysis. Forward-directivity characteristics are positive fault normal to fault
parallel response spectral ratios for long periods, and a reasonably well-defined velocity
pulse in the fault normal direction.
Simplified sine-pulses matching the dominant pulses of the selected ground
motions were developed using the characterization illustrated in Figure 4.7. In addition to
the time domain parameters of the pulse, the period corresponding to a clear and global
peak of the pseudo-velocity response spectra was also calculated for each recording (see
Figure 4.10). Table 4.4 details the methodology used to obtain the parameters in Figure
4.7. Subscripts N and P are used to describe the fault-normal and fault-parallel
components, respectively. While forward-directivity motions generally have well defined
velocity pulses in the fault-normal direction, the fault-parallel direction is not always
pulse-like. Nonetheless, the fault-parallel motion was also fit using the simplified sine-
pulse representation. The simplified representation of these motions have the effect of
filtering out the high frequency content of the pulses. The amplification of velocities in a
site response analysis is affected mainly by intermediate period energy. Consequently,
the simplified sine-pulse representation of the fault parallel motions constitutes a
reasonable scenario when these are used as input motions for a site response analysis. For
characterizing the fault parallel direction, only the time interval in which the pulses occur
for the fault normal component is considered. In most cases, this time interval contains
the peak velocity values in the fault parallel direction. Tables 4.5 and 4.6 lists the
114
selected recordings with their corresponding simplified sine-pulse parameters. Table 4.7
lists the stations satisfying the geometrical conditions of forward-directivity that were
excluded from the analysis for not showing the characteristics of forward-directivity
motions.
Previous studies of structural response indicated that pulse period is an important
parameter to characterize the response of structures to pulse-type motions. In order to
simplify the analysis of the empirical database, a single pulse period associated with each
recording is desired. This simplification is justified in large part by the fact that the
energy of forward-directivity pulses is constrained to a relatively narrow band of
frequencies (Somerville 2000).
Three options were considered for the pulse period:
a) The period of the pulse with the maximum amplitude.
b) The weighted average period, which is calculated as the sum of the periods of each
half-cycle of motion weighted by the pulse amplitude and divided by the number of
half-cycles of motion.
c) The period corresponding to a clear and global peak (i.e. maximum value) in the
pseudo-velocity response spectra (Tv-p).
Figure 4.8 shows the relationship between the period of the pulse with maximum
amplitude and the weighted average period. The mean ratio of these two parameters is
0.99 with a standard deviation of 0.19. In almost all cases, the weighted average period is
within 20% of the period of the maximum pulse. Given the large uncertainty associated
with predictions of Tv, these two parameters can be used interchangeably. Note that the
115
definition of pulse period in Table 4.4 uses either the zero crossing time or the time at
which velocity is equal to 10% of the peak velocity for this pulse. This is necessary for
pulses in which the pulse is preceded by a small drift in the velocity time history (Figure
4.9). A degree of subjectivity is involved in this definition and can lead to some
variations in the estimates of Tv. However, as will be shown later, the uncertainty
associated with predicting Tv is much larger than the possible errors in estimating Tv from
zero crossings. On the other hand, a parameter such as Tv-p is relatively unambiguous
(Figure 4.10). Table 4.8 shows the ratio of Tv-p to Tv for events for which there were
multiple near-fault recordings. For a velocity time-history defined by a one-cycle sine
pulse, Tv-p is approximately equal to Tv (i.e. Tv-p = 0.95Tv). For the recorded time
histories, the two measures of pulse period coincide for events characterized by one or
two dominant pulses of motion in the velocity time histories (i.e. Imperial Valley and
Northridge). For events with more complex velocity time histories (i.e. Loma Prieta and
Kobe) Tv-p is significantly lower than Tv. The overall average of the ratio between Tv and
Tv-p is 0.84 with a standard deviation of 0.28. A relationship between the two parameters
is plotted in Figure 4.11. Due to potential differences in the magnitude of Tv and Tv-p,
values of both parameters are obtained for the stations listed in Table 4.5. Given that the
emphasis of this work is the time-domain characterization of near-fault forward-
directivity motions, the parameter Tv will be emphasized. The choice also allows a better
match to recorded motions in the time domain.
The coincidence of Tv and Tv-p indicates that the velocity pulse in the ground
motion contains energy in a narrow period band. Figure 4.12 shows the normalized
power spectral density (PSD) of four ground motions. The El Centro records from the
116
Imperial Valley earthquake have equal values of Tv-p and Tv. The PSD of these records
shows a clear peak at a frequency close to the frequency of the pulse. The fact that the
peak in the PSD and the pulse frequency (1/Tv) do not coincide is because the pulse is not
a perfect sine pulse. Also included in Figure 4.12 are the Los Gatos Presentation Center
(LGPC) record from the Loma Prieta earthquake and the Pacoima Dam Downstream
record from the Northridge earthquake. These records have significantly lower Tv-p than
Tv. Even tough the PSD of these records has a dominant peak, the motions also contain
energy at higher frequencies. This implies that the motion is not dominated by a single
pulse. The fact that the peak in the spectral velocity plot does not coincide with the
period of the largest pulse is not surprising for these types of records.
4.3.3 Statistical Evaluation of Equivalent Pulse parameters in the Fault Normal
Direction
Number of Significant Pulses
The number of significant pulses in the velocity time-history is an important
parameter for structural response. Multiple cycles of motion can dramatically increase
the damage potential of the ground motions (Alavi and Krawinkler 2000). In the present
study, the number of cycles of motion (referred to as the number of significant pulses) is
defined as the number of half-cycle velocity pulses that have an amplitude at least 50% of
the peak ground velocity of the ground motion (Table 4.4). For evaluating the number of
significant velocity pulses, only the fault-normal component of motion is considered.
Table 4.9 lists the number of significant pulses for the recordings in each of the
earthquakes included in this study. Figures 4.13 to 4.23 show the velocity-time histories
117
for all the records used in the analysis. A tabulation of 15 recorded near-fault motions by
Somerville (1998) rendered a similar proportion of number of pulses, where the number
of pulses was defined using the equivalent motion model of Krawinkler (1998).
The 50% level chosen as a cut-off is arbitrary, however, the values listed in Table
4.9 are useful as a guideline for estimating the number of significant velocity pulses in an
earthquake. The number of pulses within the initial pulse sequence with a 33% cut-off is
also included in Table 4.9 to illustrate the sensitivity of number of significant pulses to
the cut-off level. It is important to note that, in general, each earthquake event has a well-
defined pulse sequence for nearly all of its near-fault motions. For example, most of the
recordings in the Imperial Valley earthquake consist of two significant velocity pulses or
a full velocity cycle of motion (Figure 4.15). With the exception of the OSAJ motion, all
the records in the Kobe earthquake consist of four significant velocity pulses or two full
velocity cycles (Figure 4.22). This might be expected for faults that have a relatively
uniform slip distribution or earthquakes where slip is concentrated over a single zone.
For these types of earthquakes, stations that are close to each other will be equidistant to
regions of high slip. Moreover, path effects are minimized for stations in the near-fault
region. For an earthquake with highly non-uniform slip, such as the Northridge
earthquake, the type of pulse sequence observed depends on the instrument distance
relative to the asperities. In fact, Somerville (1998) suggests that the number of half sine
pulses in the velocity time-history might be associated with the number of asperities in a
fault. From the point of view of ground motion prediction, this implies that the prediction
of number of significant velocity pulses in a given earthquake is associated with the
118
determination of slip distribution in the causative fault. This of course is a difficult thing
to estimate a priori.
Pulse Period
The equivalent pulse period (Tv) of the velocity time-history and the period
corresponding to the peak of the pseudo-velocity response spectra (Tv-p) were evaluated
for all the sites listed in Table 4.5. In addition, each recording site was classified either as
rock (r) or soil (s). The rock or soil classification is along the lines of the simplified
classification schemes described in Chapter Three (Abrahamson and Silva 1997). In
terms of the geotechnically-based classification system introduced in Chapter Three, the
rock category includes competent rock sites (Site B) and shallow stiff soil/weathered rock
sites (Site C), while the soil category comprises deep stiff clay sites (Site D) and some
soft clay sites (Site E). Note that the grouping of competent rock sites and shallow stiff
soil/weathered rock sites into a single category increased the standard deviations in the
prediction of site response for motions at intermediate to long distances (see Chapter
Three). In light of this, the classification presented herein seems counter-intuitive.
However, results in Chapter Six indicate that for the study of long period velocity pulses,
shallow stiff soil/weathered rock sites do not modify significantly the input rock velocity
pulse. This observation holds only for long pulse periods. Given that the present analysis
concentrates on velocity pulses that typically have long periods, it was deemed
appropriate to follow this simplified rock vs. soil site classification scheme.
Somerville (1998) performed a regression analysis using data from 15 recorded
time histories augmented by 12 simulated time histories. The records correspond to a
119
magnitude range of 6.2 to 7.5 and distances of 0 to 10 km. The pulse parameters modeled
were the period and amplitude of the largest cycle of motion in the velocity time-history.
The relationship obtained is
log10Tv = -2.5 + .425 Mw (4.3)
where Tv is the period of the largest cycle of motion (Somerville 1998). In a larger study
of slip distributions using slip models for 15 earthquakes, Somerville et al. (1999) provide
justifications to the use of self-similar scaling relationships to constrain fault parameters.
In a self-similar system, events of different sizes cannot be distinguished except by a scale
factor. Using this self-similar scaling model, the magnitude scaling in the above
relationship is 0.5 and the resulting equation is
log10Tv = -3.0 + .5 Mw (4.4)
The period of the velocity pulse is associated with the rise time of slip in the fault,
which measures the duration of slip at a single point on the fault. Somerville et al. (1999)
performed a regression analysis on the slip duration or rise time using slip models for 15
earthquakes. The relationship obtained is
log10TR = -3.34 + 0.5 Mw (4.5)
120
where TR is the rise time. This relationship is also constrained by the self similar scaling
model. Joining Equations 4.4 and 4.5, the relationship between pulse period and rise time
is (Somerville 1998)
Tv = 2.2 TR (4.6)
The relationship between pulse duration and rise time can also be inferred from
the physics of the fault rupture phenomenon. If a fault is modeled as a point and
propagation effects are ignored, the duration of motion would be equal to the rise time
(Somerville 1998). Fault finiteness and propagation effects contribute to widening the
pulse. Rise time is then, in essence, a lower bound for pulse period.
A similar study relating pulse period to moment magnitude was presented by
Alavi and Krawinkler (2000) on the same data set used by Somerville (1998). Alavi and
Krawinkler defined the pulse period as the predominant period in a velocity response
spectrum plot (Tv-p). The relationship obtained is
log10Tv-p = -1.76 + 0.31 Mw (4.7)
Equations 4.6 and 4.7 do not account for the potential effect of local site
conditions on the pulse period. Moreover, a measure of the uncertainty associated with
the estimate of Tv and Tv-p is not provided.
121
The records listed in Table 4.5 are used to develop a relationship between Tv and
Mw. A linear relationship between the logarithm of pulse period and moment magnitude
is assumed, consistently with the relationship by Somerville (1998). The number of
records for each earthquake varies from just one in some cases to 13 for the Imperial
Valley earthquake. To avoid a relationship that is controlled by the few events with a
large number of records, a random effects model is used (Abrahamson and Silva 1997).
The random effects model partitions the standard error associated with each data point
into an inter-event term (defines that part of the overall standard error resulting from
scatter between events) and an intra-event term (defines that part of the overall standard
deviation resulting from scatter within each event). Thus, the relationship for pulse
period becomes
ln(Tv)ij = a + bMw + ηi + εij (4.8)
where (Tv)ij is the pulse period of the jth recording from the ith event, a and b are the
model parameters, ηi is the inter-event term, and εij represent the intra-event variations.
The relationship is valid for moment magnitudes 6.1 to 7.4 and for distances less than 20
km. The inter-event and the intra-event error terms are assumed to be independent
normally distributed random variables with variances τ2 and σ2, respectively. The
standard error associated with the estimate of Tv is then
σtotal2 = τ2 + σ2 (4.9)
122
The algorithm presented by Abrahamson and Youngs (1992) is used for the
regression analysis. The analysis is performed both on the whole data set, and on the
subsets of rock and soil motions separately. The values of the model parameters resulting
from the analyses are presented in Table 4.10 and illustrated in Figure 4.24. Note that
this relationship was developed with data from earthquakes with moment magnitudes
ranging from 6.1 to a maximum of 7.4. Equation 4.8, along with the parameters in Table
4.10, should be used only within this magnitude ranges. Moreover, the amount of data at
higher magnitudes is limited, so this predictive relationship should be applied with
caution. Finally, all the data are restricted to within 20 km from the fault plane, so the
resulting relationship should not be used for greater distances.
The random effect term associated with inter-event variations can be used to
evaluate the deviation from the median relationship for individual events. For given
model parameters, ηi is given by (Abrahamson and Youngs 1992)
ni
τ 2
∑ yij − µ ij
j =1
ηi = (4.10)
niτ 2 + σ 2
where ni is the number of records for event i, yij is the jth recording for the ith event, µij is
the pulse period predicted by the model (Equation 4.8), and the summation index j
represents a summation over all the recordings of event i. Equation 4.10 illustrates how
the random effect model partitions the error into inter and intra-event terms. For an event
123
with a single recording (ni = 1), the percentage of the residuals that is assigned to the
inter-event term is given by the ratio
τ2/(τ2 + σ2) (4.11)
If the number of recordings for a single event is large (ni >> 1), the inter-event term
becomes the mean residual for event i. Table 4.11 gives the values of ni for each of the
11 earthquakes in the data set. Observe that the event term for the Imperial Valley
earthquake is much larger than those for the other events. This in particular illustrates the
need for an adequate way of partitioning the standard error between intra and inter-event
terms. If the inter-event error term is ignored and error is partitioned individually among
all recordings, the Imperial Valley records would control the regression data.
A comparison of the relationship developed herein with those of Somerville
(1998) and Alavi and Krawinkler (2000) is shown in Figure 4.25. The definition of pulse
period of Somerville is similar to that use in this study (Tv). On the other hand, the pulse
period of Alavi and Krawinkler (2000) is the period corresponding to a clear maximum in
the pseudo-velocity response spectra (Tv-p). For comparison purposes, Figure 4.25
includes a regression line for Tv-p from this study. The parameters for the regression are
included in Table 4.10. For lower magnitudes, the relationships developed in this study
render lower pulse periods than those of Somerville and Alavi and Krawinkler. For larger
magnitudes (Mw > 7.0), the regression line for Tv matches that of Somerville, while the
regression line for Tv-p matches the Alavi and Krawinkler line.
124
The relationship illustrated in Figure 4.24 has important implications both for
design and for the further evaluation of the existing database of near-fault motions. The
regression analysis predicts longer periods at soil sites than at rock sites for lower
magnitude earthquake events. This difference diminishes as magnitude increases and
disappears for large magnitudes. From the point of view of structural design, the longer
periods predicted at soil sites for earthquakes of low to intermediate magnitudes might
lead to different design considerations when dealing with near-fault ground motions.
From the point of view of ground motion prediction, Figure 4.24 illustrates the
importance of considering local site effects in the evaluation of near-fault ground
motions. This becomes particularly relevant for the development of future attenuation
relationships that include predictions of near-fault effects.
The significant difference in pulse periods predicted for different magnitude
earthquakes also raises important questions for ground motion estimation. For many
short-period structures, the expected large pulse periods from large magnitude
earthquakes may not produce significant levels of damage for these structures. Lower
magnitude earthquake events may result in velocity pulses with periods closer to the
natural period of short-period structures, which are often more common in urban building
stocks. In this case, the lower magnitude earthquake may result in larger levels of
damage associated with the velocity pulse. Following the same reasoning, ground
motions recorded at near-fault sites for large magnitude earthquakes cannot be assumed
to be “worst-case scenarios” for the design of all structures. It is likely that in the future
development of attenuation relationships, accounting for near-fault effects could result in
125
larger spectral accelerations for lower magnitude earthquakes at periods of 1 to 2 seconds,
especially at soil sites.
The standard deviations of the model in Equation 4.8 are relatively high. Figure
4.26 shows the predictive relationships for the median and the plus and minus one
standard deviation values for soil and rock. The standard deviation associated with the
prediction of Tv is larger than the predicted difference between soil and rock sites for the
high magnitude (long period) range. For these cases, however, the regression analysis
indicates a diminished difference between pulse period in soil and rock. However, the
difference in pulse period between soil and rock sites at low magnitudes is significant and
compares with the standard deviation of the estimates.
Additional evidence of period elongation due to site effects can be obtained by
looking at paired stations for individual events. However, in the database of near-fault
motions, the number of paired (rock and soil) stations that are relatively close to each
other and have similar pulse sequences are limited. For the present analysis, the Gilroy
motions from the 1989 Loma Prieta earthquake are selected. Parameters for these
stations are included in Table 4.5 and are repeated for this comparison in Table 4.12.
Figure 4.27 shows the time history and pseudo-velocity response spectral plots for
these motions. The observed fault-normal pulse sequence has the same characteristics, an
initial half-cycle pulse followed by two full cycles of motion. The corresponding pulse
periods, however, vary significantly depending on the soil type, as indicated in Table
4.12, with soil sites having larger velocity pulse periods than rock sites.
126
Peak Ground Velocity
The random effects model (Abrahamson and Youngs 1992) was also used to
evaluate the PGVs listed in Table 4.5. PGV is affected by both magnitude and distance.
Attenuation relationships for PGV typically use elaborate functional forms to match the
data over all distance ranges (e.g. Campbell 1997). For near-fault ground motions,
however, the data, by definition, are restricted to relatively short distances, and the
functional forms can be simplified. Somerville (1998) proposed the use of a bilinear
relationship between the logarithm of PGV, magnitude, and the logarithm of distance. To
avoid unrealistic predictions of PGV at short distances, Somerville (1998) used a distance
cut-off at 3 km. A similar relationship was later proposed by Alavi and Krawinkler
(2000). For the study of near-fault data, however, it is desirable to obtain predictions of
PGV at close distances to the fault as well. For this reason, the following functional form
was used:
ln(PGV) = a + b Mw + c ln (R2 + d2) (4.12)
where PGV is in units of cm/s. This functional form results in a nearly zero slope at close
distances to the fault. The relationship becomes linear at larger distances. Other
functional forms, including a functional relationship on magnitude for the distance
scaling term c were attempted without significant improvements on the match of the data.
The functional form given by Equation 4.12 was chosen based on its simplicity.
127
The robustness of the predictions of Equation 4.12 depends on the distribution of
the data both with distance and with magnitude (Figure 4.28). Observe that data are
scarce for larger magnitude earthquakes, thus magnitude scaling is controlled by data for
earthquakes with Mw ranging from 6.5 to 6.9. The parameters of the attenuation
relationships are listed in Table 4.13 and Table 4.14; and the resulting attenuation
relationships for soil and rock are illustrated in Figure 4.29. This attenuation relationship
is valid only within the distance and magnitude ranges of the data, that is, distances < 20
km and magnitudes of 6.1 to 7.4. Observe that the distance scaling term, c, is only
slightly larger for rock data set than for the soil data set, implying that the PGV is not
largely affected by intensity of motion. However, the magnitude scaling term in the rock
attenuation relationship is much larger than in the soil attenuation relationship, implying
that the ratio of PGV in soil to PGV in rock decreases as magnitude increases (Figure
4.30). Since magnitude is related to pulse period, this implies that for longer period
motions, there is lesser amplification of PGV in soils. However, due to the large scatter
in the data, the low amount of rock sites in the data set, and the poor distribution of data
across magnitude, caution must be exercised when interpreting these observations.
Numerical simulations of near-fault site effects are needed to complement the empirical
database due to the lack of data. Chapter Six deals in length with the effects of soil on
PGVs.
Figure 4.29b compares the relationship in Equation 4.12 with the relationships
developed by Somerville (1998) and Alavi and Krawinkler (2000). These researchers
augmented the empirical database with 12 simulated ground motions. The regression
analysis in Equation 4.12 differs from the other relationships mainly in the magnitude
128
scaling term. Somerville (1998) and Alavi and Krawinkler (2000) propose a much
stronger variation of PGV with distance. The variation cannot be attributed to the
addition of the simulated time histories because Somerville (1998) indicates that the PGV
of the recorded time histories grows more rapidly with magnitude than for the simulated
time histories. The differences are likely due to the amount of data included. The current
work includes a larger database.
4.3.4 Characteristics of the fault parallel component of motion
The previous analysis dealt with the identification of the salient characteristics of
the fault normal component of near-fault ground motions. As previously indicated, this
component is typically larger than the fault-parallel component, and thus is commonly
used as input for structural response studies. The fault parallel component, however, can
also be significant as shown for example in Figure 4.6. The different types of pulses
resulting from varying intensities and frequency characteristics of fault parallel ground
motions are best observed by looking at a particle velocity-trace plot of the horizontal
component of motion. In this section, the different pulse shapes from near-fault ground
motions are evaluated.
No particular trend in the ratio of peak ground velocities in the fault parallel and
fault normal directions (PGVP/N) with respect to varying magnitude or distance is
observed. However, a trend of decreasing PGVP/N with increasing fault parallel velocity
is inferred from the available data (Figure 4.31). The trend is stronger for rock motions.
With the exception of one outlier, the PGVP/N ratio in rock is lower than 0.5 for fault
normal PGV larger than about 50 cm/s. Although the variability in the PGVP/N ratio is
129
large both for rock and soil sites across all values of fault normal PGV, it is important to
point out that for large values of fault normal PGV, this variability is larger for soil sites
than for rock sites. This variability might be due to the effects of site amplification. For
fault normal PGV values greater than 100 cm/s, the PGVP/N ratios in soil are generally
larger than in rock, while for lower values of fault normal PGV, no particular trend with
site condition is observed. Over all the records, the PGVP/N ratio is about the same
whether rock or soil records are taken. This average value is 0.64 with a standard
deviation of 0.26.
While the fault-normal component of velocity is generally well characterized by
simple pulses, the fault parallel pulse sometimes shows more irregular velocity time
histories. Nonetheless, a clearly defined pulse in the fault parallel direction exists in
many cases. Moreover, this pulse normally occurs during the same window of time as the
fault normal pulse occurs. Table 4.15 gives the ratio of pulse period in the fault parallel
direction to pulse period in the fault normal direction for different earthquakes. As can be
observed in these tables, the ratios have a mean value of 0.75 and a standard deviation of
0.33. In general, the fault parallel period is shorter than the period in the fault normal
direction.
4.3.5 Development of bi-directional simplified pulses for use as baseline input
motions
The preceding sections dealt with the development of simplified representations
of fault normal pulses (Section 4.3.3), and the description of general characteristics of
fault parallel pulses (Section 4.3.4). This section deals with the development of bi-
130
directional pulses, that is, when both fault-normal and fault-parallel components of
motion are defined simultaneously. As previously indicated, typical structural analysis
procedures deal only with the fault normal component of motion, this one being
considered the controlling component for forward-directivity motions. The recordings on
Figure 4.6, however, illustrate that forward-directivity motions can also have significant
peak ground velocities in the fault parallel direction, particularly when lower input
velocities are involved. From the point of view of site response, the fault-parallel
component of motion can lead to larger levels of strain in the soil column, thus leading to
differences in site response. For this reason, a definition of bi-directional input pulses is
desired. Note that prescribing a fault-normal and fault-parallel velocity results in a given
pulse shape in the horizontal velocity-trace plot. For ease of notation, the term “pulse
shapes” will be used to define the shape of velocity pulses in a fault-normal versus fault-
parallel horizontal velocity plot.
The bi-directional horizontal velocity pulses developed in this section are used as
input motions for the site response analyses presented in Chapter Six. Ideally, input
pulses would be taken only from records on outcropping rock. The available number of
rock records, however, is relatively small and excluding soil sites proved to be too
restrictive. For this reason, the analysis of pulse shapes will be performed for all records
independently of their site classification. Note, however, that the pulse parameters used
in baseline input ground motions (e.g. PGV, PGVP/N, Tv) will be estimated from the
regression curves for rock motions.
131
A selected number of the records listed in Table 4.5 were chosen for a more
detailed look at the pulse shape. A list of these motions is given in Table 4.16. The
selected motions are those that have all of the characteristics of forward-directivity
motions, that is:
- Most of the energy is concentrated on the initial pulse. This can be
determined by examining the plot of arias intensity versus time (Husid plot).
If most of the energy is concentrated in the velocity pulse, the slope of the
Husid plot is large for the time duration of the pulse. Quantitatively, the
condition stated above (most of the energy is concentrated on the initial pulse)
can be translated into the condition that at least 50% of the Arias Intensity
occurs before the end of the pulse.
- Positive fault normal to fault parallel spectral ratios for periods in the
neighborhood of the pulse period (in general, ratios are positive for periods
longer than one second).
Moreover, motions that have pulses not amenable to fitting with simple sine
pulses were not included in this list. Note that this creates an automatic bias towards
simpler pulses. This is inevitable if simplicity in the equivalent pulses is desired.
Motions that are not fit well by simple sine pulses (such as the Gilroy Gavilan College
record for the Loma Prieta earthquake) are also used without modification as input
motions for the analyses in Chapter Six. This is done in order to verify that the selection
of simple pulses does not adversely affect findings regarding site response.
132
Seven motions with a dominant half-cycle of motion (with amplitude at least two
times larger than the amplitude of the remaining pulses) are shown in Figure 4.32. For
these eight motions, the ratio of fault parallel pulse period to fault-normal pulse period
ranges from 0.24 to 1.12, with an average value of 0.5. The ratio of peak velocities in the
fault-parallel to the fault-normal direction ranges from 0.21 to 0.7, with an average of 0.3.
Figure 4.33 presents motions with a pulse consisting of a full cycle of motion.
Fifteen such motions were studied in detail. In general, the positive and negative
amplitude of the half-cycle of motion are equal. Moreover, the periods of each half cycle
are within 10% of each other in most of the motions. The ratios of fault parallel
amplitude to fault normal amplitude, as well as the period ratios (Tv,FP/Tv,FN) fall within
the ranges of all the motions as presented in Section 4.3.4. Simplified representations for
the dominant velocity pulses of some of these motions are shown in Figure 4.34.
In light of the preceding discussion, the simplified pulses in Figure 4.35 were
developed. These pulses have fault parallel period ratios (Tv,FP/Tv,FN) consistent with the
records in Figure 4.32 and 4.33. Beyond the definition of the fault parallel and fault
normal ratios, the time delay (toff) between the initiation of the pulse in each direction is
also needed for the full definition of the pulse shape. The data do not suggest that there is
a particular trend for this parameter. The parameter was arbitrarily selected so that either
of the following three conditions occur: a) both motions start at the same time (toff = 0), b)
the fault parallel motion initiates when the fault normal is at a maximum (toff = TvN/4), or
c) the resulting maximum amplitude in the fault normal and fault parallel motion are
reached simultaneously. Preliminary site response analyses with these pulses indicated
133
that Pulses 6, 7 and 8 cover the range of fault normal period elongation and amplification
of peak ground velocity. Consequently, these pulses were selected as the baseline input
motions for the analyses in Chapter Six.
Beyond the pulse shape, the pulse parameters must also be chosen for the input
motions. These were selected from the regression analyses for rock motions. In general,
the following criteria was used:
• Fault normal PGV ranges from 75 cm/s to 300 cm/s. The latter extreme value is
a judgmental choice to cover potential extreme values that can result from large
magnitude earthquakes. Records from the 1999 Chi-Chi earthquake in Taiwan
showed peak ground velocities in this range of values. These velocity peaks,
however, include slip effects (fling-step) and are not due solely to the effects of
forward-directivity (Abrahamson, personal comm.). Nonetheless, the value of
300 cm/s was chosen to a reasonable assumption for an upper limit on
velocities. Note that a PGV of 300 cm/s is also the median prediction of the
attenuation relationships for PGV in rock sites for a magnitude of 7.5 and a
distance of 0 km, and the plus one standard deviation prediction for a magnitude
of 7.0 and a distance of 1 km.
• The fault period in the fault normal direction was varied from 0.6 to 4.0
seconds. The lower value was also used by Somerville et al. (1997) in the
empirical analysis of directivity effects. Moreover, it corresponds to the
estimated period for an earthquake with a magnitude of 6.2. The upper limit
was taken as a reasonable period beyond which site amplification effects will
134
not be very significant for typical soil profiles (exception being very deep soft
soil deposits). Moreover, this period is beyond the range of interest for most
typical structures.
• The PGVP/N ratio was varied from 0.25 to 1.0. The lower value applies mainly
to large velocity input motions. A ratio higher than 0.5 is considered reasonable
only for motions lower than about 150 cm/s.
4.4 SUMMARY AND FINDINGS
Near-fault strong motion recordings that satisfied the geometric conditions for
forward directivity (Somerville et al. 1997) were selected and analyzed. The motions
were parameterized in the time domain by means of the number of significant pulses,
pulse amplitudes, and pulse periods of their velocity time-history. Several
characterizations of the pulse period were evaluated and compared. The representation of
pulse period by the period of the pulse with largest amplitude was selected. Some
important findings include:
• Near-fault, forward-directivity motions can be adequately represented by
simplified time histories consisting of one or a few sine-pulses. The number of
pulses is likely related to slip distributions in the causative fault, and
consequently is difficult to predict.
• Pulse period is a function of moment magnitude. For earthquakes with Mw =
6.1, the pulse period is about 0.6 s and increases to about 6.0 s for Mw = 7.5.
Most of the energy in forward-directivity ground motions is concentrated on the
135
narrow-period band centered on the pulse period. Consequently, lower
magnitude events might result in more damaging ground motions for typical low
period structures. Relatively high standard deviations are associated with the
prediction of pulse period (Figure 4.26).
• Local site conditions have an important effect on pulse period. Longer periods
occur at soil sites than at rock sites for events with magnitudes lower than about
Mw = 7.0. The difference diminishes as the magnitude increases. For events
with Mw ≈ 7.5, the pulse periods at rock and soil sites are approximately the
same (see Figure 4.24).
• Peak ground velocities in the near-fault region vary significantly with magnitude
and distance. The attenuation relationships developed (Equation 4.12 and Table
4.13) has relatively high standard deviation. Median peak ground velocities for
soils are larger than median peak ground velocities at rocks sites for low
magnitude events. The difference diminishes as the magnitude increases. This
observation, however, relies on a data-set with a small number of rock
recordings. Moreover, the difference between peak ground velocities in rock
and in soil is obscured by the large scatter in the data.
Simplified pulse representations based on the analysis of the database were
developed for use as input ground motions in site response analysis. Chapter Five
introduces the site response methodology that is used in Chapter Six for the site response
analyses.
136
BLANK
137
Table 4.1. Modification to ground motion parameters to account for directivity effects.
Parameters X, Y, θ, and φ are defined in Figure 4.2, along with exclusion zones for strike
slip faults. For coefficients and standard deviations see Somerville et al. (1997) (adapted
from Somerville et al. 1997).
Ground Motion Description Equation Range of Applicability

Parameter
Amplitude Bias in average Strike-Slip faults: Moment Magnitude (Mw): 6.5 –
Factor: Ratio of horizontal response y = C1+C2 Xcosθ 7.5
data/model spectral acceleration Distance (R): 0 – 50 km
(log) with respect to Dip-Slip faults: C1, C2 functions of period
Abrahamson and Silva y = C1+C2 Ycosφ
(1997)
Duration factor: Bias in duration of Strike-Slip faults: 6.5 ≤ Mw ≤ 7.5
Ratio of acceleration with y = C1+C2 Xcosθ 0 ≤ R ≤ 20 km
data/model respect to Abrahamson
(D0.05-0.75) and Silva (1997) Dip-Slip faults:
y = C1+C2 Ycosφ
Strike Natural logarithm of the y = cos2ξ [C1 + 6.0 ≤ Mw ≤ 7.5
Normal/Average ratio of strike normal to C2 ln(R + 1) + 0 ≤ R ≤ 50 km
Amplitude average horizontal C3(Mw-6)] ξ = θ for strike-slip, φ for dip-
spectral acceleration slip. 0 < ξ < 90°
C1, C2, C3 function of period.
Given separately for cases in
which dependence on ξ is
included, and cases in which
dependence on ξ is ignored.
138
Table 4.2. Near-source factors from the 1997 Uniform Building Code (UBC).
(a) Short-period factor (Na)
Closest Distance to Known Seismic Source1

Seismic Source
Type ≤ 2 km 5 km ≥ 10 km
A 1.5 1.2 1.0
B 1.3 1.0 1.0
C 1.0 1.0 1.0
(b) Intermediate-period factor (Nv)
Closest Distance to Known Seismic Source1

Seismic Source
Type ≤ 2 km 5 km 10 km ≥ 15 km
A 2.0 1.6 1.2 1.0
B 1.6 1.2 1.0 1.0
C 1.0 1.0 1.0 1.0
(c) Description of seismic source types
Seismic Source Definition

Seismic
Source Description Maximum
Slip Rate, SR
Type Moment
(mm/year)
Magnitude, Mw
Faults that are capable of
producing large magnitude events
A Mw ≥ 7.0 SR ≥ 5
and that have a high rate of seismic
activity
Mw ≥ 7.0 SR > 5
All faults other than Types A and
B Mw < 7.0 SR > 2
C
Mw ≥ 6.5 SR < 2
Faults that are not capable of

producing large magnitude
C earthquakes and that have a Mw < 6.5 SR ≤ 2
relatively low rate of seismic
activity
1
The closest distance to seismic source shall be taken as the minimum distance between the site and the
surface projection of the fault plane. The surface projection need not include portions of the source at
depths of 10 km or greater.
139
Table 4.3. Earthquakes included in the study of near-fault ground motions. Fault
parameters are obtained from Somerville et al. (1997).
Event # Earthquake Date Moment Mechanism1 Strike Dip

Magnitude
1 Parkfield 6/27/66 6.1 SS 317 90
2 San Fernando 2/9/71 6.6 TH 290 50
3 Imperial Valley 10/15/79 6.5 SS 143 90
4 Morgan Hill 4/24/84 6.2 SS 154 90

Superstition
5 11/24/87 6.6 SS 127 90
Hills (B)
6 Loma Prieta 10/17/89 7.0 OB 128 70
Erzincan,
7 3/13/92 6.7 SS 300 86
Turkey
8 Landers 6/28/92 7.3 SS 3512 90
9 Northridge 1/17/94 6.7 TH 122 40
10 Kobe 1/17/95 6.9 SS 50 85
11 Kocaeli 8/17/99 7.4 SS 90 90
1
SS: Strike-slip, OB: Oblique-slip, TH: Thrust
2
The Landers Earthquake occurred on a fault with multiple segments. The Lucerne record was rotated to
the indicated strike corresponding to the orientation of highest fault normal peak ground velocity.
140
Table 4.4. Parameters used to define the simplified sine-pulse ground motions.
Parameter Abbreviation Methodology to obtain parameter

Number of significant N Number of pulses (half-cycle pulses) in the velocity-
pulses. time history with amplitudes at least 50% of the peak
ground velocity of the record.
Pulse Period. Tv,i For each half sine pulse, Tv,i = 2 (t2 – t1), where t1 and t2
are either the zero-crossing time, or the time at which
velocity is equal to 10% of the peak velocity for the
pulse if this time is significantly different than the zero
crossing time. Tv corresponding to the pulse with
maximum amplitude is the overall representative
velocity pulse period.
Predominant Period Tp-v Period corresponding to a clear and global peak in the
from Pseudo-velocity pseudo-velocity response spectra at 5% damping.
response spectra.
Pulse Amplitude. Ai For each half sine pulse, the peak ground velocity in
the time interval [t1,t2].
Peak ground velocity PGV Maximum velocity, defined by the maximum value of
Ai. Note, however, that in very few exceptions, the
maximum value of Ai in the fault parallel direction does
not occur concurrently with the fault normal pulse.
Ratio of fault parallel PGVP/N Defined by the ratio of maximum AP divided by
to fault normal maximum AN, where the subscripts P and N denote
amplitude fault-parallel and fault-normal motions respectively.
Time delay between toff Time of initiation of fault parallel pulse minus the time
fault normal and fault of initiation of fault normal pulse.
parallel pulse
141
Table 4.5. Stations included in the analysis of near-fault ground motions.
R2 Site PGA PGV T v5 Tv-p5

Station Agency Station # Event 1 RDI3 PGVP/N5
(km) Condition4 (g) (cm/s) (s) (s)
Cholame #2 CDMG 1013 1 0.1 1.08 s 0.47 75 -6 0.67 0.66
Temblor CDMG 1438 1 9.9 1.08 r 0.29 17.5 1.07 0.44 0.4
Pacoima Dam CDMG 279 2 2.8 1.04 r 1.47 114 0.33 1.44 1.15
Brawley Airport USGS 5060 3 8.5 1.48 s 0.21 36.1 0.99 3.01 3.1
EC County Center FF CDMG 5154 3 7.6 1.48 s 0.22 54.5 0.79 3.78 3.4
EC Meloland Overpass FF CDMG 5155 3 0.5 1.48 s 0.38 115 0.24 2.82 2.9
El Centro Array #10 USGS 412 3 8.6 1.48 s 0.23 46.9 0.84 3.93 3.8
El Centro Array #5 USGS 952 3 1 1.48 s 0.53 91.5 0.54 3.37 3.2
El Centro Array #6 USGS 942 3 1 1.48 s 0.44 112 0.58 3.65 3.4
El Centro Array #7 USGS 5028 3 0.6 1.48 s 0.46 109 0.41 3.73 3.3
El Centro Differential Array USGS 5165 3 5.3 1.48 s 0.44 59.6 0.86 4.18 3.0
Holtville Post Office USGS 5055 3 7.5 1.48 s 0.26 55.1 0.78 4.28 4.2
Westmorland Fire Sta CDMG 5169 3 15.1 1.48 s 0.1 26.7 0.28 3.93 4.7
Coyote Lake Dam CDMG 57217 4 0.1 1.17 r 1.00 68.7 1.02 0.73 0.725
Gilroy Array #6 CDMG 57383 4 11.8 1.17 r 0.61 36.5 0.29 1.00 1.18
El Centro Imp. Co. Cent CDMG 01335 5 13.9 1.48 s 0.31 51.9 0.70 1.85 1.25
Parachute Test site USGS 5051 5 0.7 1.48 s 0.42 107 0.43 2.11 1.83
Gilroy - Gavilan Coll. CDMG 47006 6 11.6 1.48 r 0.41 30.8 0.86 1.14 0.39
Gilroy - Historic Bldg. CDMG 57476 6 12.7 1.48 s 0.29 36.8 0.65 1.33 1.45
Gilroy Array #1 CDMG 47379 6 11.2 1.48 r 0.44 38.6 0.74 1.16 0.40
Gilroy Array #2 CDMG 47380 6 12.7 1.48 s 0.41 45.6 0.61 1.41 1.45
Gilroy Array #3 CDMG 47381 6 14.4 1.48 s 0.53 49.3 0.69 1.46 0.48
LGPC UCSC 16 6 6.1 1.48 r 0.65 102 0.5 2.14 0.40
Saratoga – Aloha Ave CDMG 58065 6 13 1.48 s 0.38 55.5 0.78 2.31 1.55
Saratoga – W Valley Coll. CDMG 58235 6 13.7 1.48 s 0.4 71.3 0.84 1.71 1.15
Erzincan 95 7 2 1.23 s 0.49 95.5 0.48 2.27 2.2
Lucerne # SCE 24 8 1.1 1.48 r 0.78 147 0.21 5.39 4.15
Jensen Filter Plant # USGS 0655 9 6.2 1.14 s 0.62 104 0.9 1.99 2.8
142
R2 Site PGA PGV T v5 Tv-p5
Station Agency Station # Event 1 RDI3 PGVP/N5
(km) Condition4 (g) (cm/s) (s) (s)
LA Dam USGS - 9 2.6 1.14 r 0.58 77 0.25 1.24 1.32
Newhall – Fire Sta # CDMG 24279 9 7.1 1.16 s 0.72 120 0.42 2.04 0.69
Newhall – W. Pico Canyon USC 90056 9 7.1 1.15 s 0.43 87.7 0.85 2.19 2.0
Rd.
Pacoima Dam (downstr) # CDMG 24207 9 8 1.16 r 0.48 49.9 0.46 0.61 0.43
Pacoima Dam (upper left) # CDMG 24207 9 8 1.16 r 1.47 107 0.43 0.89 0.72
Rinaldi Receiving Sta # DWP 77 9 7.1 1.13 s 0.89 173 0.29 1.31 1.05
Sylmar – Converter Sta # DWP 74 9 6.2 1.14 s 0.8 130 0.72 2.87 1.05
Sylmar – Converter Sta East # DWP 75 9 6.1 1.14 s 0.84 116 0.67 2.64 3.0
Sylmar – Olive View Med FF CDMG 24514 9 6.4 1.16 s 0.73 123 0.44 1.76 2.35
#
KJMA (Kobe) - 10 0.6 1.14 r 0.85 96 0.56 1.91 0.86
Kobe University CEOR - 10 0.2 1.48 r 0.32 42.2 0.92 1.59 1.32
OSAJ - 10 8.5 1.48 s 0.08 19.9 0.86 3.83 1.18
Port Island (0 m) CEOR - 10 2.5 1.12 s 0.38 84.3 0.30 1.91 1.32
Arcelik Kandilli - 11 17 s 0.21 42.3 0.31 6.82 5.45
Duzce ERD - 11 11.9 s 0.37 52.5 0.82 1.92 1.36
Gebze ERD - 11 17 r 0.26 40.7 0.84 5.04 4.5
Yarimka Kandilli - 11 4.4 s 0.32 87.1 1 6.00 3.80
1
See Table 4.3
2
Closest distance to the fault plane (Sadigh et al. 1993)
3
Equation 4.1
4
Soil(s) or Rock ®
5
Defined in Table 4.4
6
The Cholame #2 record in the Parkfield earthquake triggered only in one direction (15o from the fault normal direction)
143
Table 4.6. Pulse parameter for stations included in this study.
2 2
N2, N3 A
N AP2 TvN2 TvP2 toff
Station Event 1 (50%) (33%) (cm/s) (cm/s) (s) (s) (s)
Cholame 1 2 2 -56.9, 20, -20, 75 0.73, 0.2, 0.2, 0.67
12.2, -5.2, 8.19, -16.3, 11, -7.2, 11.4, -17.5, 1.41, 0.31, 0.44, 0.37, 0.44, 0.29, 0.4, 0.5,
Temblor 1 4 5 0.62
14.2 5.7 0.31 0.25
Pacoima Dam 2 1 3 45.3, -114, 46.2 15.3, 38, -17.1 1.25, 1.44, 1.17 1.21, 1.47, 0.49 -0.17
15.6, -35.8, -9.6, - 0.72, 2.14, 0.68, 1.55,
Brawley Airport 3 2 4 36.1, -24.4 2.56, 3.89 0.44
11.6, 11.3 1.89
EC County Center FF 3 2 4 35.9, -54.5 -31.5, 42.9 3.98, 3.78 1.21, 1.76 1.79
EC Meloland Overpass FF 3 1 3 -49.7, 115, -50.2 -27.1, -8.93, 22.3 3.78, 2.82, 3.22 1.46, 1.01, 2.72 1.81
El Centro Array #10 3 2 5 31.2, -46.9 -11.7, 11, -29.6, 39.3 3.38, 3.93 1.26, 0.65, 1.91, 1.96 0.48
El Centro Array #3 3 2 3 21.3, -41.1, 19 -16.7, 45.4, -32.7 1.96, 4.5, 3.25 1.89, 2.76, 2.19 0.29
-34.5, 32.6, 5.6, 8.2, - 1.06, 1.28, 0.49, 0.6,
El Centro Array #4 3 2 3 77.8, -71.7 4.31, 4.38 1.44
40.1 1.59
El Centro Array #5 3 2 3 75.9-91.4 49, -43.2, -29.5, 23.6 3.97, 3.37 3.38, 1.62, 1.77, 0.66 -0.3
El Centro Array #6 3 2 2 112, -97.7 31.1, -64.6, 54.3 3.65, 3.85 1.46, 3.32, 2.29 -0.05
4.66, 1.71, 1.03, 1.66,
El Centro Array #7 3 3 3 -55.7, 109, -69.6 -32.7, 18, -41, 0, 44.5 3.98, 3.73, 2.9 -0.16
3.02
-51.9, 44.6, -12.7,
El Centro Array #8 3 2 3 47.7, -48.5 4.08, 3.98 0.98
40.9, 34
El Centro Differential Array 3 2 4 52.5, -59.6 -49.9, 51.4 3.63, 4.18 1.41, 1.76 1.48
Holtville Post Office 3 2 3 36.2, -55.1, 24.3 42.8, -35 5.94, 4.28, 3.88 2.87, 3.73 3.88
Westmorland Fire Sta 3 4 2 26.7, -17.6 7.5, -5.6, 7.1 3.93, 5.14 0.5, 2.92, 2.22 0.91
-40.2, 67.2, -44.1,
Coyote Lake Dam 4 4 5 19.1, -68.7, 42, -16.1 1.12, 0.55, 0.98, 0.5 1.04, 0.9, 0.96, 0.34 -0.19
55.5
Gilroy Array #6 4 3 5 36.5, -22.2, 19.8 -10.6, 10.6, -4.13 1, 1.13, 0.88 0.88, 0.85, 0.81 0.15
21.9, -51.9, 0, -22.1, -10.5, -8.9, 36.1, 6, - 1.47, 1.85, 0.15, 0.56, 1.82, 0.47, 1.07, 0.44,
El Centro Imp. Co. Cent 5 1 4 0.17
25.4 25.5 2.21 2.28
Parachute Test site 5 2 2 -107, 80.9 45.8, -45.6 2.11, 2.36 1.47, 2.41 0.13
Gilroy – Gavilan Coll. 6 2 6 -30.8, 0, -14.9 15.2, 10.5, -26.6 1.16, 0.05, 0.64 0.74, 0.78, 0.82 0.08
20.2, -24.1, 14.5, -
Gilroy – Historic Bldg. 6 2 4 35.8, -36.8, 0, 14.5 1.62, 1.33, 0.47, 1.09 1.43, 0.98, 1.26, 0.67 0.13
20.9
Gilroy Array #1 6 4 7 -38.6, 0, -14.1 16.4, 14.1, -28.7 1.16, 0.12, 0.6 0.88, 0.76, 0.73 0.02
-45.6, 33.6, 16.6, - 19.4, -9, -27.6, 23.3, - 1.36, 1.31, 0.55, 0.96,
Gilroy Array #2 6 4 6 1, 0.6, 1.11, 1.06, 0.6 0.05
33.6, 37 7 1.81
Gilroy Array #3 6 2 5 -49.3 34.1 1.46 1.56 -0.18
LGPC 6 6 10 -70.8, 102 -41.2, 51.6 2.92, 2.14 2.74, 1.33 0.26
144
2 2
N2, N3 AN AP2 TvN2 TvP2 toff
Station Event 1 (50%) (33%) (cm/s) (cm/s) (s) (s) (s)
Saratoga – Aloha Ave 6 3 5 -31.5, 55.5, 43 16.3, 0, -43.3 3.02, 2.31, 1.81 4.5, 1.96, 2.01 -0.08
Saratoga – W Valley Coll. 6 2 3 71.3 -29.1, 60 1.71 1.76, 2.46 -0.15
Erzincan 7 3 3 -59, 95.5, -58 -17.8, 45.9, -36.2 1.43, 2.27, 2.52 1.59, 1.71, 2.11 -0.35
Lucerne # 8 1 2 -145, 55.7 29.7, -19.2 5.39, 4.43 4.68, 1.76 -1.59
104, -84.7, 62.8, - 36.8, -60.7, 61, -94, 1.11, 1.06, 2.67, 1.64,
Jensen Filter Plant # 9 3 6 1.99, 3.2, 2.7, 2.2, 3.6 0
44.9, 38.8 90.8 2.23
LA Dam 9 1 2 77, -35.5 19, -8 1.24, 1.55 1.39, 1.43 -0.25
Newhall – Fire Sta # 9 3 5 68.6, 37.7, 120 49.9, -19.3, 42.7 2.04, 0.68, 0.95 1.05, 0.25, 0.4 0.08
Newhall – W. Pico Canyon Rd. 9 2 2 82.8, -87.7 -74.7, 48.7 2.92, 2.19 1.71, 2.51 0.92
Pacoima Dam (downstr) # 9 1 3 49.9, -19, 24.7 12, -23, 3 0.61, 0.32, 0.51 0.38, 1.1, 0.25 0.18
18.9, -27.2, 25.5, -
Pacoima Dam (upper left) # 9 2 3 107, -72.7, 46 0.89, 0.94, 0.2 0.52, 0.48, 0.36, 0.67 0.04
46.5
Rinaldi Receiving Sta # 9 1 2 173, -81.1, 33 -17.3, 23, -13, -50 1.31, 1.15, 0.61 0.36, 0.54, 0.33, 2.06 -0.04
75.1, -39.9, 61.5, -
Sylmar – Converter Sta # 9 3 6 130, 69, 33.2, 93.8 2.87, 2.23, 1.25, 1.51 1.7, 1.89, 2.27, 1.43 -0.32
93.3
37.6, -39.5, 40.5, -
Sylmar – Converter Sta East # 9 2 2 116, 37.3, 25.4, 66.6 2.64, 2.53, 1.32, 1.32 1.04, 1.73, 2.53, 1.7 0
78.3
Sylmar – Olive View Med FF # 9 2 4 123, -56.3, 62.9 -22.3, 54.4, -40.6 1.76, 2.72, 2.85 0.83, 0.78, 0.32 -0.28
-73.4, 67.6, -88.7, -14.4, 13.4, -34.2,
KJMA (Kobe) 10 7 10 1.03, 0.87, 0.85, 1.91 1.32, 0.37, 0.37, 0.73 0.02
95.7 53.4
38.3, -20, 10.8, -29.7, 28.8, -39.6, 16.3, - 1.85, 1.62, 0.49, 1.55, 1.02, 1.21, 1.74, 0.53,
Kobe University 10 4 8 -0.02
42.9 9.05, 26.4 1.59 1.62
OSAJ 10 2 3 18.9-19.9, 5.7 -10.4, 16, -17.1 4.16, 3.83, 0.58 3.7, 2.64, 1.83 0.32
-48.9, 58.2, -33.6, -9.6, 16.9, -19.4, 25.6, 1.86, 1.16, 0.97, 1.91, 1.26, 1.16, 1.46, 2.17,
Port Island (0 m) 10 5 6 1.13
84.3, -76.4 -18.3 2.92 2.42
Arcelik 11 2 2 -42.2, 30.6 12.8, -13.1 6.82, 5.16 11.2, 4.24 -3.9
-32.4, 37.8, -10.3,
Duzce 11 4 5 29.2, -46.5, 52.5, 42.3 1.61, 1.46, 1.92, 0.7 2.47, 1.02, 0.26, 0.83 0.66
22.4
Gebze 11 2 2 13.1, -40.7, 31.6 -19.2, 34.3, -9.8 5.76, 5.04, 4.46 7.2, 5.5, 2.5 -0.32
Yarimka 11 2 4 87, -42, 47.1 32.2, -48, 87 6, 3.9, 3.7 4.3, 2.9, 5.9 -0.4
1
See Table 4.3
2
Defined in Table 4.4 (N is the number of half-cycle pulses with peak velocities at least 50% of the PGV of the entire record).
3
The 33% level is added to illustrate how different cut-off values result in a different number of pulses
145
Table 4.7. Ground motion recordings satisfying geometric requirements for forward-
directivity conditions not included in this study. Listed are only stations for earthquakes
with Mw ≥ 6.5 and R < 20 km.
Earthquake Ground Motion Station

Imperial Valley Parachute Test Site
Imperial Valley El Centro Array #1
Kobe Fukushima
Kobe Nishi-Akashi
Kobe Shin-Osaka
Kobe Amagasaki
Kobe Kobc
Kocaeli Izmit
Loma Prieta Gilroy Array #6
Loma Prieta Lexington Dam
Loma Prieta Watsonville
Nahanni, Canada Site 3 - Battlement Creek
Northridge Canyon Country - W Lost Canyon
Northridge Dam toe
Northridge Jensen Generator Building
Northridge Sepulveda V.A.
Superstition Hills(B) 11369 Westmorland Fire Sta
Superstition Hills(B) 5060 Brawley
146
Table 4.8. Values of Tv-p/Tv for multiple-record events.
Earthquake Average Std. Dev.

Imperial Valley .96 .11
Loma Prieta .64 .36
Northridge .90 .34
Kobe .57 .23
147
Table 4.9. Number of half-cycle pulses (N) by event for the recordings considered in this
study. Value in parenthesis is the number of half-cycle pulses that corresponds to a cut-
off value of 33% of the PGV (as opposed to 50% used to define N).
Number of Records with given number of half-cycle

Number of
Earthquake Year pulses (N)
Records
1 pulse 2 pulses 3 pulses > 3 pulses
Parkfield 66 2 0 (0) 1 (1) 0 (0) 1 (1)
San Fernando 71 1 1 (0) 0 (0) 0 (1) 0 (0)
Imperial Valley 79 13 1 (0) 10 (1) 1 (7) 1 (5)
Morgan Hill 84 2 0 (0) 0 (0) 1 (0) 1 (2)
Superstition 87 2 1 (0) 1 (1) 0 (0) 0 (1)
Hills(B)
Loma Prieta 89 8 0 (0) 4 (0) 1 (1) 3 (7)
Erzincan, Turkey 92 1 0 (0) 0 (0) 1 (1) 0 (0)
Landers 92 1 1 (0) 0 (1) 0 (0) 0 (0)
Northridge 94 10 3 (0) 4 (4) 3 (2) 0 (4)
Kobe 95 4 0 (0) 1 (0) 0 (1) 3 (3)
Kocaeli, Turkey 99 4 0 (0) 3 (2) 0 (0) 1 (2)
Totals 48 7 (0) 24 (10) 7 (13) 10 (25)
148
Table 4.10. Parameters from the regression analyses for the period of the pulse of
maximum amplitude, Tv, and the period corresponding to the maximum pseudo-velocity
response spectral value, Tv-p (Equation 4.8).
a) Tv
Data Set a b σ τ σtotal Var(σ2) Var(σ2) Cov(σ2,τ2)
All Motions -8.33 1.33 0.36 0.40 0.54 0.0008 0.0078 -0.0003
Rock -11.10 1.70 0.31 0.41 0.51 0.0029 0.0140 -0.0018
Soil -5.81 0.97 0.32 0.40 0.51 0.0008 0.0100 -0.0003
b) Tv-p
Data Set a b σ τ σtotal Var(σ2) Var(σ2) Cov(σ2,τ2)
All Motions -6.92 1.08 0.48 0.45 0.66 0.0028 0.0154 -0.0009
Rock -9.53 1.42 0.37 0.61 0.71 0.0062 0.0555 -0.0041
Soil -5.66 0.91 0.41 0.45 0.61 0.0022 0.0181 -0.0008
149
Table 4.11. Inter-event error term from the random effects model for the attenuation
relationship for pulse period.
All Motions Soil Rock

Earthquake No. of No. of No. of
Records
ηi Records
ηi
Records
ηi
Erzincan, Turkey 1 -0.03 1 -0.053 0 -
Imperial Valley 13 0.92 13 0.76 0 -
Kobe 4 -0.077 2 0.0662 2 -0.07
Kocaeli 4 -0.032 3 0.0498 1 0.076
Landers (B) 1 0.15 0 - 1 0.23
Loma Prieta 8 -0.40 5 -0.3879 3 -0.25
Morgan Hill 2 -0.066 0 - 2 0.30
Northridge 10 -0.197 7 -0.0923 3 -0.37
Parkfield 2 -0.294 1 -0.3203 1 -0.07
San Fernando 1 -0.057 0 - 1 0.146
Superstition Hills 2 0.057 2 -0.0238 0 -
150
Table 4.12. Comparison of pulse periods between rock and soil stations for recordings in
the Gilroy area in the 1989 Loma Prieta earthquake.
R Site Tv Tv-p
Station
(km) Condition (s) (s)
Gilroy#1 11.2 r 1.16 0.40
Gilroy Gavilan College 11.6 r 1.14 0.39
Gilroy#2 12.7 s 1.41 1.45
Gilroy#4 16.1 s 1.33 1.0
151
Table 4.13. Parameters from the regression analyses for peak ground velocity (Equation
4.12).
Data Set a b c d σ τ σtotal Var(σ2) Var(τ2) Cov(σ2,τ2)

All
2.44 0.50 -0.41 3.93 0.47 0.41 0.62 0.0026 0.011 8.2e-4
Motions
Rock 1.46 0.61 -0.38 3.93 0.53 0.25 0.59 0.023 0.019 -0.0120
Soil 3.86 0.30 -0.42 3.93 0.43 0.41 0.59 0.014 0.0026 8.6e-4
152
Table 4.14. Inter-event error term from the random effects model for the attenuation
relationship for peak ground velocity.
All Motions Soil Rock

Earthquake No. of No. of No. of
Records
ηi Records
ηi Records
ηi
Erzincan, Turkey 1 -0.045 1 -0.067 0 -
Imperial Valley 13 -0.009 13 -0.113 0 -
Kobe 4 -0.553 2 -0.422 2 -0.145
Kocaeli 4 -0.076 3 -0.031 1 -0.019
Landers (B) 1 0.020 0 - 1 0.026
Loma Prieta 8 0.086 5 0.142 3 -0.019
Morgan Hill 2 -0.012 0 - 2 0.047
Northridge 10 0.046 7 0.053 3 0.107
Parkfield 2 -0.214 1 -0.114 1 -0.099
San Fernando 1 0.127 0 - 1 0.085
Superstition Hills 2 0.116 2 *.079 0 -
153
Table 4.15. Ratios of pulse period in the fault parallel to pulse period in the fault normal
direction.
Earthquake Tv,FP/Tv,FN
Imperial Valley .60
Morgan Hill .88
Loma Prieta .86
Erzincan, Turkey .75
Landers .87
Northridge .89
Kobe .69
154
Table 4.16. Ground motions included in the determination of typical pulse shapes.
Earthquake Ground Motion Site Site Condition
Erzincan Erzincan Soil

Imperial Valley Brawley Airport Soil
Imperial Valley El Centro – Meloland Overpass Soil
Imperial Valley El Centro #10 Soil
Imperial Valley El Centro County Center Soil
Imperial Valley Holtville Post Office Soil
Kobe Port Island Soil
Landers Lucerne Rock
Loma Prieta Gilroy – Gavilan College Soil
Loma Prieta Gilroy – Historic Building Soil
Northridge LA Dam Rock
Northridge Newhall – West Pico Canyon Rd. Soil
Northridge Pacoima Dam Dwnstr. Rock
Northridge Pacoima Dam, Upper Left Ab. Rock
Northridge Rinaldi Receiving Station Soil
Northridge Sylmar – Olive View Hospital Soil
San Fernando Pacoima Dam Rock
Superstition Hills El Centro, Imperial County Center Soil
Superstition Hills Parachute Test Station Soil
155
Rupture Front
Healing front
S wave front
0
finished slipping
DEPTH (km)
about to slip
slipping
5
hypocenter
10
S waves travelling right

S waves travelling left
Figure 4.1 Schematic diagram of rupture directivity effects for a vertical strike-slip fault.
The rupture begins at the hypocenter and spreads at a speed that is about 80% of the shear
wave velocity. The figure shows a snapshot of the rupture front at a given instant.
Resulting time-histories close to and away from the hypocenter are represented by strike-
normal velocity recordings of the 1993 Landers earthquake at Joshua Tree and Lucerne
respectively (from Somerville et al. 1997).
156
Figure 4.2. Definition of parameters used in defining rupture directivity conditions
(adapted from Somerville et al. 1997).
157
3 3
Xcosθ = 1.0
Spectral Acceleration
Spectral Acceleration
2 2
0.75 Ycosφ =1.0
Factor
Factor
1
0.50 1
0.0
0.0
0 0
0.01 0.1 Period (s) 1 10 0.01 0.1 Period (s) 1 10
(a) Average response spectra ratio, showing dependence on period and on

directivity function.
2 2
1.8 Mw = 7.5 1.8

T = 3.0 s
FN/Average
FN/Average
1.6 R = 0 km 1.6 M = 6.0

R = 10 km M = 7.0
1.4 1.4
R = 50 km M = 7.5
1.2 1.2
1 1
0.8 0.8
0 2 4 6 0.1 1 10 100
(b) Strike-normal to average horizontal response spectral ratio for maximum

forward-directivity conditions (Xcosθ = 1).
Figure 4.3. Predictions from the Somerville et al. (1997) relationship for varying
directivity conditions.
158
Normalized Acceleration
Normalized Acceleration
1.5 1.5
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
1.5 1.5
Normalized Velocity
Normalized Velocity
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
Normalized Displacement
Normalized Displacement
1.5 1.5
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
-1.5 -1.5
0 0.5 1 1.5 2 0 0.5 1 1.5 2
(a) Krawinkler and Alavi (1998) (b) Sasani and Bertero (2000)
Figure 4.4. Simplified pulses used by other researchers.
159
Pacoima Dam (San Fernando EQ.) Pacoima Dam Dwnst. (Northridge EQ.)
100 60
40
Velocity (cm/s) 50
Velocity (cm/s)
20
0 0
-20
-50
-40
-100 -60
0 2 4 6 8 10 0 1 2 3 4 5
Time (s) Time (s)
Gilroy Gavilan College (Loma Prieta EQ.) KJMA (Kobe EQ.)

50 150
100
50
Velocity (cm/s)
Velocity (cm/s)
0 0
-50
-100
-50 -150
0 1 2 3 4 5 0 2 4 6 8 10
Time (s) Time (s)
Figure 4.5. Recorded and simplification of the dominant pulses for selected fault-normal
velocity time-histories. Thick lines correspond to motions representing a simplification
of the dominant pulses.
160
Meloland Overpass, Imperial Valley EQ.
100 Thick: Fault Normal 100
FP Velocity (cm/s)
Thin: Fault Parallel
Velocity (cm/s)
50 50
0 0
-50 -50
-100 -100
0 5 10 15 20 -100 0 100
West Pico Canyon Road, Northridge EQ.
FP Velocity (cm/s)
50 50
Velocity (cm/s)
0 0
-50 -50
0 5 10 15 20 -50 0 50
Time (s) FN Velocity (cm/s)
Figure 4.6. Fault normal (FN) and fault parallel (FP) velocity time-histories and
horizontal velocity-trace plots for two near-fault recordings. Both recordings have
significant fault normal velocities, but the Meloland Overpass recording from the
Imperial Valley earthquake has much lower fault parallel velocities than the West Pico
Canyon Road record from the Northridge earthquake.
161
NN = 3
AN,1
Fault Normal Velocity
AN,3
½ TvN,2
½ TvN,3
½ TvN,1
AN,2
NP = 2
toff
AP,1
Fault Parallel Velocity
½TvP,2
½TvP,1
AP,2
Figure 4.7. Parameters needed to define the fault parallel and fault normal components of
simplified velocity pulses. Subscripts N and P indicate fault normal and fault parallel
motions, respectively.
162
8
Weighted Average Period,(s) 7 Average Period = 0.9647Tv 1
2
R = 0.91
6 1
5
3
Regression Line
2 +- 20%
1
0
0 1 2 3 4 5 6 7 8
Period of Maximum Pulse, Tv (s)
Figure 4.8. Comparison of two different measures of pulse period: weighted average
period and period of maximum pulse (Tv).
163
100
80 Tv,1 = 2(t2 – t1)

60 t1 = time for which
Velocity = 0.1 PGV
40
FN Velocity (cm/s)
20
0.1 PGV
0
-0.1 PGV
-20
t1 = zero
-40 crossing time
-60
t2 = zero crossing time
-80
-
1 2 3 4 5 6 7 8 9 10
Time (s)
Figure 4.9. Determination of pulse period (Tv) for cases in which the pulse is preceded
by a small drift in the velocity time-history.
164
200
180
5% damping
160
140
Velocity (cm/s)
120
100
80
60
40
20
0
-2 -1 0 1
10 10 10 10
Period (s)
Tv-p
Figure 4.10. Determination of period corresponding to the peak pseudo-velocity

response spectral value. The pseudo-velocity response spectra shown is for the Erzincan
record of the Erzincan, Turkey, earthquake.
165
8
7 1
Tv-p = 0.8501Tv
6 2 1
R = 0.7469
5
Tv-p (s)
3
+- 20% from 1:1 line
2
Regression Line
1
0
0 1 2 3 4 5 6 7 8
Period of Maximum Pulse, Tv (s)
Figure 4.11. Comparison of the period of the maximum pseudo-velocity response

spectral value (Tv-p) with the period of the maximum pulse, Tv.
166
LGPC, Loma Prieta EQ. El Centro #4, Imp. Valley EQ.
1 1
Tv-p = 0.40 s Tv-p = 4.0 s
0.8 Tv = 2.14 s 0.8 Tv = 4.31 s
Normalized PSD
Normalized PSD
Tv-p/Tv = 0.2 Tv-p/Tv = 0.9
0.6 0.6
0.4 0.4
0.2 0.2
0 -1 0 1
0 -1 0 1
10 10 10 10 10 10
Pacoima Dam Dwnstr., Northridge EQ. El Centro #8, Imp. Valley EQ.
1 1
Tv-p = 0.43 s
Normalized PSD
0.8 Tv = 0.61 s 0.8 Tv-p = 4.0 s

Normalized PSD
Tv-p/Tv = 0.7 Tv = 3.98 s

Tv-p/Tv = 1.0
0.6 0.6
0.4 0.4
0.2 0.2
0 -1 0 1
0 -1 0 1
10 10 10 10 10 10
Figure 4.12. Normalized power spectral densities of velocity time histories for selected
ground motions. The two ground motions on the left have lower Tv-p than Tv, while those
on the right have roughly coinciding values of Tv-p and Tv.
167
Temblor Cholame 02
20
Rock Soil
50
Velocity (cm/s)
Velocity (cm/s)
10
0 0
-10
-50
-20
0 2 4 6 8 10 0 2 4 6 8 10
Time (s) Time (s)
Figure 4.13. Velocity time-histories for the 1966 Parkfield earthquake. The Temblor
record corresponds to the fault normal direction. The seismograph at the Cholame 02
station recorded only acceleration in one direction that corresponds to 15° degrees from
the fault normal direction. Dashed lines correspond to 50% and 33% of the PGV.
168
Pacoima Dam
100 Rock
Velocity (cm/s)
50
-50
-100
0 2 4 6 8 10
Time (s)
Figure 4.14. Fault normal velocity time-histories for the Pacoima Dam record in the
1971 San Fernando earthquake. Dashed lines correspond to 50% and 33% of the PGV.
169
El Centro - Meloland Overpass El Centro Array #7
100 Soil 100 Soil
50 50
0 0
-50 -50
-100 -100
0 5 10 15 20 0 5 10 15 20
El Centro Array #5 El Centro Array #6
100
Soil 100 Soil
50 50
0 0
-50 -50
-100
-100
Velocity (cm/s)
0 5 10 15 20 Velocity (cm/s) 0 5 10 15 20
50 Soil Soil
50
0 0
-50
-50
0 5 10 15 20 0 5 10 15 20
El Centro Diff. Array Holtville Post Office
50 Soil 50 Soil
0 0
-50 -50
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)
Figure 4.15a. Fault normal velocity time-histories for the 1979 Imperial Valley
earthquake. Dashed lines correspond to 50% and 33% of the PGV.
170
El Centro - County Center Brawley Airport
40
50 Soil Soil
20
0 0
-20
-50
-40
0 5 10 15 20 0 5 10 15 20
50 50
Soil Soil
Velocity (cm/s)
0 Velocity (cm/s) 0
-50 -50
0 5 10 15 20 0 5 10 15 20
Westmorland Fire Station Time (s)
Soil
20
-20
0 5 10 15 20
Time (s)
Figure 4.15b. Fault normal velocity time-histories for the 1979 Imperial Valley
171
Gilroy #6 Coyote Lake Dam
Velocity (cm/s) 40
Velocity (cm/s)
Rock Rock
50
20
0 0
-20
-50
-40
0 5 10 15 0 5 10 15
Time (s) Time (s)
Figure 4.16. Fault normal velocity time-histories for the 1984 Morgan Hill earthquake.
Dashed lines correspond to 50% and 33% of the PGV.
172
Parachute Test Site El Centro - Imp. Co. Cent.
Velocity (cm/s)
100
Velocity (cm/s)
Soil 50 Soil
50
0 0
-50
-100 -50
5 10 15 20 25 5 10 15 20 25
Time (s) Time (s)
Figure 4.17. Fault normal velocity time-histories for the 1987 Superstition Hills
173
LGPC Gilroy #1
100 40
Rock Rock
50 20
0 0
-50 -20
-100 -40
0 5 10 15 0 5 10 15
Gilroy Gavilan College Gilroy #2
40 50
Rock Soil
20
0 0
-20
-40 -50
Velocity (cm/s)
Velocity (cm/s)
0 5 10 15 0 5 10 15
Gilroy - Historic Building Saratoga - Aloha Av.
40
Soil 50 Soil
20
0 0
-20
-50
-40
0 5 10 15 0 5 10 15
Saratoga - West Valley Coll. Gilroy #3
50
Soil Soil
50
0 0
-50
-50
0 5 10 15 0 5 10 15
Time (s) Time (s)
Figure 4.18. Fault normal velocity time-histories for the 1989 Loma Prieta earthquake.
Dashed lines correspond to 50% and 33% of the PGV.
174
Erzincan
100
Soil
Velocity (cm/s)
50
-50
-100
0 2 4 6 8 10
Time (s)
Figure 4.19. Fault normal velocity time-histories for the Erzincan record in the 1992
Erzincan, Turkey, earthquake. Dashed lines correspond to 50% and 33% of the PGV.
175
Lucerne
Velocity (cm/s)
Rock
100
-100
5 10 15 20
Time (s)
Figure 4.20. Fault normal velocity time-histories for the Lucerne record in the 1992
Landers earthquake. Dashed lines correspond to 50% and 33% of the PGV.
176
LA Dam Pacoima Dam Dwnstr.
50
Rock Rock
50 V
el
0 o 0
cit
Velocity (cm/s)
-50 y
-50
0 2 4 6 8 10 0 2 4 6 8 10
Pacoima Dam - Upper Left Ab. Time (s)
100 Rock
50
-50
-100
0 2 4 6 8 10
Time (s)
Figure 4.21a. Fault normal velocity time-histories for rock records from the 1994
Northridge earthquake. Dashed lines correspond to 50% and 33% of the PGV.
177
Sylmar Converter Station East Sylmar Converter Station
100 Soil Soil
100
50 50
0 0
-50 -50
-100
-100
0 2 4 6 8 10 0 2 4 6 8 10
Jensen Filtration Plant Sylmar - Olive View Hospital
100 Soil Soil
100
50 50
0 0
-50 -50
-100
-100
Velocity (cm/s)
Velocity (cm/s)
0 2 4 6 8 10 0 2 4 6 8 10
Rinaldi Receiving Station Newhall - West Pico Canyon Road
Soil Soil
100 50
0 0
-100 -50
0 2 4 6 8 10 0 2 4 6 8 10
Newhall Fire Station Time (s)
Soil
100
50
-50
-100
0 2 4 6 8 10
Time (s)
Figure 4.21b. Fault normal velocity time-histories for soil records from the 1994
Northridge earthquake. Dashed lines correspond to 50% and 33% of the PGV.
178
Kobe University KJMA
50 100
Rock Rock
50
0 0
Velocity (cm/s)
-50
Velocity (cm/s)
-50 -100
0 5 10 15 20 0 5 10 15 20
Port Island OSAJ
20
Soil Soil
50 10
0 0
-50 -10
-20
0 5 10 15 20 0 5 10 15 20
Time (s) Time (s)
Figure 4.22. Fault normal velocity time-histories for records from the 1995 Kobe, Japan,
179
Yarimka Duzce
Soil 50 Soil
50
0 0
Velocity (cm/s)
Velocity (cm/s)
-50
-50
0 5 10 15 20 25 0 5 10 15 20 25
Arcelik Gebze
50 50
Soil Rock
0 0
-50 -50
0 5 10 15 20 25 0 5 10 15 20 25
Time (s) Time (s)
Figure 4.23. Fault normal velocity time-histories for records from the 1999 Kocaeli,
Turkey, earthquake. Dashed lines correspond to 50% and 33% of the PGV.
180
10.0
Pulse Period,T v (s)
Rock
Soil
1.0 All Data
Rock
Soil
0.1
6 6.5 7 7.5 8
Moment Magnitude, Mw
Figure 4.24. Attenuation relationship for pulse period (Tv).
181
6
5
Alavi and
Krawinkler (2000)
Pulse Period (s)
4 Somerville (1998)
3 Tv, Regression
Curve
2 Tv-p, Regression
Curve
0
6 7 8
Moment Magnitude (Mw)
Figure 4.25. Comparison of results form regression analysis with relationships proposed
by other researchers. The definition of pulse period of Somerville (1998) is similar to
that used in this study (Tv). On the other hand, the pulse period of Alavi and Krawinkler
(2000) is the period of the maximum pseudo-velocity response spectral value (Tv-p).
182
10
Soil
8
Rock
Pulse Period (s)
sd
0
6 7 8
Moment Magnitude (Mw)
Figure 4.26. Attenuation relationship for pulse period (Tv) showing one standard
deviation band.
183
Gilroy #1
40 40
Rock
20 20
0 0
-20 -20
-40 -40
0 2 4 6 8 10 -40 -20 0 20 40
Gilroy Gavilan College
40 40
Rock
Fault Parallel Velocity (cm/s)

20 20
0 0
Velocity (cm/s)
-20 Thick: Fault Normal -20

-40 -40
0 2 4 6 8 10 -40 -20 0 20 40
Gilroy #2
50 50
Soil
0 0
-50 -50
0 2 4 6 8 10 -50 0 50
Gilroy #4
40 40
Soil
20 20
0 0
-20 -20
-40 -40
0 2 4 6 8 10 -40 -20 0 20 40
Time (s) Fault Normal Velocity (cm/s)
Figure 4.27a. Velocity time-histories and velocity-trace plots for sites in the Gilroy area
recorded in the 1989 Loma Prieta earthquake.
184
200
Gilroy #1: Rock
180 Gilroy Gavilan College: Rock
5% damping Gilroy #2: Soil
Gilroy #4: Soil
160
Spectral Velocity (cm/s)
140
120
100
80
60
40
20
0 -2 -1 0 1
10 10 10 10
Period (s)
Figure 4.27b. Pseudo-velocity response spectra for the sites in Figure 2.27a. Observe the
significant difference in frequency content at long periods between the soil and the rock
sites.
185
Soil
Rock
7.5
6.5
-1 0 1
10 10 10
Distance (km)
Figure 4.28. Distribution of near-fault sites considered in this study.
186
All D ata
Soil; Mw = 6.0
Soil; Mw = 7.5
Rock; Mw = 6.0
2
10 Rock; Mw = 7.5
Model: Mw = 6.1 and 7.4
1
10
-1 0 1
10 10 10
Soil
Peak Ground Velocity (cm/s)
2
10
1
10
-1 0 1
10 10 10
Roc k
2
10
1
10
-1 0 1
10 10 10
Distanc e (km)
Figure 4.29. Attenuation relationship for PGV in the near-fault region.
187
Rock
3
10
Equation 4.12
Alavi and Krawinkler 2000)
Somerville (1998)
Peak Ground Velocity (cm/s)
Increasing Magnitude
2
10 Mw=7.4
Mw=6.1
1
10
2 4 6 8 10 12 14 16 18 20
Distance (km)
Equation 4.12:
ln(PGV) = 2.44 + 0.5 Mw –0.41 ln(R2 + 3.932)
Somerville (1998):
ln(PGV) = -2.31 + 1.15 Mw – 0.5 ln(R)
Krawinkler and Alavi (2000):

ln(PGV) = -5.11 + 1.59 Mw – 0.58 ln(R)
Figure 4.29b. Comparison of results from regression analysis for PGV with relationships
proposed by other researchers for a database of near-fault, forward-directivity motions.
188
1.3
PGVrock = 75 cm/s
PGVrock = 100 cm/s
1.2
PGVrock increases
PGVsoil/PGVrock
1.1
1.0
0.9
6.5 7 7.5
Figure 4.30. Dependence of ratio of PGV from soil to rock on magnitude and distance.
189
1.2
y = -0.3222Ln(x) + 1.9185
1 2
R = 0.4449
0.8
Rock Soil
PGVP/N
0.6
0.4
0.2
0
0 50 100 150 200
Fault Normal PGV (cm/s)
Figure 4.31. Relationship between the ratio of fault parallel to fault normal peak ground
velocity (PGVP/N) to fault normal peak ground velocity (PGV). Regression line and
equation are for Rock sites.
190
Pacoima Dam, San Fernando EQ.
100 Rock 100
50 50
0 0
-50 -50
-100 -100
0 1 2 3 4 5 -100 0 100
Lucerne, Landers EQ.
Rock

100 100
0 0
Velocity (cm/s)
-100 -100
0 5 10 15 -100 0 100
Pacoima Dam Dwnstr., Northridge EQ.
50 50
Rock
0 0
-50 -50
0 2 4 6 -50 0 50
El Centro, ICC, Superstition Hills. EQ.
50 Soil 50
0 0
Thick: Fault Normal

-50 -50
0 5 10 15 20 -50 0 50
Figure 4.32a. Selected motions with one dominant half-cycle pulse (N = 1).
191
El Centro - Meloland Overpass, Imp. Valley EQ.
100 Soil 100
50 50
0 0
-50 -50
-100 -100
0 2 4 6 8 10 -100 0 100
LA Dam, Northridge EQ.
Rock

50 50
0 0
Velocity (cm/s)
-50 -50
0 5 10 15 -50 0 50
Rinaldi Receiving Station, Northridge
Soil
100 100
0 0
Thick: Fault Normal

-100 -100
0 2 4 6 8 10 -100 0 100
Figure 4.32b. Selected motions with one dominant half-cycle pulse (N = 1).
192
El Centro #6, Imperial Valley EQ.
100 Soil 100
50 50
0 0
-50 -50
-100 -100
0 2 4 6 8 10 -100 0 100
Brawley Airport, Imperial Valley EQ.
40 40
Soil

20 20
0 0
Velocity (cm/s)
-20 -20
-40 -40
0 5 10 15 -40 -20 0 20 40
Gilroy - Historic Building, Loma Prieta EQ.
40 40
Soil
20 20
0 0
-20 -20
-40 -40
0 2 4 6 8 10 -40 -20 0 20 40
Newhall, West Pico Canyon Rd., Northridge
Soil
50 50
0 0
Thick: Fault Normal

-50 -50
0 5 10 15 -50 0 50
Figure 4.33a. Selected motions with a dominant full cycle of motion (N = 2).
193
Parachute Test Station, Superstition Hills EQ.
100 Soil 100
50 50
0 0
-50 -50
-100 -100
0 5 10 15 20 -100 0 100
Pacoima Dam, Upper Left Abb., Northridge EQ.
100 Rock 100

50 50
0 0
Velocity (cm/s)
-50 -50
-100 -100
0 2 4 6 8 10 -100 0 100
Soil
50 50
0 0
-50 Thick: Fault Normal -50

0 2 4 6 8 10 -50 0 50
50 50
Soil
0 0
-50 -50
0 5 10 15 -50 0 50
Figure 4.33b. Selected motions with a dominant full cycle of motion (N = 2).
194
90056 Newhall - W. Pico Canyon Rd.
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Normalized Fault Parallel Velocity

5060 Brawley Airport
1 1
Normalized Velocity
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
942 El Centro Array #6
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
57476 Gilroy - Historic Bldg.
1 1
0.5 0.5
0 0
Thick: Fault Normal

-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Time (s) Normalized Fault Normal Velocity
Figure 4.34a. Simplification of the dominant pulses in the recordings in Figure 4.33a.
195
5051 Parachute Test site
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1

24207 Pacoima Dam (upper left)
1 1
Normalized Velocity
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
1 1
0.5 0.5
0 0
Thick: Fault Normal

-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Figure 4.34b. Simplification of the dominant pulses in the recordings in Figure 4.33b.
196
Set 1
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1

Set 2
1 1
0.5 0.5
Normalized Velocity
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Set 3
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Set 4
1 1
0.5 0.5
0 0
Thick: Fault Normal
-0.5 Thin: Fault Parallel -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Figure 4.35a. Simplified sine-pulse representation of near-fault ground motions. The

fault parallel PGV is set to 50% of the fault normal PGV.
197
Set 5
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1

Set 6
1 1
Normalized Velocity
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Set 7
1 1
0.5 0.5
0 0
-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Set 8
1 1
0.5 0.5
0 0
Thick: Fault Normal

-0.5 -0.5
-1 -1
0 1 2 3 4 5 -1 0 1
Figure 4.35b. Simplified sine-pulse representation of near-fault ground motions. The

fault parallel PGV is set to 50% of the fault normal PGV.
198
CHAPTER 5
SITE RESPONSE ANALYSIS METHODOLOGY
5.1 INTRODUCTION
Site response analysis has been an important element of seismic risk assessment
since the pioneering work of Seed and coworkers in the late 1960s. The advent of
computers and the increased experimental database on the cyclic behavior of soils has led
to the development of increasingly more sophisticated site response methods. The ability
of these methodologies to reproduce observed behavior has been well documented for a
number of earthquakes (e.g. Seed et al. 1991, Chang and Bray 1995, Borja et al. 1999).
This chapter presents the implementation of a site response analysis within the context of
the finite element method. The site response methodology is then used in Chapter 6 for
the analysis of site response to near-fault ground motions.
For a given soil profile, the site response problem consists in estimating the
ground motion at a specified depth at the site in response to a ground motion prescribed at
another depth at the site. Typically, the ground motion is assumed to be composed of
vertically propagating, horizontally polarized (SH) shear waves traveling through a
horizontally layered soil profile. This assumption reduces the site response analysis to the
solution of a one-dimensional wave propagation problem. The representation of the
nonlinear response of the soil is typically accounted for in some fashion. The effects of
soil non-linearity on seismic ground motions have long been recognized by the
199
engineering profession, and lately have also been accepted by leading seismologists (Chin
and Aki 1991, Silva et al. 1988, among others). The site response problem is generally
solved either by equivalent-linear analysis (e.g., Schnabel et al. 1972; Hudson et al.
1994), or using non-linear time domain solutions (e.g., Wang et al. 1990, Borja et al.
1999).
In the equivalent-linear analysis, both the secant shear modulus and equivalent
viscous damping values are calculated for an assumed strain level. The wave propagation
problem is solved assuming linear behavior for the assumed values of shear moduli and
damping for each layer. The procedure is iterated until the assumed strain level
corresponds to a specified percentage of the maximum strain calculated for each soil
layer. At each step, the linear 1-D wave propagation analysis is solved in the frequency
domain using a closed form solution to the wave equation (Schnabel et al. 1972). This
methodology involves important assumptions. The frequency domain approach together
with the assumption of linearity at each iteration step implies the assumption that the
ground motion is stationary. Moreover, a unique set of properties (e.g, stiffness and
damping) is used for the entire time history at each layer. The equivalent-linear
procedure has been validated for a number of case studies, primarily using the computer
code SHAKE (Schnabel et al. 1972), and its later edition SHAKE91 (Idriss and Sun
1992). However, it is important to point out that validations in general have been done
using "normal" ground motions, that is, ground motions at reasonable levels of intensity
that do not exhibit typical near-fault characteristics. Seed et al. (1991) indicate that the
difference between the equivalent-linear method and non-linear analyses could be
significant with higher levels of shaking. Moreover, peak strains in near-fault ground
200
motions occur during a single or a limited number of high intensity stress cycles. Strains
observed during these cycles are likely not representative of strain levels during the
remainder of the motion. Assuming soil properties at only a fraction of peak strain during
these cycles may render unreasonable estimates of soil stiffness and damping during near-
fault pulses.
Non linear computer programs differ in two main aspects: the tools used for the
solution of the equation of motion (i.e. the wave propagation equation) and the definition
of soil stress-strain behavior. Methods that have been applied to the solution of site
response problems include elastic wave propagation methods (e.g. SHAKE, Schnabel et
al. 1972), the method of characteristics (e.g. CHARSOIL, Streeter et al. 1974), finite
difference (e.g. Martin and Seed 1982), and finite element (e.g. DYNA1D, Prevost 1989;
QUAD4M, Hudson et al. 1994). Soil non-linearity is addressed by introducing
mathematical relationships that model the stress-strain behavior of the soil. A brief
summary of cyclic non-linear models was presented in Chapter 2. The most general
models able to represent soil behavior are based on plasticity theory. Plasticity-based
models permit the evaluation of soil response along a variety of stress paths and can be
easily expanded into three-dimensional stress space to model simultaneously the three
components of motion.
A site response method that is used for the analysis of near-fault, forward-
directivity ground motions must take into consideration their particular characteristics.
These motions are dominated by a small number of long-period pulses best observed in
velocity or displacement-time histories (see Chapter 4). These strong pulses of motion
201
likely generate large strain levels in soils, particularly for softer and deeper soils that have
site periods on the order of one to three seconds. Forward-directivity pulses also control
peak response values affected by long-period motions (i.e. PGV and peak ground
displacement, PGD). In addition, the strain levels generated by uni-directional shaking
are lower than those generated by multi-dimensional shaking. Hence, site response
analyses using all components of motion are more likely to provide reasonable results
than uni-directional analyses. This is important also for near-fault ground motions
because the analysis of the near-fault ground motion database indicated that, in some
cases, significant fault parallel velocities in phase with the fault normal velocity pulse are
observed. Thus, in summary, a site response analysis for near-fault ground motions
should be able to:
• Handle significant levels of strains
• Account for bi-directional input motions
• Deal with highly non-stationary input motions (i.e., time domain analysis)
These requirements are best met by finite element analysis in conjunction with an
appropriate constitutive model. This chapter presents the development of the site
response methodology used in Chapter 6 for the analysis of near-fault ground motions.
Section 5.2 summarizes the constitutive model used for the analysis. Section 5.3 presents
the implementation of this constitutive model into the finite element analysis program
GeoFEAP (Espinoza et al. 1995). Finally, a validation of the site response methodology
is presented in Section 5.4.
202
5.2 CONSTITUTIVE MODEL
5.2.1 Selection of a constitutive model
A constitutive model is defined by a set of equations that relate the strains in a
material to the current state of stress and temperature through a set of internal variables.
The mechanical response of soil is generally very complex, and constitutive models that
aim at describing a comprehensive range of soil behavior (such as dilatancy, small strain
non-linearity, stress anisotropy, etc) are generally complicated and involve a large number
of input parameters. Conversely, relatively simple models might over-simplify important
aspects of soil response. In general, all constitutive models share the same basic
requirements (Luccioni 1999):
1. Assumptions inherent to the formulation of the constitutive laws should be
justified with respect to both the material and problem to be modeled. The
response predicted by constitutive models should match observed field and
laboratory response of soil. The latter requirement implies that particular
aspects of soil behavior relevant to the problem in question are addressed. For
example, a model that assumes a linear elastic region might be adequate when
dealing with the prediction of stresses in a slope, but is inadequate for dealing
with the intermediate strain cyclic behavior of soil.
2. Procedures to estimate model parameters should be provided along with the
formulation of the constitutive model.
203
3. Numerical implementation of these constitutive laws into numerical codes
should ensure that when the converged state is achieved, governing equations
and energy and mass balance laws are satisfied.
A large number of constitutive models satisfying the above requirements have
been proposed. A brief summary of some of the models currently available for the cyclic
response of soils is presented in Chapter 2. In this section, a model capable of modeling
cyclic soil behavior is sought. The model is applied to site response problems involving
generalized site profiles (see Chapter 6). A number of the simplifying assumptions
involved in the generation of the site profiles have implications on the requirements for
the constitutive model. These assumptions and their implications on model development
include:
1. The generalized site profiles represent average site conditions for a given site.
Soil models that are used for these generalized profiles should capture general
aspects of soil behavior, but it is not required that the model be able to match
exactly each possible response of a particular type of soil.
2. Undrained soil behavior is assumed. Moreover, pore pressure increase during
shaking is assumed to be insignificant. These assumptions imply that the
constitutive model can be developed in a total stress framework. The use of an
effective stress framework in a constitutive model in general leads to wider
predictive capabilities at the cost of increased complexity in the definition of the
model. The use of a total stress soil model permits simplifying assumptions, such
as the uncoupling of the deviatoric and volumetric components soil response.
204
This, in turns, permits the uncoupling of site response analysis to incoming shear
and compressive waves. The uncoupling of horizontal shear waves and
compressive waves has been verified with models that do account for the
interaction between volumetric and deviatoric behavior of soils (EPRI 1993). The
validity of the assumption that no significant pore pressure is generated can be
questioned for the case of near-fault motions, in particular due to the potentially
large pore pressures that may develop due to the large strains observed in a soil
profile subject to near-fault velocity pulses. However, with the exception of very
loose to medium dense, saturated sands and silts, and probably some special clays,
the small number of pulses involved in near-fault ground motions may preclude
the rapid development of sufficient pore pressures that significantly alter site
response. In any case, a first-look at the near-fault site response problem
considering soils not prone to the development of significant pore pressures
during loading provides insight for these types of soils. The uncoupling of
volumetric and deviatoric behavior permits the use of models using a constant
void ratio, essentially, the void ratio at small strains.
The soil constitutive model used should be able to accommodate the large strains
expected in near-fault ground motions. Moreover, the model should be able to account
for soil yielding for the cases in which the soil's shear strength becomes an issue (such as
for soft clays, or Site E in Table 3.1). For this study, the model presented by Borja and
Amies (1994) is selected from those available at the start of this research. This model
conforms to the requirements identified previously while balancing simplicity and
robustness in its implementation.

205
The model by Borja and Amies (1994) is a multiaxial constitutive model
developed for clays. The model expands on the concept of bounding surface plasticity
with a vanishing elastic region (Dafalias and Popov 1977) by presenting general criteria
for loading and unloading that are applicable to general stress states. The concept of the
vanishing elastic region allows for the modeling of plastic strains developed at small
strains. This section presents a summary of the model by Borja and Amies (1994) that
includes the kinematic hardening of the boundary surface (Borja et al. 1999). The model
presented herein differs from the mentioned models (Borja and Amies 1994, Borja et al.
1999) solely in the introduction of Rayleigh damping to model small strain viscous
damping. The presentation, however, is slightly different from that in the aforementioned
papers.
5.2.2 Mathematical development
General Description
Assume the existence of a point Fo in three dimensional stress space defined by
the stress tensor σo (Figure 5.1). This point defines the stress state where the soil
experienced the most recent unloading, where unloading is defined as the condition in
which the direction of the load step causes the instantaneous hardening modulus to
increase (Borja and Amies 1994). The unloading condition is presented with larger detail
later in this section. The point Fo is located within a bounding surface defined by the
function B. The soil is assumed to behave elastically at point Fo and to follow a linear
kinematic hardening plasticity law at the bounding surface B. In between the point Fo
206
and the bounding surface B, the hardening modulus is interpolated between the values of
infinity at Fo (corresponding to elastic behavior) and Ho at the surface B (corresponding
to a linear, kinematic hardening plasticity law). A yield surface F is defined inside the
bounding surface passing through a point defined by the current stress tensor σ. The
vanishing elastic region corresponds to the limit when the size of F is zero. The
interpolation of the hardening modulus H' is generated from well-accepted one-
dimensional models for soils (Borja and Amies 1994). A formal mathematical
development of the model is now introduced. For further details of the model, see Borja
and Amies (1994) and Borja et al. (1999).
Model Development
Assume that strains within the soil are infinitesimal. Furthermore, assume an
σinv)
additive decomposition of the Cauchy stress tensor, σ, into an inviscid component (σ
σvis):
and a viscous component (σ
σ = σinv + σvis (5.1)
Laboratory tests have shown that damping at large strain levels is essentially
frequency independent and can be attributed to the hysteretic behavior of soils. On the
other hand, recent experimental results with damping at very small strains indicated that
at these strain levels, soils experience damping levels that can not be attributed to
hysteretic mechanisms alone (Lanzo and Vucetic 1999). The mechanism that controls
energy dissipation at these strain levels has been attributed to rate-of-loading effects
(Lanzo and Vucetic 1999). Typically, damping at low strains is low and can be modeled
207
as equivalent viscous damping. Viscous damping is typically represented as a fraction of
critical damping, where critical damping corresponds to the smallest damping value that
inhibit oscillations completely. The viscous term in Equation 5.1 is assumed to follow
Rayleigh damping, that is (Cook et al. 1989),
σ vis = Dε (5.2)
where ε is the strain tensor and D is a rank-four damping tensor given by:
[ ]
D = α 1 C e + α 2 [M ] (5.3)
where Ce is the rank four small-strain elasticity tensor and M is the mass matrix. The use
of Rayleigh damping results in frequency dependent damping. The constants α1 and α2 in
Equation 5.3 are calculated for a selected value of critical damping ratio, ξ, using two
reference frequencies, ω1 and ω2, at which the resultant damping ratio is matched to the
desired damping ratio ξ (Figure 5.2). The constants α1 and α2 are determined by
2ξ
α=
ω1 + ω 2
(5.4)
2ω ω ξ
α 2 = 1 2 = ω 1ω 2α 1
ω1 + ω 2
The constants α1 and α2 control the values of stiffness and mass proportional
damping, respectively (Equation 5.3). Figure 5.2 illustrates the dependence of mass and
stiffness proportional damping on frequency. Mass-proportional damping is zero at high
frequencies. On the other hand, stiffness proportional damping results in over-damping at
high frequency. The use of both mass and stiffness proportional damping allows a fit for
208
the desired value of damping ratio within a selected range of frequencies (i.e., close to
and between ω1 and ω2). Between ω1 and ω2, the resulting damping is lower than the
desired critical damping ratio. Away from this range, the system is over-damped.
Hudson et al. (1994) proposed setting ω1 to the natural frequency of the soil deposit and
letting ω2 = n ω1, where n is an odd integer, such that n is the closest odd integer greater
than ωi/ ω1, where ωi is the predominant frequency of the input earthquake motion. The
choice of n as an odd integer is motivated by shear beam studies showing that the
frequencies of higher modes are odd multiples of the frequency of the fundamental mode
of the beam. In general, ω1 and ω2 can be chosen so as to cover a desired range of
frequencies.
The inviscid term in Equation 5.1 is represented in terms of the bounding surface
plasticity model outlined previously. The stress-strain relations are described using an
incremental formulation that relates infinitesimal increments of stress and strain. For rate
independent models, these equations can be written either in rate or differential format.
When these relations are written in rate format, time serves only to indicate precedence in
the history of events. Additive decomposition of elastic and plastic strains is assumed:
εD = εD e + εD p (5.5)
where ε is the total strain tensor and the dot indicates differentiation with respect to time.
The generalized Hooke's law is used to relate elastic strains with the stress tensor, thus,
σ = C e : (ε − ε p ) (5.6)
where Ce is the rank four elasticity tensor.
209
Plastic strains are associated with the yield surface F and the boundary surface B.
These surfaces are translated cylinders in stress space, and are given by:
F = ζ F :ζ F − r2 = 0 (5.7)
where ζF = σ' - α, α is a deviatoric back stress tensor representing the center of F, and r is
the radius of the cylinder; and
B = ζB : ζB − R2 = 0 (5.8)
where ζB = σ' - β, β is a deviatoric back stress tensor representing the center of B, and R
is the radius of the bounding surface B.
The yield surface F is always contained in B. This implies that B = 0 if and only
if F = 0. For both the yield surface and the bounding surface, an associative flow rule is
used to define the direction of plastic strain increments, therefore
ε p = λFnˆ F + λBnˆ B (5.9)
where λF and λB are the consistency parameters associated with the yield surface and the
bounding surface respectively, and n̂ F and n̂ B are unit normal tensors to each yield
surface ( n̂ F = ∂F/∂σ
σ = ζF/ ζ F and n̂ B = ∂B/∂σ
σ = ζB/ ζ B ). Equation 5.9 is similar to
generalized flow rules used for multiple yield surface plasticity models (Pestana 1994).
For a point in the surface of F, the unit normal n̂ F is equal to ζF/r. Note that n̂ F is
ill defined in the limiting case of the vanishing elastic region when the radius of the yield
surface F tends to zero. Taking the limit as r goes to zero, n̂ F if defined by:
210
ζ σ
ˆ F = lim  F  =
n (5.10)
r →0  r  σ
A restriction on the consistency parameters is placed such that
λB λ F = 0 (5.11)
which implies that at any given time, only one of the two surfaces (F and B) is the active
yield surface, in the sense that plastic flows occurs only in that surface. Loading
conditions in traditional rate-independent plasticity models are given by the Kuhn-Tucker
loading/unloading (complementarity) conditions (Simo and Hughes 1998). For any given
yield surface f, these conditions are given by:
λ > 0, f ≤ 0, λf = 0 (5.12)
where λ is the consistency. Equation 5.12 is used to define loading and unloading when
the bounding surface is the active yield surface (i.e. λB ≠ 0). These loading conditions,
however, are ill defined for the yield surface F due to the assumption of a vanishing
elastic region (Borja and Amies 1994). The loading/unloading condition for states of
stress within the bounding surface is presented later.
Associative flow is assumed for the hardening rules for parameters β and α (note
that ∂F/∂α β = - n̂ B ). The hardening rules are

α = - n̂ F and ∂B/∂β
αC = −λ F d Fnˆ F
(5.13)
βC = −λ d nˆ
B B B
211
where dB and dF are generalized plastic moduli coefficients (Simo and Hughes 1998).
The plastic moduli coefficients are chosen to so that the uniaxial stress σ = (3/2)1/2 σ ′
is related to the uniaxial plastic strain ε p = (2/3)1/2 ε p by the hardening modulus H', so
that
σ = H ' ε p (5.14)
where H' is the uniaxial plastic modulus. For this effect, dB and dF are chosen to be equal
to 2/3 H'. The functional form of H' defines the stress-strain behavior of the soil.
The consistency conditions are applied in two mutually exclusive cases, (a) when
the stresses are inside the bounding surface, that is, when the active yield surface is F and
the model behaves as a bounding surface plasticity model (the necessary and sufficient
condition for this case is λB = 0); and (b) when the stresses are on the bounding surface
(i.e., B = 0). In the latter case, the bounding surface acts as a yield surface. The
application of the consistency conditions completes the required steps for the full
definition of the constitutive model and leads to the relationship between stresses and
strains. Both cases are presented separately below.
case i) Stresses are inside the bounding surface (λB = 0). The consistency condition is
defined by expanding F using Taylor series and neglecting terms higher than
linear,
FD = 2ξ F : ξD F = 0 (5.15)
dividing by r and assuming r ≠ 0 we get
212
ˆ F : (σD − αD ) = 0
n (5.16)
Equation 5.6, relating the stress rate tensor to the strain rate tensor, can
now be rewritten using the associative flow rule (Equation 5.9) with λB = 0, and
observing that n̂ F is a deviatoric tensor,
σ = K tr(ε )1 + 2G( ε ′ − λ F n
ˆF) (5.17)
where K is the elastic bulk modulus, G is the elastic shear modulus, 1 is the rank-
two identity tensor, and tr(⋅) is the trace operator. Equation 5.17 assumes that the
elastic behavior of the soil is isotropic. By virtue of the vanishing elastic region,
the total deviatoric strain tensor εD ′ is directed along the unit normal vector n̂ F .
Using Equation 5.16 in conjunction with the hardening rule (Equation 5.13), the
consistency parameter λF can be expressed as
1
λF = ε′ (5.18)
H'
1+
3G
From Equation 5.17 and 5.17, and the fact that the strain rate is oriented
parallel to n̂ F , the rate constitutive equation is obtained:
−1
 3µ 
σC = K tr(εC )1 + 2G 1 +  εC ′ (5.19)
 H′ 
Note that the assumption of a vanishing elastic region enters the derivation
solely through the limit in Equation 5.10.
213
case ii) The stress state is on the bounding surface (λF = 0, λB ≠ 0). At this point,
assume that the yield surface and the bounding surface coincide at the stress
point σ'. This assumption must be assured in the implementation by making
sure that loading inside the bounding surface continues until the bounding
surface is encountered. This assumption ensures that the total strain is aligned
with the normal vector n̂ B . The same condition was obtained inside the
bounding surface by using the vanishing elastic region assumption (Equation
5.10). The consistency parameter λB is then obtained in an identical fashion as
its counterpart inside the bounding surface ,λF (Equations 5.15 – 5.20), and is
given by
1
λB = εD (5.20)
H′
1+
3G
The resulting stress-strain difference equation is identical to Equation
5.19. The difference in both cases (i.e. when either F or B are the active yield
surface) lies on the definition of the hardening modulus H' that is presented later.
Note that the previous development could have been followed making use of a
single yield surface F, but with different hardening rules that specify separately when α
changes (in the present development, when λF ≠ 0) and when β changes (i.e., when λB ≠
0).
214
Hardening Function
The hardening modulus H' can be defined in such a way as to fit accepted one-
dimensional cyclic stress-strain relationships. Along with the functional definition of H',
the unloading conditions must also be clearly defined, since plastic deformations grow
proportionally with distance from the point of last unloading (Borja and Amies 1994).
The hardening modulus is obtained from an interpolation between the elastic value (H' =
infinity) at the last unloading point (elastic nucleus), to a limiting value of Ho at the
bounding surface. Following the concept of bounding surface plasticity models (Dafalias
and Popov 1977), an image point σ̂ on B is defined such that (Borja and Amies 1994):
ˆζ = ζ + κ (σ′ − σ′ ) (5.21)
B B o
where ˆζ B = σ
ˆ ′ − β , and σo' is the stress tensor at Fo (Figure 5.1). The dimensionless scalar
quantity κ satisfies the condition.
ˆζ = ζ + κ (σ ′ − σ ′ ) (5.22)
B B o
An exponential hardening modulus is chosen for this work. This model was
validated by Borja et al. (1999 and 2000) using events recorded at the Lotung and Gilroy
downhole arrays. The model is expressed by:
H ′ = hκ m (5.23)
where h and m are material parameters. For other interpolation functions for H', see
Borja and Amies (1994).
215
The loading and unloading conditions when the stress is on the bounding surface
(λB ≠ 0) are given by the Kuhn-Tucker conditions (Equation 5.12). Within the bounding
surface, a different condition must be applied because the singularity due to zero radius of
the elastic region renders the loading condition ill defined. Borja and Amies (1994)
define unloading as the condition in which the direction of the load step causes the
hardening modulus to increase. The loading condition is then postulated as (see Borja
and Amies 1994 for details)
−
(1 + κ )ζ B + κ (1 + κ )(σ′ − σ′o ) : ε′ > 0 (5.24)
ζ B : (σ′ − σ′o ) + κ σ′ − σ′o
2
Upon unloading, the position of the reference point Fo must be shifted to the
current position defined by the stress tensors σ'. Equation 5.24 implies that the
loading/unloading condition depends solely on the direction of the strain tensor εD ′ . This
is a highly desirable condition given that the finite element program GeoFEAP is strain
driven.
5.3 FINITE ELEMENT IMPLEMENTATION
5.3.1 General
The finite element method has been used extensively to solve numerous problems
in geotechnical engineering (e.g. Britto and Gunn 1987, Zienkiewicz et al. 1999). The
finite element method solves a boundary value problem by prescribing interpolation
functions between nodal values for the unknown variables. The solution is obtained by
216
minimizing an energy function or some error measure over the whole domain. The
present implementation is developed within the finite element program GeoFEAP
(Espinoza et al. 1995). GeoFEAP is a multi purpose, nonlinear finite element code
developed as a combined effort between the Computational Mechanics and Geotechnical
groups at the University of California at Berkeley.
For this implementation, the site is assumed to be composed of horizontal layers
of finite thickness and infinite lateral extent resting on an elastic half-space. The
condition of zero initial static shear stresses is assumed. Moreover, the seismic waves are
assumed to originate in the elastic half-space of the soil column and travel in the vertical
direction through the soil column. These assumptions imply that all the variables are
functions of depth and time only. The problem is thus reduced to one spatial dimension,
and the site can be represented by a vertical soil column. The boundary value problem
can then be defined as: Given a prescribed displacement function (of time) at the base of
the soil column, and given a constitutive relations that determines the stress-strain
behavior of the soil, solve for the time-displacement function at any other point in the
column for the given boundary conditions.
The boundary conditions are zero strain in the horizontal direction and zero stress
at the soil surface. The boundary condition for the base of the soil column must be
prescribed such that it acts as an elastic, semi-infinite half space (Section 5.3.3).
Although the problem is one-dimensional in space, the prescribed displacement time
function can have arbitrary spatial orientation, thus each node for each soil element is
allowed three degrees of freedom (rotational degrees of freedom are ignored).
217
The boundary value problem can also be stated in terms of the relative
displacements of the soil with respect to the base. In this case, the boundary condition at
the base is fixed and an inertial body force equal to ρ uDDg , where uDDg is the ground
acceleration, is applied to the soil column (Zienkiewicz and Taylor 1989). The numerical
computations for both cases are identical if the same initial conditions are assumed. The
advantage of the relative displacement formulation is that the input is prescribed in terms
of accelerations, which are often more accurately known (Zienkiewicz et al. 1999).
The governing equations of the dynamic boundary value problem are the balance
laws of linear momentum (i.e. Zienkiewicz and Taylor 1989). The dynamic problem
involves spatial and temporal independent variables. Typically, the time variable is
discretized and is solved using some type of finite difference approximation, while the
spatial problem is solved using the finite element method. Similarly, the solution can be
posed in a space-time domain in what is commonly referred to as space-time finite
element (Zienkiewicz and Taylor 1989). In this implementation, the former method is
used. The method chosen for advancing the solution in time was proposed by Hilber,
Hughes, and Taylor and is commonly known as the HHT method (Hilber et al. 1977).
This time integration scheme is second order accurate and unconditionally stable. The
HHT method is part of a larger family of single step algorithms presented by Katona and
Zienkiewicz (1985). The parameters in the HHT method can be used to control the
amount of numerical damping introduced by the time integration scheme.
The finite element approximation to the balance of linear momentum equations is
presented in detail in finite element textbooks (e.g., Zienkiewicz and Taylor 1989). In
218
this implementation, an iso-parametric formulation using three dimensional "brick"
elements with tri-linear shape functions is used. At any given time, the finite element
approximation together with the time stepping scheme yield a system of nonlinear
algebraic equations for the nodal displacements, F(u), where u is the vector of nodal
displacements. This system of equations represents the balance of internal and external
(applied) forces. In the site response problem, external forces correspond to the applied
inertial forces. Thus, from the global point of view, the finite element problem is load-
controlled. The system of equations F(u) is solved using a Newton-Raphson scheme
(Taylor 1998). The effectiveness of the solution depends on the "quality" of the tangent
used in the solution. Simo and Taylor (1985) have shown that the Newton-Raphson
method in combination with the algorithmic or 'consistent' tangent presents a very
efficient way of solving the nonlinear system of equations F(u). The tangent matrix
∂F/∂u is assembled by the finite element code from the element stiffness matrices. Both
element stiffness matrices and the internal forces are calculated at the element level for a
given value of the displacement node vector. Thus, at the element level, the problem
becomes strain (displacement) controlled. The integration of the constitutive equations at
the element level is presented in Section 5.3.2.
5.3.2 Local integration of the constitutive equations
The incremental solution of boundary value problem requires the integration of
the constitutive equations. In the context of the finite element method, displacements are
primary variables determined iteratively from the global balance laws. Hence, total strain
increments are considered as input data and the integration of the constitutive model
219
reduces to finding the stress state at time n+1 for the given strain increment and the
converged values at time n. The stresses found are used to determine the internal stress
vector used to solve the global set of equations. Moreover, the algorithm at the local
element level must provide the consistent tangent (Simo and Taylor 1985), the element
mass matrix, and the element damping matrix. The type of the mass matrix used in this
implementation is the consistent mass matrix (Zienkiewicz and Taylor 1989). The
element damping matrix is obtained using Equation 5.3. The consistent tangent arises
from the linearization of the stress increment with respect to the strain increment. The
presentation in this section is reduced to the integration of the constitutive equations and
the consistent tangent. The definition of the internal stress vector is presented in finite
element textbooks (e.g. Zienkiewicz and Taylor 1989 and Cook et al. 1989).
Let σn, β n, and σon denote the known stress tensor, the back-stress tensor, and the
last unloading point at time n. For a given strain increment ∆εε, the local integration
algorithm must return the updated values of the stress tensor, σn+1, and the hardening
variable β n. The first step in the iteration is to check whether the current state of stress is
on the bounding surface, let
B = ζBn : ζBn − R2 (5.25)
where B = 0 defines the bounding surface (Equation 5.8), and ζBn is the algorithmic
counterpart of ζ. In the case of B = 0, the bounding surface is the active yield surface (λB
≠ 0 and λF = 0) and the integration of the constitutive model is done through a radial
220
return mapping algorithm. The radial return mapping algorithm is presented in Box 5.1
(from Simo and Hughes 1998).
If B < 0, the current state of stresses is inside the bounding surface and the
integration of the constitutive model and the plastic strains are obtained from the
bounding surface plasticity model. The loading/unloading condition is determined using
the direction of the strain increment ∆εε. The condition for unloading is obtained by
taking the limit of Equation 5.24 as ∆t → 0 (Borja and Amies 1994),
−
(1 − κ n )ζ Bn + κ n (1 + κ n )(σ′n − σ′on ) : ∆ε ′
> tolo (5.26)
ζ Bn : (σ′n − σ′on ) + κ n σ′n − σ′no ∆ε ′
where tol0 is a numerical tolerance parameter. Upon each unloading, the stresses at the
last unload point, σon are reset to σn and the value of κ is set to infinity (i.e., an infinite
plastic modulus). In a strict sense, unloading occurs anytime the left hand side of
Equation 5.26 is larger than zero. Within a finite element implementation, however, the
iterative nature of the solution sometimes produces unloading at extremely small values
of strain. This unloading (termed 'numerical' unloading) can lead to instabilities in the
convergence to the solution. After a reload cycle, the instantaneous plastic modulus H' is
now interpolated between the new unload point at σ ′o and the bounding surface. When
the unloading step is small (i.e. with numerical unloading), the result is observed as a
jump in the stress-strain loop (Figure 3). For this reason, tol0 is set to a positive value <<
1. From the physical point of view, this is equivalent to establishing a small elastic
region around σo where the material behaves elastically upon unloading. The value of
tol0 is determined solely from convergence requirements of the global finite element
221
algorithm. Note, however, that tol0 can also be assigned a physical significance. That is,
tol0 is equivalent to the strain level that induces the soil to 'feel' unloading. This value
could be related to the elastic threshold strain (Pestana, personal comm.).
After the load/unload condition has been established, the integration of the
constitutive equation is performed following the algorithm of Borja and Amies (1994),
which is included in Box 5.2. If Equation 5.26 indicates a condition of numerical
unloading, the material unloads following the tangent stiffness at σ, but the stress reversal
point Fo is not changed from its previous value.
Equation 1 in Box 5.2 is obtained by applying a generalized trapezoidal rule to
Equation 5.19. The parameter β is the trapezoidal integration parameter (i.e., explicit if β
= 0, implicit otherwise). The parameters ψ and κ in Equations 3 and 4 in Box 5.2 are
solved using a Newton iteration with initial estimates ψo = 2G and κn+1 = κn. If κ
converges to a negative value, it implies that the final stress state is outside the bounding
surface. In this case, the strain step is partitioned such that the first step takes place inside
the bounding surface, and the second step corresponds to a kinematically hardening
plasticity model with the bounding surface as the active yield surface. The strain tensor is
then given by:
∆ε = ∆ε B : ∆ε F (5.27)
where ∆εεF = γ ∆εε is the strain within the bounding surface, and ∆εεB = (1 - γ) ∆εε is the
strain step corresponding to plastic yielding on the bounding surface. The parameter γ
corresponds to an intermediate time tn+γ between tn and tn+1 corresponding to the strain
222
step ∆εεF. It is assumed that the strain step is applied uniformly over the time step. The
value of γ is obtained following the algorithm in Box 5.3. After the partition, the stresses
σn+ γ are obtained using the algorithm in Box 5.2 with ∆εε = ∆εεF = γ ∆εε. The strain step
∆εεB is then taken using the algorithm in Box 5.1 with σn+γ as the initial condition. The
tangent modulus Cep is obtained by a linear interpolation of the tangent moduli returned
by Box 5.1 and Box 5.2:
F : (α ∆ ε ) + C B : [( 1 − α ) ∆ ε ]
C ep : ∆ ε = C ep ep
(5.28)
[
= (α )C ep
F + (1 − α )C B
ep
]
: ∆ε
where Cep
F is the consistent tangent corresponding to the strain step inside the bounding
surface (Box 5.2), and Cep

B is the tangent for the kinematically hardening strain step (Box
5.1).
The algorithm in Box 5.2 is implicit for values of β larger than zero. Similarly,
the radial return mapping algorithm in Box 5.1 is obtained by applying the fully implicit
backward Euler integration scheme to the constitutive equation. Fully implicit algorithms
guarantee linear stability, also called A-Stability. Non-linear stability, or B-stability, can
be identified by combining the return mapping algorithm with a time discretization of the
algorithmic weak form of the finite element approximation (Simo and Hughes 1998).
Unfortunately, non-linear stability analyses become very cumbersome and have been
done only for relatively simple models (e.g., Simo and Govindjee 1991).
223
5.3.3 Radiation Boundary Conditions
General considerations
Figure 5.4 presents a schematic diagram of the site response problem. Note that
for simplicity, the column analyzed is called 'soil', and the half-space where the motion is
prescribed is called 'rock'. This terminology is used only for convenience. Clearly, the
input motion could be as easily be prescribed within the soil, as well as the site response
analyses could be performed for a rock column. An incoming stress wave is applied to
the soil column at point A. This stress wave travels through the soil and is reflected at the
free surface of the soil. The reflected wave hits the soil-rock interface and is both
transmitted across and reflected from the base rock. The actual energy of the reflected
and transmitted waves is a function of the impedance contrast (ρsVs/ρrVr) between the soil
and the rock layers. The boundary conditions prescribed in the finite element analysis
should account for the correct distribution of the reflected and transmitted energy at the
rock-soil boundary.
In Figure 5.4, both point A (underneath the soil column), and point B (at the free-
surface) are subject to the same incoming vertically propagating seismic shear waves.
Both points, however, are subject to different boundary conditions. Point A receives the
stress waves that are reflected from the soil surface. On the other hand, the boundary
condition at point B is that of a stress-free boundary. The incoming waves at the free
rock surface are fully reflected into the half space. The effect of the full reflection is that
the incoming and reflected wave have constructive interference and the displacements at
the surface are doubled. The process of modifying the free surface rock motion (typically
224
called the outcropping motion) to represent the motion underneath the soil column (called
the within motion) is called deconvolution. Deconvolution consists essentially in
eliminating the 'free surface' effect and in removing the energy corresponding to the wave
reflected from the surface. In the context of the finite element analysis, the deconvolution
process is accomplished by the appropriate definition of the boundary conditions at the
base of the soil column. In this section, the corresponding boundary conditions are
presented.
When the prescribed input motion in a site response analysis corresponds to an
outcropping motion, a number of conditions must be satisfied so that the deconvolution
process results in the actual motion at the base of the soil column. These conditions
apply in particular when site response analyses are performed to obtain a ground motion
at a particular location. In the present work, site response analyses are preformed for
generic site conditions (Chapter 6) and the considerations provided below are not
necessarily critical in all cases. However, these conditions are presented because of their
importance in the general context of site response analysis.
1) The materials underneath the soil column are assumed to behave as an elastic,
homogeneous half space. If this assumption does not hold in the field, a number
of phenomena which are not accounted for in the analysis can take place. Some of
the energy that is transmitted into the rock from downward traveling waves could
be reflected once more into the soil from larger impedance contrasts at depth.
This situation is likely to arise when the input motion is specified at a relatively
soft layer. At this soft layer, a large proportion of the downward traveling waves
225
is transmitted into the underlying layer. This energy, however, will likely be
reflected back in part from basement rock at lower depths. This consideration
affects long period waves in particular, because these are able to travel larger
distances without scattering and they are attenuated less rapidly by material
damping. Conversely, when the material underlying the soil column is much
stiffer than the overlying material, most of the energy of the downward traveling
waves is reflected back into the soil and the effects of potential impedance
contrasts at greater depths are reduced.
2) The properties of the material underneath the soil column (point A in Figure 4) are
the same as the materials at the free surface (point B). This assumption is
commonly made in site response analyses. The stiffness of geo-materials,
including rock, is a function of confining stress. For this reason alone, it is
unlikely that this condition is satisfied in the field. In addition, rock that is
exposed to a free surface is more likely weathered and has larger joint spacing
than rock that is covered by soil. Both these effects imply softer conditions for
rock at B than at A.
3) The energy of the outcropping motion corresponds solely to vertically propagating
body waves. This condition is also not likely satisfied for most recordings at rock
sites. Silva (1988) found that about 75% of the power (87% of the energy) in a
free surface motion could be attributed to vertically propagating shear waves at
frequencies up to 15 Hz (Kramer 1996). Improvements in the deconvolution
226
process can be realized by appropriate pre-filtering of the recorded ground motion
(Silva 1988).
Mathematical development
Let u = u(z,t) denote the relative horizontal displacement field in the soil column
with respect to the free field condition (input motion). Moreover, assume that only elastic
behavior exists. Under this condition, the laws of balance of momentum and the
constitutive relation σxz = G ∂u/∂z yield the one dimensional wave equation
∂ 2u 1 ∂ 2u
= (5.29)
∂z 2 Vs2 ∂t 2
where Vs is the shear wave velocity (Vs2 = G/ρ). The solution to this equation is given by
u = u I ( z − Vs t ) + uO ( z + Vs t ) (5.30)
where and uI and uo are two waves traveling in the upward (incoming) and downward
direction (outgoing), respectively. At the rock-soil interface (point A), the boundary
condition should represent only an outgoing wave (recall that u is the relative
displacement with respect to the free field condition, and the displacement at the free
field is the displacement of the upward wave). Then we have (Zienkiewicz et al. 1999)
u = u O ( z + Vs t ) (5.31)
Using this condition, we observe that:
227
∂u ∂u
= u ′Vs and = u′ (5.32)
∂t ∂z
du (.)
where u ′ = .
d (.)
Therefore, on the boundary, we have:
∂u 1 ∂u
= (5.33)
∂z Vs ∂t
The tangential shear stress at the boundary then becomes
∂u ∂u
σ xz : G = ρVs (5.34)
∂z ∂t
This is equivalent to the requirement that a viscous dashpot acts on the boundary.
A similar result is obtained for compressive (P) waves. Representation or the boundary
conditions in the manner presented above was suggested independently by Lysmer and
Kuhlmeyer (1969) and Zienkiewicz and Newton (1969). Lysmer and Kuhlmeyer (1969)
show that this type of boundary condition results in appropriate energy transmitting
properties.
5.3.4 Model Parameters
The parameters of the constitutive model and the parameters used in its numerical
implementation are listed in Table 5.1. The methods used to obtain the appropriate value
of the parameters are also included in Table 5.1. The bounding surface plasticity model is
fully defined by two elastic constants, model parameters h and m, and the radius of the
228
bounding surface R. In addition, the plastic modulus Ho defines the behavior of the
model after the bounding surface has been reached.
Parameter h and m are obtained by matching the shear modulus reduction (G/Gmax,
where Gmax is the small strain modulus) and material damping versus shear strain curves.
These curves are typically obtained from tests using radial loading, paths such as cyclic
triaxial, resonant column, and torsional shear tests (Borja et al. 1994). Using the stress
path defined by cyclic simple shear tests, the relationship between the moduli ratio and
the corresponding amplitude of shear strain-stress curves, χ, is given by (Borja et al.
1999):
−1
 R + θ G χ − ξ 
m

3 2θGmaxγ   2 max  
1= ∫ h  + H o  dξ (5.35)
2χ 0
  ξ  
  
where θ = G/Gmax.
Equation 5.35 can be made to pass through two points in the measured shear
modulus reduction versus shear strain curve, and the parameters h and m can be defined
for a given value of R. Once the modulus reduction curve is matched, the predicted
damping curve should be checked to ensure it is within reasonable values for the soil in
question. More often than not, a perfect match of both shear modulus reduction and
damping is not possible, and a compromise between matching the damping curve or
matching the shear modulus curve must be accepted. Figure 5.5 and 5.6 shows the
influence of parameters h and m on shear modulus reduction curves. In general, an
increase in m causes the curvature of shear modulus reduction curves to increase, while
229
an increase in parameter h causes a shift to the right of both shear modulus reduction and
damping curves.
The parameter R, the radius of the bounding surface, is defined by the shear
strength of the soil. The value of R is given by R = 8 3 su, where su is the unconfined
compressive strength of the soil. Similarly, R can be defined by R = 2 τf, where τf is the
failure strength of soil in a simple shear tests. In terms of the shear modulus reduction
curve, R controls the abscissa of the stress-strain curve at large strains. The parameters R,
h, and m are inter-related. Ideally, the shear modulus reduction curves should be defined
independently of the strength of the soil.
The parameter Ho defines the plastic modulus after the soil has reached the
bounding surface. Ho can be determined from the slope of the stress-strain curve at large
strains. Soil typically do not exhibit work hardening, consequently, as they reach failure
the stress-strain curve becomes parallel to the strain axes. In this case, a very low value
of Ho can be assigned. With a low value of Ho, the radial loading component of the
constitutive model becomes nothing else than a numerical tool to deal with soil failure.
Figure 5.7 shows typical stress strain curves predicted by the model at different
amplitudes for regular cyclic loading. Figure 5.8 shows the stress-strain curves
experienced by one element subject to an earthquake loading.
5.3.5 General comments
This section presented the implementation of the bounding surface plasticity
model described in Section 5.2 into a finite element program. An evaluation of the ability
230
of the model to represent observed soil behavior is presented in the following section.
One problem with the soil constitutive model presented here, which is common to many
other nonlinear models, is that variations in the model parameters that are used to shift
shear modulus reduction curves to the right also shift damping curves in the same
direction. Thus, a simultaneous match of recorded shear modulus reduction and damping
curves is not always possible. Pestana and Lok (2000) proposed a hysteretic model based
on the perfect hysteretic model formulated by Hueckel and Nova (1979). Pestana and
Lok used model parameters to describe separately small strain and large strain non-
linearity. In addition, a separate parameter is used to control the rebound modulus, which
describes the slope of the stress-strain curve at the point of strain reversal (Lok 1999). By
modifying this parameter, the shear modulus reduction and damping curves can be
adjusted in opposite directions allowing a better match of both curves (Figure 5.9). Some
soils exhibit strain softening behavior when approaching yield. The model implemented
herein cannot model this behavior. Consequently, problems controlled by strain softening
soil response should not be modeled using this implementation.
5.4 VALIDATION
5.4.1 General
The constitutive model presented in Section 5.2 has been previously validated
through its application to site response analyses for events recorded in the Lotung,
Taiwan and Gilroy, California, downhole arrays (Borja et al. 1999, 2000). Section 5.3
presented an implementation of the model into the finite element program GeoFEAP. In
231
this section, a validation of this implementation is presented with particular emphasis on
the ability of the implementation in capturing long period motions typically associated
with near-fault ground motions. Moreover, the case studies presented serve to illustrate
the stress-strain behavior predicted by the constitutive model. Four case studies are
analyzed. The first case study corresponds to the same one performed by the developers
of the constitutive model (Borja et al. 1999), where the constitutive model is validated for
a particular finite element implementation. The same case study is reconsidered here to
validate the implementation of the model into the finite element program GeoFEAP. The
remaining case studies correspond to an events recorded in the Chiba downhole array in
Japan (Katayama et al. 1990), and a series of shake-table analysis performed at the
University of California at Berkeley (Meymand 1998 ).
5.4.2 Lotung Array
General
The Lotung site is located approximately 30 miles south-east of Taipei, Taiwan.
This site was one of the reference sites used by the Electric Power Research Institute in an
extended study on ground motion estimation (EPRI 1993). The site is the location of the
Large-Scale-Soil-Structure Test (LSST) facility, which was constructed in 1985 jointly by
Taipower and EPRI. The site is instrumented with two downhole arrays and includes
1/12- and 1/4-scale models of nuclear containment structures (EPRI 1993). The two
downhole arrays are located at 3 and 49 m from the containment structures.
Accelerographs are placed at 0, 6, 11, 17 and 47 m, and have recorded a number of
earthquakes since their installment in 1986. The Lotung site has been used in numerous
232
site response studies (i.e. EPRI 1993, Chang et al. 1990, Borja et al. 1999, Borja et al.
2000, among others). The study of Borja et al. (1999, 2000) used the earthquake of May
20, 1986 (Mw = 6.5, R = 68 km) to validate the constitutive model presented in Section
5.2. In this section, the same analysis is repeated to validate the implementation of the
Borja soil constitutive model into the finite element program GeoFEAP. The array
furthest from the containment structures is used throughout validations to represent a
free-field soil response.
Site profile and model parameters
The LSST site is located within the SMART-1 strong motion array. It is situated
on a flat plain in a basin of triangular shape that is 15 km wide and 8 km long (EPRI
1993). The soils at the site generally consist of Holocene and Pleistocene deposits of
interlayered silt-sand and sandy-silt with some gravel overlying a thicker layer of silty-
clay. Bedrock is estimated at approximately 400 m below ground surface (EPRI 1993).
Groundwater is located within a few feet of the ground surface. The shear-wave velocity
profile at the site is presented in Figure 5.10 from the data presented in the
aforementioned EPRI report (1993). The shear-wave velocity profile was originally
reported by Anderson and Tang (1989). The shear wave velocity profile was validated
through elastic analysis of small amplitude earthquakes (EPRI 1993).
Zeghal et al. (1995) used data from recorded events in the LSST array to
backcalculate the shear modulus reduction and damping curves. The parameters h and m
were obtained by matching the shear modulus reduction curve at G/Gmax values of 0.9 and
0.5. The parameters R and Ho are obtained from the moduli ratio curve at 5% shear
233
strain. The resulting shear modulus reduction curves are shown in Figure 5.11 and the
model parameters are given in Table 5.2. Resonant column and cyclic torsional tests on
"undisturbed" soil from the Lotung site suggested that the shear modulus reduction ratio
curves agree well with the upper bound curve for sands proposed by Seed and Idriss
(1970), while the damping curve plots in between the upper and lower bound curves
(EPRI 1993). The EPRI curve as well as the Seed and Idriss curves are included in
Figure 5.11 for comparison. Further details on the properties of the soil at the site are
given in Borja et al. (1999), Zeghal et al. (1995), and EPRI (1993).
Results
The event of May 20, 1986, generally known in the literature as the LSST 7 event,
produced peak ground accelerations at the surface of 0.16 g and 0.21g in the east-west
and north-south directions, respectively. The motion recorded at a depth of 47 m is used
as the input motion in the analysis. A finite element mesh with 47 brick elements, 1 m
thickness, was used. Consistent with the presentation of the finite element approach to
site response, the analysis is performed with only one spatial dimension assuming body
waves propagating in the vertical direction. All nodes in the brick elements at a given
depth are constrained to have equal displacements. With this constraint, the brick
elements behave as one-dimensional stick elements with linear interpolation functions.
The finite element analysis is performed using relative displacements (with respect to the
free-field), and the input ground motion is entered in the form of inertial forces
throughout the finite element model. Since the input motion was recorded at the base of
234
the soil column, the prescribed boundary condition is zero relative displacements (with
respect to the input displacements used in the analysis).
The recorded and the predicted particle acceleration and velocity plots are shown
in Figure 5.12. Peak accelerations and velocities are predicted accurately by the site
response analysis. Time histories of velocity and acceleration at all depths are presented
in Figures 5.13 and 5.14. The matching of both accelerations and velocities at all depths
in the downhole array is excellent.
The same analysis was performed for an uni-directional input motion using
independently the north-south and east-west components of motion. The predicted peak
response values are similar for the uni-directional and bi-directional analyses. Figure
5.15 compares the velocity response spectra for both these analyses. Around the one
second period where maximum spectral velocities are concentrated, the bi-directional
analysis gives a marginally better fit to the recorded motion in the East-West and North-
South directions.
General Observations
The results shown in the previous section agree well with the recorded data.
Similar findings were presented by Borja et al. (1999 and 2000). The LSST event was
also used to perform a series of parametric studies on the influence of different model
parameters on the results. An extended parametric study is not presented here for reasons
of brevity, but a number of conclusions are listed. Observations by Borja and coworkers
in the above papers are also included to present a more complete picture of the
constitutive model used.
235
a) A small increase in the viscous ratio did not result in significantly different
motions at the surface. On the other hand, a decrease in ξ led to motions with
high frequency noise. Borja et al. (2000) indicate that viscous damping has the
effect of suppressing high frequency noise. Results indicated that variations in the
viscous damping affected mainly the initial portion of the ground motion. After
the first acceleration peak, hysteretic damping controls and viscous damping does
not have a significant effect on the results. The suppression of high frequency
noise, however, proved to be important beyond just the high frequency region of
the spectrum. High frequency noise causes a larger number of unloads in the soil.
Upon each unload, the stiffness in the soil changes thus affecting the solution.
b) Borja et al. (2000) repeated the analysis for shear modulus reductions curves
covering the upper and lower bounds of the data processed by Zeghal et al.
(1995). Peak accelerations are under-predicted when the lower bound curves are
used, but are unchanged when the upper bound curve is used.
c) The effectiveness of the program is a function of the convergence rate of the finite
element program for each time step. Ideally, the convergence of the global
Newton-Raphson iteration should be quadratic for starting points in the
neighborhood of the solution. Table 5.3 illustrates the convergence for an element
test for stress levels that remain within the bounding surface. The residual norm
is the norm of the internal minus the external forces. The one element test permits
the direct evaluation of the element consistent tangent. Note that convergence is
better than linear, but it is not quadratic. The cause for this is that the model is
236
slightly strain-step dependent due to the trapezoidal integration of the constitutive
equation and the exponential nature of the function defining H'. The lack of
quadratic convergence does not affect the accuracy of the implementation.
Moreover, for stress states on the bounding surface, the convergence is quadratic
as expected for the radial return algorithm applied to J2 type plasticity (Simo and
Hughes 1998).
d) The parameter tol0 controls the size of the elastic nucleus around an unloading
point. It was found that variations of tol0 within a certain range do not affect the
results. The range is from about 10-8 (in strain units) to the strain defined by the
'elastic threshold' of the model. For this study, the elastic threshold strain
corresponds to the strain level after which the shear modulus reduction curve is
less than 0.99.
e) Changes in the time step do not significantly affect the results. In general, smaller
time steps do increase the energy of high frequency motions. For these small time
steps, viscous damping becomes more important. The time step is in general
determined from the data sampling of the input motion. The value of ∆t also
establishes an upper limit to the frequencies that can be represented in a time
history (Nyquist frequency). The element size also provides a limit to the
frequencies that are affected by the finite element analysis. Frequencies higher
than those corresponding to a wavelength of twice the element length are not
captured by the analysis. The element size used (1 m) more than suffices for the
frequencies of typical earthquake motions (0.2 – 20 Hz) for typical values of soil
237
stiffness. The sensitivity of the results to mesh size were also tested by increasing
the number of elements in the mesh. Results were not affected by element size.
f) The model used herein uses total stresses; thus, it cannot predict pore pressure
generation. Although pore water pressure sensors were not installed during the
LSST 7 event, they were installed shortly thereafter. A later earthquake in
November 1986 of a slightly higher magnitude and similar distance (LSST16)
generated a pore pressure ratio of about 25%. These levels are unlikely to cause
significant degradation of soil shear strength and stiffness (Borja et al. 1999).
g) The model, as presented, can also deal with vertically propagating pressure (P)
waves that normally generate vertical motion. Borja et al. (1999) present an
analysis of the vertical motion on the LSST 7 event resulting in a good match with
observed motion. Those results were duplicated in this study. For the remainder
of the work, vertical motions are excluded. Although the vertical ground motions
can be potentially significant (EPRI 1993), the study of near-fault motions
presented herein is constrained to the more important horizontal motions.
A number of other site response analyses have been performed on the LSST 7 data
and are included on the EPRI study on ground motion estimation (1993). Borja et al.
(2000) compares the results of these studies with the predictions of the bounding surface
plasticity model. Borja's model performs satisfactorily when compared with other
nonlinear models such as DESRA (Lee and Finn 1991), SUMDES (Li et al. 1992), and
TESS (Pyke 1992). The equivalent linear model SHAKE91 (Idriss and Sun 1992) has
predictions that are as good as these predictions with the current model, although peak
238
values are slightly under-predicted by SHAKE91 (Borja et al. 2000). Further
comparisons with SHAKE91 are performed for the Chiba downhole array.
5.4.3 Chiba Downhole Array
Generalities
The Chiba downhole array is located approximately 30 km east of Tokyo on the
Chiba Plain. The Chiba array data are made available by the Institute of Industrial
Science of the University of Tokyo. The array consists of 44 three component
accelerographs located in 14 boreholes (Katayama et al. 1990). The accelerograms are
placed at different depths in each of the boreholes. The borehole at the center of the array
has accelerographs at depths of 1, 5, 10, 20, and 40 m and it is used in this study to
represent free-field soil response. Previous site response analyses at this site using
equivalent-linear analysis (e.g., Katayama et al. 1990) and a system identification
approach (Glaser and Gaskings 1998) resulted in a good match of recorded motions
indicating that typical site response analysis assumptions (i.e., vertically propagating
shear waves, no significant pore pressure generation) hold for this site.
Soil Profile and Model Parameters
The Chiba array is placed in a geologically and topographically simple location.
Fifteen boreholes were drilled and logged. The logs indicated spatially consistent
horizontal profiles. The site consists of approximately 3 to 5 m of loam, underlain by 2 to
4 m of sandy clay which is in turn underlain by a stiffer sand layer (Katayama et al. 1990).
The water table is located at a depth of 5 m. The sand layer contains thin clay layers and
239
has a stiffness that increases with depth. The inferred shear wave velocity profiled used
in the analysis is shown in Figure 5.16 (Katayama et al. 1990). The Vucetic and Dobry
(1991) damping and shear modulus reduction curves for clay with PI=30 are used to
match the shear modulus reduction curve for the clay layers. For the sand layers, the
confining pressure-dependent curves by Iwasaki et al (1978) are used. Unit weights are
assumed using reasonable values for the soil types in the profile.
The model and numerical parameters used in the analysis are listed in Table 5.4.
The shear modulus reduction and damping curves used are shown in Figure 5.17. In
matching the shear modulus reduction and damping curves, the parameter R was used as a
matching parameter and was not used matched to the soil strength. This allows a better
matching of both shear modulus reduction and damping curves for strains up to 1%. For
larger strains, the model over-predicts the strength of the soil. However, the largest
strains observed in the profile for these ground motions were only on the order of 0.1%.
Results
The Chibaken-Toho-Oki earthquake (MJMA = 6.7) was used for this study. The
recorded and predicted time histories of acceleration and velocity are shown in Figures
5.18 and 5.19. Peak ground accelerations at the surface in the direction of maximum
intensity (North-South) are under-predicted by about 15% (0.33 g compared to 0.27 g).
In the East-West direction, the peak ground acceleration is predicted to within 10% (0.2 g
for the recorded motion, 0. 18 for the predicted motion). The general shape of the
acceleration time history is predicted relatively well with the exception of the under-
prediction of some PGAs late in the motion. Velocities, on the other hand, are predicted
240
to within .5 % in the north south direction, and to within 5% in the east-west direction
(Figure 5.19). Prediction of velocities is important because near-fault ground motions
have high energies at periods controlled by velocities. Note, however, that the velocities
observed at Chiba are lower than velocities expected at near-fault sites. However,
downhole records of near-fault soil response not involving liquefaction (i.e. Port Island
Array in the 1995 Kobe earthquake) are not available. The Chiba records represents the
highest intensity records available at this time.
A site response analysis using the equivalent linear code SHAKE91 (Idriss and
Sun 1992) was also performed. Figure 5.20 shows the pseudo-velocity response spectra
in the North-South direction for both analysis (the GeoFEAP analysis is performed using
only the North-South directions to simulate the one-directional SHAKE91 analysis).
Both methods yield almost identical results at all periods except for the PGA. Figure
5.21 shows the maximum strain profile predicted by both methods. The profiles show
similar patterns and predict the same level of strains at all depths except near the surface,
where SHAKE91 predicts larger strains.
5.4.4 Analysis of Shaking Table Tests
General Description of Experiments
A series of scaled physical model experiments were performed at the U.C.
Berkeley/PEER Center Earthquake Simulator Laboratory to examine the seismic response
of soil-pile-superstructure interaction. The experiments are described in detail in
Meymand (1998) and Lok (1999). Each test included a borehole with accelerographs at
241
different depths located at a relatively large distance from the pile group. It is assumed
that this array reflects free-field response.
The soil used in the analyses is a model soil prepared to simulate the response of
San Francisco Bay Mud at the model scale. The model soil consisted of 72% kaolinite,
24% bentonite, and 4% type C fly ash by weight, and has a plasticity index of 75. The
soil has an undrained shear strength of 100 psf and a shear wave velocity of
approximately 32 m/s. The model container is 7.5 feet in diameter and 7 feet in height.
The container was filled with 6 feet of the model soil overlying 6 inches of sand. The
flexible-wall container was designed to provide pseudo-free-field conditions such that the
soil column can deform horizontally in pure shear mode (Meymand 1998, Lok 1999). An
evaluation of the response of the soil column, particularly as it pertains to representing
true free-field response, is included in Lok (1999). Lok was able to predict reasonably
well spectral accelerations at long periods using both an equivalent-linear model
(SHAKE91) and a fully nonlinear model (Lok 1999). This indicates that the free field
assumptions is likely acceptable, and influence of the boundary conditions generated by
the container, if any, is constrained to high frequencies.
The shaking table tests were performed using the Yerba Buena Island record from
the Loma Prieta earthquake and the Port Island record from the 1995 Kobe earthquake as
input motions. The motions were scaled both in the time domain (to comply with model
scaling laws) and in amplitude, to simulate earthquakes at varying intensities. In this
study, the tests with the Port Island motion are used. The Port Island record from the
242
1995 Kobe earthquake contains forward-directivity effects. Table 5.6 lists the input
motion PGA for the analyses performed in this study.
Soil Properties
The shear-wave velocity profile of the model soil in the container is shown in
Figure 5.22. The shear-wave velocity profile was obtained before the testing sequence
was initiated. The stiffness of the model soil column could be expected to decrease in
response to strong shaking due to shear modulus reduction and inelastic effects; and
increase with periods of rest due to thixotropy (Meymand 1998). Sine-sweep tests were
performed prior to each test to determine the natural period of the soil column prior to
each test. Meymand (1998) and Lok (1999) used the sine-sweep tests to calibrate the
shear wave velocity profiles for each individual site response analyses. The base-line
shear-wave velocity profile was multiplied by a constant to obtain the same first-mode
frequency measured in the sine-sweep tests. This, somewhat simplistically, implies that
the effects of stiffness degradation and thixotropy occurred uniformly across the entire
soil depth. This would not be expected due to the effects of boundary conditions and
shear strain concentrations during shaking. The shear wave velocity profile used by
Meymand (1998) and Lok (1999) in equivalent linear analyses was also increased by a
factor of 30%. This increase was justified on the grounds that the strains induced in the
soil by the sine sweep tests were high enough to cause a modulus degradation of this
magnitude, thus they did not capture the small-strain stiffness of the soil. This
justification relies both on the assumption that strains (and consequently moduli
degradation) are constant through the profile and also on the equivalent-linear assumption
243
of an 'equivalent strain' for the whole time history. In the present analysis, this 30%
increase was ignored and the baseline shear-wave velocity profiles modified to match sine
sweep tests were used.
The measured shear modulus reduction and damping curves are presented in
Figure 5.23 along with the curves predicted using model parameters calibrated to match
these curves. The model parameters used to match these curves are given in Table 5.5.
As in the analysis of the Chiba data, the parameter R was used to match shear modulus
reduction curves at strains lower than 10%. Thus, soil strength is overestimated for
strains larger than 10%. Nonetheless, maximum strains observed in the analysis are
lower than 4%. The properties of the bottom sand layer in the profile were not very well
defined by experiments. A subsequent parametric study, however, showed that results
were relatively insensitive to the shear-wave velocity of this layer. The shear-wave
velocity value estimated by Meymand (1998) as a 'best match' was used in the analyses.
Results
The predicted and observed velocity response spectra for all the runs in Table 5.6
are shown in Figures 5.24 to 5.27. Observe that in general, the shape of the velocity
response spectra is predicted well. However, at periods corresponding to peak response
spectral velocities, the model underpredicts peaks in the velocity response spectrum,
except for the test with the highest intensity (Figure 5.27). In the worst case, the under-
prediction is about 25 % (Test 2.53), however, in most cases the peak spectral velocities
are predicted to within 15 %. At high frequencies, the model fails to predict a peak in the
response spectra. Table 5.6 lists the recorded and calculated peak ground motion values.
244
Clearly, the match of velocity values is better than the acceleration values, reflecting the
better match at longer periods. The spectral velocities presented in Figure 5.24 to 5.27
are at the model scale. Velocities at field scale must be multiplied by the scaling factor
8 (Meymand 1998). The input velocities used in these analysis correspond to the
ranges of peak ground velocities expected at near-fault, forward directivity sites.
Discussion
Note that the high frequency peak in the recorded response spectra is missing for a
depth of 18 inches (Figure 5.24 to 5.26). At this depth, response spectral values have a
better match for the high frequency region of the spectrum. The better match is a result of
the fact that this depth lacks the high energy content at high frequencies that is observed
at other depths. Not coincidentally, this depth corresponds to a nodal point (i.e., a point
of zero displacements) for the second natural frequency of the system, that is, the
frequency corresponding to a wavelength equal to 4/3H, where H is the height of the soil
column. This observation implies that the finite element model does not capture well the
standing wave that develops at the second mode.
Another reason for the relatively poor match at high frequencies may be the
effects of viscous damping. Note that the time-scaling of the motion results in input
motions with uncharacteristically high energy at low frequencies. At these frequencies,
the effect of viscous damping is much higher. The high frequency energy present in the
motion might also be caused by the container and not a characteristic of soil-column
response.
245
Finally, observe that predictions are better for the fault normal direction than for
the fault parallel direction (Figures 5.25 – 5.28). This might occur because the pile
groups show much more intense motions in this direction (i.e., the direction of larger
intensity). Although the model was unable to predict high frequency response of the soil
in the shaking table experiments, velocities were predicted with relative accuracy. Since
near-fault ground motions are evaluated mainly on the long-period content, the results are
encouraging, especially because the input motion used has the characteristics of near-
fault, forward-directivity motions that are targeted in this work.
5.4.5 General Comments
In general, the case studies indicate that the finite element site response analysis
methodology can adequately model bi-directional site response. In particular, the
implementation showed ability to predict amplifications in the velocity-controlled region
of the spectra. The site response implementation is used in Chapter 6 to study site
response to near-fault ground motions.
246
BLANK
247
BOX 5.1. Integration of radial return algorithm (from Simo and Hughes 1998).
1. Assume the stress states is initially on the bounding surface B. Let:

σ trn + 1 = 2G∆ε′n + 1 (1)
ζ trB = σ trn +1 − β n (2)

2. Load/unload check
f ntr+ 1 = ζ trn + 1 − R (3)
if f ntr+1 < 0, then set unloading point σo to current stress. Proceed with Box 5.2.
3. Calculate
∆ε ζ trn +1
λB = ˆ =
; n (4)
H′ ζ trn +1
1+
3G
4. Update stresses and hardening variables.
2
β n+1 = β n + H o λB nˆ (5)
3
σ n +1 = σ n + Ktr (∆ε )1 + 2G (∆ε − λBn
ˆ B) (6)
5. Elasto-Plastic Tangent
 1 
Cep
n + 1 = K 1 ⊗ 1 + 2Gθ n + 1  I −
ˆ B ⊗n
1 ⊗ 1  − 2G θn + 1n ˆB (7)
 3 
where I is the rank-four symmetric identity tensor, and θn+1 and θ n+1 are given by
2GλB 1
θn +1 = 1 − and θn + 1 − (1 − θ n + 1 )
tr
ζ n +1 H′
1+
3G
248
BOX 5.2. Integration of bounding surface plasticity model (Borja and Amies 1994).
1. From Equation 5.19, applying a trapezoidal integration:
 1− β β 
∆σ′ + 3G +  ∆σ′ = 2G∆ε′ (1)
 H n′ H n′ +1 
where H' =h κm + Ho
2. Use the fact that ∆σ

σ' is directed along ∆εε', and obtain
1− β β 
ψ + 3Gψ  +  = 2G (2)
 H n′ H n′ + 1 
were ψ is a positive scalar defined such that ∆σ
σ' = ψ ∆εε'
3. Apply Equation 5.22
R = ζ n + ψ∆ε ′ + κ n +1 (σ′n + ψ∆ε′ − σ′on ) (3)
4. Solve for ψ and κ from Equations 2 and 3

5. Apply the stress increment
σ n+1 = σ n + Ktr (∆ε )1 + ψ∆ε ′ (4)
249
BOX 5.3. Decomposition of strain step.
1. Let:
∆ε = ∆ε B + ∆ε F (1)
where ∆εεF is the strain that loads the soil to the bounding surface at time tn+γ, and ∆εεB is
the strain the soil experiences after it has reached the bounding surface.
2. Solve for
2G
ψγ = (2)
1− β β 
1 + 3G + 
 H n′ H o 
3. Apply Equation 5.22 at time tn+γ:
σ n+γ = σ n + ψ γ γ∆ε − β n = R (3)
where ∆εεF = γ∆εε. Now solve for α
− ζ n : ∆ε + (ζ n : ∆ε )2 + ∆ε : ∆ε(R 2 − ζ n : ζ n )
γ = (4)
ψ γ ∆ε : ∆ε
250
Table 5.1. Parameters needed for the model implementation.
Parameter Function Procedure to obtain it

Shear modulus obtained from shear-
wave velocity logs (G = ρVs2)
Poisson's ratio estimated or
Shear Any pair of elastic calculated from
Modulus (G) parameters. Describe
1  2 − V p Vs 
Elastic 2 2
Poisson the elastic behavior of
Parameters ν =  ,
Ratio (ν) the soil at unload 2  1 − V p2 Vs2 
points.
where Vs and Vp are the shear wave
and compression wave velocities,
respectively.
Material
Density (ρ) Measured from boring log samples
Density
Interpolation function
Matching modulus degradation
h, m for the hardening
Exponential curves (Equation 5.35).
modulus.
Model
Kinematic hardening
Parameters Obtained from tangential shear
Ho parameter of the
modulus at large strains.
bounding surface.
R= 8 3 su ; R = 2 τf
Strength Radius of bounding where su is the soil strength obtained
R from unconfined compression tests,
Parameter surface.
and τf is the soil strength from simple
shear tests.
ξ is damping ratio at very small
Determine the viscous
strains, obtained from cyclic tests at
ξ component of stress-
small strain levels (e.g., Lanzo and
strain relationship.
Vucetic 1999).
Small Strain
Frequency band where
Damping
ξ is matched. Outside ω1, ω2 typically chosen to encompass
ω1, ω2 this band, the model the natural period of the soil and the
results in damping predominant period of input motion.
higher than ξ.
Trapezoidal integration Choose β :> 0 for implicit integration
β parameter. (Box 5.2).
Determines radius of
region around
unloading point where
the soil remains elastic.
Numerical Its function is to
parameters minimize 'numerical Elastic threshold from laboratory
tol0 unloading', that is, modulus reduction curves (i.e. strain
unloading in the soil level at which G/Gmax < 1).
due to infinitesimal
strain steps taken while
the finite element
algorithm searches for a
solution.
251
Table 5.2. Model and numerical parameters used for analysis of Lotung array. The
shear-wave velocity profile is given in Figure 5.10.
Parameter Value
h/Gmax 0.63
m 0.97
R/Gmax 0.0015
Ho 0
ρ 1.95
v 0.48
ξ 2.0%
ω1 4
ω2 0
β 0.5
tol0 0.0001 %
Time step (∆t) 0.04 s
252
Table 5.3. An example of the rate of convergence of the Residual Norm in the finite
element implementation.
Iteration Residual Norm

1 1
2 5e-2
3 1e-4
4 3e-7
5 8e-10
6 2e-12
convergence
7
(< 1e-16)
253
Table 5.4. Model and numerical parameters used in the analysis of the Chiba downhole
array. The shear-wave velocity profile is given in Figure 5.16.
(a) Model parameters
Sand Sand
Parameter Clay
1 ksc < σv' < 3 ksc σv' > 3 ksc
h/Gmax .088 0.366 0.126
m 1.0768 1.035 1.159
R/Gmax .005 .002 0.01
Ho 0 0 0
ξ 1% 1% 1%
ρ 1.92 – 2.0 1.92-2.0 1.45-1.76
ν 0.49* 0.49 *
0.49*
*
For incompressible (undrained) deformation, ν should be set to 0.5. The value 0.49 is used to avoid
numerical errors associated with incompressible deformations.
(b) Numerical Parameters
Parameter Value
ω1 11
ω2 33
β 0.5
tol0 0.0001 %
254
Table 5.5. Model and numerical parameters used in the analysis of the Shaking Table
tests. Shear-velocity profile is given in Figure 5.23.
Parameter Model Soil
h/Gmax 0.0613
m 0.859
R/Gmax .02
Ho 0
ρ 1.51
ν 0.49*
ξ 2%
ω1 20
ω2 0
β 0.5
tol0 0.0001 %
*
255
Table 5.6 Peak response values for the Shaking Table Runs.
PGA PGA PGV PGV

Input PGA
Test recorded calculated recorded calculated
(g)
(g) (g) (cm/s) (cm/s)
2.16 0.1 0.15 0.10 8.3 7.9
2.53 0.25 0.61 0.36 27.7 17.7
2.55 0.7 1.22 0.62 55.6 41.4
2.58 1.0 1.27 0.68 67.9 57.6
256
σ1
σ'
H'= ∞ σ'
σo B
F
Fo
Ho < H' < ∞ σ3

σ2
H' > Ho
Figure 5.1. Schematic representation of bounding surface plasticity model. Fo is the

unloading point, B is the bounding surface and, F is the yield (loading) surface. Contours
of constant H' are centered about Fo, where H' is infinite, decreasing to zero on B
(adapted from Borja and Amies 1994).
257
Fraction of Critical Damping
Total Damping
target ξ
Stiffness Proportional
Damping
Mass Proportional
Damping
ω1 ω2
Frequency
Figure 5.2. Fraction of critical damping versus frequency for Rayleigh damping. ξ is the
target fraction of critical damping.
258
1 1
0.9 0.9
0.8 0.8
Normalized Shear Stress
Normalized Shear Stress

0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0 0.5 1 1.5 2 0 0.5 1 1. 5 2
Shear Strain (%) Shear Strain (%)
a) b)
Figure 5.3. Stress-strain loops for two different loading paths. The dotted line is for
monotonic loading. The solid line is for monotonic loading with an unload cycle of
magnitude ∆ε at 0.5% strain. a) Numerical unloading (∆ε ≈ 1e-6%). b) True unload (∆ε ≈
0.2 %).
259
Reflected
Wave
Soil
A z B
Rock
x
Incoming
Wave
Figure 5.4. Schematic representation of site response problem.
260
1
0.8
Moduli Ratio
0.6
0.4 Decreasing h
0.2
0 -4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain
30
h = 0.01 Gmax
25 h = 0.1 Gmax Decreasing h
h = 0.5 Gmax
Damping Ratio
20 h = 1 Gmax
h = 5 Gmax
15
10
0 -4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain
Figure 5.5. Influence of hardening parameter h on modulus reduction and damping

curves. R/Gmax = .02, m = 1, Ho = 0 (adapted from Borja and Amies 1994).
261
1
0.8
Moduli Ratio
0.6
0.4 Decreasing m
0.2
0 -4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
30
m=0.5
25 m=1
m=2
Damping Ratio
20 m=4
m=8
15 Decreasing m
10
0 -4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
Figure 5.6. Influence of hardening parameter m on modulus reduction and damping

curves. R/ Gmax = .02, h/ Gmax = 1, Ho = 0 (adapted from Borja and Amies 1994).
262
1
0.8
0.6
0.4
Normalized Stress
0.2
-0.2
-0.4
-0.6
-0.8
-1
-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1
Strain (%)
Figure 5.7. Representative stress-strain loops at different amplitudes. R/ Gmax = .02,

h/ Gmax = 1, m = 1, Ho = 0.
263
0.8
0.6
0.4
Normalized Stress
0.2
-0.2
-0.4
-0.6
-0.8
-0.4 -0.3 -0.2 -0.1 0 0.1 0.2 0.3
Strain (%)
Figure 5.8. Finite element response when subjected to cyclic seismic loading R/Gmax =
.02, m = 1, h/Gmax = 1, Ho = 001. The loading was the KJMA station in the Kobe
earthquake.
264
τ
Gmax
Increasing c
(a)
Modulus Degradation for Nonlinear Cyclic Model

1
G/Gmax
0.5
c
0
-4 -2 0
10 10 10
Damping for Nonlinear Cyclic Model
25
Damping Ratio (%)
20
15
10
5 c
0
-4 -2 0
10 10 10
Shear Strain (%)
(b)
Figure 5.9. Influence of the slope of the rebound modulus in the model by Pestana and
Lok. (a) Increase in parameter c reduces the slope of the stress-strain curve at the
rebound point. (b) The resulting stress-strain curves can be adjusted in opposite
directions (from Lok 1999).
265
Vs (m/s)
0 100 200 300 400
0
10
20
Depth(m)
30
40
50
Figure 5.10. Shear wave velocity (Vs) profile for Lotung Site (Borja et al. 1999, EPRI
1993).
266
1
0.8
0.6
G/Gmax
0.4
0.2
0
-4 -3 -2 -1 0
10 10 10 10 10
Strain (%)
40
Seed and Idriss (1970) Sand
Curves
EPRI Curves
30
Damping Ratio (%)
Model Prediction
20
10
0
-4 -3 -2 -1 0
10 10 10 10 10
Strain (%)
Figure 5.11. Modulus degradation and damping curves for the Lotung site.
267
Recorded data GeoFEAP
North-South Acceleration (g)
North-South Acceleration (g)

0.2 0.2
0 0
-0.2 -0.2
-0.2 0 0.2 -0.2 0 0.2

East-West Acceleration (g) East-West Acceleration (g)
North-South Velocity (cm/s)
North-South Velocity (cm/s)
20 20
0 0
-20 -20
-20 0 20 -20 0 20
East-West Velocity (cm/s) East-West Velocity (cm/s)
Figure 5.12. Comparison of recorded and computed ground motions at the surface for the
Lotung array.
268
Surfac e Sur face
0 .2
0.2
Acc. (g)
Acc. (g)
0.1
0
0
-0.1
-0 .2
5 10 15 20 5 10 15 20
Elev 6 Elev 6
0 .2
0.2
Acc. (g)
Acc. (g)
0.1
0
0
-0.1
-0 .2
5 10 15 20 5 10 15 20
Elev 11 Elev 11
0 .2
0.2
Acc. (g)
Acc. (g)
0.1
0
0
-0.1
-0 .2
5 10 15 20 5 10 15 20
Elev 17 Elev 17
0 .2
0.2
Acc. (g)
Acc. (g)
0.1
0
0
-0.1
-0 .2
5 10 15 20 5 10 15 20
Elev 47 Elev 47
0 .2
Thick: Record ed 0.2
Thin: Calcula ted
Acc. (g)
Acc. (g)
0.1
0
0
-0.1
-0 .2
5 10 15 20 5 10 15 20
Time(s ) Time(s)
(a) East-West direction (b) North-South direction
Figure 5.13. Recorded and computed acceleration time histories in the Lotung array.
Thick lines are recorded accelerations. The input motion is at Elev. 47.
269
Surfac e Sur face
20 20
Vel.(cm/s)
Vel.(cm/s)
0 0
-2 0 -20
5 10 15 20 5 10 15 20
Elev 6 Elev 6
20 20
Vel.(cm/s)
Vel.(cm/s)
0 0
-2 0 -20
5 10 15 20 5 10 15 20
Elev 11 Elev 11
20 20
Vel.(cm/s)
Vel.(cm/s)
0 0
-2 0 -20
5 10 15 20 5 10 15 20
Elev 17 Elev 17
20 20
Vel.(cm/s)
Vel.(cm/s)
0 0
-2 0 -20
5 10 15 20 5 10 15 20
Elev 47 Elev 47
20 Thick: Recor ded 20

Vel.(cm/s)
Vel.(cm/s)
Thin: Calcula ted
0 0
-2 0 -20
5 10 15 20 5 10 15 20
Time(s ) time(s)
(a) East-West direction (b) North-South direction
Figure 5.14. Recorded and computed velocity time histories in the Lotung array. Thick
lines are recorded velocities. The input motion is at Elev. 47.
270
East - West
70
Velocity Response Spectra
60
50
40 Recorded
(cm/s)
Bi-directional
30
Uni-directional
20
10
0
0.01 0.1 1 10
Period (s)
North - South
70
Velocity Response Spectra
60
50
40 Recorded
(cm/s)
Bi-directional
30
Uni-directional
20
10
0
0.01 0.1 1 10
Period (s)
Figure 5.15. Comparison of uni-directional and bi-directional analyses for the Lotung
array.
271
Vs (m/s)
0 200 400 600
0
10
15
Depth(m)
20
25
30
35
40
Figure 5.16. Shear wave velocity (Vs) profile for the Chiba site (Katayama et al. 1990).
272
1
0.8
0.6
G/Gmax
0.4
0.2
0 -4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (% )
40
PI =30 (Vucetic an d Do bry 1 991)
35 San ds, C P = 1-3 k sc (Iwasaki 1978)
San ds, C P > 3 ksc (Iwa saki 1978 )
30 Mod el: C lay
Damping Ratio (%)
Mod el: Sand, CP = 1-3 ksc

Mod el: Sand, CP > 3 ksc
25
20
15
10
5
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (% )
Figure 5.17. Modulus degradation and damping curves for soils in the Chiba site. Thin
lines correspond the curves predicted by the model. Model parameters are matched to
curves from the indicated references.
273
Surface Sur face
Acc. (g) 0 .2 0.2
Acc. (g)
0 0
-0 .2 -0.2
5 10 15 5 10 15
Elev 10 Elev 10
0 .2 0.2
Acc. (g)
Acc. (g)
0 0
-0 .2 -0.2
5 10 15 5 10 15
Elev 20 Elev 20
0 .2 0.2
Acc. (g)
Acc. (g)
0 0
-0 .2 -0.2
5 10 15 5 10 15
Elev 40 Elev 40
Thick: Recor ded
0 .2 Thin: Calcula ted 0.2
Acc. (g)
Acc. (g)
0 0
-0 .2 -0.2
5 10 15 5 10 15
Time(s) Time(s)
a) East-West direction b) North-South direction
Figure 5.18. Recorded and computed acceleration time histories in the Chiba array.
Thick lines are recorded accelerations. The input motion is at Elev. 40. The
accelerogram at 5 m depth did not trigger for this event.
274
Surface Sur face
15 15
10 10
Vel.(cm/s)
Vel.(cm/s)
5 5
0 0
-5 -5
- 10 -10
- 15 -15
5 10 15 5 10 15
Elev 10 Elev 10
15 15
10 10
Vel.(cm/s)
Vel.(cm/s)
5 5
0 0
-5 -5
- 10 -10
- 15 -15
5 10 15 5 10 15
Elev 20 Elev 20
15 15
10 10
Vel.(cm/s)
Vel.(cm/s)
5 5
0 0
-5 -5
- 10 -10
- 15 -15
5 10 15 5 10 15
Elev 40 Elev 40
15 15
10 Thick: Record ed 10
Thin: Ca lculat ed
Vel.(cm/s)
Vel.(cm/s)
5 5
0 0
-5 -5
- 10 -10
- 15 -15
5 10 15 5 10 15
Time(s) Time(s)
a) East-West direction b) North-South direction
Figure 5.19. Recorded and computed velocity time histories in the Chiba array. Input
motion is at Elev. 40.
275
50
45
40
35
Velocity (cm/s)
30
Recorded
25
SHAKE91
20 GeoFEAP
15
10
0
0.01 0.1 1 10
Period (s)
Figure 5.20. Acceleration response spectra for the analyses of the North-South
component of motion in the Chiba downhole array.
276
Strain (% )
0 0.05 0.1
0
10
Depth(m)
SHAKE91
20
GeoFEAP
30
40
Figure 5.21. Strains predicted by SHAKE91 (Idriss and Sun 1992) and the GeoFEAP
analysis for the North-South component of the Chiba downhole array.
277
Vs (m/s)
0 100 200 300
0 0
-8
0.2
Vs Profile
0.4 -18
Accelerograph
0.6 Locations
-30
0.8
Depth(m)
-48
1.2
1.4
-60
1.6 -66
1.8 -72
-75
2
Figure 5.22. Shear-wave velocity profile of the model clay soil used in the Shaking
Table Test 2.53, including location of accelerograms.
278
1
0.8
0.6
G/Gmax
0.4
0.2
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
40
Young Bay Mud (Isenhower and Stokoe 1981)
Laboratory Curve
30
Damping Ratio (%)
Predicted Curve
20
10
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
Figure 5.23. Modulus degradation and damping curves for model soil used in shaking
table test.
279
20 20
Elev 0
10 10
0 0
20
Elev -8
10
20 20
Elev -18
10 10
0 0
20
Elev -30
Velocity Response Spectra (cm/s)
10
20 20
Elev -48
10 10
0 0
20
Elev -60
10
20
Elev -66
10 Thick: Recorded
Thin: Calculated
0
20
Elev -72
10
20 20
Elev -75
10 10
0 0
0.01 0.1 1 10 0.01 0.1 1 10
a) Fault Normal Direction b) Fault Parallel Direction
Figure 5.24. Velocity response spectral for recorded and calculated motions in Test 2.16.
The thick line corresponds to the recorded motions. The input motion is at Elev. –75.
280
Elev 0
50 50
0 0
Elev - 8
50
Elev - 18
50 50
0 0
Elev - 30
50
Elev - 48
50 50
0 0
Elev - 60
50
Elev - 66
50
Thic k: Recor ded
Thin: Calculated
0
Elev - 72
50
Elev - 75
50 50
0 0
0.01 0.1 1 10 0.0 1 0.1 1 10
281
Elev 0
1 00 100
0 0
Elev - 8
1 00
Elev - 18
1 00 100
0 0
Elev - 30
1 00
Elev - 48
1 00 100
0 0
Elev - 60
1 00
Elev - 66
1 00
Thic k: Recor ded
Thin: Calculated
0
Elev - 72
1 00
Elev - 75
1 00 100
0 0
0.01 0.1 1 10 0.0 1 0.1 1 10
282
2 00 200
Elev 0
1 00 100
0 0
2 00
Elev - 8
1 00
2 00 200
Elev - 18
1 00 100
0 0
2 00
Elev - 30
1 00
2 00 200
Elev - 48
1 00 100
0 0
2 00
Elev - 60
1 00
2 00
Elev - 66
1 00
Thic k: Recor ded
Thin: Calculated
0
2 00
Elev - 72
1 00
2 00 200
Elev - 75
1 00 100
0 0
0.01 0.1 1 10 0.0 1 0.1 1 10
283
CHAPTER 6
SITE RESPONSE ANALYSIS OF NEAR-FAULT
GROUND MOTIONS
6.1 INTRODUCTION
The analysis of near-fault, forward-directivity ground motions presented in
Chapter Four indicated that, for some motions, velocity pulse-period and amplitude
depend on local site conditions. Only a limited amount of recorded data is available to
evaluate site effects in the near-fault region. However, the understanding of the possible
effects of site conditions on near-fault motions can be enhanced by the use of numerical
simulations. This chapter presents a numerical evaluation of site response to forward-
directivity motions. The trends in the recorded data of near-fault ground motions
presented in Chapter Four are used to validate the results of the analyses presented herein.
The Borja and Amies (1994) constitutive model was employed in this near-fault
seismic site response study. This model is implemented in the program GeoFEAP, as
described in Chapter Five. The advantages of using a bi-directional time-domain
formulation for the analysis of near-fault motions were discussed in Chapter Five. The
numerical parameters used in the validation of the model (Table 5.4b in Section 5.4) are
used for the baseline analyses presented in this chapter. Radiation boundary conditions
(Lysmer and Kuhlemeyer 1969) are used to model the elastic half-space underlying the
finite element model of the soil profile.

284
A set of generalized site profiles for the primary site conditions presented in Table
3.1 is first developed. A set of recorded near-fault rock motions is then used as input for
seismic site response analyses using the generalized profiles. In order to validate the use
of simplified pulses for site response studies, the results of these studies are compared to
analyses using simplified representations of the same input motions. The simplified
pulses developed in Chapter Four are then used to perform an analysis of the effect of site
conditions on the characteristics of rock velocity pulses. Insights from these analyses
provide the basis for the findings presented at the end of this chapter.
6.2 GENERALIZED SITE PROFILES
6.2.1 General
The generalized site profiles were selected to match average shear wave velocity
profiles associated with the primary site categories developed in this study and listed in
Table 3.1. Namely, the generalized site profiles represent the Shallow, Very Stiff Soil
(Site C), Deep Stiff Soil (Site D), and Soft Clay (Site E) groups. To avoid the
complication of the influence of significant pore pressure generation from loose saturated
sand sites that might liquefy, loose to medium-dense saturated cohesionless soils (Site F)
were excluded from this analysis. The nonlinear stress-strain properties for each site
condition were selected to represent generic soils commonly found in each of these three
soil profiles (i.e. sites C, D, and E).
The shear wave velocity profiles selected are based on average shear wave
velocity profiles from a database compiled by Walter Silva (personal comm.). This
285
database consists of shear wave velocity profiles from a large number of sites, which are
located largely within California. These sites are classified according to the 1997
Uniform Building Code shear wave velocity-based classification system. The 1997 UBC
site categories C, D, and E correspond approximately to this study's site categories C, D
and E, respectively (see Chapter Three). The mean and standard deviation of the
“average profile” for each site condition are obtained by averaging shear wave velocities
at each depth for all the profiles within the same group. Therefore, the standard
deviations represent the spatial variability of shear wave velocity across a horizontal
plane. When the number of profiles involved is relatively large, the resulting average
profiles vary smoothly with depth. The number of profiles for each site condition is given
in Table 6.1. Results with equivalent linear analyses have shown that the average profile
captures the median response of these site conditions to a wide range of earthquake
ground motions. The use of average profiles allows for the evaluation of trends in the
response, as well as to quantify the average response of a particular soil class. However,
the use of average profiles precludes an estimate of the standard deviations associated
with ground motion values in the soil deposits. The following paragraphs describe in
detail the characteristics of each of the generalized site profiles.
Deep Stiff Soil (Site D)
Figure 6.1 shows the shear wave velocity profile selected for the Deep Stiff Soil
site. The average and the one standard deviation band shear wave velocity profiles for the
Site D category (180 m/s < Vs ≤ 360 m/s) from the Silva database are also shown. The
shear wave velocity in the upper 10 m of the profile was selected to match the average
286
profile of the Silva database. The variation of shear wave velocity with depth was
assumed to be proportional to σ'm(z)n, where σ'm is the mean effective stress at depth z.
The value of n was taken to be 0.25 (Hardin and Drnevich 1972). The shear wave
velocity profile obtained in this manner is approximately near or within one-half of the
standard deviation of the mean of the Silva profile. For larger depths, the value is lower
than the mean of the Silva profile because the latter includes also shear wave velocity
measurements of underlying rock deposits. The depth of the profile was varied from 30
to 200 m to study the effect of variations of depth on site response to velocity pulses. The
density of the soil is assumed to be 1.90 Mg/m3 for the upper 30 m and 1.95 Mg/m3 for
the deeper soil layers. The soil column is assumed to overlie an elastic half-space with a
shear wave velocity of 1200 m/s and a density of 2.4 Mg/m3. A linear transition, which
represent weathered rock, with a thickness of 3 m is placed between the soil and the
elastic half-space.
The selected nonlinear stress-strain properties of the soil in the profile correspond
to soils of low to medium plasticity. The soils were assumed to contain sufficient fines to
preclude significant pore pressure generation. The shear modulus reduction vs. strain and
equivalent viscous damping vs. strain curves were selected to lie within the range of both
the curves for sand from Seed and Idriss (1970), and the curves for clays with plasticity
indices (PI) of 15 and 30 from Vucetic and Dobry (1991). The shear modulus reduction
and damping curves within this range are appropriate for a relatively wide range of soil
types. Three sets of shear modulus reduction and damping curves were used. The sets
were selected to represent a generic low-plasticity soil, a clay with a PI of 15, and a clay
with a PI of 30. The parameters for the Borja and Amies model used in each of these sets
287
are listed in Table 6.2. The resulting shear modulus reduction and material damping
curves are shown in Figure 6.2. A large value of model parameter R was used to give a
better match to both the shear modulus reduction and damping curves while preventing
yielding of the soil at low strains. The implicit assumption is that the soil will not fail,
therefore the results must be checked to verify that the resulting stresses in the profile
remain within acceptable ranges and that the assumption holds. The strength profile of
the clayey soil was developed assuming an su/σv' ratio of 0.8, where su is the undrained
shear strength as measured by an unconfined triaxial compression test and σv' is the
vertical effective stress. The water table was arbitrarily assumed to be 10 m deep. A
value of su = 150 kPa corresponding to an unconfined compressive strength (qu) of 300
kPa was used as a lower bound for the soil strength. A qu value of 300 kPa corresponds
to a Very Stiff Soil and is consistent with the generalized stiff soil profile developed
herein, assuming that the water table at 10 m produces desiccation and overconsolidation
in the upper clay layers. The same shear strength profile was assumed to be applicable to
the Generic Low-Plasticity Soil. The static shear strength was multiplied by a factor of
1.4 (Lefebvre and Leboeuf 1987) to account for rate effects during rapid earthquake
loading. The resulting dynamic shear strength as a function of depth is shown in Figure
6.3.
It is important to note that the assumption that the soil will not fail due to the
shear induced by the ground motion is justified for the purpose of this study. Soil
yielding during earthquake ground motion can result in significant ground deformation,
particularly if the soil is subject to large static stress levels. However, soil yielding also
results in a large level of damping in the soil due to plastic dissipation of energy, reducing
288
the intensities of ground motion at the surface. Moreover, the input motions required to
develop yielding will vary largely depending on soil type and a generalization based on a
single value of soil shear strength would not be realistic.
Soft Clay Profile (Site E)
A large number of metropolitan areas in the world have clay deposits that are
characterized by high sensitivity, low stiffness, and low strengths, such as Boston Blue
Clay and San Francisco Bay Mud. Seismic site response issues associated with these
deposits could have a significant economic impact. The response of soft clay sites to
earthquake ground motions is highly site-dependent, which implies that the response of a
generic soft clay profile can only be used as an indication of possible trends in site
response for similar soil types.
Soil profiles from the San Francisco Bay with thick layers of Holocene Bay Mud
are used to develop a generic “soft clay” profile. These profiles typically consist of a
layer of artificial fill over Holocene Bay Mud overlying a thicker layer of Pleistocene Bay
Clay. A shear wave velocity of 110 m/s was selected for the Holocene Bay Mud. This
value is the average of a number of shear wave velocity measurements on normally
consolidated Holocene Bay Mud sites in the San Francisco Bay region. For the
Pleistocene Bay Clay, typical clay shear wave velocity values of 200 to 400 m/s were
used. The resulting generic shear wave velocity profile is shown in Figure 6.4.
Undrained shear strength values for the soft clay were obtained by using an su /σv'
value of 0.3. This value is typical of Bay Mud, which is a medium plasticity silty clay. A
minimum value of su = 25 kPa was used for soils near the surface of the profile. The
289
strength profile for the Pleistocene Bay Mud was obtained using an su /σv' value of 0.8.
The water table was assumed to be at the surface. As with the Stiff Soil profile, a
dynamic loading multiplicative factor of 1.4 was used to account for the increased rate of
loading experienced during seismic events. The resulting dynamic shear strength profile
is shown in Figure 6.5. The nonlinear stress-strain properties for the Holocene Bay Mud
were obtained by matching the shear modulus reduction and material damping vs. strain
curves with the curves presented by Sun et al. (1988). Two different sets of model
parameters were developed to conform to different assumptions regarding the behavior of
the soft clay under seismic loading. The first set of parameters was developed under the
assumption that stresses in the soil will not reach the yield strength of the soil. An
artificially high strength value was used to better match the shear modulus reduction and
damping curves at low strains for this "non-yielding soil" case. However, preliminary site
response analyses indicated that the maximum shear stresses induced in the soil deposit
undergoing high intensity (near-fault, forward-directivity) motions exceeded the dynamic
strength of the soft clay. For this reason, a second set of curves was developed to match
the soil strength and the shear modulus reduction and damping curves over a wider range
of strains. A parametric study showed that for the cases in which the soil yields, the
parameter that controls soil behavior is the yield strength of the soil. Moreover, the peak
ground velocity (PGV) at the surface is not very sensitive to relatively large variations in
the nonlinear small strain behavior of the soil. The value of the Borja and Amies (1994)
soil model parameter R was varied with respect to depth to match the value of su at each
depth, according to the strength profile shown in Figure 6.5. The variation of R with
respect to depth results in different shear modulus reduction and damping curves at
290
different depths. The resulting curves are shown in Figure 6.6. The model parameters are
listed in Table 6.3. For strain levels larger than 0.1 %, the equivalent viscous damping in
the model with variable strength is much higher than the damping expected in a soft clay
such as Holocene Bay Mud. However, when stresses in the soil reach the yield strength
of the material, energy dissipation is controlled by plastic yielding and large damping
levels can be expected.
Nonlinear model parameters for the Pleistocene Bay Clay were obtained by fitting
the shear modulus reduction and damping curves to the range determined by the Vucetic
and Dobry (1991) curves for clays with a PI between 15 and 30. The resulting curves are
shown in Figure 6.7 and the model parameters are listed in Table 6.3. The parameter R
was set to an artificially high value assuming that the soil would not reach failure stresses.
When this assumption did not hold, values of R consistent with the dynamic shear
strength profile shown in Figure 6.5 were used. Equivalent viscous damping values for
large strains are not realistic for values of R lower than 0.003. For these soils, however,
energy dissipation is controlled by plastic yielding. For low values of R, the shear
modulus reduction curve at low strains falls outside the range of the Vucetic and Dobry
curves for PI between 15 and 50. However, as indicated above, when the soil reaches
failure response is controlled by the yield strength of the soil and is relatively insensitive
to the shear modulus reduction curves at low strains.
Shallow, Very Stiff Soil Profile (Site C)
The average shear wave velocities obtained from the Silva database for profiles
classified by the 1997 UBC as Site C (360 m/s < Vs ≤ 760 m/s) were used to develop a
291
profile corresponding to a Shallow, Very Stiff Soil site category (Site C). The shear wave
velocity profile used is shown in Figure 6.8. The shear modulus reduction and material
damping curves selected are those for the Generic Low-Plasticity Soil profile used in the
Stiff Soil profile (Figure 6.2). The soil was assumed not to fail during earthquake
loading.
6.3 COMPARISON OF RESULTS FOR RECORDED AND
SIMPLIFIED NEAR-FAULT MOTIONS
6.3.1 General
The simplified velocity pulses developed in Chapter Four generally capture the
low frequency energy associated with the effects of forward-directivity conditions.
However, these simplified representations do not contain significant energy in the high
frequency range, which is also generally present in recorded ground motions. Although
local soil conditions can alter the underlying rock motions throughout a wide range of
frequencies, velocities tend not to be significantly affected by high frequencies. Hence, it
is appropriate for this examination of near-fault site effects to focus on the low frequency
energy represented by the simplified velocity pulses.
In the following section, the simplified velocity pulses are used to conduct a
parametric study of the influence of site response on this type of input motions. The
implicit assumption is that site response to the simplified pulses is similar to site response
to recorded near-fault ground motions. This assumption is verified in this section by
292
comparing site response analyses to recorded near-fault ground motions to the site
response to the corresponding simplified velocity pulses.
Four ground motions recorded on rock sites were selected for this study. These
motions are the Pacoima Dam record from the San Fernando earthquake, the Gilroy
Gavilan College record from the Loma Prieta earthquake, the Pacoima Dam Downstream
record from the Northridge earthquake, and the KJMA record from the Kobe earthquake.
A stiff soil profile (Figure 6.1) with a depth of 45 m with the Generic Low-Plasticity Soil
parameters (Figure 6.2, Table 6.2) was used for these site response analyses. Three
different sets of input motions were developed for each of the ground motions:
a) the full ground motion record,
b) the time interval of the record containing the fault normal velocity pulse, and
c) the simplified representation of the velocity pulse.
The period and the amplitude of the half-sine pulses of the simplified ground
motions are given in Table 4.6. A comparison of the frequency content for the input
ground motions for the three cases considered above (i.e. the full recorded ground
motion, the time interval of the recorded ground motion containing the velocity pulse, and
the simplified velocity time-history) is shown in Figure 6.9. With the exception of the
Pacoima Dam record for the San Fernando earthquake, the power spectral density of the
full time-history is similar to that of the recorded pulse alone. This implies that for these
records, most of the energy is concentrated in the velocity pulse. The Pacoima Dam
record, however, has a secondary pulse that arrives late in the time-history. The energy in
the high frequency range corresponds to this secondary pulse. The effect of secondary
293
pulses is addressed in a later section. A comparison of the frequency content of the
simplified pulse and the recorded pulse shows that the simplified pulse does not include
the higher frequency content of the motion
The input (rock) velocity time-histories and the resulting output (soil) velocity
time-histories at the top of the soil column are shown in Figures 6.10 to 6.13. A
comparison between the results for the recorded pulse and those for the simplified pulse
shows that, with the exception of the KJMA record, the output motion for the simplified
and the real pulses are similar, both in period and amplitude. This implies that the
exclusion of higher frequencies resulting from the use of simplified velocity time-
histories does not significantly affect output (soil) velocities. The same inference can be
made by observing the particle velocity-trace plots (Figure 6.14). The output particle
velocity-traces for the simplified velocity time-histories are remarkably similar to the
output for the recorded pulses. Results are also listed in Table 6.4. With the exception of
the KJMA record, the soil PGVs for the simplified input motion differ at most by 10%
from the soil PGVs calculated for the recorded time-histories.
The KJMA input motion is the only record for which the site response to the
simplified velocity pulse yielded results that were significantly different from the site
response to the recorded motion. The KJMA record indicates the potential problems
associated with determining pulse period from zero crossings. The simplified sine pulse
does not match well the recorded pulse (Figure 4.5); the recorded pulse could also have
been fit with a sine pulse with a lower period (Tv ≈ 0.9 s). As a result, the simplified
294
pulse is affected more significantly by the soil column, both in the amplification of PGV
and in the elongation of the pulse period.
The results shown in Figures 6.10 to 6.14 and in Table 6.4 can also be used to
evaluate the effect of the soil column properties on input near-fault ground motions.
While the input PGV of the recorded ground motion is amplified for the 1971 Pacoima
Dam, 1989 Gilroy, and 1994 Pacoima Dam Downstream records, it is de-amplified for
the 1995 KJMA record. Also significant is the elongation of the pulse period due to site
effects. A similar trend is observed with the pulse period, which is longer on the top of
the soil column for all the input motions except the KJMA. As indicated before, the
KJMA record could as well be represented by a shorter period, thus the apparent
shortening of the pulse period is likely due to the poor measure of input pulse period.
The PGV de-amplification might be due to strains induced in the soil in the first cycles of
motions. These strains results in a reduction of the profile stiffness. The revised input
pulse period (Tv = 0.9) is likely shorter than the degraded natural period of the profile,
consequently PGV de-amplification is observed. For the same reason, the longer-period
simplified input motion results in a PGV amplification. The concept of degraded natural
period and its effect on site response discussed at length in Section 6.4
The simplified velocity pulses do not always provide realistic representations of
the accelerations observed in near-fault ground motions. This discrepancy occurs because
accelerations have high frequency content and, as indicated before, the simplified pulses
do not contain high frequencies. In some cases the peak ground acceleration develops
later in the time-history and does not coincide with the forward-directivity velocity pulse,
295
as illustrated by the Pacoima Dam record for the San Fernando earthquake (Figure 6.15).
The simplified representation of the forward-directivity pulses should therefore not be
used to evaluate site effects on peak ground accelerations.
It is also important to point out that the sensitivity of the response to higher
frequencies is a function of the natural period of a site (i.e. of profile depth and stiffness).
Shallower, stiffer sites respond to higher frequencies and thus these sites might be more
affected by the high frequencies that are excluded from the simplified pulses. However,
forward-directivity velocity pulses occur at long periods; thus, the effect of shallow stiff
profiles on the pulse characteristics is minimal (see Section 6.4).
Figure 6.16 shows the maximum strains developed in the soil profile for the
recorded velocity pulses and the corresponding simplified velocity pulses. The maximum
strains observed in both cases are of the same order of magnitude. For some of the input
motions, the strains associated with the simplified pulses are slightly larger. One possible
explanation is that these larger strains are observed because the simplified pulses
primarily excite the lower frequency modes of the finite element mesh. Moreover,
recorded pulses also introduce energy into higher frequency modes. These higher modes
might lead to non-constructive interference resulting in lower strains. Induced strains are
important because they control the nonlinear response of the soil. Based on the observed
results, it is reasonable to conclude that the simplified pulses lead to reasonable estimates
of strains in the soil profile.
The results in this section can be summarized as follows:
296
a) The majority of the energy of pulse-like forward-directivity motions is
generally concentrated within a narrow low-frequency band.
b) Simplified velocity pulses can be used to evaluate the effects of local site
conditions on the amplitude and period.
c) Simplified velocity pulses do not consistently lead to good approximations of
peak ground accelerations. Hence, these simplified pulses should not be used
to evaluate the effects of site response on accelerations.
6.4 SITE RESPONSE TO SIMPLIFIED PULSES
6.4.1 General
Results presented in Section 6.3 indicate that using simplified representations of
input rock velocity pulses can capture salient aspects of seismic site response to near-fault
ground motions containing forward-directivity. Simplified pulses that represent the
existing ground motion database of both rock and soil motions were presented in Chapter
Four (Section 4.3.5). The simplified velocity pulses identified as Sets 6, 7, and 8 (Figure
4.35) are used as input motions in the study of site response for the generalized profiles
presented in Section 6.2.
The use of simplified representations facilitates a systematic study of the influence
of pulse parameters on site response. The pulse parameters and the site conditions are
varied systematically to study the effect of site response on the characteristics of the
velocity pulse. Table 6.5 gives the different combinations of pulse parameters and site
conditions used in the site response analyses. The pulse parameters were chosen to
297
represent typical values of rock pulses as described in Chapter Four. The results are
presented first for the Stiff Soil profiles (Sites C and D) and then for the Soft Soil profile
(Site E).
Before discussing the results of the nonlinear analyses, it is instructive to analyze
the response of a uniform, linear soil column to the simplified pulses. Input motion Set 6
is applied to an elastic soil deposit with a uniform shear wave velocity of 120 m/s and a
depth of 30 m (natural period: Tn = 4H/Vs = 1.0 s). A target damping ratio of 5%
(Rayleigh Damping) is used to simulate the hysteretic damping in soil. The site response
analysis is carried out using the finite element program GeoFEAP with linear-elastic brick
elements. Figure 6.17a presents the calculated PGV at the top of the soil column as a
function of input velocity pulse period. This type of plot is sometimes referred to as a
"shock spectrum" (Chopra 1995). The maximum amplification of PGV occurs at an input
period of 1.1 s. This period is only slightly longer than the undamped natural period of
the site. Figure 6.17b shows the ratio of the pulse period on the top of the soil column to
the period of the input motion. Observe that when the input pulse period is less than the
undamped natural period of the site, the pulse period at the top of the soil column is
longer than the input pulse period.
In frequency domain, equivalent-linear analysis, the equation of motion is solved
for an equivalent strain level that is constant over time for each soil layer. The constant
strain level allows the definition of a "degraded site period", which is the natural period
corresponding to the equivalent strain level used in the analysis. This degraded site
period corresponds to the first resonant frequency of the site. The amplification of input
298
ground motions is a maximum at the first natural frequency and peaks at the subsequent
higher order frequencies of the site (Figure 6.18). In nonlinear analysis, the concept of a
constant site period over time is ill-defined, because the tangent shear modulus (and
hence the shear wave velocity) changes as strain levels in the soil change during seismic
loading. The concept of degraded site period, however, can be used as a conceptual tool
to help understand the response of a site.
An increase in the intensity of the input motion results in larger strains and hence
a decrease in soil stiffness with a consequent increase of the degraded site period. The
decrease in soil stiffness leads to a larger impedance contrast between the rock, which
behaves almost linearly, and the soil. This larger impedance contrast leads to larger
amplifications of the input motion (Figure 6.18a). An increase in strain levels in the soil,
however, also results in an increase in hysteretic damping levels. Larger damping levels
damping can lead to lower velocities (Figure 6.18b). The amplification or de-
amplification of motion results from a balance of the effects of a change in natural period,
the effects of increased damping, and the effects of the larger impedance contrast. For
non-forward-directivity motions, the effect of increased damping at large input motion
intensities is significant, particularly at high frequencies. On the other hand, damping is
not as significant when the input motion is pulse-like (Chopra 1995), such as is the case
with forward-directivity motions. Moreover, damping has less of an effect when the
input motion is a long period motion. Therefore, site response to pulse-type input
motions is most likely controlled by the larger impedance contrast and the shift in the
natural period of the site that has degraded dynamic stiffness due to its response to intense
near-fault input motions.

299
6.4.2 Stiff Soil Profiles (Sites C and D)
General
The shear wave velocity profile and the nonlinear properties of the Deep Stiff Soil
profile (Site D) and the Shallow, Very Stiff Soil profile (Site C) were presented in Section
6.2. Three different sets of model parameters corresponding to different shear modulus
reduction and damping curves were also presented in Section 6.2. This section presents
the results of site response analyses using the shear modulus reduction and damping
curves for the soil termed "Generic Low-Plasticity Soil." The influence of the input pulse
parameters is first evaluated. The effects of profile stiffness are then evaluated by
comparing results of the 30 m deep Stiff Soil profile to those of the Very Stiff profile
(Site C). The effect of varying depth to bedrock for the Stiff Soil profiles is then
evaluated. Finally, the influence of the nonlinear stress-strain properties of the soil is
evaluated by comparing the analysis results for the Generic Low-Plasticity Soil to those
for the clays with PI of 15 and 30.
Influence of input pulse period and input PGV
The particle velocity plots at the top of the soil column for the Deep Stiff Soil
profile with a depth of 60 m are given in Figures 6.19 to 6.21 for three different input
motion sets. The influence of both the input pulse period (Tv) and the intensity of the
input motion (PGV) on the resulting velocity plots is evident. It is apparent that the
effects of input pulse period and PGV are interdependent. For example, for an input
pulse period of 1.0 s, velocity amplification decreases as the input PGV increases.
300
However, for an input pulse period of 4.0 s, velocity amplification increases with
increasing input PGV. Figure 6.22a presents the ratio of PGV in the soil to PGV in the
rock plotted as a function of input pulse period. PGV soil to rock ratios for this profile
range from 1.0 to 2.0 for the cases considered. By analogy with the results for a linear-
elastic soil profile (Figure 6.17), the degraded (strain-compatible) period of a site can be
identified with the period at which the largest PGV amplification occurs. The period at
which maximum PGV amplification is observed increases with increasing input motion
intensity for all three input motion sets. This is due to the larger strains induced by the
more intense motions. The shift in degraded site period with increasing intensity is well
illustrated by the PGV soil to rock ratios for the 30 m deep Stiff Soil profile shown in
Figure 6.23.
The input pulse period and input PGV also influence the period of the resultant
velocity pulse. The ratio of the resultant (soil) pulse period to the input (rock) pulse
period is plotted in Figure 6.21b as a function of input pulse period. The pulse period in
the soil is larger than the pulse period in the rock for low input pulse period values. The
same trends were observed in the empirical database of near-fault ground motions.
Influence of soil stiffness
The influence of soil stiffness on site response is evaluated by comparing the site
response results for the 30 m deep Stiff Soil profile (natural site period of 0.44 s) with the
results for the 30 m deep Very Stiff Soil profile (natural site period of 0.22 s). Both these
profiles have the same depth to bedrock and the same nonlinear stress-strain soil
properties, but they differ in shear-wave velocity profiles (Figures 6.1 and 6.8). Figure
301
6.24a compares the amplification of PGV in the soil column for these two soil profiles.
For the Very Stiff Soil profile, maximum amplification of PGV for the cases considered
occurs at an input pulse period of 0.6 s, while essentially no PGV amplification is
observed for periods longer than 2.0 s. Conversely, the Stiff Soil profile displays
maximum PGV amplification for input pulse periods ranging from 1.5 to 2.5 s, reflecting
the longer natural site period. For periods in this range, PGV amplifications are larger
than the amplifications for the Stiff Soil profile. Amplification of PGV in the Very Stiff
Soil profile is nearly independent of the input PGV. For this profile, variations in the
input PGV affect the resulting PGV soil to rock ratio. This difference in response is due
to the different strain levels induced in the profile. At the larger strains in the Stiff Soil
profile, the shear modulus reduction curve is more sensitive to changes in strain.
The effect of these two site profiles on the period of the velocity pulse is
illustrated in Figure 6.24b. The stiffer profile (Very Stiff Soil profile) has very little
influence on the pulse period, except for the case of an input pulse period of 0.6 s.
Conversely, the softer profile (Stiff Soil profile) causes significant elongation of the input
pulse period for all periods lower than 2 seconds.
Influence of soil profile depth
Figure 6.25 shows the soil to rock PGV ratios for all the Stiff Soil Profiles for
input motion Set 6. The depth to bedrock of these profiles varies from 30 m to 200 m.
The ratio of peak ground velocity in the soil (output) to the rock (input) varies from 0.6 to
2.3 depending on the input pulse period and the profile depth. For the conditions
considered in this study, the input pulse period for which maximum amplification occurs
302
increases with profile depth. This is a direct consequence of the larger site periods for
deeper profiles. Also, observe that in general, PGV amplification decreases with
increasing input motion intensity for input motions with short pulse periods. The more
intense motions result in a degraded site period that is much longer than the input pulse
period, hence resulting in lower amplifications. Conversely, for input motions with long
periods, the longer degraded site periods resulting from more intense ground motions are
closer to the input pulse period, thus resulting in PGV amplifications.
Figure 6.26 shows the ratio of pulse period on soil to pulse period on rock. As the
depth of the profile increases (and consequently the natural period of the sites increase),
the soil pulse period increases. In general, when the input motion is shorter than the
degraded period of the site, the soil has a tendency to oscillate at its degraded period,
resulting in an elongation of the pulse. For longer site periods (e.g., deeper profiles), the
resultant pulse period is consequently longer. However, if the input pulse period is much
longer than the degraded site period, the wavelength of the input motion is much larger
than the profile depth and the soil column moves more or less as a rigid body.
Influence of shear modulus reduction and damping curves
Three different soil types are used in the site response analyses to the same set of
input motions. These soils are the Generic Low-Plasticity Soil used in the previous
analyses and two soils whose nonlinear stress-strain behavior is "matched" to the Vucetic
and Dobry (1991) curves for PI of 15 and 30 (see Section 6.2). Figure 6.27 presents the
resulting ratios of PGV in the soil to PGV in the rock for the three soil types. For any
given input soil period and velocity, the PGV soil to rock ratios vary with soil type. For
303
an input pulse period of 0.6 s, the PGV soil to rock ratios are larger for soils with larger PI
(i.e. PI = 30 soil > PI = 15 soil > low-plasticity soil). This reflects the fact that if two
soils with different PI have the same elastic shear modulus, for a given strain level the
soils with larger plasticity have higher stiffness. The higher stiffness implies that the soil
profile has a lower degraded site period and amplifies more shorter input pulse periods.
Conversely, at longer input pulse periods, the PGV of the low-plasticity soil is larger than
the PGV in the other two soils. These results highlight the importance of a good
characterization of shear modulus reduction and damping curves in site response
analyses. The effects of different shear modulus degradation curves are important
especially for low input motion intensities. For larger intensities, high strains are induced
and degraded site periods are closer for the three profiles.
The stress-strain properties of the soils also affect the pulse period of the soil
(Figure 6.28). Velocity pulse periods are larger for the Generic Low-Plasticity Soil. This
reflects the higher large-strain stiffness of the soils with higher plasticity (Figure 6.2).
The variation in pulse periods with site condition, however, is not very significant.
Influence of fault parallel motion and pulse shape
In the following paragraphs, the effects of the fault parallel input motion and the
pulse shape on site response are examined. The effects of varying the input fault parallel
to fault normal velocity ratio (PGVp/n) is first considered. The sensitivity to pulse shape,
as defined in Chapter Four, is then considered. Figure 6.29 shows the PGV soil to rock
ratios in the 45 m deep Stiff Soil for two input motion sets and for varying values of the
ratio of input fault parallel to fault normal velocity (PGVp/n). The amplification of PGV is
304
consistently higher for input pulses with higher PGVp/n ratios. The larger fault parallel
input motion leads to increased strain levels in the soil. The larger strains induce a
reduction in the shear modulus which in turn increases the impedance contrast and shifts
the degraded site period towards the long period energy contained in these near-fault
forward-directivity input motions. These changes lead to larger amplifications of
velocities for most of the cases studied herein. The influence of the fault parallel motion
on the amplification of fault normal velocities is also a function of the frequency content
of the fault parallel motion. When this frequency content is such that it leads to larger
amplifications, the strains developed in the profile are consequently larger and the
influence on fault normal velocities increases. For a PGV of 75 cm/s, it is reasonable to
use a fault parallel PGV of up to 75% of the fault normal value. In this case, the largest
fault normal amplification is 1.8 for Set 7. For comparison, the largest amplification for
the same input motion set but a fault parallel PGV of 25% of the fault normal value is 1.5,
and if the fault parallel component is ignored in analysis, the largest calculated
amplification is only 1.4. This increase is important and justifies the use of both fault
normal and fault parallel input motions, even for cases in which the fault normal motion
is significantly larger than the fault parallel motion.
The PGVp/n ratios also affect period elongation when the input pulse period is
lower than the degraded site period. The larger strains for the higher PGVp/n ratios imply
larger degraded periods, thus the pulse periods on the top of the soil column are longer for
higher PGVp/n. This effect decreases as the period of the input motion approaches the site
period. This is illustrated for the site with profile depth of 45 m in Figure 6.30.
305
Figure 6.31 compares the resulting PGV ratios for three different shapes of the
input pulses for all the Stiff Soil profiles. Recall that “pulse shape” was defined as the
shape of the particle velocity-trace plot of the pulse (Chapter Four). Maximum PGV
amplification occurs for Set 6. This shape is also the pulse shape that leads to largest
strains. In general, larger values of PGV are associated with larger shear strain levels in
the soil. This is not necessarily the case when the input motion is not pulse-like. For
non-pulse type motions, the damping associated with large strain levels has a more
significant effect in attenuating the ground motions because of the increased number of
loading cycles in the soil. This is accentuated by the fact that non-forward-directivity
motions often contain significant energy at higher frequencies, which are more affected
by damping.
Observations on resulting stresses and strains
The soil model used for the Site D profile (Section 6.2) included the assumption
that the soil does not reach the failure stresses defined by Figure 6.3. The soil shear
strength was never exceeded by the calculated stresses even for the input motion Set 6,
which yields the maximum stresses. Figure 6.32 shows the maximum strains calculated
for each depth in the profile for a number of runs on the 45 m deep Stiff Soil profile. The
strains shown are the vectorial sum of the fault normal and fault parallel strains. For an
input PGV of 75 cm/s, shear strains induced in the soil profile are largest when the input
period is 1 s to 2 s (Figure 6.32a). Note that for a given PGV, the lower the pulse period,
the larger the acceleration; thus a motion with a higher acceleration (i.e. the one with Tv =
0.6) develops a lower level of strains. On the other hand, for motions with equal
306
accelerations, strains are larger for larger velocities (Figure 6.32b). In general, the strain
level calculated in the profile is generally associated with the amount of PGV
amplification and consequently with the site period.
Figure 6.33 shows the maximum shear strains developed in all the Stiff Soil
profiles for a given input motion. This figure also includes the Shallow, Very Stiff Soil
profile (Site C) in addition to the Stiff Soil profiles. For the shorter input pulse periods,
larger shear strains occur in the shallow profiles. As the period of the input motion
increases, larger strains are observed for progressively deeper profiles. This is consistent
with the observations presented throughout this chapter.
6.4.3 Soft clay profile
The particle velocity plots for input motion Set 6 for the Soft Clay profile (Site E)
are shown in Figures 6.34 to 6.36. The magnitude of the ground deformation at the top of
the soft clay profile is a function of both the period and the intensity of the input rock
velocity pulse. In general, the effects of input pulse period and input PGV on the
amplifications of velocities are interdependent. The effect of the intensity of the input
motion is more pronounced for the Soft Clay profile than for the Stiff Soil profiles. A
significant consideration is that the amplification of velocities can sometimes also result
in a change of the input pulse from a full cycle to a half cycle of motion for short input
pulse periods.
The shear stresses and strains induced in the soil profile are shown in Figure 6.37.
In the soft clay (elevation 3 to 27 m), the shear stresses are bounded by the strength of the
soft clay, i.e., in all the cases studied the soft clay reached failure. The occurrence of
307
failure explains the large strains in the soft clay. For large intensity motions (input PGV
= 300 cm/s), failure is also reached in the underlying Pleistocene Clay deposit. As
indicated previously, a PGV of 300 cm/s represents an upper bound for predicted
velocities at rock sites. It is interesting to observe that in the soft clay layer, larger strains
are observed for input motions with longer input pulse periods; while in the Pleistocene
Clay layer, strains are larger for motions with intermediate input pulse periods. A
parametric study on the nonlinear stress-strain properties of the Soft Clay indicated that
the failure stress is the parameter that controls response in the Soft Clay deposit. In fact,
large variations in the shear modulus reduction and damping curves in the low strain
region did not affect either the predicted peak ground velocities at the top of the soil
column, or the resultant period of the velocity pulse. It is important to realize that
equivalent linear programs, such as SHAKE91, do not account for soil failure and thus
cannot be expected to correctly predict site response for the Soft Clay profiles for intense
near-fault ground motions.
A number of studies from paired stations in rock and soft clay have shown the
tendency of soft clays to attenuate high frequency motions and to amplify long period
motions. In particular, soft clays have a strong tendency to attenuate accelerations when
these are larger than about 0.4 g (Idriss 1991, Figure 2.1). Figure 6.38 shows the
maximum displacement, acceleration, and velocity values observed in the soft clay profile
for an input PGV of 160 cm/s. For the input pulse periods shorter than 4 s, there is a
decrease in maximum accelerations in the soft clay layer (3 – 20 m depth) with respect to
the underlying clay. Velocities are more sensitive to longer period energy. A decrease in
308
velocity with respect to the underlying clays is observed in the soft clay layer when the
input period is 0.6 s, but input periods of 2.0 s or more result in slightly larger velocities.
Figure 6.39 compares the PGV ratios (soil/rock) for the Soft Clay profile and a
Stiff Soil profile of equal depth (60 m). The level of amplification in both profiles is
approximately equal for a PGV of 75 cm/s. The soft clay profile would be expected to
have larger velocities because of the higher impedance contrast with the underlying
bedrock. However, yielding of the soft clay deposits results in a degraded site period that
is much larger than the period of the input motions, reducing the expected amplification.
Moreover, yielding in the soil introduces large levels of damping that attenuate the input
pulse. This effect is intensified for larger input motions (PGV = 160 cm/s and PGV = 300
cm/s). For the larger motions, the velocities in the soft clay are lower than in the stiff
soil. Figure 6.40 compares the amplification of PGV in the Stiff Soil with the
amplifications in the Soft Clay profile for all the input motion sets. For the Deep Stiff
Soil, the PGVs in the soil are generally larger than in the rock. Conversely, the Soft Clay
(Site E) de-amplifies input PGVs which are higher than a crossover value. This crossover
value is a function of the input period. The elongation of the input pulse period is slightly
larger in the soft clay site than in a stiff soil deposit of the same depth (Figure 6.41). This
occurs because the degraded period of the soft clay site is larger than that of the stiff soil
deposit.
309
6.4.4 General Observations
Pulse Shape
The results presented so far have dealt primarily with the fault normal component
of motion. This component typically controls the response of structures to forward-
directivity motions. However, it is also important to consider the effects of site
amplification on the fault parallel component of motion. This section presents some
findings related to the change of pulse shape due to site effects.
If the frequency content of the two perpendicular directions of motion is different,
site response will affect each direction in different proportions, altering the ratio of fault
parallel to fault normal velocity. Typical values of fault parallel to fault normal PGVs in
the entire database of forward-directivity motions are shown in Figure 4.26. The input
motions used in the analyses are defined by input PGVp/n ratios ranging from 0.0 to 0.75.
The variation in PGVp/n ratios due to site effects is shown in Figure 6.42. For the
Shallow, Very Stiff Soil (Site C), the PGVp/n ratios of the output motions are independent
of input pulse period and its values are equal to the PGVp/n ratios of the input motions.
For the Deep Stiff Soil with 60 m depth (Site D) and the Soft Clay (Site E) sites, PGVp/n
ratios tend to increase with increasing input period for the range of values considered in
this study. The changes in PGVp/n ratio reflect the varying frequency content of the fault
parallel motions. The fault parallel pulses in the simplified input pulse Set 7 have shorter
pulse periods. Therefore, when the fault normal period is longer than the degraded site
period the fault parallel pulse period matches the degraded site period, resulting in a
larger amplification of PGV in the fault parallel direction. Note that in nonlinear analysis,
310
the fault normal and fault parallel components cannot be separated. In fact, large
amplifications of fault parallel motions (resulting in a decrease of PGVp/n ratios) are not
observed when site response analyses are performed with only the fault parallel
component of motion. The amplifications of fault parallel motion are associated with the
shear modulus reduction induced by large fault normal velocities. In recorded near-fault
ground motions, the period of the fault parallel pulse is generally shorter than the period
of the fault normal pulse (see Chapter Four); thus, the results obtained herein lie within
the bounds of the empirical data. If the initial PGVp/n ratio is high, as some of the low
intensity motions in the database indicate, site amplification can lead to PGVp/n ratios
larger than one. This is an important observation with respect to seismic demand
estimation for structural analysis in the near-fault region. High PGVp/n ratios are also
observed in some of the recordings in the strong motion database.
Number of Cycles of Motion
In the dynamic analysis of nonlinear structures, the number of cycles of motion is
an important input parameter. Site response can affect the number of cycles of motion in
two ways:
a) Energy is trapped in the softer soil layers. This case applies in particular to sites with
a large impedance contrast. At a large scale, this phenomenon causes the basin effects
observed in a number of earthquakes (e.g., Aki and Larner 1970, Bard and Bouchon
1985). At individual sites, the trapping of energy in softer soil layers is responsible
for the longer duration of strong shaking in soil sites than in rock sites (Abrahamson
and Silva 1997). The results presented herein do not indicate that there are velocities
311
of significant magnitude beyond the initial velocity pulses. With pulse-type motions,
the initial pulse of motion typically has the largest amplitude. The amplitude of the
ground motion resulting from the trapping of vertically-propagating waves in the soil
profile is generally lower than the amplitude of the initial pulse.
b) When the period of the input pulse is significantly shorter than the natural period of
the site, input motions with a full cycle of motion tend to be one-sided pulses. This
occurs because the soil elongates as a result of the initial pulse and does not capture
the second half of the pulse (Figures 6.18 to 6.20 for Tv = 0.6, and Figures 6.34 to
6.36 for Tv ≤ 1 s). This effect has to be considered when evaluating a database for
estimating the number of cycles of near-fault ground motions.
Consecutive Pulses
The analyses presented in this chapter were conducted using input pulses with a
full cycle of motion in the fault normal direction. Although this condition characterizes a
large number of records in the empirical database, some near-fault recordings, such as the
Gilroy records for the Loma Prieta earthquake and the KJMA record for the Kobe
earthquake, have two consecutive full cycles of pulse-like motion. The number of cycles
is generally related to the slip distribution in the causative fault. On occasions in which
the slip distribution is uneven, the pulses of motion for near-fault, forward-directivity
motion may arrive at separate times. To account for both of these factors, additional
analyses were performed with input motions consisting of two consecutive cycles of
motion at different offsets. Peak amplification normally occurs in the first full cycle of
motion. This implies that the results presented in this chapter in terms of the effects on
312
peak values of velocity are also valid for ground motions with more than two consecutive
cycles, as long as the largest velocity corresponds to the first cycle of motion. However,
near-fault motions containing consecutive pulses (i.e. extended durations) will
undoubtedly affect the nonlinear response of structures that have yielded.
A further look at pulse shape
The shape of the pulse in a particle-velocity plot is also affected by the delay
between the fault normal and fault parallel pulses. The effects of changing the delay
between the fault normal and fault parallel pulses on site response were explored. A fault
normal pulse with an amplitude of 160 cm/s and a full cycle of motion with a period of
2.0 s was used as input motion for site response analyses of the 60 m deep Stiff Soil
deposit. The amplitude of the fault parallel motion was assumed to have a value of 66%
of the fault normal amplitude, with a full cycle of motion and a period of 75% of the fault
normal period. The offset between the fault normal and fault parallel pulses was varied
between a range of two pulse periods. The maximum difference in fault normal PGVs in
the soil is about 13%. Maximum amplifications occur when the fault parallel pulse leads
the fault normal pulse. Considering the existing level of uncertainty in the prediction of
PGVs, the effect of the offset is not considered significant at this time. Similar
conclusions apply to the velocity pulse period.
313
6.5 FINDINGS
General findings
This chapter presented the results of ground motion analyses to near-fault ground
motions using simplified velocity pulses. The use of simplified representations of pulse
motions was validated by comparing the results of site response analyses to recorded
ground motions with the results of site response analyses to simplifications of the
recorded ground motions.
The surface PGV values obtained from the site response analyses of the Stiff Soil
deposits using simplified velocity pulses are plotted against the input PGVs in Figure
6.43. Ratios of soil to rock PGV generally lie between 1.0 and 2.0. The findings related
to PGV amplification in Stiff Soil can be summarized as follows:
• PGV amplification is a function of both the period and the intensity of the input
pulse (Figure 6.44). The period for which maximum amplification of PGV
occurs can be associated with the degraded natural period of the site (Figure
6.23).
• The degraded natural period of a site increases with increasing profile depth, it
decreases with increasing profile stiffness, and it increases with increasing input
motion intensity. Consequently, sites with long degraded site periods (i.e. Stiff
Soil sites deeper than 60 m, Soft Clay sites, or Shallow Stiff Clay sites subject to
intense shaking) tend to amplify input pulses with periods longer than 2.0 s.
Conversely, sites with short degraded site periods (i.e. Shallow, Very Stiff Soil
314
sites, or sites subject to low intensity input motions) tend to amplify input pulses
with short periods (≤ 1 s) and are not affected by long pulse periods.
• The nonlinear stress-strain properties of the soil have an important effect on the
PGV amplification. For a given strain level, soils that exhibit lesser shear
modulus reduction tend to amplify motions with shorter input pulse periods
more than soils with larger shear modulus reduction.
• Amplification of input pulses with long periods is observed for input PGVs as
large as 300 cm/s when the input pulse period is longer than 1 s.
• The mid-period amplification factor for Site D in the 1997 UBC (180 m/s < Vs
≤ 360 m/s) in Zone 4 is 1.6. This factor agrees with the range of Stiff Soil PGV
amplification obtained in the site response analyses for input pulse amplitudes
lower than 300 cm/s (Figure 6.43). Mid-period amplification factors can be
compared to velocity amplification factors, because mid-period spectral values
are generally controlled by velocities.
• PGV amplification for recorded ground motions (Section 6.3) is comparable to
PGV amplifications obtained from the simplified pulses (Figure 6.43).
In addition, site response also affects the period of the velocity pulse. Figure 6.45
illustrates the ratios of the calculated velocity pulse period to the input pulse periods.
Figure 6.45 also includes the ratio of the median regression curves for pulse period in soil
and rock obtained from the database of near-fault ground motions (Chapter Four). The
trends in the results from the numerical simulations agree well with the regression line
315
developed using empirical data. Some observations relating to the influence of site
response on the velocity pulse period include:
• When the input pulse period is shorter than the degraded natural period of the
site, the soil column tends to elongate the velocity pulse, because of its tendency
to oscillate at its natural site period. This results in a significant velocity pulse
period increase for short input pulse periods.
• Input pulse periods that are significantly longer than the degraded site period
have much longer wavelengths than the profile depth. Consequently, the soil
column moves more or less as a rigid body, and the period (i.e. and amplitude)
of the input pulse is not affected significantly by the response of the soil
column.
• For the recorded near-fault ground motions presented in Section 6.3, the
magnitudes of the calculated pulse periods lie in the same range as the
calculated pulse periods for simplified velocity pulses with comparable
amplitudes (Figure 6.46).
Figure 6.40 showed a summary of the results of the analyses for the Soft Clay
profile to a number of the simplified velocity pulses. Contrary to the results obtained for
the Stiff Soil profiles, PGV amplifications in the Soft Clay profile diminish as the
intensity of the input motion increases. Some important findings relating to the response
of soft clay deposits to the simplified velocity pulses are:
• The Soft clay profile used in this study (static su = 25 kPa) reaches its yield
strength even for input velocity pulse amplitudes as small as 75 cm/s. This
316
value is well within the expected range of velocities for near-fault ground
motions (Chapter Four).
• The response of the Soft Clay profile is controlled by the value of the yield
strength of the soil. The analyses indicated that soft soils are expected to yield
at input velocities typical of near-fault ground motions. Thus, the evaluation of
the dynamic shear strength of soft clays is important for the proper evaluation of
seismic site response to near-fault motions. The mid-period amplification factor
for Site E ( Vs ≤ 180 m/s) in the 1997 UBC in Zone 4 is 2.4, a value larger than
the estimated PGV amplification factors from the site response analyses. The
large code factor is likely due to extrapolation from equivalent linear analyses
that do not account for soil failure.
Limitations
As indicated in Chapter Five, the finite element implementation of site response
used in this study assumes that horizontal shear waves propagate vertically. This
assumption is commonly made in most site response analyses, and it applies to sites away
from the fault. For these sites, non-vertically travelling waves are refracted as they
propagate from the source to progressively softer soils, resulting in the observed
vertically-propagating waves. However, the short travel paths to near-fault sites could
prevent a complete reorientation of the wave front causing some of the energy to be
originated from non-vertically-propagating waves. In addition, constructive interference
with long-period surface waves originating from basin effects might contribute to a large
portion of the long-period energy of recorded motions, particularly near basin edges.
317
The evaluation of site response effects on velocity pulses was achieved by
comparing the recorded soil motions to the simplified velocity pulses used as input
motions in the site response analyses. These simplified velocity pulses are assumed to be
outcropping rock motions. This assumption implies that the simplified pulses already
capture the effects of the geological structures underlying the soil profile. In general,
site response affects the input ground motions only at periods shorter than about 1.5 to 2
times the degraded site period. Input motions with longer periods have wavelengths that
are much larger than the profile depth and thus are unaffected by the profile. However,
these long period motions are affected by the deeper velocity structure underlying the site.
As mentioned, it is assumed that the outcropping velocity pulses already capture these
effects.
318
Table 6.1. Number of profiles for each Site Type (based on UBC classification scheme,
Table A3) in the database of shear wave velocity profiles of Silva (personal comm.).
Site Type Number of Profiles

B 35
C 227
D 343
E 42
319
Table 6.2. Model parameters used for the soils considered in the Stiff Soil profiles. The
shear-wave velocity profile is given in Figure 6.1.
Generic Low-
Parameter Clay, PI = 15 Clay, PI = 30
Plasticity Soil
h/Gmax 0.032 0.126 0.126
m 1.4 1.16 1.16
R/Gmax .005 .005 .01
Ho 1e-6 1e-6 1e-6
ξ 1% 1% 1%
ρ=(Mg/m ) 3
1.9 – 1.95 1.9 - 1.95 1.9 - 1.95
ν 0.49* 0.49* 0.49*
*
320
Table 6.3. Model parameters used for the soils considered in the Soft Clay profiles. The
shear-wave velocity and density profile is given in Figure 6.4.
Holocene Bay Mud

Holocene Bay Mud Variable Strength Clay, PI = 30
Parameter
Low Strain curve Curve Variable Strength Curve
h/Gmax 0.218 1.7 0.8

m 0.97 1.1 0.8
R/Gmax 0.0.014 8 3s u /Gmax** 8 3s u /Gmax**
Ho 1e-6 1e-6 1e-6
ξ 1% 1% 1%
ν 0.49 0.49 0.49*
*
**
su profile given in Figure 6.5.
321
Table 6.4. Input motions and results of site response analyses on recorded near-fault
ground motions (PCD = Pacoima Dam , San Fernando earthquake; GIL = Gilroy Gavilan
College, Loma Prieta earthquake; PAC= Pacoima Dam Downstream, Northridge
earthquake; KJM = KJMA, Kobe earthquake).
(a) Input Motions

Fault Normal Fault Parallel
PGA (g) PGV (cm/s) Tv (s) PGA (g) PGV (cm/s) Tv (s)
Full Recorded
1.47 114 1.44 0.78 38.0 1.47
Ground Motion
Recorded
PCD 0.78 113 1.44 0.36 39.7 1.47
Velocity Pulse
Simplified
0.51 114 1.44 0.22 36.2 1.47
Velocity Pulse
Full Recorded
0.29 30.8 1.16 0.41 26.6 0.82
Ground Motion
Recorded
GIL 0.29 31.6 1.16 0.41 26.3 0.82
Velocity Pulse
Simplified
0.25 31.1 1.16 0.33 26.7 0.82
Velocity Pulse
Full Recorded
0.48 49.9 0.61 0.31 23.1 1.1
Ground Motion
Recorded
PAC 0.48 52.7 0.61 0.31 23.5 1.1
Velocity Pulse
Simplified
0.52 51.4 0.61 0.20 22.4 1.1
Velocity Pulse
Full Recorded
0.85 95.7 1.91 0.55 53.3 0.73
Ground Motion
Recorded
KJM 0.85 96.8 1.91 0.55 53.5 0.73
Velocity Pulse
Simplified
0.68 96.5 1.91 0.47 53.2 0.73
Velocity Pulse
(b) Calculated soil response

PGA (g) PGV (cm/s) Tv (s) PGA (g) PGV (cm/s) Tv (s)
Full Recorded
0.47 159 1.78 0.36 65.8 1.60
Ground Motion
Recorded
PCD 0.48 166 1.80 0.34 63.5 1.66
Velocity Pulse
Simplified
0.48 164 1.84 0.45 87.2 1.72
Velocity Pulse
Full Recorded
0.26 39.5 1.49 0.36 37.3 1.70
Ground Motion
Recorded
GIL 0.33 41.4 1.52 0.39 38.0 1.70
Velocity Pulse
Simplified
0.28 45.5 1.56 0.27 38.8 1.68
Velocity Pulse
Full Recorded
0.31 51.6 1.04 0.30 25.9 1.06
Ground Motion
Recorded
PAC 0.31 55.3 1.08 0.32 26.0 1.00
Velocity Pulse
Simplified
0.33 61.8 1.06 0.36 35.8 1.02
Velocity Pulse
Full Recorded
0.50 79.0 1.68 0.52 79.6 1.24
Ground Motion
Recorded
KJM 0.50 77.7 1.40 0.54 77.6 1.24
Velocity Pulse
Simplified
0.47 136 1.92 0.64 73.9 1.56
Velocity Pulse
322
Table 6.5. Values of parameters used in the parametric study of site response to
simplified pulse motions.
Parameter Values
Tv,N 0.6, 1.0, 2.0, 4.0
Fault Normal PGV (cm/s) 75, 160, 300
PGVp/n 0, 0.25, 0.5, 0.75

Input Motions Sets (Pulse
Set 6, Set 7, and Set 8 (Figure 4.35)
Shapes)
Shallow Very Stiff Soil, Deep Stiff Soil (varying profile depth),
Vs profile
and Soft Clay (Section 6.2)
Depth to bedrock For Stiff Soil profile: 30 m, 45 m, 60 m, 100 m, 200 m.
323
Vs (m/s)
0 200 400 600 800 1000
0
20
40
60
80
Depth(m)
100
120
140
160
180
200
Selected Profile
Median and +_ one standard deviation from 343 profiles
Figure 6.1. Selected shear wave velocity profile for the Stiff Soil profile. Included in the
graph are the median and ± one standard deviation for a database of 343 profiles (Silva,
personal comm.) classified as Site D by the Uniform Building Code (e.g. 180 m/s < Vs ≤
360 m/s).
324
1
Modulus Reduction 0.8
0.6
0.4
0.2
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
30
Damping (%)
20
10
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
Sand Curves (Seed and Idriss 1970)

PI = 15 (Vucetic and Dobry 1991)
Generic Low-Plasticity model curve
PI = 15 model curve
PI = 30 model curve
Figure 6.2. Shear modulus degradation and equivalent viscous damping curves for the
soils used in the Stiff Soil profile.
325
Undrained Strength, s u (kPa)
0 500 1000 1500 2000 2500
0
20
su/σv' = 1.4 (0.8)
40
60
80
Depth (m)
100
120
140
160
180
200
Figure 6.3. Dynamic undrained shear strength for the Stiff Soil profile.
326
Vs (m/s) Density (Mg/cm3)
0 500 1000 1500 1.5 2 2.5
0 0
Fill
10 10
Soft
Clay
20 20
Stiff Clay
Depth (m)
30 30
40 40
50 50
60 60
Rock
70 70
Figure 6.4. Generic Soft Clay shear wave velocity and density profiles. The profile
represents typical Bay Mud sites from the San Francisco Bay region.
327
Undrained Strength, su (kPa)
0 100 200 300 400 500 600

0
10
20
Depth (m)
30
40
50
60
Figure 6.5. Dynamic shear strength profile for the generic Soft Clay site.
328
1
0.6
0.4
0.2
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
40
Young Bay Mud (Isenhower and Stokoe 1981)
Low-strain model
30 Large-strain model
Damping (%)
20
10
0
-4 -3 -2 -1 0 1
10 10 10 10 10 10
Strain (%)
Soft Clay soil. The Large-Strain model curve is given for the average strength value of
the clay. The curves shift to the right for an increase in strength, and to the left for a
decrease in strength.
329
1
0.6
0.4
0.2
0
-4 -3 -2 -1 0
10 10 10 10 10
Strain (%)
40
PI = 30
30 PI = 50
Upper bound model curve
Damping (%)
Lower bound model curve

20
10
0
-4 -3 -2 -1 0
10 10 10 10 10
Strain (%)
Pleistocene clay used in the generic Soft Clay profile. The upper- and lower-bound
model curve corresponds to the upper bound and lower bounds, respectively, for the
strength parameter R. The upper bound curve is used when soil yielding is not of
concern.
330
Vs (m/s)
0 500 1000 1500
0
10
Depth (m)
15
20
25
30
Selected Profile
Average and +_ one standard deviation for 227 profiles
Figure 6.8. Selected shear wave velocity profile for the Very Stiff Soil profile (Site C).
Included in the graph are the median and ± one standard deviation for a database
including 227 profiles (Silva, personal comm.) classified as Site C by the Uniform
Building Code (e.g. 360 m/s < Vs ≤ 760 m/s).
331
Pacoima Dam, SF Eq. Gilroy Gavilan College
1 1
Normalized PSD
Normalized PSD
0.5 0.5
0 0
-2 0 2 -2 0 2
10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)
Pacoima Dam, North. Eq. KJMA

1 1
Normalized PSD
Normalized PSD
0.5 0.5
0 0
-2 0 2 -2 0 2
10 10 10 10 10 10
Frequency (Hz) Frequency (Hz)
Recorded Ground Motion
Recorded Velocity Pulse
Simplified Velocity Pulse
Figure 6.9. Power spectral densities, PSD (normalized by the maximum value of the
PSD) for the ground motions used in the site response analyses. PSD are given for the
full recorded ground motion, for the time interval of the ground motion containing the
velocity pulse, and for the simplified representation of the velocity pulse.
332
Recorded Velocity Time History
200 200
Velocity (cm/s)
100 100
0 0
-100 -100
-200 -200
0 5 10 0 5 10
200 200
Velocity (cm/s)
100 100
0 0
-100 -100
-200 -200
0 5 10 0 5 10
200 200
Output(Soil)
Velocity (cm/s)
100 100 Input (Rock)
0 0
-100 -100
-200 -200
0 5 10 0 5 10
Time (s) Time (s)
Figure 6.10. Site response analyses results. Pacoima Dam record for the 1971 San
Fernando Earthquake.
333
50 50
Velocity (cm/s)
0 0
-50 -50
0 1 2 3 4 5 0 1 2 3 4 5
50 50
Velocity (cm/s)
0 0
-50 -50
0 1 2 3 4 5 0 1 2 3 4 5
50 50
Output(Soil)
Velocity (cm/s)
Input (Rock)
0 0
-50 -50
0 1 2 3 4 5 0 1 2 3 4 5
Time (s) Time (s)
Figure 6.11. Site response analyses results. Gilroy Gavilan College record for the 1989
Loma Prieta Earthquake.
334
50 50
Velocity (cm/s)
0 0
-50 -50
2 3 4 5 6 7 2 3 4 5 6 7
50 50
Velocity (cm/s)
0 0
-50 -50
0 1 2 3 4 5 0 1 2 3 4 5
50 50 Output(Soil)
Velocity (cm/s)
Input (Rock)
0 0
-50 -50
0 1 2 3 4 5 0 1 2 3 4 5
Time (s) Time (s)
Figure 6.12. Site response analyses results. Pacoima Dam downstream record for the
1994 Northridge Earthquake.
335
100 100
Velocity (cm/s)
0 0
-100 -100
0 5 10 0 5 10
100 100
Velocity (cm/s)
0 0
-100 -100
0 5 10 0 5 10
100 100
Velocity (cm/s)
0 0
Output(Soil)
-100 -100
Input (Rock)
0 5 10 0 5 10
Time (s) Time (s)
Figure 6.13. Site response analyses results. KJMA record for the 1995 Kobe
Earthquake.
336
Pacoima Dam, San Fernando EQ
200 200
100 100
0 0
-100 -100
-200 -200
-200 0 200 -200 0 200
Gilroy Gavilan College, Loma Prieta EQ.
50 50
0 0
-50 -50
-50 0 50 -50 0 50
Pacoima Dam Downstream, Northridghe EQ
50 50
0 0
-50 -50
-50 0 50 -50 0 50
KJMA, Kobe EQ
100 100
50 50
Input
0 Output 0
-50 -50
-100 -100
-100 0 100 -100 0 100

Fault Normal Velocity (cm/s) Fault Normal Velocity (cm/s)
Recorded Simplified
Figure 6.14. Site response analyses results. Input (rock) and output (soil) particle
velocity-trace plots.
337
1
0.5
Acceleration (g)
-0.5
-1
0 5 10 15
Time (s)
100
50
Velocity (cm/s)
-50
-100
0 5 10 15
Time (s)
Figure 6.15. Fault-normal acceleration and velocity time-histories for the Pacoima Dam
record of the 1971 San Fernando earthquake. The dominant velocity pulse is highlighted
to show that the maximum accelerations do not coincide with the fault-normal velocity
pulse.
338
Gilroy Gavilan College, Loma Prieta EQ Pacoima Dam, San Fernando EQ
0 0
-10 -10
Depth (m)
Depth (m)
-20 -20
-30 -30
-40 -40
-50 -50
0 0.1 0.2 0.3 0.4 0 0.5 1 1.5
Shear Strain (%) Shear Strain (%)
Pacoima Dam Dwnstr., Northridge EQ KJMA, Kobe EQ

0 0
-10 -10
Depth (m)
Depth (m)
-20 -20
-30 -30
-40 -40
-50 -50
0 0.1 0.2 0.3 0.4 0 0.5 1 1.5
Shear Strain (%) Recorded Pulse Shear Strain (%)
Simplified Pulse
Figure 6.16. Calculated maximum strains using the recorded velocity pulse and the
simplified velocity pulse as input motions.
339
Linear Elastic Soil Deposit, Tsoil = 1 s
2.5
Amplitude of first half cycle
Amplitude of second half cycle
2
Maximum Amplitude
PGVsoil/PGVrock
1.5
0.5
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Input Pulse Period, Tv (s)
(a) Amplification of PGV.
4
Period of first half cycle
Period of second half cycle
3 Period of pulse with Maximum Amplitude
Tv,soil/Tv,rock
0
0 0.5 1 1.5 2 2.5 3 3.5 4
Input Pulse Period, Tv (s)
(b) Ratio of soil pulse period to rock pulse period.
Figure 6.17. Response of a linear-elastic soil column subject to the simplified velocity
pulses. The natural period of the soil deposit is 1.0 s.
340
8
Amplification Factor
Impedance ratio = 0
6
0.3
4
0.5
2
0
0 1 2 3 4 5 6 7
ωH/Vs
kH
(a) Amplification factors for a soil with zero damping for varying
impedance contrast.
14
Amplification Factor
12
ξ = 5%
10
8 10%
6 20%
4
2
0
0 5 10 15 20
ωH/Vs
(b) Amplification factors for an elastic soil on rigid bedrock for

varying levels of soil damping.
Figure 6.18. Amplification of input motions by a linear-elastic profile (ω is the

frequency of the input motion, Vs is the shear wave velocity of the elastic soil, and H is
the soil depth (from Kramer 1996).
341
Tv,rock = 0.6 s Tv,rock = 1.0 s Tv,rock = 2.0 s Tv,rock = 4.0 s
PGVrock = 75 cm/s
FP Vel. (cm/s)
100 100 100 100
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

FN Vel. (cm/s) FN Vel. (cm/s) FN Vel. (cm/s) FN Vel. (cm/s)
PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200
-200 0 200 -200 0 200 -200 0 200 -200 0 200

PGVrock = 300 cm/s
500 500 500

FP Vel. (cm/s)
Input (rock)
Output (soil) 0 0 0
-500 -500 -500

-500 0 500 -500 0 500 -500 0 500
FN Vel. (cm/s) FN Vel. (cm/s) FN Vel. (cm/s)
Figure 6.19. Results from the site response analyses for the 60 m deep Stiff Soil profile.
Particle velocity-trace plots for the input (rock) and output (soil) motions. Input motion
set is Set 6. PGVp/n of input motion is 0.5.
342
PGVrock = 75 cm/s
FP Vel. (cm/s)
100 100 100 100
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200
-200 0 200 -200 0 200 -200 0 200 -200 0 200

PGVrock = 300 cm/s
500 500 500

FP Vel. (cm/s)
Input (rock) 0 0 0
Output (soil)
-500 -500 -500

-500 0 500 -500 0 500 -500 0 500
343
PGVrock = 75 cm/s
FP Vel. (cm/s) 100 100 100 100
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200
-200 0 200 -200 0 200 -200 0 200 -200 0 200

PGVrock = 300 cm/s
500 500 500

FP Vel. (cm/s)
Input (rock) 0 0 0
Output (soil)
-500 -500 -500

-500 0 500 -500 0 500 -500 0 500
344
Set 6 Set 7 Set 8
3 3 3
2.5 2.5 2.5

PGVsoil/PGVrock
PGVsoil/PGVrock
PGVsoil/PGVrock
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 2 4 0 2 4 0 2 4
Tv,FN (s) Tv,FN (s) Tv,FN (s)
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
(a) PGV soil to rock ratios.
Set 6 Set 7 Set 8

3 3 3
2.5 2.5 2.5

Tv,soil/Tv,rock
Tv,soil/Tv,rock
Tv,soil/Tv,rock
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 2 4 0 2 4 0 2 4
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
(b) Ratio of pulse period in soil to pulse period in rock.
Figure 6.22. Results of site response analyses for the 60 m deep Stiff Soil site for three
different input motion sets. Results are plotted with fault normal input pulse period in the
abscissa.
345
3
PGVrock = 75 cm/s
PGVrock =160 cm/s
2.5 PGVrock = 300 cm/s
PGVsoil/PGVrock
1.5
0.5
0 0.5 1 1.5 2 2.5 3 3.5 4
Tv,FN
Figure 6.23. Site response analyses results for the 30 m deep Stiff Soil profile. The input
motion is Set 6, with a PGVp/n ratio of 0.5. Observe how the period at which maximum
amplification of PGV occurs shifts to the right with increasing input motion intensity.
Also, note that the level of PGV amplification decreases slightly with increasing input
motion intensity.
346
Very Stiff Soil Site Stiff Soil Site
2 2
PGVsoil/PGVrock
PGVsoil/PGVrock
1.5 1.5
1 1
0.5 0.5
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) Tv,FN (s)
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
(a) PGV soil to rock ratios
Very Stiff Soil Site Stiff Soil Site

2 2
Tv,soil/Tv,rock
Tv,soil/Tv,rock
1.5 1.5
1 1
0.5 0.5
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) Tv,FN (s)
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
(b) Ratio of pulse period in soil to pulse period in rock.
Figure 6.24. Comparison of site response studies for two profiles with equal depth to
bedrock (30 m) and varying profile stiffness. Input motion is Set 6 with a PGVp/n ratio of
0.5.
347
30 m 45 m 60 m
3 3 3
2.5 2.5 2.5

PGVsoil/PGVrock
PGVsoil/PGVrock
PGVsoil/PGVrock
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 2 4 0 2 4 0 2 4
100 m 200 m
3 3
2.5 2.5
PGVrock = 75 cm/s
PGVsoil/PGVrock
PGVsoil/PGVrock
PGVrock = 160 cm/s

2 2 PGVrock = 300 cm/s
1.5 1.5
1 1
0.5 0.5
0 2 4 0 2 4
Tv,FN (s) Tv,FN (s)
Figure 6.25. Comparison of site response analyses results for the Stiff Soil profile. The
depth to bedrock is varied from 30 m to 200 m. The input motion is Set 6 with a PGVp/n
ratio of 0.5.
348
30 m 45 m 60 m
5 5 5
4 4 4
Tv,soil/Tv,rock
Tv,soil/Tv,rock
Tv,soil/Tv,rock
3 3 3
2 2 2
1 1 1
0 2 4 0 2 4 0 2 4
100 m 200 m
5 5
4 4
Tv,soil/Tv,rock
Tv,soil/Tv,rock
PGVrock = 75 cm/s
3 3 PGVrock = 160 cm/s
PGVrock = 300 cm/s
2 2
1 1
0 2 4 0 2 4
Tv,FN (s) Tv,FN (s)
Figure 6.26. Ratio of pulse period in soil to pulse period in rock for site response
analyses on Stiff Soil profiles with varying depth to bedrock. The input motion is Set 6
with a PGVp/n ratio of 0.5.
349
PGV = 75 cm/s PGV = 160 cm/s
3 3
2.5 2.5
PGVsoil/PGVrock
PGVsoil/PGVrock
2 2
1.5 1.5
1 1
0.5 0.5
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) Tv,FN (s)
Generic Low PI soil
PI = 15 Clay
PI = 30 Clay
Figure 6.27. Comparison of site response analyses for soils with different shear modulus
reduction and damping curves. Results for input motion Set 6 with a PGVp/n ratio of 0.5.
350
Generic Low-plasticity soil PI = 15 Clay PI = 30 Clay
3 3 3
2.5 2.5 2.5

Tv,soil/Tv,rock
Tv,soil/Tv,rock
Tv,soil/Tv,rock
2 2 2
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 2 4 0 2 4 0 2 4
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
Figure 6.28. Comparison of calculated ratio of soil pulse period to rock pulse period.
Results for site response analyses for soils with different shear modulus reduction and
damping curves. Input motion is Set 6, with PGVp/n ratios of 0.5.
351
Input Motion Set 7 Input Motion Set 8
2 2
PGVsoil/PGVrock
PGVsoil/PGVrock
1.5 1.5
1 1
0.5 0.5
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) Tv,FN (s)
PGVp/n = 0
PGVp/n = 0.25
PGVp/n = 0.75
Figure 6.29. Comparison of site response analyses results to input motions with varying
levels of fault parallel velocity. Results for a 45 m deep Stiff Soil for PGVp/n ratios of
0, 0.25 and 0.75. Input fault-normal PGV is 75 cm/s.
352
3
2.5 PGVp/n = 0
PGVp/n = .25
PGVp/n = .75
Tv,soil/Tv,rock
1.5
0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Tv,FN (s)
Figure 6.30. Comparison of site response analyses to input motions with different fault-
parallel velocities. Results for the 45 m deep Stiff Soil profile, Input motion Set 6.
353
30 m 45 m 60 m
2.5 2.5 2.5
2 2 2
PGVsoil/PGVrock
PGVsoil/PGVrock
PGVsoil/PGVrock
1.5 1.5 1.5
1 1 1
0.5 0.5 0.5

0 2 4 0 2 4 0 2 4
100 m 200 m
2.5 2.5
2 2 Set 6
PGVsoil/PGVrock
PGVsoil/PGVrock
Set 7
Set 8
1.5 1.5
1 1
0.5 0.5
0 2 4 0 2 4
Tv,FN (s) Tv,FN (s)
Figure 6.31. Comparison of site response analyses results for the different input motion
sets. The analyses are for the Stiff Soil profiles with varying depth. The input PGVp/n
ratio is 0.25 for Sets 7 and 8, and 0.5 for Set 6.
354
0
-10
Tv = .6 s; PGA = 0.8 g
Tv = 1.0 s; PGA = 0.48 g
Depth (m)
-20
Tv = 2.0 s; PGA = 0.24 g
Tv = 4.0 s; PGA = 0.12 g
-30
-40
-50
0 1 2
Strain (%)
(a) PGV = 75 cm/s
0
-10
Tv = 1 s; PGV = 75 cm/s
Depth (m)
-20 Tv = 2.0 s; PGV = 160 cm/s

Tv = 4.0 s; PGV = 300 cm/s
-30
-40
-50
0 5 10
Strain (%)
(b) PGA = 0.5 g
Figure 6.32. Comparison of strain levels for runs with (a) equal PGV, and (b) equal
PGA. Stiff Soil profile with a depth of 45 m. Input motion is Set 6 with PGVp/n = 0.5.
355
0 0
Tv = 0.6 s
-50 -50
Tv
Depth (m)
Depth (m)
-100 -100
Very Stiff Soil (30 m)
Stiff Soil (30 m)
-150 Stiff Soil (60 m) -150
Stiff Soil (100 m)
Stiff Soil (200 m)
-200 -200
0 2 4 6 0 2 4 6
Max. Strain (%) Max. Strain (%)
0 0
Tv
-50 -50
Depth (m)
Depth (m)
-100 -100
Tv
-150 -150
-200 -200
0 2 4 6 0 2 4 6
Max. Strain (%) Max. Strain (%)
Figure 6.33. Calculated strains for various Stiff Soil profiles. Input motion is Set 6 with
PGVp/n ratios of 0.5, and input PGV of 160 cm/s.
356
PGVrock = 75 cm/s
100 100 100 100

FP Vel. (cm/s)
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200

-200 0 200 -200 0 200 -200 0 200 -200 0 200
PGVrock = 300 cm/s
400 400 400

FP Vel. (cm/s)
200 200 200
Input (rock) 0 0 0
Output (soil) -200 -200 -200
-400 -400 -400
-400-200 0 200 400 -400-200 0 200 400 -400-200 0 200 400
Figure 6.34. Results of site response analyses. Particle velocity-trace plots of input and
output motions. Soft Clay site, input motion is Set 6 with PGVp/n ratios of 0.5.
357
PGVrock = 75 cm/s
100 100 100 100

FP Vel. (cm/s)
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200

-200 0 200 -200 0 200 -200 0 200 -200 0 200
PGVrock = 300 cm/s
400 400 400

FP Vel. (cm/s)
200 200 200
Input (rock) 0 0 0
Output (soil) -200 -200 -200
-400 -400 -400
-400-200 0 200 400 -400-200 0 200 400 -400-200 0 200 400
358
PGVrock = 75 cm/s
100 100 100 100

FP Vel. (cm/s)
0 0 0 0
-100 -100 -100 -100
-100 0 100 -100 0 100 -100 0 100 -100 0 100

PGVrock = 160 cm/s
200 200 200 200

FP Vel. (cm/s)
0 0 0 0
-200 -200 -200 -200

-200 0 200 -200 0 200 -200 0 200 -200 0 200
PGVrock = 300 cm/s
400 400 400

FP Vel. (cm/s)
200 200 200
Input (rock) 0 0 0
Output (soil) -200 -200 -200
-400 -400 -400
-400-200 0 200 400 -400-200 0 200 400 -400-200 0 200 400
359
0 0
Tv = 0.75 s
-10 -10 Tv = 1.0 s
Tv = 4.0 s
-20 -20 Strength Profile
Depth (m)
Depth (m)
-30 Tv = 0.75 s -30
Tv = 1.0 s
-40 Tv = 4.0 s -40
-50 -50
-60 -60
-70 -70
0 10 20 30 0 500 1000
Max. Strain (%) Max. Stress (kPa)
(a) PGV of input motion = 160 cm/s
0 0
Tv = 1.0 s
-10 -10 Tv = 2.0 s
Tv = 4.0 s
-20 -20 Strength Profile
Depth (m)
Depth (m)
-30 -30
Tv = 1.0 s
-40 -40
Tv = 2.0 s
Tv = 4.0 s
-50 -50
-60 -60
-70 -70
0 10 20 30 0 500 1000
Max. Strain (%) Max. Stress (kPa)
(b) PGV of input motion = 300 cm/s
Figure 6.37. Calculated stresses and strains for the Soft Clay profile. Input motion is Set
6 with PGVp/n = 0.5.
360
0 0 0
-20 -20 -20

Depth (m)
-40 -40 -40
-60 -60 -60
-80 -80 -80

0 200 400 0 200 400 0 1 2
Displacement (cm) Velocity (cm/s) Acceleration (g)
Tv = .6
Tv = 2.0
Tv = 4.0
Figure 6.38. Calculated profiles of displacement, velocity, and acceleration for the Soft
Clay site. Input motion is Set 6 with PGVp/n of 0.5 and input PGV of 160 cm/s.
361
Stiff Soil Site, 60 m depth Soft Clay Site
2 2
PGVsoil/PGVrock
PGVsoil/PGVrock
1.5 1.5
1 1
0.5 0.5
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) PGVrock = 75 cm/s Tv,FN (s)
PGVrock = 160 cm/s
PGVrock = 300 cm/s
Figure 6.39. Comparison of results of site response analyses for two profiles with equal
depth to bedrock but different soil profiles. Ratios of PGV in soil to PGV in rock. Input
motion is Set 6 with PGVp/n of 0.5.
362
Stiff Soil
500
2
400
1
1
1
PGVsoil (cm/s)
300
200
100
0
0 100 200 300 400 500
PGVrock (cm/s)
Soft Clay Site

500
Tv = 0.6
Tv = 1.0
2 Tv = 2.0
400 Tv = 4.0
1
1
1
300
PGVsoil (cm/s)
200
100
0
0 100 200 300 400 500
PGVrock
(cm/s)
Figure 6.40. Comparison of site response analyses for the Soft Clay profile and the Stiff
Soil profile with equal depth (60 m). Results shown are for all input motion sets with
PGVp/n values ranging from 0.25 to 0.5.
363
Stiff Soil Site, 60 m depth Soft Clay Site
Tv,soil/Tv,rock 4 4
Tv,soil/Tv,rock
3 3
2 2
1 1
0 1 2 3 4 0 1 2 3 4
Tv,FN (s) Tv,FN (s)
PGVrock = 75 cm/s
PGVrock = 160 cm/s
PGVrock = 300 cm/s
Figure 6.41. Comparison of results of site response analyses for two profiles with equal
depth to bedrock but different soil profiles. Ratios of pulse period in soil to pulse period
in rock. Input motion is Set 6 with PGVp/n of 0.5.
364
Very Stiff Soil Stiff Soil (45 m depth) Soft Soil
1.5 1.5 1.5
1 1 1
PGVp/n
Set 7
0.5 0.5 0.5
0 0 0
0 2 4 0 2 4 0 2 4
1.5 1.5 1.5
1 1 1
Set 8
PGVp/n
0.5 0.5 0.5
0 0 0
0 2 4 0 2 4 0 2 4
Input PGVp/n = 0.25
Input PGVp/n = 0.75
Figure 6.42. Effect of site response on the PGVp/n ratio. Results for three different
profiles and for input motion sets 7 and 8. PGV of input motion is 75 cm/s.
365
600
2
Pulse Motions
Tv = 0.6 s 1
500 Tv = 1.0 s
1
Tv = 2.0 s
Tv = 4.0 s 1
400
PGVsoil (cm/s)
300
Actual Records
200
× Pacoima Dam, SF EQ.
○ Gilroy Gav. Coll. LP EQ.
100 □ Pacoima Dam Dwnst.

Northridge EQ.
∆ KJMA, Kobe EQ.

0
0 100 200 300 400 500 600
PGVrock (cm/s)
Figure 6.43. Calculated amplification of peak ground velocity. Results for site response
analyses for all the Stiff Clay profiles and all the simplified velocity poulse input motion
sets. Large symbols are the predicted PGV amplification for the Stiff Soil site with a
depth of 45 m using the indicated records as input motions.
366
500 1.5
1 1
400 1
PGVsoil (cm/s)
Tvp = 0.6 s
300
Tvp = 1.0 s
Tvp = 2.0 s
200 Tvp = 4.0 s
100
0
0 100 200 300 400 500
PGVrock (cm/s)
Figure 6.44. Relationship between PGV in soil and PGV in rock. Results for the 45 m
deep Stiff Soil deposit. Input motion is Set 6 with PGVp/n ratio of 0.5. Observe the
dependence of PGV soil to rock ratio on pulse period (Tvp) and intensity of motion
(PGVrock).
367
5.5
4.5
PGVrock = 75 cm/s
4 PGVrock = 160 cm/s
PGVrock = 300 cm/s
3.5
Tv,soil/Tv,rock
3 Regression line for

recorded near-fault
2.5 ground motions
2
1.5
0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Tv,FN (s)
Figure 6.45. Calculated ratios of pulse period in the soil over pulse period in rock.
Results from site response analyses on Stiff Clay profiles using all simplified velocity
pulse input motions sets and PGVp/n ratios ranging from 0.25 to 0.75. The heavy line
represents the ratio of the average pulse period in soil over the average pulse period in
rock from the regression analysis of the near-fault ground motion database.
368
Stiff Clay Sites
5.5
5
Pulse Actual Records
×
PGVrock = 75 cm/s
4.5
PGVrock = 160 cm/s Pacoima Dam, SF EQ.
(PGVrock = 114 cm/s )
4
○ Gilroy Gav. Coll. LP EQ.
(PGVrock = 31 cm/s)
3.5
Tv,soil/Tv,rock
3
□ Pac. Dam D.. North.
EQ. (PGVrock = 50
cm/s )
2.5
2
∆ KJMA, Kobe EQ.
(PGV 96 / )
1.5
0.5
0 0.5 1 1.5 2 2.5 3 3.5 4 4.5
Tv,FN (s)
Figure 6.46. Calculated ratio of pulse period in soil to pulse period in rock for the 45 m
deep Stiff Soil profile. Results are shown for all the input motion sets with input PGV
lower than 300 m/s. Results for the indicated ground motions are included for
comparison with the results of the simplified velocity pulses.
369
CHAPTER 7
CONCLUSION
7.1 SUMMARY
This dissertation presented an evaluation of site effects on seismic ground
motions, with a particular emphasis on near-fault, forward-directivity ground motions.
The contributions of this work include the following:
1) The database of ground motions from the 1989 Loma Prieta and the 1994
Northridge earthquakes was used to evaluate a proposed new site
classification system to account for site amplification. The proposed site
classification system is based on a general characterization of the site that
includes depth to bedrock. This site classification system was used to evaluate
the influence of depth to bedrock and stiffness on site amplification. Site
amplification factors for different site conditions were proposed.
2) The database of near-fault ground motion records containing forward-
directivity effects was evaluated. Recorded near-fault forward-directivity
motions were represented by simplified velocity time-histories consisting of a
sequence of half-cycle sine pulses. The database was used to develop
attenuation relationships for the period and amplitude of the pulse. The
effects of local site conditions on these parameters were evaluated and
370
relationships between pulse period on rock and pulse period on soil were
proposed.
3) A nonlinear site response methodology using a bounding surface plasticity
model developed by Borja and Amies (1994) was implemented in the finite
element program GeoFEAP. The numerical simulations can represent the
two-dimensional response of a soil column to bi-directional motions (i.e.
fault-normal and fault-parallel components). The implementation was
evaluated with an emphasis on its ability to predict long-period site response.
The long-period site response was emphasized because near-fault ground
motions are characterized by long-period pulses of motion.
4) Simplified velocity time-histories representative of near-fault, forward-
directivity ground motions were developed to be used as input motions in site
response analyses. The time histories are developed both for the fault-normal
and fault-parallel components of motion.
5) Site response analyses using pulse-like simplified velocity time-histories as
input motions were performed. The ability of the simplified time-histories to
capture amplification of velocities and site effects on the input pulse motions
was validated by comparing results to site response analyses of recorded
ground motions. Site response analyses were performed on generalized site
profiles representing average site conditions for different site categories.
Conclusions regarding the influence of site conditions on the input velocity
pulses were presented.
371
7.2 FINDINGS
7.2.1 General
The primary findings in this dissertation are presented in this section.
Observations regarding the characterization of site response using the proposed
classification system are first presented. The results from the empirical analyses of near-
fault ground motions are then discussed. Finally, conclusions regarding site response to
near-fault ground motions are summarized.
7.2.2 Characterization of site response
A site classification scheme based on a general geotechnical characterization of
the site and the depth to bedrock was introduced. This site classification scheme was
evaluated with data from the 1989 Loma Prieta and 1994 Northridge earthquakes. The
main findings can be summarized as follows:
• The use of the proposed site classification scheme in ground motion estimation
results in a reduction in the standard deviation associated with median
estimates of ground motion when compared with a simpler "rock vs. soil"
classification system.
• Simplified classification systems commonly used in attenuation relationships
divide sites into "rock" and "soil" categories. The "rock" category groups
competent rock sites and weathered soft rock/shallow stiff soil sites. Results
showed that the response of these two site conditions is significantly different.
For an improved site categorization system for defining site-dependent ground

372
motions, the "rock" category should be divided along the indicated sub-
categories.
• Current attenuation relationships use the generic "rock" category as the baseline
site condition. The "rock" sites in the current database of strong ground
motions are dominated by weathered soft rock/shallow stiff soil sites; hence,
rock attenuation relationships likely reflect the response of these types of soils.
This, in turn, emphasizes the need to review the database of ground motions to
redefine the baseline site condition. Competent rock motions have spectral
acceleration values (at 5% damping) that are on average 30% less than those
predicted by widely used "rock" attenuation relationships.
• The standard deviations resulting from the proposed classification system are
comparable with the standard deviations obtained using a more burdensome
average shear wave velocity classification system. This illustrates that an
estimate of profile depth can compensate for a less exhaustive estimate of soil
stiffness. Profile depth is an important parameter for the estimation of seismic
site response.
7.2.3 Empirical study of near-fault, forward-directivity ground motions
Near-fault ground recordings that satisfied geometric conditions for forward-
directivity were selected. These recordings are characterized by long-period pulses of
motion that can be best observed in velocity or displacement time-histories. The
recordings in the database were characterized by a pulse period and their peak ground
velocity (PGV). The database was used for a statistical analysis of these parameters.
373
Regression analyses for pulse period as a function of magnitude and pulse amplitude as a
function of magnitude and distance were presented. The findings of the empirical study
can be summarized as follows:
• Near-fault, forward-directivity motions can be adequately represented by
simplified time-histories consisting of one or a few sine-pulses. The number of
pulses is likely related to slip distributions in the causative fault, and
consequently is difficult to predict.
• Pulse period is a function of moment magnitude (Mw). For earthquakes with a
magnitude of 6.1, the pulse period is about 0.6 s; and increases to about 6.0 s
for a magnitude of 7.5. Relatively high standard deviations are associated with
the median prediction of pulse period.
• Most of the energy in forward-directivity ground motions is concentrated in the
narrow-period band centered on the pulse period. Consequently, low
magnitude events might result in more damaging ground motions for typical
low period structures. Hence, care must be exercised when seismic demand is
estimated using deterministic analyses for Maximum Credible Earthquakes.
• Local site conditions have an important effect on pulse period. Longer periods
occur at soil sites than at rock sites for events with magnitudes lower than
about Mw = 7.0. The difference diminishes as the magnitude increases. For
events with Mw ≈ 7.5, the pulse periods at rock and soil sites are approximately
the same.
374
• PGVs in the near-fault region vary significantly with magnitude and distance.
Estimates of these velocities have large standard deviations. Median PGVs for
soils are larger than median PGVs for rock sites for low magnitude events. The
difference diminishes as the magnitude increases.
7.2.4 Site response analyses to near-fault ground motions
The finite element implementation of the bounding surface plasticity model by
Borja and Amies (1994) was evaluated with recordings at the Lotung array in Taiwan,
and at the Chiba array in Japan. The implementation reproduced accurately the surface
accelerations and velocities recorded in these arrays. In addition, a series of shaking table
tests performed at U.C. Berkeley (Meymand 1998, Lok 1999) using a near-fault recording
as the input motion were also evaluated. The implementation was able to represent well
the long-period site response and the PGVs for these experiments for PGVs as large as
190 cm/s in field scale (model scale values of PGV of up to 70 cm/s).
Site response analyses were performed on recorded near-fault ground motions and
on a series of simplified bi-directional velocity pulses developed from a statistical
analysis of the database of near-fault, forward-directivity ground motions. Generic soil
profiles used to represent different site categories were used in the analyses. The main
findings can be summarized as follows:
• In most cases, site response to simplified representation of recorded near-fault,
forward-directivity ground motions rendered results equivalent to those from
site response analyses to the corresponding recorded ground motions.
375
However, not all near-fault motions can be represented by simplified sine-pulse
velocity time-histories.
• PGV amplification is a function of both the period and the intensity of the input
pulse. The period for which maximum amplification of PGV occurs can be
associated with the degraded natural period of the site.
• The degraded natural period of a site increases with increasing profile depth, it
decreases with increasing profile stiffness, and it increases with increasing
input motion intensity. Consequently, sites with long degraded site periods
(i.e. Stiff Soil sites deeper than 60 m, Soft Clay sites, or Shallow Stiff Clay
sites subject to intense shaking) tend to amplify input pulses with periods
longer than 2.0 s. Conversely, sites with short degraded site periods (i.e.
Shallow, Very Stiff Soil sites, or sites subject to low intensity input motions)
tend to amplify input pulses with short periods (≤ 1 s) and are not significantly
affected by long pulse periods.
• The nonlinear stress-strain properties of the soil have an important effect on
PGV amplification. For a given strain level, soils that exhibit lesser shear
modulus reduction tend to amplify motions with shorter input pulse periods
more than soils with relatively more shear modulus reduction at the given strain
level.
• For near-fault ground motions, the response of soft clays is controlled by its
dynamic shear strength (i.e. soil model yield stress). For the cases studied in
this work, the soft clay reached its dynamic shear strength for input velocity
376
pulse amplitudes as low as 75 cm/s. This value is well within the expected
range of velocities for near-fault ground motions. Hence, the evaluation of the
shear strength of soft clays is important for the proper evaluation of site
response to near-fault ground motions.
• Results indicated that PGVs in stiff soils are amplified by a factor ranging from
one to two. The amplification is observed for input PGVs as large as 300 cm/s
when the input pulse period is longer than 1 s. For the cases considered in this
study, soft soils amplify PGVs when the input pulse period has amplitudes
lower than about 150 cm/s. For higher intensities of the input motion, PGVs at
the surface of the Soft Soil profile used in this study decrease with respect to
the input PGV.
• The mid-period amplification factor for Site D in the 1997 UBC (180 m/s < Vs
≤ 360 m/s) in Zone 4 is 1.6. This factor agrees with the range of Stiff Soil
PGV amplification obtained in the site response analyses for input pulse
amplitudes lower than 300 m/s. Mid-period amplification factors can be
compared to velocity amplification factors because mid-period spectral values
are generally controlled by velocities. The amplification for Site E in the UBC
( Vs ≤ 180 m/s) in Zone 4 is 2.4. This factor is larger than the amplification
factors obtained from the site response analyses for the Soft Clay site (Figure
6.40). The large code factors are likely due to extrapolations from equivalent
linear analyses that ignore the effect of soil failure.
377
• Site effects change the period of the velocity pulse. Typically, the calculated
pulse period is longer than the period of the input velocity pulse. An increase
in pulse period occurs when the input pulse period is shorter than the degraded
natural period of the site. In these cases, the soil column tends to oscillate at
the natural site period; hence, a significant increase in velocity pulse period is
observed.
• Input pulse periods that are significantly longer than the degraded site period
have much longer wavelengths than the profile depth. Consequently, a soil
column subject to these input pulses moves more or less as a rigid body and the
period and amplitude of the input pulse is not affected by the soil column.
• The median value of the calculated ratios of soil to rock pulse period for all the
cases considered in this study agree with the ratios estimated from the
regression analysis of the ground motion database of near-fault, forward
directivity ground motions.
7.3 RECOMENDATIONS FOR FURTHER RESEARCH
While the work in this thesis has increased the understanding of seismic site
response, it also identified a number of issues that warrant further investigation:
• Inclusion of additional earthquakes to the database of ground motions used in
this study for the evaluation of site response. The extended database should be
used in the development of updated attenuation relationships that use the
geotechnical site categories presented in this study.
378
• The data from the recent 1999 Chi-Chi, Taiwan, earthquake contains a large
number of near-fault recordings. Properly processed records from this and
other recent events should be included in the database of near-fault ground
motions for further analyses.
• Seismological fault modeling is gaining acceptance as an alternative method
for the prediction of ground motions, particularly for the large-and short
distance range where there are not a large number of recorded ground motions.
The important conclusions regarding site effects on near-fault ground motions
presented in this study should be included in the seismological modeling of
these types of motions.
• Soil constitutive models able to match simultaneously shear modulus
degradation and equivalent viscous damping curves are necessary for the
proper evaluation of site response. For near-fault ground motions, it is
important that these models represent the large strain behavior of soils while
also satisfying the extended Masing rules (Pyke 1979) for irregular cyclic
loadings. Models such as the model by Pestana and Lok (Lok 1999) should be
implemented and used for the evaluation of near-fault site response.
• The work in this dissertation did not include response of liquefiable soils. Soil
liquefaction, however, is a considerable problem and issues relating to the
particular characteristics of near-fault ground motions, such as the influence of
the reduced number of cycles on liquefaction potential, are not well
understood. Both empirical and analytical studies using effective stress soil
379
models and coupled mechanical-pore-water dissipation analysis should be
performed to address these issues.
• Near-fault, forward-directivity ground motions typically have long-period
energy. Basin effects affect ground motions in the same period-range. Three-
dimensional models including soil non-linearity should be used to evaluate the
effect of basin topography on long-period pulses.
• The relationship between earthquake magnitude and pulse period, together with
the narrow-banded nature of near-fault ground motions, imply that for certain
period bands, larger intensities could be expected from earthquakes of lower
magnitudes. The effects of this observation on building codes and on the use
of Maximum Credible Earthquakes for design should be evaluated.
380
REFERENCES
Abrahamson, N. A. and Shedlock, K. M. (1997). "Overview." Seismological Research

Letters, Vol. 68(1), 9-23.
Abrahamson, N. A. and Silva, W. J. (1997). "Empirical Response Spectral Attenuation

Relationships for Shallow Crustal Earthquakes." Seismological Research Letters,
Vol. 68(1), pp. 94-127.
Abrahamson, N. A. and Somerville, P. G. (1996). "Effects of the Hanging Wall and

Footwall on Ground Motions Recorded during the Northridge Earthquake."
Bulletin of the Seismological Society of America, Vol. 86, No. 1B, pp. S93-S99.
Abrahamson, N. A. and Sykora, D. (1993). "Variation of ground motion across

individual sites." Proceedings of the 4th DOE Natural Phenomena Hazards
Mitigation Conference, LLNL CONF-9310102.
Abrahamson, N. A., and Youngs, R. R. (1992). "A stable algorithm for regression
analyses using the random effects model." Bulletin of the Seismological Society
of America, 82(1), 505-510.
Aki, K. and Larner, L. L. (1970). "Surface motion of a layered medium having an

irregular interface due to incident plane SH waves." Journal of Geophysical
Research, 75, 933-954.
Alavi, B. and Krawinkler, H. (2000). "Consideration of near-fault ground motion effects

in seismic design." Proceedings, 12th World Conference on Earthquake
Engineering, New Zealand.
Anderson, D. G., and Tang, Y. K. (1989). “Summary of soil characterization programs

for the Lotun Large Scale Seismic Testing Experiment.” Proceedings,
EPRI/NRC/TPC workshop on seismic soil structure interaction analysis
techniques using data from Lotung, Taiwan. Palo Alto, Electric Power Research
Institute, NP6154.
Anderson, J. G., Lee, Y., Zheng, Y., and Day, S. M. (1996). "Control of strong motion by
the upper 30 meters." Bulletin of the Seismological Society of America, 86(6),
1749-1759.
Applied Technology Council (ATC) (1978). Tentative Provisions for the Development
of Seismic Regulations for Buildings, ATC3-06 Report, Redwood City, CA.
381
Bard, P. Y., and Bouchon, M. (1980). "The seismic response of sediment-filled valleys.
Part I. The case of incident SH waves." Bulletin of the Seismological Society of
America, 58, 367-398.
Bardet, J.P. (1989). "Prediction of deformations of Hostun and Reid Bedford sands with
a simple bounding plasticity model," in Constitutive Equations for Granular Non-
cohesive Soils, A. Saada and G. Bianchini, Eds., 131-148, Balkema.
Bazant, Z. P. and Krizek, R. J. (1976). "Endochronic constitutive law for liquefaction of

sands." Journal of Engineering Mechanics Division, ASCE 102(EM2) 225-238.
Benjamin, J. R. and Cornell, A. (1970). Probability, statistics, and decision for civil
engineers, Mc Graw-Hill, New York.
Boatwright, J. and Seekins, L. C. (1997). "Response Spectra from the 1989 Loma Prieta,
California, Earthquake Regressed for Site Amplification, Attenuation, and
Directivity," U.S. Geological Survey, MS 977, Menlo Park, CA.
Boore, D. M., and Brown, L. T. (1998). "Comparing Shear-Wave Velocity Profiles from
Inversion of Surface-Wave Phase Velocities with Downhole Measurements:
Systematic Differences Between the CSX Method and Downhole Measurements
at Six USC Strong-Motion Sites." Seismological Research Letters, Vol. 69(3),
pp. 222-229.
Boore, D. M., Joyner, W. B., and Fumal, T. E. (1993). "Estimation of response spectra
and peak accelerations from Western North American earthquakes." Interim
Report, Part II. USGS Open File Report 93-509.
Boore, D. M., Joyner, W. B., and Fumal, T. E. (1994a). "Estimation of response spectra
and peak accelerations from Western North American earthquakes." Interim
Report, Part II. USGS Open File Report 94-127.
Boore, D. M., Joyner, W. B., and Fumal, T. E. (1994b). "Ground motion estimates for
strike- and reverse- slip faults, provided to the Southern California Earthquake
Center, and widely distributed as an insert in Boore et al. 1994a.
Boore, D. M., Joyner, W. B., and Fumal, T. E. (1997). "Equations for Estimating
Horizontal Response Spectra and Peak Acceleration from Western North
American Earthquakes: A Summary of Recent Work." Seismological Research
Letters, Vol. 68(1), pp. 128-153.
Borcherdt, R. D. (1994). "Estimates of Site-Dependent Response Spectra for Design

(Methodology and Justification)." Earthquake Spectra, Vol. 10(4), pp. 617-653.
382
Borja, R. I., Chao, H. Y., Montáns, F. J., Lin, C. H. (1999). "Nonlinear ground response
at Lotung LSST site." Journal of Geotechnical and Geoenvironmental
Engineering, ASCE 125, 187-197.
Borja, R. I., Chao, H. Y., Sama, K. M., and Masada, G. M. (2000). "Modelling non-
linear ground response of non-liquefiable soils." Earthquake Engineering and
Structural Dynamics, 29, 63-83.
Borja, R.I., and Amies, A.P. (1994). "Multiaxial cyclic plasticity model for clays."
Journal ofGeotechncial Engineering, 120(6) 1051-1070.
Bray, J. D., Riemer, M. F., Gookin, W. B. (1999)."On the dynamic characterization of

soils." Proceedings, 2nd International Conference on Earthquake Geotechnical
Engineering, Vol. 3, 847-856.
Britto, A. M., and Gunn, M. J. (1987). “Critical state soil mechanics via finite elements.”
Halsted Press, E. Horwood, New York.
Campbell, K. W. (1990). "Empirical prediction of near-source soil and soft-rock ground

motion for the Diablo Canyon power plant site, San Luis Obispo County, CA."
Dames and Moore Report to Lawrence Livermoore National Laboratory, Sept.
Campbell, K. W. (1997). "Empirical Near-Source Attenuation Relationships for

Horizontal and Vertical Components of Peak Ground Acceleration, Peak Ground
Velocity, and Pseudo-Absolute Acceleration Response Spectra." Seismological
Research Letters, Vol. 68(1), pp. 154-179.
CDMG. 1:250,000 Scale Local Geologic Map Series.
Chang, C.-Y.; et al. (1990). “Equivalent linear versus nonlinear ground response analyses
at Lotung seismic experiment site.” Proceedings, Fourth U.S. National
Conference on Earthquake Engineering, Proceedings, Earthquake Engineering
Research Inst., El Cerrito, California, Vol. 1, 327-336
Chang, S. W., Bray, J. D., and Seed, R. B. (1996). "Engineering Implications of Ground
Motions from the Northridge Earthquake." Bulletin of the Seismological Society
of America, Vol. 86(1), Part B Suppl., pp. 270-288.
Chang, W. S., and Bray, J. D. (1995). "Seismic Response of Deep, Stiff Soil Deposits in
the Oakland, California Area during the Loma Prieta Earthquake." Report No.
UCB/GT/95-06. Geotechnical Engineering, Department of Civil and
Environmental Engineering, University of California, Berkeley, California.
383
Chang, W. S., Bray, J. D., Gookin, W. B., and Riemer, M. F. (1997). "Seismic Response
of Deep Stiff Soil Deposits in the Los Angeles, California Area during the 1994
Northridge Earthquake." Report No. UCB/GT 97/01, Geotechnical Engineering,
Department of Civil and Environmental Engineering, University of California,
Berkeley, California.
Chin, B. H.and Aki. K. (1991). "Simultaneous study of the source, path, and site effects
on strong ground motion during the Loma Prieta earthquake: a preliminary result
on pervasive nonlinear site effects." Bulletin of the Seismological Society of
America, 81(5), 1859-1884.
Chopra, Anil K. (1995). Dynamics of structures, theory and applications to earthquake

engineering. Prentice Hall, Englewood Cliffs, New Jersey.
Cook, R. D., Malkus, D. S., and Plesha, M. E. (1989). "Concepts and applications of
fnite element analysis." John Wiley and Sons, New York, NY.
Crouse, C. B. and McGuire, J. W. (1996). "Site response studies for purpose of revising
NEHRP seismic provisions." Earthquake Spectra, 12, 407-439.
CSMIP (1992). "CSMIP Strong-Motion Records from the Landers, California

Earthquake of June 28, 1992." California Department of Conservation, Division
of Mines and Geology, Office of Strong Motion Studies, Report OSMS 92-09.
Dafalias, Y. F. (1986). "Bounding surface plasticity, I: Mathematical foundation and

hypoplasticity." Journal of Engineering Mechanics, ASCE, 112 966-987.
Dafalias, Y. F. and Herrmann, L. R. (1982). "Bounding surface formulation of soil

plasticity," in Soil Mechanics – Transient and Cyclic Loads, Pande, G. N. and
Zienkiewicz, O. C., Eds., John Wiley and Sons, New York, NY.
Dafalias, Y. F. and Popov E. P. (1975). "A simple constitutive law for artificial graphite-
like materials." Transactions 3rd SMIRT, C 1/5, London, U.K.
Dafalias, Y. F. and Popov, E. P. (1976). "Plastic internal variables formalism of cyclic

plasticity." Journal of Applied Mechanics, 98(4) 645-650.
Dafalias, Y. F., and Popov, E. P. (1977). “Cyclic loading for materials with a vanishing
elastic region.” Nuclear Engineering and Design, 41, 2, Apr. 1977, pages 293-302.
Darragh R. B. and Idriss, E. M. (1998). "Shear-wave velocities and design response

spectra – an examination using strong motion data from the Gilroy array:
Preliminary results." CSMIP 97, Seminar on Utilization of Strong Motion Data:
Los Angeles, CA. Proceedings, CDMG, SOC1997, 71-87.
384
Darve, F., and Labanieh, S. (1982). "Incremental constitutive law for sands and clays:
simulation of monotonic and cyclic tests." International Journal of Numerical and
Analytical Methods in Geomechanics, 6, 243-275.
Day, S. M. (1996). "RMS response of a one-dimensional half-space to S waves."

Bulletin of the Seismological Society of America, 86, 363-370.
Dibblee, T.W. 1:24,000 Scale Geologic Maps. Dibblee Geological Foundation Map
Series.
Dickenson, S. E. (1994). "Dynamic Response of Soft and Deep Cohesive Soils during
the Loma Prieta Earthquake of October 17, 1989." A dissertation submitted in
partial satisfaction of the requirements for the degree of Doctor of Philosophy in
Civil Engineering of the University of California at Berkeley.
Dobry, R., Borcherdt, R. D., Crouse, C. B., Idriss, I. M., Joyner, W. B., Martin, G. R.,
Power, M. S., Rinne, E. E., and Seed, R. B. (2000). "New site coefficients and
site classification system used in recent building seismic code provisions.",
Earthquake Spectra, 16(1), 41-67.
Dobry, R., Martin, G. M., Parra, E., and Bhattacharyya, A. (1994). "Development of Site-
Dependent Ratios of Elastic Response Spectra (RRS) and Site Categories for
Building Seismic Codes." Proceedings of the NCEER/SEAOC/BSSC Workshop
on Site Response during Earthquakes and Seismic Code Provisions, University of
Southern California, Los Angeles, November 18-20, edited by G. M. Martin.
Dobry, R., Martin, G. M., Parra, E., and Bhattacharyya, A. (1994). "Development of Site-
Dependent Ratios of Elastic Response Spectra (RRS) and Site Categories for
Building Seismic Codes." Proceedingseedings of the NCEER/SEAOC/BSSC
Workshop on Site Response during Earthquakes and Seismic Code Provisions,
University of Southern California, Los Angeles, November 18-20, edited by G. M.
Martin.
Dobry, R., Ramos, R., and Power, M. S. (1997). "Site Factors and Site Categories in
Seismic Codes: A Perspective." In Proceedings of the FHWA/NCEER Workshop
on the National Representation of Seismic Ground Motion for New and Existing
Highway Facilities, Technical Report NCEER-97-0010. Friedland, I. M., Power,
M. S., and Mayes, R. L., Eds.
Duke, C. M. and Leeds, D. J. (1962). "Site Characteristics of Southern California Strong

Motion Earthquake Stations." Report No. UCLA-ENG 62-55.School of
Engineering and Applied Science, University of California, Los Angeles,
California.
385
Duke, C. M., Johnson, J. A., Kharraz, Y., Campbell, K. W., and Malpiede, N. A. (1971).
"Subsurface Site Conditions and Geology in the San Fernando Earthquake Area."
Report No. UCLA-ENG 72-06. School of Engineering and Applied Science,
University of California, Los Angeles, California.
EPRI (1993). Guidelines for Determining Design Basis Ground Motions. Electric Power
Research Institute, EPRI TR-012293s, Palo Alto, CA
Espinoza, Bray, J. D., R.D., Soga, K., and Taylor, R.L. (1995) “GeoFEAP: Geotechnical
Finite Element Analysis Program,” Report No. UCB/GT/95-05, Department of
Civil Engineering, University of California, Berkeley
Finn, W. D. L. (1982). "Dynamic response analysis of saturated sands," in Soil

Mechanics – Transient and Cyclic Loads, Pande, G. N. and Zienkiewicz, O. C.,
Eds., John Wiley and Sons, New York, NY.
Finn, W. D. L. (1988). "Dynamic analysis in geotechnical engineering." Proceedings,

Earthquake Engineering and Soil Dynamics II – Recent Advances in Ground
Motion Evaluations. Geotechnical Special Publication 20, ASCE 523-591.
Finn, W. D. L. (1991) "Geotechnical aspects of microzonation." Proceedingseedings.

4th Int. Conference On Seismic Zonation; Vol. 4, 199-259.
Fumal, T. E. (1991). "A Compilation of the Geology and Measured and Estimated Shear-
Wave Velocity Profiles at Strong-Motion Station that Recorded the Loma Prieta,
California, Earthquake." USGS Open-File Report 91-311.
Fumal, T. E., Gibbs, J. F., and Roth, E. F. (1981). "In-Situ Measurements of Seismic
Velocity at 19 Locations in the Los Angeles, California Region." USGS Open-
File Report 81-399.
Velocity at 22 Locations in the Los Angeles, California Region." USGS Open-
File Report 82-833.
Velocity at 16 Locations in the Los Angeles, California Region," USGS Open-
File Report 84-681.
Gaskins, L. and Glaser, S. (2000). “Consistency of ground motion estimation made using
system identification.” Bulletin of the Seismological Society of America, 89(4),
992-1008.
386
Geomatrix Consultants (1993). "Compilation of Geotechnical Data for Strong-Motion
Stations in the Western United States." Report to Lawrence Livermore National
Laboratory, Project No. 2256.
Gibbs, J. F., and Fumal, T. E. (1993). "Seismic Velocities and Geologic Logs from
Borehole Measurements at Seven Strong-Motion Stations that Recorded the Loma
Prieta Earthquake, Part IV." USGS Open-File Report 94-552.
Gibbs, J. F., Fumal, T. E. Powers, T. J. (1994). "Seismic Velocities and Geologic Logs
from Borehole Measurements at Seven Strong-Motion Stations that Recorded the
Loma Prieta Earthquake." USGS Open-File Report 94-222.
Gibbs, J. F., Fumal, T. E., and Roth, E. F. (1980). "In-Situ Measurements of Seismic
Velocity at 27 Locations in the Los Angeles, California Region," USGS Open-
File Report 80-378.
Gibbs, J. F., Fumal, T. E., Boore, D. M., and Joyner, W. B. (1992). "Seismic Velocities
and Geologic Logs from Borehole Measurements at Seven Strong-Motion
Stations that Recorded the Loma Prieta Earthquake." USGS Open-File Report
92-287.
Gibbs, J. F., Fumal, T. E., Powers, T. J. (1993). "Seismic Velocities and Geologic Logs
from Borehole Measurements at Seven Strong-Motion Stations that Recorded the
Loma Prieta Earthquake." USGS Open-File Report 93-376.
Gibbs, J. F., Tinsley, J. C., and Joyner, W. B. (1996). "Seismic Velocities and Geological
Conditions at Twelve Sites Subjected to Strong Ground Motion in the 1994
Northridge, California, Earthquake," USGS Open-File Report 96-740.
Hardin, B. O., and Drnevich, V. P. (1972). "Shear modulus and damping in soils:
measurements and parameter effects." Journal of Soil Mechanics and Foundation
Engineering Division, ASCE 98(SM6) 603-624.
Haskell, N. A. (1960). "Crustal reflection of plane SH waves." Journal of Geophysical

Research, 65:4147-4150.
Hilber, H. M., Hughes, T. J. R., Taylor, R. L. (1977). “Improved numerical dissipation

for time integration algorithms in structural dynamics.” Earthquake Engineering
and Structural Dynamics, 5, 3, July-Sept. 1977, pages 283-292
Hirai, H. (1987). "An elastoplastic constitutive model for cyclic behavior of sands."
International Journal of Numerical and Analytical Methods in Geomechanics, 1,
513-520.
387
Hudson, M., Idriss, I.M., and Beikae, M. (1994) “QUAD4M: A Computer Program to
Evaluate the Seismic Response of Soil Structures using Finite Element Procedures
and Incorporating a Compliant Base,” Center for Geotechnical Modeling, Dep. of
Civil & Env. Engng, University of California, Davis.
Hueckel, T. and Nova, R. (1979) “Some Hysterical Effect on the Behavior of Geological
Media,” Int. Journal of Solids and Structures, 15(3), 625-642.
Idriss, I. M. (1990). "Responseof softsoil sites during earthquakes." Proceedings. of the

Symposiumto Honor Prof. H. B. Seed, Berkeley, 273-289.
Idriss, I. M. and Sun, J. I. (1992). "User's Manual for SHAKE91, A computer Program
for Conducting Equivalent Linear Seismic Response Analyses of Horizontally
Layered Soil Deposits Program Modified based on the Original SHAKE Program
Published in December 1972 by Schnabel, Lysmer and Seed."
Idriss, I. M., (1991). "Procedures for Selecting Earthquake Ground Motions at Rock
Sites. A Report to the National Institute of Standards and Technology,"
University of California at Davis, September, revised March 1993.
Isenhower, W. M. and Stokoe, K. H. II (1981). "Strain-rate dependent shear modulus of

San Francisco Bay mud." International Conference on Recent Advances in
Geotechnical Earthquake Engieneering, University of Missouri,, Rolla, pp. 597-
602.
Iwan, W,. D. (1967). "On a class of models for the yielding behavior of continuous and
composite systems." Journal of Applied Mechanics, 34, 612-617.
Iwasaki, T., Tatsuoka, F., and Takagi, Y. (1976). "Dynamic Shear Deformation
Properties of Sand for Wide Strain Range." Report of Civil Engineering Institute,
No. 1085, Ministry of Construction, Tokyo, Japan.
Iwasaki, T., Tatsuoka, F., Takagi, Y. (1978). “Shear moduli of sands under cyclic
torsional shear loading.” Soils and Foundations, 18, 1, Mar. 1978, pages 39-56.
Joyner W. B., and Chen, A. T. F. (1975). "Calculation of nonlinear ground response in

earthquakes." Bulleting of the Seismological Society of America, 65(5) 1315-
1336.
Joyner, W. B., Boore, D. M. (1981). "Peak acceleration and velocity from strong-motion
records including records from the 1979 Imperial Valley, CA, earthquake."
Bulleting of the Seismological Society of America, 71, 2011-2038.
Joyner, W. B., Boore, D. M. (1982). "Prediction of earthquake response spectra." USGS

Open File Report 82-977.
388
Kanai (1950). "Effect of solid viscosity of surface layer on earthquake motions and the
nature of surface layer." Bulletin of the Earthquake Research Institute, 34:53-88.
Katayama, T.; et al. (1990). “A strong motion database for the Chiba seismometer array
and its engineering analysis.” Earthquake Engineering & Structural Dynamics,
19, 8, Nov. 1990, pages 1089-1106.
Katona, M. G., and Zienkiewicz, O. G. (1985). ""A unified set of single step algorithms
Part 3: The Beta-m method, a generalization of the Newmark scheme."
International Journal for Numerical Methods in Engineering, 21, 1345-1359.
Kondner, R. L. (1963). "Hyperbollic stress-strain response: Cohesive soils." Journal of

the Soil Mechanics and Foundation Division, ASCE, 89(1) 115-143.
Kramer, S. L. (1996). "Geotechnical Earthquake Engineering." Prentice Hall, Upper

Saddle River, New Jersey.
Krawinkler, H. and Alavi, B. (1998). "Development of improved design procedures for

near-fault ground motions." SMIP 98, Seminar on Utilization of Strong Motion
Data: Oakland, CA.
Krieg, R. D. (1975). "A practical two-surface plasticity theory." Journal of Applied

Mechanics, 42, 641-646.
Lanzo, G. and Vucetic, M. (1999). "Effect of soil plasticity on damping ratio at small
cyclic strains." Soils and Foundations, 39(4) 131-141.
Lee, M. K. W. and Finn, W. D. L. Finn (1978). "DESRA-2, Dynamic effective stress

response analysis of soil deposits with energy transmitting boundary including
assessment of liquefaction potential." Soil Mechanics Series No. 36, Department
of Civil Engineering, University of British Columbia, Vancouver, Canada.
Lee, M. K. W. and Finn, W. D. L. Finn (1991). "DESRA-2C: Dynamic effective stress

response analysis of soil deposits with energy transmitting boundary including
assessment of liquefaction potential." University of British Columbia, Faculty of
Applied Science.
Lefebvre, G., and LeBoeuf, D. (1987). “Rate effects and cyclic loading of sensitive
clays.” Journal of Geotechnical Engineering, 113(5), 476-489.
Li X. S., Shen, C. K., Wang, Z. L. (1998). "Fully-coupled inelastic ground response

analysis for 1986 Lotung Earthquake Events." Journal of Geotechnical
Engineering, ASCE, 124(7).
389
Lii. X.S., Wang, Z. L, and Shen, C. K. (1997). “SUMDES. A nonlinear procedure for
response analysis of horizontally layered sites subjected to multi-directional
earthquake loading. Department of Civil Engineering, University of California at
Davis.
Lok, M. H. (1999). "Numerical modeling of seismic soil-pile-superstructure-interaction

in soft clay." A dissertation submitted in partial satisfaction of the requirements
for the degree of Doctor of Philosophy in Civil Engineering of the University of
California at Berkeley.
Lok, T. M. H., and Pestana, J. M. (2000). “An enhanced hysteretic model for site
response analysis.” Proceedings, 12th World Conference on Earthquake
Engineering. New Zealand.
Luccioni, X. L. (1999). :Numerical development and implementation of a constitutive

model for clays with applications to deformation around a deep excavation. A
dissertation submitted for the partial fulfillment of the Ph.D. requirement.
University of California at Berkeley.
Lysmer J. M. and Kuhlmeyer, A. M. (1969). "Finite dynamic model for infinite media."
Journal of the Engineering Mechanics Division, ASCE, 95(EM4) 859-877.
Martin, G. R., ed. (1994). Proceedings of the 1992 NCEER/SEAOC/BSSC Workshop on

Site Response during Earthquakes and Seismic Code Provisions. University of
Southern California, Los Angeles, november 18-20. National center for
Earthquake Engineering Research Special Publication NCEER-94-SP01, Buffalo,
NY.
Martin, P. P., and Seed, H. B. (1982). "One-dimensional dynamic ground response

analyses." Journal of Geotechnical Engineering Division, ASCE, 108(7), 935-
952.
Matasovic, N. and Vucetic, M. (1993). Seismic Response of Composite Horizontally-

Layered Soil Deposits. Department of Civil Engineering, University of
California, Los Angeles.
Meymand, P. J. (1998). " Shaking Table Scale Model Tests of Nonlinear Soil-Pile-
Superstructure Interaction In Soft Clay. A dissertation submitted in partial
satisfaction of the requirements for the degree of Doctor of Philosophy in Civil
Engineering of the University of California at Berkeley.
Mohraz, B. (1976), "Earthquake response spectra for different geologic conditions

Bulletin of the Seismological Society of America, 66, 915-935.
390
Mróz, Z. (1967). "On the description of anisotropic work-hardening." Journal of Mech.
Phys. Solids, 15, 163-175.
Mróz, Z., Norris, B. A., and Zienkiewicz, O. C. (1978). "Application of an anisotropic

hardening modeling in the analysis of elasto-plastic deformation.", Geotechnique,
29(1), 1-34.
Mróz, Z., Norris, B. A., and Zienkiewicz, O. C. (1981). "An anisotropic critical state
model for soils subjected to cyclic loading." Geotechnique, 31, 451-469.
Newmark, N. M., Rosenblueth, E. (1971). Fundamentals of Earthquake Engineering.

Prentice-Hall, Inc. Inglewood Cliffs, NJ.
PEER (1998). Workshop meeting of Task 3: Characterization of Site Response, Joint

PEER-PG&E Research group. Held in San Francisco, April.
Pestana, J. M. (1994). "A unified constitutive model for clays and sands." A dissertation
submitted in partial fulfillment of the requirements for the degree of Doctor of
Science in Civil Engineering. Massachusetts Institute of Technology.
Pestana, J. M. and Whittle, A. J. (1999). "Formulation of a unified constitutive model for

clays and sands." International Journal of Numerical Analysis Methods in
Geomechanics, 23, 1215-1243.
Prevost, J. H. (1977). "Mathematical modeling of monotonic and cyclic undrained clay

behavior." International Journal of Numerical Analysis Methods in
Prevost, J. H. (1989). "DYNA1D: A computer program for nonlinear site response

analysis, technical documentation." Technical Report No. NCR-89-0025,
National Center of Earthquake Engineering Research, Sunny at Buffalo, NY.
Pyke, R. M. (1979). "Nonlinear models for irregular cyclic loadings." Journal of

Geotechnical Engineering Division, ASCE, 105(GT6) 715-726.
Pyke, R. M. (1992). "TESS: A computer program for nonlinear ground response

analyses." TAGA Engin. Systems and Software, Lafayette, CA.
Ramberg, W. Osgood, W. T. (1943). "Description of stress-strain curves by three

parameters." Tech. Note 902, National Advisory Committee of Aeronautics.
Rathje, E., Idriss, I. M., Rathje, E., Somerville, P. (2000). "Strong-ground motions and
Site Effects." To appear in Earthquake Spectra in a special publication on the
1999 Kocaeli Earthqake.
391
Rosenblueth, E.and Herrera, I. (1964). "On a kind of hysteretic damping." Journal of the
Engineering Mechanics Division, ASCE, 90(EM4) 37-47.
ROSRINE (1998). URL: http://rccg03.usc.edu/rosrine/. Resolution of Site Response

Issues in the Northridge Earthquake.
Sadigh, K., Chang, C. Y., Egan, J. A., Makdisi, F., and Youngs, R. R. (1997).
"Attenuation Relationships for Shallow Crustal Earthquakes Based on California
Strong Motion Data." Seismological Research Letters, Vol. 68(1), pp. 180-189.
Sadigh, K., Chang, C. Y., Egan, J. A., Makdisi, F., and Youngs, R. R. (1997).
"Attenuation relationships for shallow crustal earthquakes based on California
strong motion data." Seismological Research Letters, Vol. 68(1), 180-189.
Sadigh, K., Chang, C-Y, Abrahamson, N. A., Chiou, S. J., and Power, M. (1993).
"Specification of Long Period Motions: Updated Attenuation Relations for Rock
Site Conditions and Adjustment Factors for Near-Fault Effects." Proceedings,
ATC 17-1, pp. 59-70.
SAS Institute, Inc. (1995). JMP User's Guide, Version 3.1. SAS Institute Inc. Cary, North
Carolina.
Sasani, M. and Bertero, V. V. "Importance of severe pulse-type ground motions in

performance-based engineering: Historical and critical review." Proceedings, 12th
World Conference on Earthquake Engineering, New Zealand.
Schnabel, P. B., Lysmer, J., and Seed, H. B. (1972). "SHAKE: A computer program for
earthquake response analysis of horizontally-layered sites," Report No. EERC-
72/12. Earthquake Engineering Research Center, University of California at
Berkeley, Berkeley, CA.
Schofield, A. N. and Wroth, C. P. (1968). Critical State Soil Mechanics. McGraw-Hill,

London.
Scott, R. F. (1985). "Plasticity and constitutive relations in soil mechanics." Journal of

Geotechnical Engineering, ASCE 111(5) 563-605.
Seed, H. B. and Idriss, I. M. (1969). "The influence of soil conditions on ground motions
during earthquakes." Journal of the Soil Mechanics and Foundation Engineering
Division, ASCE, No. 94, 93-137.
Seed, H. B., and Alonso, J. L. (1974). "Soil-Structure Interaction Effects in the Caracas
Earthquake of 1967." Report No. 73-23, Earthquake Engineering Research Center,
Engineering Department, University of California, Berkeley, California.
392
Seed, H. B., and Idriss, I. M. (1970). "Soil Moduli and Damping Factors for Dynamic
Response Analyses," Report No. EERC 70-10, Earthquake Engineering Research
Center, University of California, Berkeley, California.
Seed, H. B., and Idriss, I. M. (1982). "Ground Motions and Soil Liquefaction during
Earthquakes." In Engineering Monographs on Earthquake Criteria, Structural
Design, and Strong Motion Records, V. 5., Earthquake Engineering Research
Institute.
Seed, H. B., and Idriss, I. M. (1983). "Ground motions and soil liquefaction during
earthquakes." EERI Monograph Series, Earthquake Engineering Research
Institute, Berkeley, CA.
Seed, H. B., Romo, M. P., Sun, J. J., Lysmer, J. (1987). "Relationships Between Soil
Conditions and Earthquake Ground Motions in Mexico City in the Earthquake of
September 19, 1985." Report No. UCB/EERC-87/15, Earthquake Engineering
Research Center, College of Engineering, University of California, Berkeley,
California.
Seed, H. B., Ugas, C., and Lysmer, J., (1976). "Site dependent spectra for earthquake-
resistant design." Bulletin of the Seismological Society of America, 66, 221-244.
Seed, H. B., Wong, R. T., Idriss, I. M., and Tokimatsu, K. (1984). "Moduli and dynamic
factors for dynamic analyses of cohesionless soils." Earthquake Engineering
Research Center, University of California at Berkeley, UBC/EERC-84/14.
Seed, R. B., Dickenson, S. E., and Mok, C. M. (1991). "Seismic Response Analysis of
Soft and Deep Cohesive Sites: A Brief Summary of Recent Findings."
Proceedings, CALTRANS First Annual Seismic Response Workshop,
Sacramento, California, December 3-4.
Sezawa and Kanai (1932). "Possibility of free oscillations of strata excited by seismic
waves." Bulletin of the Earthquake Research Institute, 10:273.
Shannon and Wilson, Inc. and Agbabian Associates (1980). "Geotechnical Data From
Accelerograph Stations Investigated during the Period 1975-1979, Summary
Report." Prepared for U.S. Nuclear Regulatory Commission, NUREG/CR-1643.
Silva, W. (1998). Surface geology based strong motion amplification factors for the San
Francisco Bay area. Prepared by Pacific Engineering and Analysis, El Cerrito,
CA.
Silva, W. J. (1988). ""Soil response to earthquake ground motion." EPRI Report NP-
5754, Electric Power Research Institute, Palo Alto, CA.
393
Silva, W. J. (1998). Ground Motion Database. Personal Communication.
Silva, W. J. and Stark, C. L. (1992). "Source, path, and site ground motion model for the
1989 M 6.9 Loma Prieta earthquake, CDMG draft final report.
Silva, W. J., Turcotte, T., and Moriwaki, Y. (1988). Soil Response to Earthquake Ground
Motion. Electric Power Research Institute, RP-2556-07.
Silverstein, B. L. (1979). "Geologic Description of Selected Strong-Motion

Accelerograph Sites, Part III." USGS Open File Report No. 79-1619.
Simo, J. C. and Hughes, P. J. R. (1998). Computational Inelasticity. Interdisciplinary

Applied Mathematics. (IAM) Springer.
Simo, J. C., and Govindjee, S. (1991). “Nonlinear B-stability and symmetry preserving
return mapping algorithms for plasticity and viscoplasticity,” International
Journal for Numerical and Analytical Methods in Engineering, 31, 151-176.
Simo, J. C., and Hughes, T. J. R. (1998). “Computational Inelasticity.” Interdisciplinary

applied mathematics, v. 7. Springer, New York.
Simo, J. C., and Taylor, R. L. (1986). “Return mapping algorithm for plane stress
elastoplasticity.” International Journal for Numerical Methods in Engineering 22
649-670.
Simo, J.C., and Taylor, R.L. (1985). “A consistent return mapping algorithm for plane
stress elastoplasticity.” UCB/SESM-1985/04, Department of Civil Engineering,
University of California, Berkeley.
Somerville, P. G. (1997). "Engineering characteristics of near fault ground motion."

SMIP 97, Seminar on Utilization of Strong Motion Data: Los Angeles, CA.
Somerville, P. G. (1998). "Development of an improved ground motion representation

for near-fault ground motions." SMIP 98, Seminar on Utilization of Strong
Motion Data: Oakland, CA.
Somerville, P. G., Irikura, K., Graves, R., Sawada, S., Wald, D., Abrahamson, N.,
Iwasaki, Y., Kagawa, T., Smith, N., and Kowada, A. (1999). "Characterizing
Crustal Earthquake Slip Models for the Prediction of Strong Ground Motion."
Seismological Research Letters, 70(1), 59-80.
Somerville, P. G., Smith, N. F., Graves, R. W., and Abrahamson, N. A. (1997).

"Modification of Empirical Strong Ground Motion Attenuation Relations to
Include the Amplitude and Duration Effects of Rupture Directivity."
Seismological Research Letters, Vol. 68(1), pp. 199-222.
394
Spudich, P., Fletcher, J. B., Hellwev, M., Boatwright, J., Sullivan, C., Joyner, W.B.,
Hanks, T. C., Boore, D. M., McGarr, A., Baker, L. M., and Lindh, A. G. (1997).
"SEA96 – A new predictive relation for earthquake ground motions in extensional
tectonic regimes." Seismological Research Letters, Vol. 68(1), 190-198.
Stewart, J. P., and Stewart, A. F. (1997). "Analysis of Soil-Structure Interaction Effects

on Building Response from Earthquake Strong Motion Recordings at 58 Sites."
Report No. UCB/EERC-97/01, Earthquake Engineering Research Center,
University of California at Berkeley, Berkeley, California.
Stewart, J.P., Bray, J. D., Seed, R.B., and Sitar, N. (1994). "Preliminary Report on the
Principal Geotechnical Aspects of the January 17, 1994 Northridge Earthquake,"
Report No. UCB/EERC 94-08, Earthquake Engineering Research Center,
University of California, Berkeley, California.
Streeter, B. L., Wylie, E. B., and Richart, F. E. (1974). "Soil motions computations by
characteristics methods." Journal of the Geotechnical Engineering Division,
ASCE, 100(GT3) 247-263.
Su, F, Koyanagi, S., Zeng, Y. Mayeda, K. Teng, T., and Aki, K. (1990). "A recursive
stochastic inversion of site effects using coda waves." EOS, Transactions of the
American Geophysical Union, 71, 1475.
Sun, J. I., Golesorkhi, R., and Seed, H. B. (1988). "Dynamic Moduli and Damping Ratios
for Cohesive Soils," Report No. UCB/EERC-88/15, Earthquake Engineering
Research Center, College of Engineering, University of California, Berkeley,
California.
Sun, J. I., Golesorkhi, R., and Seed, H. B. (1988). "Dynamic moduli and damping rations
for cohesive soils." ReportNo. UCB/EERC-88/15, Department ofCivil
Engineering, University of California, Berkeley.
Taylor, P. W. and Larkin, T. J. (1978). "Seismic response of nonlinear soil media."

Journal of the Geotechnical Engineering Division, ASCE, 104(GT3) 369-183.
Taylor, R. L. (1998). FEAP: A Finite Element Analysis Program. Department of Civil

and Environmental Engineering, University of California at Berkeley, Berkeley,
CA.
Taylor, R.L. (1998) “FEAP – A Finite Element Analysis Program,” Department of Civil
Engineering, University of California, Berkeley.
Thiel, C. C., and Schneider, J. F. (1993). "Investigation of Thirty-three Loma Prieta

Earthquake Strong Motion Recording Sites." California Universities for Research
in Strong Motion Seismograph Sites in California.
395
Thomson, W. T. (1950). "Transmission of elastic waves through a stratified solid
medium." Journal of Applied Physics.. 21.
Toro, G. R., Abrahamson, N. A., and Schneider, J. F. (1997). "Model of strong ground
motions from earthquakes in Central and Eastern North America: Best Estimates
and Uncertainties." Seismological Research Letters, Vol. 68(1), 41-57.
Trifunac, M. D., and Todorovska, M. I. (1996). Nonlinear Soil Response - 1994

Northridge, California, Earthquake." Journal of Geotechnical Engineering, ASCE,
122 (9), 725-735.
Vucetic, M. and Doroudian, M. (1995). "The Characteristics of Earthquake Ground

Motions for Seismic Design. Task H-5: Geotechnical Site Database for Southern
California Digitized Boring Logs." Southern California Earthquake Center,
University of Southern California, Los Angeles, California.
Vucetic, M., and Dobry, R. (1991). "Effect of Soil Plasticity on Cyclic Response."
Journal of Geotechnical Engineering, ASCE, Vol. 117(1).
Wang, Z. L., Dafalias, Y. F. and Shen, C. K. (1990). "Bounding surface hypo-plasticity

model for sands." Journal of Engineering Mechanics, ASCE, 116(5) 993-1001.
Whitman, R., ed. (1992). Proceedings of the Site Effects Workshop, October 24.
ReportNCEER-92-0006, National Center for Earthquake Engineering Research,
Buffalo, NY.
Whittle, A. J., and Kavvadas, M. J. (1994). "Formulation of MIT-E3, constitutive model

for overconsolidated clays." Journal Geotechnical Engineering, ASCE 120(10)
173-198.
Wills, C. J. (1998). "Differences in Shear-Wave Velocity Due to Measurement Methods:

A Cautionary Note." Seismological Research Letters, Vol. 69(3), pp. 216-221.
Youngs, R. R. (1993). "Soil amplification and vertical and horizontal ratios for analysis
of strong motion data from active tectonic regions, Appendix 2C in Guidelines
for Determining Design Basis Ground Motions (see EPRI 1993).
Youngs, R. R., Chiou, S. –J., Silva, W. J., Humphrey, J. R. (1997). "Strong ground
motion attenuation relationships for subduction zone earthquakes." Seismological
Research Letters, Vol. 68(1), 58-73.
Zeghal, M., Elgamal, A. W., Tang, H. T., and Stepp, J. C. (1995). "Lotung downhole
array. II: Evaluation of soil nonlinear properties." Journal of Geotechnical
Engineering, ASCE, 121(4), 363-378.
396
Zienkiewicz, O. C. and Taylor, R. L. (1989). The Finite Element Method. Volumes I and
II. McGraw-Hill Book Company, London.
Zienkiewicz, O. C., and R. E. Newton. "Coupled vibrations of a structure submerged in a

compressible fluid" Proceedings International Symposium on Finite Element
Techniques, Stuttgart 1-15, May 1969.
Zienkiewicz, O. C., Chan, A. H. C., Pastor, M., Schrefler, B. A., Shiomi, T. (1999).
Computational Geomechanics with Special Reference to Earthquake Engineering.
John Wiley and Sons, New York, NY.
Zienkiewicz, O. C., Chang, C. T., and Hinton, E. (1978). "Nonlinear seismic response
and liquefaction." International Journal of Numerical Analysis Methods in
397
APPENDIX A
List of Ground Motion Sites with Corresponding Site Classification
398
Table A-1. Ground motion stations showing site classification, Northridge Earthquake.
(1) (2) (5)
Station Name Agency Sta. # Surface Seed & UBC 97 This Quality Depth IR Conrey's Source
(3)
Geology Dickenson (Table A-3) Study (m) depth
(Table A-4) (Table 1) (ft)
Alhambra - Fremont School CDMG 24461 Holocene B1? C C3? Fair >9 0? SCEC
Anacapa Island # CDMG 25169 Miocene AB2? D? C2? Poor - - - Geol.
Anaheim - W Ball Rd USC 90088 Holocene C3? D D1S? Fair >23.5 - 7000 SCEC
Anaverde Valley - City R # CDMG 24576 Holocene AB2? C? C?2 Poor - - - Geol.
Antelope Buttes # CDMG 24310 Jurassic- AB1 B C?1 Poor - - - Geol. ,T.
Cretaceous
Arcadia - Arcadia Av USC 90099 Holocene C3? D D?1S? Fair >40 - - SCEC
399
Arcadia - Campus Dr. USC 90093 Holocene C3? D D?1S? Fair 19 1.8 - SCEC
Arleta - Nordhoff Fire Sta # CDMG 24087 Holocene C3 D D1S Good >150 - - SCEC,ROSRINE
Baldwin Park - N. Holly Ave USC 90069 Holocene AB2 C C2 Good 21.5 5.5 - SCEC
Bell Gardens - Jaboneria USC 90094 Holocene C3 D D1S? Fair >29 - 6000 SCEC
Beverly Hills - 12520 Mulhol USC 90014 Miocene AB1 C C1 Fair 22(w?) 3.3 - SCEC
Beverly Hills - 14145 Mulhol USC 90013 Miocene AB? D C1 Fair 24(w?) 4.0 - SCEC
Big Tujunga, Angeles Nat F USC 90061 Mesozoic AB2 C C1 Fair 21.5(w) 3.3 - SCEC
Brea - S. Flower Ave. USC 90087 Pleistocene B2? C2? D? D?2C Fair >35 - - SCEC
Brentwood V.A. Hospital USGS 638 Pleistocene C2? D? D?2S? Poor >30? - - SCEC, Geol.
(1) (2) (5)
(3)
(1) (2) (5)
(3)
Buena Park - La Palma USC 90086 Holocene C3? D D1S (F?) Poor >30 - 7000 SCEC
Burbank - Howard Rd. USC 90059 Cretaceous A1 B C?1 Good 6 - 9 (w) 3.0 - SCEC
+
Camarillo CDMG 25282 Holocene C2 D D1C? Fair >30 - - SCEC
Canoga Park - Topanga Can USC 90053 Holocene C2 D C3 Good 50 (to soft 2.2 - SCEC, G.e.a.
rock)
Canyon Country - W Lost USC 90057 Holocene B1? C3? D? D?1S Poor >24 - - SCEC
Cany AB1?
Carson - Catskill Ave USC 90040 Pleistocene C3 D D2S Fair >22 - 3000 SCEC
400
Carson - Water St. USC 90081 Holocene F F F Fair 220 4.0 3000 SCEC,ROSRINE
Castaic - Old Ridge Route # CDMG 24278 Miocene AB2 C C?1 Fair 8 (w) - - USGS, D.e.a.,
CSMIP
Compton - Castlegate St USC 90078 Holocene F F F Poor - - 5000 SCEC
Covina - S. Grand Ave. USC 90068 Pleistocene B D C3 Fair 25.5 3.0 - SCEC
Covina - W. Badillo USC 90070 Holocene B D C3 Fair 23 2.2 - SCEC
Downey - Birchdale USC 90079 Holocene C3 D D1S Good 120 3.0 6000 SCEC
Downey - Co Maint Bldg # CDMG 14368 Holocene B1? C2? D D1S? Fair >120? - 9000 SCEC, CSMIP,
C3? G.
Duarte - Mel Canyon Rd. USC 90067 Holocene? AB2 (F?) C (F?) C2? Good 17 8.0 - SCEC
(1) (2) (5)
(3)
El Monte - Fairview Av USC 90066 Holocene F F F Fair 31 2.8 - SCEC
Elizabeth Lake # CDMG 24575 Holocene AB2? D? C? C?2 Poor - - - Geol.
Featherly Park - Pk Maint CDMG 13122 Holocene B1? C3? D? D?1S? Poor - - - G.
Bldg #
Garden Grove - Santa Rita USC 90085 Holocene C2 D D?1C Fair >30 - 8000 SCEC
Glendale - Las Palmas USC 90063 Pleistocene AB2 D C3?? Fair 33 2.6 - SCEC
Glendora - N. Oakbank USC 90065 Holocene B C C3? Fair 41 2.0 - SCEC
Hacienda Hts - Colima Rd USC 90073 Holocene B2 D C3?? Fair 34? 2.0 - SCEC
401
Hemet - Ryan Airfield # CDMG 13660 Holocene C2? C3? D? D1C? Poor - - - Geol.
Hollywood - Willoughby Ave USC 90018 Holocene C2 D D2C Good 100 - 1000 SCEC
Huntington Bch - Waikiki USC 90083 Holocene C2 D D1C? Check - - 4000 Geol.
Huntington Beach - Lake St CDMG 13197 Holocene- C2? D? D2S? Poor - - 3000 G.
# Pleistocene
Inglewood - Union Oil # CDMG 14196 Pleistocene C3 D? D2S? Poor - - 4000 G.
Jensen Filter Plant # USGS 655 Holocene F F F Good >93 - - G.e.a., Stew.
LA - 116th St School # CDMG 14403 Pleistocene C2 C? D2C Poor 152 - - SCEC, D&L,G
LA - Baldwin Hills # CDMG 24157 Pleistocene C2 D D2C? Good >85 - 2000 SCEC,G,
D&L,ROSRINE
LA - Centinela St USC 90054 Pleistocene AB2 D C3 Fair 22 3.6 5000 SCEC
(1) (2) (5)
(3)
GeologyDickenson (Table A-3) Study (m) depth
LA - Century City CC North CDMG 24389 Pleistocene C2 D? D2C? Fair >120 2.6 - S&W,G
#
LA - Chalon Rd USC 90015 Jurassic? AB1 C C1 Good 15 (w) >2.0 - SCEC
LA - City Terrace # CDMG 24592 Pliocene AB1? A1? B? C? C?1 Fair <10? - 0 CSMIP,
SCEC,S&S
LA - Cypress Ave USC 90033 Holocene AB2 C C2 Fair 21 3.5 - SCEC
LA - E Vernon Ave USC 90025 Holocene C2?C3? D D1C? Fair >30 - 3000 SCEC
LA - Fletcher Dr USC 90034 Holocene C3?B1? D D?1S Fair - - - SCEC
LA - Hollywood Stor FF # CDMG 24303 Holocene C2 D D1C Good 103.6 2.8 1000 USGS, Chang
402
LA - N Faring Rd USC 90016 Jurassic? AB1 D C1 Good 20 (w)
LA - N Westmoreland USC 90021 Holocene AB2 C C2 Good 22 3.8 - SCEC
LA - N. Figueroa St. USC 90032 Holocene AB2 C C2 Fair 24 2.0 - SCEC
LA - Obregon Park # CDMG 24400 Holocene F? F? F? Poor - <1000 Geol., SCEC
LA - Pico & Sentous # CDMG 24612 Holocene B1? D D1S Fair 120 - <1000 SCEC, D&L
LA - S Grand Ave USC 90022 Holocene C3 D D?1S Good >24.5? - 3000 SCEC
LA - S. Vermont Ave USC 90096 Holocene C2 D D1C Good >150 - 2000 SCEC, S&S
LA - Saturn St USC 90091 Holocene F F F Good - - 2000 SCEC
LA - Temple & Hope # CDMG 24611 Miocene AB1 D? C1 Good 64 (w) 2.4 0 CSMIP,
D&L,S&S
(1) (2) (5)
(3)
LA - UCLA Grounds CDMG 24688 Pleistocene AB2 C C2 Good 20 (w) 2.0 - S&S
LA - Univ. Hospital # CDMG 24605 Miocene AB1 D? C1 Good 20? - - S&S
LA - W 15th St USC 90020 Pleistocene C3 D D2S Fair - - 1000 SCEC
LA - Wonderland Ave USC 90017 Cretaceous A1 B B Good 4 (w) 2.7 - SCEC
La Crescenta - New York USC 90060 Pleistocene AB2 D C3 Good 30 2.0 - SCEC
LA Dam USGS 0 Pliocene AB1 C C1 Good 0 - - G.e.a.
La Habra - Briarcliff USC 90074 Pleistocene C2?C3? D D2C? Fair - - 2000 SCEC
403
La Puente - 504 Rimgrove USC 90072 Holocene B D D?1S? Fair - - - SCEC

Ave
Lake Hughes #1 # CDMG 24271 Holocene C2 C D2C Good 260 - - G., USGS, S&W
Lake Hughes #12A # CDMG 24607 Paleocene A1 C C?1 Good 10 5 - USGS, S&W, G.
Lake Hughes #4 - Camp CDMG 24469 Mesozoic A1 B B Good 5 (w) - - G., S&W
Mend #
Lake Hughes #4B - Camp CDMG 24523 Mesozoic A2 B B Good 6 (w) - - G., S&W
Mend #
Lake Hughes #9 # USGS 127 Precambria A1 B C?1 Good 8 (w) - - USGS, S&W, G.
n
Lakewood - Del Amo Blvd USC 90084 Holocene C3? D D1S? Fair >25 - 8000 SCEC
Lancaster - Fox Airfield CDMG 24475 Holocene C3 D D1S Good - - - S&S

Grnds
Lawndale - Osage Ave USC 90045 Pleistocene C2? D D2C? Poor - - 3--- SCEC
(1) (2) (5)
(3)
LB - City Hall # CDMG 14560 Pleistocene C3 D? D2S? Fair - - 3000 SCEC, Geol.
LB - Rancho Los Cerritos # CDMG 14242 Pleistocene C3? D? D2S? Poor - - 5000 SCEC
Leona Valley #1 # CDMG 24305 Pleistocene A1? B? B Fair - - - Geol., CSMIP
Leona Valley #2 # CDMG 24306 Holocene AB2? C? C2?? Poor - - - Geol.
Leona Valley #3 # CDMG 24307 Holocene? A1? B? B? Poor - - - CSMIP
Leona Valley #4 # CDMG 24308 Pleistocene AB2? C? C1? Poor - - - CSMIP, Geol.
Leona Valley #5 - Ritter # CDMG 24055 Holocene AB2? C? C2? Poor - - - Geol.
404
Leona Valley #6 # CDMG 24309 Holocene AB2? C? C?1 Poor - - - Geol.
Littlerock - Brainard Can # CDMG 23595 Mesozoic A1? B? B? Poor - - - Geol., CSMIP
Malibu - Point Dume Sch # CDMG 24396 Pleis? AB1? C? C1? Poor - - - Geol., SCEC
Miocene?
Manhattan Beach - USC 90046 Holocene C3? D D1S Fair >22 - 3000 Geol.
Manhattan
Mojave - Hwys 14 & 58 # CDMG 34093 D? D?1?S? Poor
Mojave - Oak Creek Canyon CDMG 34237 D? D?1?S? Poor

#
Montebello - Bluff Rd. USC 90011 Holocene F F F Good >21.5 - 1000? SCEC
Moorpark - Fire Sta # CDMG 24283 Holocene C3-2? C?D? D1S Poor 150 - - Geol.
Mt Baldy - Elementary Sch # CDMG 23572 Holo. near AB?A? B?C? C?1 Fair - - - Geol., CSMIP
Cret.
(1) (2) (5)
(3)
Mt Wilson - CIT Seis Sta # CDMG 24399 Mesozoic AB2 B C1 Fair 23 (w) 11.5 - Geol., CSMIP,
D&L,G.,S&S
N. Hollywood - Coldwater USC 90009 Holocene AB2 C C2 Good 17 4.0 - SCEC
Can
Neenach - Sacatara Ck # CDMG 24586 Holocene C? C? D1S? Poor
Newhall - Fire Sta # CDMG 24279 Holocene B1 D C3 Good 55 2.8 - ROSRINE
Newhall - W. Pico Canyon USC 90056 Holocene AB2? D? C2? Check - - - Geol.
Rd.
Newport Bch - Irvine Ave. CDMG 13160 Pleistocene B1? C?D? D?2S? Poor - - - Geol.
F.S. #
Newport Bch - Newp & CDMG 13610 Miocene AB1 C? C1 Fair - - - Geol.,
Coast # CSMIP,S&S
405
Northridge - 17645 Saticoy USC 90003 Holocene C3 D D1C Good 81 2.5 - SCEC
St
Pacific Palisades - Sunset USC 90049 Holocene AB2 D C2 Good 21 4.3 - SCEC
Blvd
Pacoima Dam (downstr) # CDMG 24207 Mesozoic A1 B? B Fair 0 3 - CSMIP, Geol.
Pacoima Dam (upper left) # CDMG 24207 Mesozoic A1 B B Fair 0 3 - D.e.a., G.
Pacoima Kagel Canyon # CDMG 24088 Miocene AB1 C C1 Good 20 (w) 2 - G., ROSRINE
Palmdale - Hwy 14 & CDMG 24521 Holocene- B1? C C3? Fair >30,<60 Hig - S&S, SCEC
Palmdale # Pleistocene ? h?
Pardee - SCE 0
Pasadena - N Sierra Madre USC 90095 Holocene - AB2 D C3? Good 34 2.0 SCEC
Pleistocene
Phelan - Wilson Ranch # CDMG 23597 Holocene C2?C3? D? D1S? Poor - - - Geol.
(1) (2) (5)
(3)
Playa Del Rey - Saran USC 90047 Holocene F F F Fair - - 3000 SCEC
Point Mugu - Laguna Peak # CDMG 25148 Miocene AB1? C? C?1 Poor - - - Geol.
Port Hueneme - Naval Lab. CDMG 24281 Holocene C3 D D1S Fair >30 - - Geol., SCEC
#
Rancho Cucamonga - Deer CDMG 23598 Mesozoic A1? B? B? Poor - - - Geol., CSMIP
Can #
Rancho Palos Verdes - CDMG 14404 Miocene A1? B? C1 Fair - - 0 Geol.
Hawth #
Rancho Palos Verdes - USC 90044 Miocene AB1 D C1? Poor 10.5 2.0 - Geol. ,SCEC
Luconia
Rinaldi Receiving Sta # DWP 77 Holocene AB2 C C2 Fair 40-Pico 2.0 - G.e.a.
(thin) 250 3.0
406
Riverside - Airport # CDMG 13123 Pleistocene AB2? C C2 Poor - - - Geol.
Rolling Hills Est-Rancho CDMG 14405 Pliocene AB1? C? C1 Poor - - - Geol.

Vista
Rosamond - Airport # CDMG 24092 Holocene B? D? D?1S? Poor - - - Geol.
San Bernardino - CSUSB Gr CDMG 23672 Holocene B1?C3? D? D?1S? Poor - - - Geol. ,S&S
#
San Bernardino - E & Hosp CDMG 23542 Holocene C3? D D?1S Poor - - - CSMIP, SCEC
#
San Gabriel - E. Grand Ave. USC 90019 Holocene A1 C C?2 Good 6.5 >10 0 SCEC
San Jacinto - CDF Fire Sta # CDMG 12673 Holocene C2?C3? D? D1C? Poor - - - Geol.
San Marino, SW Academy # CDMG 24401 Holocene C3? C?D? C3?? Poor >30 - - D&L,USGS,
D&L,G
San Pedro - Palos Verdes # CDMG 14159 Miocene AB1? C? C?1 Poor - - 0 Geol.
(1) (2) (5)
(3)
Sandberg - Bald Mtn # CDMG 24644 Mesozoic AB1? B?C? C?1 Poor - - - Geol.
Santa Barbara - UCSB CDMG 25091 Holocene AB1 C? C1 Good >5 - - SCEC
Goleta #
Santa Fe Spr - E. Joslin USC 90077 Holocene C? D C3?? Check 25.5 3.0 - SCEC
Santa Monica City Hall # CDMG 24538 Pleistocene B1 D D2C Good 183+ 6.0 2000 Chang, SCEC
Santa Susana Ground # USGS 5108 Cretaceous A1?AB1? B?C? C1 Fair 2 - - S&S
Seal Beach - Office Bldg # CDMG 14578 Pleistocene C3? D? D2S Fair - - 3000 Geol., S&S
Sepulveda VA # USGS 637 Pleistocene C2 C D2C Good >76 1.8 - G.e.a.

407
Simi Valley - Katherine Rd USC 90055 Holocene A1 C C2 Good 12.5 4.0 - G.e.a.
Stone Canyon # MWD 78
Sun Valley - Roscoe Blvd USC 90006 Holocene AB2 C C2 Fair - - - SCEC
Sunland - Mt Gleason Ave USC 90058 Holocene AB2 C C3 Good 24.5 2.7 - SCEC
Sylmar - Converter Sta # DWP 74 Holocene C3 D D1S Fair >92 2.9 - G.e.a.
Sylmar - Converter Sta East DWP 75 Holocene C3? D? D1S? Poor - - - G.e.a.
#
Sylmar - Olive View Med FF CDMG 24514 Holocene C2 C D1C? Good 79 2.5 - G.e.a.
#
Tarzana, Cedar Hill # CDMG 24436 Pliocene AB2 D C2? Fair - - - Geol., CSMIP,
ROSRINE
Terminal Island - S Seaside USC 90082 Holocene C2 D D1C Fair > 17 - 1000 SCEC
(1) (2) (5)
Station Name Agency Sta. # Surface
Seed & UBC 97 This Quality Depth IR Conrey's Source
(3)
Dickenson (Table A-3) Study
Geology (m) depth
Topanga - Fire Sta # USGS 5081 Holo. near AB? C? C2? Poor - - - Geol.
Mioc.
Tustin - E. Sycamore USC 90089 Holocene F F F Fair - - 1000 SCEC
Vasquez Rocks Park # CDMG 24047 Pleistocene A?AB? C?B? B? Poor - - - Geol., CSMIP
Ventura - Harbor & CDMG 25340 Holocene C2 D D1C Fair >24 - - S&S
California
Villa Park - Serrano Ave USC 90090 Pleistocene A1 C C?2 Good 7 - - SCEC
(Shallow)
West Covina - S. Orange USC 90071 Holocene AB2 D C2 Good 16 - - SCEC
Ave
Whittier - S. Alta Dr USC 90075 Pleistocene A1 C C2 Good 7 - - SCEC
408
Wrightwood - Jackson Flat # CDMG 23590 Mesozoic A1? B? B? Poor - - - Geol.
Wrightwood - Nielson Ranch CDMG 23573 Holocene AB2? C?D? C2?? Poor - - - Geol.
#
Wrightwood - Swarthout # CDMG 23574 Holocene AB2? C?D? C2?? Poor - - - Geol.
Malibu Canyon, Monte Nido USGS 5080 Miocene A?AB? B? B? Poor - - - Geol.
(4)
Fire
Point Mugu - Naval Air CDMG 25147 Holocene C? D? D1C? Poor - - - Geol.
(4)
Station
Malibu, W. Pacific Coast USC 90051 Pleistocene AB2 C C2 Good 14 4.0 - SCEC
(4)
Hwy.
Rancho Cucamonga - L&J CDMG 23497 Holocene C3 D D1S Good 210 - - S&S, G.
(4)
(1) Depth to bedrock obtained from boring log. If nothing is specified it refers to depth of soil cover over weathered rock.
(2) Estimated Impedance Ratio.
(3) Depth to base of Pliocene deposits (Conrey 1967).
(4) Ground motion sites added to the Walter Silva Database.
(5) Abbreviated source for geotechnical and geological data. Corresponds to the following references:
Chang - Chang et. al. (1997).
CSMIP - CSMIP (1992).
D.e.a. - Duke et al. (1971).
D&L - Duke and Leeds (1962).
G - Geomatrix Consultants (1993).
G e.a. - Gibbs et al. (1996).
Geol. - Local geological maps (CDMG and Dibblee).
ROSRINE - Internet Site (ROSRINE 1998).
S&S - Stewart and Stewart (1997).
S&W - Shannon and Wilson (1980).
SCEC - Vucetic and Dourodian (1995).
Stew - Stewart et. al. (1994)
T. - Trifunac and Todorovska (1996).
USGS - Fumal, Gibbs, and Roth (1981, 1982, and 1984).
409
Table A-2. Ground motion stations showing site classification, Loma Prieta Earthquake.
(1) (2) (4)
Station Name Agency Sta. # Surface Seed & UBC 97 This Quality Depth IR Source
Geology Dickenson (Table A-3) Study (m)
(Table A-4) (Table 1)
Agnews State Hospital CDMG 57066 Holocene C2 D D1C Good 300? - T&S,G
Alameda NAS Navy Holocene C4-D1-E2 E E1 Good 141 - T&S
Anderson Dam USGS 1652 Pleistocene B1? C C3? Good > 50 ? - G, 94-552
(Downstream)
APEEL 10 - Skyline CDMG 58373 Eocene AB1 C C1 Good 17(w) 1.7 79-1619,G
APEEL 2 - Redwood City USGS 1002 Holocene C4-D1-E2 E E1 Good 84.7 8.6 93-376,G,79-
1619
APEEL 2E Hayward Muir CDMG 58393 Pleistocene C2 D D2C Good 152 - 79-1619
School
APEEL 3E Hayward CDMG 58219 Pliocene AB2 C C?1? Good 12 3.6 79-1619,T&S
411
CSUH
APEEL 7 - Pulgas CDMG 58378 Eocene AB1 C C1 Good 16 (w) 1.3 G
APEEL 9 - Crystal Spring USGS 1161 Pleistocene B1 C C3?? Good >30 1.3 79-1619,G
Residence
Bear Valley Sta 10, Webb USGS 1479 Holocene - B2 D C2 Good 21 1.8 94-552
(3)
Residence Pleistocene
Bear Valley Sta 12, USGS 1481 Holocene C3 D D1C Good >60 94-552
(3)
Williams
Bear Valley Sta 5, Callens USGS 1474 Holocene B2 C C2 Good 30 2.6 94-552
(3)
Ranch
Bear Valley Sta. 7, USGS PNM Miocene A1 B B Fair 91-311
(3)
Pinnacles (Rhyolite)
Belmont - Envirotech CDMG 58262 Franciscan AB1 C C1 Good 13(w) 1.2 T&S,G
Berkeley LBL CDMG 58471 Cretaceous AB1 C C1 Fair shallow soil 1.3
?
BRAN UCSC 13 AB1? B? B? Poor
(1) (2) (4)
(1) (2) (4)
Capitola CDMG 47125 Holocene B1 D C2?? Check >183? ,16? 2.6 T&S
Corralitos CDMG 57007 Holocene AB2 C C1 Good 5 1.8 G

(soft),32(Hard)
Coyote Lake Dam CDMG 57504 Holocene AB2 C C3? Fair <50? 1.5? G
(Downst)
Foster City - 355 USGS 1515 Holocene C4-D1-E2 E E1 Good 115 High G
Menhaden
Foster City -Redwood CDMG 58375 Holocene C4-D1-E2 E E2 Good 188 93-376
(3)
Shores
Fremont - Emerson Court USGS 1686 Pleistocene B1?C3? D D?2S Good >46 - 94-222, G
412
Fremont - Mission San CDMG 57064 Pleistocene B1 D D?2S Fair 5? 78? - G

Jose
Gilroy - Gavilan Coll. CDMG 47006 Pleistocene AB2 C C2 Good 13 6 S&W, G
Gilroy - Historic Bldg. CDMG 57476 Holocene AB2 C C2?? Poor - - G
Gilroy Array #1 CDMG 47379 Franciscan A1 B B Good 1 2.85 G
Gilroy Array #2 CDMG 47380 Holocene C2 D D1C Good 165 - G,92-387,T&S
Gilroy Array #3 CDMG 47381 Holocene C2 D D1C Good >60, 480? - 92-387,T&S,G
Gilroy Array #4 CDMG 57382 Holocene C2 D D1C Good >30, 800? - G
Gilroy Array #6 CDMG 57383 Eocene- AB1 C C1 Good 1 - G

Paleocene
Gilroy Array #7 CDMG 57425 Pleistocene AB2 D C2 Good 17m 3.5 G, 92-376
Golden Gate Bridge USGS 1678 Franciscan AB1 C? C?1 Poor 5? 1.3
(1) (2) (4)
Halls Valley CDMG 57191 Holocene AB2 D C2?? Fair 17 Soft, 43 5.2 T&S,G
Hard
Hayward - BART Sta CDMG 58498 Pleistocene B2? D? D?2C Poor - - Stewart
Hayward City Hall, ground USGS 1129 Pliocene A1 C B Good 10 94-222

(3)
site (Rhyolite)
Hollister - SAGO Vault USGS 1032 Mesozoic A1 B B Fair 0 2.5 G
Hollister - South & Pine CDMG 47524 Holocene C2 D D1C? Fair > 105? G
Hollister City Hall USGS 1028 Holocene C2? D D1C Good >105 79-1619,G
Hollister Diff. Array USGS 1656 Holocene C2? D D1C? Good >106 79-1619,G
(3)
Larkspur Ferry Terminal USGS 1590 Holocene C4-D1-E2 E E2 Good 27.5 5 94-222
413
LGPC UCSC 16 AB1 C? C?1 Poor
Monterey City Hall CDMG 47377 Mesozoic AB1 C C1? Good 6s, >8(w) - T&S
Oakland - Title & Trust CDMG 58224 Pleistocene C2 D D2C Good 137 - 93-376,T&S
(3)
Oakland Outer Harbor CDMG 58472 Holocene C2 D D1C Good 150 3.8 Chang
Palo Alto - 1900 Embarc. CDMG 58264 Holocene C4-D1-E2 E E1 Good >55 - 93-376
Palo Alto - SLAC Lab USGS 1601 Pleistocene AB2 C C2 Good 12 - 94-222
Piedmont Jr High CDMG 58338 Franciscan A1 B B Good 4 9 T&S
Point Bonita CDMG 58043 Cretaceous A1 B B Good 2 (w) 2 T&S
Richmond City Hall CDMG 58505 Pliocene- C2 D C3?? Good 58 2.5 T&S
Pleist.
SAGO South - Surface CDMG 47189 Mesozoic A1 B B Good 4.5(w) 2.5 T&S
(1) (2) (4)
Salinas - John & Work CDMG 47179 Holocene B?C? D D?1C? Poor - - G
San Francisco, 1295 USGS 1675 Franciscan A1 B B Fair 7 91-311

(3)
Shafter, Fire Station
San Jose - Sta. Teresa CDMG 57563 ? A1 B? B? Poor G.
(3)
Hills
Saratoga - Aloha Ave CDMG 58065 Pliocene AB1? C C2? Poor - - G
Saratoga - W Valley Coll CDMG 58235 Pleistocene AB2? C C2? Poor 4-10? - G
SF - Cliff House CDMG 58132 Franciscan A1?AB1? C?B? B? Poor 0? -
SF - Diamond Heights CDMG 58130 Franciscan AB1 C C1 Good 4 s, 11 (w) 4.7 T&S
SF - Pacific Heights CDMG 58131 Cretaceous A1 B B Good 6-9(w) 1.5 T&S

414
SF - Presidio CDMG 58222 Franciscan AB1 C C1 Good 17.5 1.9 93-376

?
SF - Rincon Hill CDMG 58151 Franciscan A1?AB1? C? B? Poor 5 (w) T&S (may be
wrong boring)
SF - Telegraph Hill CDMG 58133 Franciscan A1?AB1? C? B? Fair 6? - T&S
SF Intern. Airport CDMG 58223 Holocene C4-D1-E2 E E1 Good 134 - T&S,G,92-287
So. San Francisco, Sierra CDMG 58539 Franciscan A1 B B Good 4-5(w) 1.3 T&S
Pt.
Sunnyvale - Colton Ave. USGS 1695 Holocene C2 D D1C? Good >60 - 94-222,G
Sunol Fire St (Calaveras USGS SNF Pleistocene B1 C D?2S Good 35(>48) 2 94-552
(3)
Array)
Treasure Island CDMG 58117 Holocene F F F Check 88 - 92-287
UCSC UCSC 15 AB2 C? C2? Poor
UCSC Lick Observatory CDMG 58135 Paleozoic AB2 C C1 Fair >13 (w) - T&S
(1) (2) (4)
Woodside CDMG 58127 Eocene AB2 C B Good 6 1.6 93-376
Yerba Buena Island CDMG 58163 Franciscan A1 B C1 Good 0? 2.5 92-287
(1) Depth to bedrock obtained from boring log. If no modifier is specified next to the depth value it refers to depth of soil cover over weathered
rock. If (w) is indicated, indicates depth of weathering (depth of weathered rock to slightly weathered or unweathered rock)
(2) Estimated Impedance Ratio.
(3) Ground motion sites added to the Walter Silva Database.
(4) Unless specifically omitted, information for all sites was also obtained from Fumal (1991). All other abbreviated source for geological data
correspond to the following references:
79-1619 (Silverstein 1979).
92-287 - (Gibbs et al. 1992).
93-376 - (Gibbs et al. 1993).
415
94-222 - (Gibbs et al. 1994).

94-552 - (Gibbs and Fumal (1993).
Chang - Chang and Bray (1995).
D&L - Duke and Leeds (1962).
G - Geomatrix Consultants (1993).
Geol. - Local geological maps (CDMG).
S&S - Stewart and Stewart (1997).
S&W - Shannon and Wilson (1980).
Stewart - Stewart, J. P. (personal comm.)
T&S - Thiel and Schneider (1993).
Table A-3. Site categories in the 1997 UBC. Vs is the average shear wave velocity
measured over the upper 100 feet.
SA Hard rock with measured shear wave velocity, Vs > 5000 ft/s.
SB Rock with 2500 ft/s < Vs ≤ 5000 ft/s
SC Very dense soil and soft rock with 1200 ft/s < Vs ≤ 2500 ft/s or with either N >
50 or su ≥ 2000 psf., where N is the average Standard Penetration blowcount
over the upper 100 ft, and su is the average undrained shear strength over the
upper 100 feet.
SD Stiff soil with 600 ft/s ≤ Vs ≤ 1200 ft/s or with 15 ≤ N ≤ 50 or 1000 psf ≤ su ≤
2000 psf.
SE A soil profile with Vs < 600 ft/s or any profile with more than 10 ft. of soft clay
defined as soil with PI > 20, wmc ≥ 40 percent and su < 500 psf.
SF Soils requiring site-specific evaluation.
1. Soils vulnerable to potential failure or collapse under seismic loading such as
liquefiable soils, quick and highly sensitive clays, collapsible weakly
cemented soils.
2. Peats and/or highly organic clays (H > 10 ft of peat and/or highly organic clay
where H = thickness of soil.
3. Very high plasticity clays (H > 25 ft with PI > 75).
4. Very thick soft/medium stiff clays (H > 120 ft).
416
Table A4. Seed et al. (1991) site classification system (from Dickenson 1994).
Site Site
4.1 General Description Site Characteristics1,2
Class Condition
(Ao) Ao Very Hard Rock. Vs (avg.) > 5,000 ft/sec in top 50 ft.
A A1 Competent rock with little or no soil 2,500 ft/sec ≤ Vs (rock) ≤ 5,000 ft/sec,
and/or weathered rock veneer. and Hsoil + weathered rock < 40 ft with Vs > 800
ft/sec (in all but the top few feet3).
AB AB1 Soft, fractured and/or weathered rock. For both AB1 and AB2:
AB2 Stiff, very shallow soil over rock and/or 40 ft ≤ Hsoil + weathered rock ≤ 150 ft, and Vs ≥
weathered rock. 800 ft/sec (in all but the top few feet3).
B1 Deep, primarily cohesionless4 soils. No "Soft Clay" (See Note 5), and
(Hsoil ≤ 300 ft.) Hcohesive soil < 0.2 Hcohesionless soil.
B B2 Medium depth, stiff cohesive soils Hall soils≤ 200 ft., and Vs(cohesive soils)
and/or mix of cohesionless with stiff > 500 ft/sec.
cohesive soils; no "Soft Clay." (See Note 5).
C1 Medium depth, stiff cohesive soils Same as B2 above, except
and/or mix of cohesionless with stiff 0 < Hsoft clay ≤ 10 ft.
cohesive soils; thin layer(s) of "Soft (See Note 5)
Clay."
C2 Deep, stiff cohesive soils and/or mix of Hsoil > 200 ft, and
cohesionless with stiff cohesive soils; no Vs (cohesive soils) > 500 ft/sec.
"Soft Clay."
C3 Very deep, primarily cohesionless soils. Same as B1 above, except
Hsoil > 300 ft.
C4 Soft, cohesive soil at small to moderate 10 ft. ≤ Hsoft clay ≤ 100 ft, and
levels of shaking. Amax,rock ≤ 0.25 g
D D1 Soft, cohesive soil at medium to strong 10 ft. ≤ Hsoft clay ≤ 100 ft, and
levels of shaking. 0.25 g ≤ Amax,rock ≤ 0.45 g, or
[0.25 g ≤ Amax,rock ≤ 0.45 g and M ≤
7.25]
E1 Very deep, soft cohesive soil. Hsoft clay ≥ 100 ft. (See Note 5)
E2 Soft, cohesive soil and very strong Hsoft clay ≥ 10 ft., and either:
(E)4 shaking Amax,rock ≥ 0.55 g, or
Amax,rock ≥ 0.45 g and M > 7.25
E3 Very high plasticity clays Hclay > 30 ft with PI > 75% and
Vs < 800 ft/sec.
F1 Highly organic and/or peaty soils. H > 20 ft. of peat and/or highly
organic soils
F2 Sites likely to suffer ground failure due Liquefaction and/or other types of
(F)7
either to significant soil liquefaction of ground failure analysis required.
other potential modes of ground
instability.
(See next page for "notes")
417
Notes for Table A-4.
1. H = total (vertical) depth of soils of the type or types referred to.
2. Vs = seismic shear wave velocity (ft/sec) at small shear strains (shear strain ≈ 10-4 %).
3. If surface soils are cohesionless, Vs may be less than 800 ft/sec in top 10 feet.
4. "Cohesionless soils" = soils with less than 30% "fines" by dry weigth;
"Cohesive soils" = soils with more than 30% "fines" by dry weight, and 15% ≤ PI (fines) ≤ 90%.
Soils with more than 30% fines, and PI (fines) < 15% are considered "silty" soils herein, and these
should be (conservatively) treated as "cohesive" soils for site clasification purposes in this Table.
(Evaluation of approximate Vs for these "silty" soils should be based either on penetration resistance or
direct field Vs measurement; see Note 8 below).
5. "Soft Clay" is defined herein as cohesive soil with: (a) Fines content ≥ 30%, (b) PI (fines) ≥ 20%, and
(c) Vs ≤ 500 ft/sec.
6. Site-specific geotechnical investigations and dynamic site response analyses are strongly
recommended for these conditions. Variability of response characteristics within this class (E) of sites
tends to be more highly variable than for classes Ao through D, and the very approximate response
projections herein should be applied conservatively in the absence of (strongly recommended) site-
specific studies.
7. Site-specific geotechnical investigations and dynamic site response analyses are recquired for these
conditions. Potentially significant ground failure must be mitigated, and/or it must be demonstrated
that the proposed structure/facility can be engineered to satisfactorily withstand such ground failure.
8. The following approaches are recommended for evaluation of Vs:
(a) For all site conditions, direct (in-situ) measurement of Vs is recommended.

(b) In lieu of direct measurement, the following empirical approaches can be used:
(i) For sandy cohesionless soils: either SPT-based or CPT-based empirical correlations may be
used.
(ii) For clayey soils: empirical correlations based on undrained shear strength and/or some
combination of one or more of the following can be used (void ratio, water content, plasticity
index, etc.) Such correlations tend to be somewhat approximate, and should be interpreted
accordingly.
(iii) Silty soils of low plasticity (PI ≤ 15%) should be treated as "largely cohesionless" soils here;
SPT-based or CPT-based empirical correlations may be used (ideally with some "fines"
correction relative to "clean sand" correlations.) Silty soils of medium to high plasticity
should be treated more like "clayey" soils as in (iii) above.
(iv) "Other" soil types (e.g. gravelly soils, rockfill, peaty and organic soils, etc.) require
considerable judgement, and must be evaluated on an individual basis; no simplified
"guidance" can appropriately be offered herein.
418
APPENDIX B
Site Visits to Selected Ground Motion Sites
419
CONTENTS OF APPENDIX B
CDMG 24461- Alhambra - Fremont School................................................................... 421

USC 99 - Arcadia - 855 Arcadia Av. .............................................................................. 422
USC 93- Arcadia - Campus Dr. ...................................................................................... 423
USC 14 - Beverly Hills - 12520 Mulholland .................................................................. 424
USC 13 - Fire Station # 99, 14145 Mulholland .............................................................. 425
USC 57: Canyon country - W. Lost Canyon ................................................................... 426
USC 65 - Glendora - N. Oakbank ................................................................................... 427
USC 34 - LA Fletcher Dr. ............................................................................................... 428
CDMG 24271 - Lake Hughes #1 - Fire Station #78 ....................................................... 429
Leona Valley sites (1-6) .................................................................................................. 430
CDMG 24305............................................................................................................... 430
CDMG 24306............................................................................................................... 430
CDMG 24307............................................................................................................... 430
CDMG 24308............................................................................................................... 430
CDMG 24055............................................................................................................... 430
CDMG 24309............................................................................................................... 430
CDMG 24396 - Malibu, Point Dume School.................................................................. 431
CDMG 14404 - Rancho Palos Verdes - Hawthorne Blvd............................................... 432
USC 44 - Rancho Palos Verdes - 30511 Lucania Dr. ..................................................... 433
CDMG 14405 - Rolling Hills Estates - Rancho Vista School ........................................ 434
CDMG 14159 - San Pedro - Palos Verdes ...................................................................... 435
CDMG 24644 - Sandberg - Bald Mountain .................................................................... 436
USC 77 - Santa Fe Spring - 11500 E. Joslin ................................................................... 437
USGS 5081...................................................................................................................... 438
CDMG 24047 - Vasquez Rock Park ............................................................................... 439
USC 51 - Malibu - St. Adams Episcopal Church............................................................ 440
420
CDMG 24461- Alhambra - Fremont School
Date: June 2, 1998
Time: 2:00 p.m.
Instrument: Could not access to school.
Description: The school is located in a flat area north of the Los Angeles basin. Some
hills are seen in the SW and NE (about 200 m to the SW and 300 m to the NE) that
correspond to units of Miocene Shale in geological maps (Dibblee).
421
USC 99 - Arcadia - 855 Arcadia Av.
Date: June 3, 1998
Time: 3:00 p.m.
Instrument: The instrument is located in private property. Could not access it nor locate
it.
Description: Building is located in a very flat region. Closest relief are the beginning of
the San Gabriel Mountains about 1 mile north.
422
USC 93- Arcadia - Campus Dr.
Date: June 3, 1998
Time: 3:30 p.m.
Instrument: Inside the school, but not located.
Description: The school is located in a very flat region. Closest relief are some hills to
the north (about 1 mile, the beginning of the San Gabriel Mountains). The site is far
enough from any topographic feature that, from observation alone, it would be assumed
that it has thick sedimentary deposits. The C2 classification may be reviewed.
423
USC 14 - Beverly Hills - 12520 Mulholland
Date: June 3, 1998
Time: 11:00 p.m..
Instrument: Located in the equipment room of Fire Station #108, about 50 m from the
southern end of an esplanade along a cliff on Mulholland Dr.
Description: The building is located next to Mulholland Dr., along a cliff in the Santa
Monica Mountains. There are canyons on both sides of the fire station, which is build on
an esplanade facing lower ranges of the Sta. Monica Mountains in the south side. On the
east and west side the slopes are relatively steep (~ 45º). There is some exposed rock in
the south of the station, a well-cemented, medium-size grained sandstone (about 100 m.
south of the station, in the slope of the south side of the station). There is a rock cut on
the north side of Mulholland Dr. showing what could be basalt as indicated in the Dibblee
map. A cut in the parking lot (a small 1 m3 manhole opening) shows what could be shale
in the south edge of the cut, and sandstone in the north side. Probably is unit Tt in
Dibblee.
From this we can infer that the site is in between Tt and Ttls (see Dibblee map for
Beverly Hills Quadrangle).
Geological Units mentioned:

Tt Middle Topanga formation—mostly interbedded gray to tan semi-friable
sandstone and gray micaceous claystone, locally includes lenses of pebbly sandstone
and pebble-cobble conglomerate
Ttls Lower Topanga formation—mostly tan, semi-friable to hard arkosic sandstone;
locally includes gray micaceous shale.
Recommended Classification from Site Visit: C1
424
USC 13 - Fire Station # 99, 14145 Mulholland
Date: June 3, 1998
Time: 10:00 a.m.
Instrument: At the time of the visit, the instrument was not installed because of work
being done in the floor of the building. Normally the instrument is located underneath a
stairwell in the office (2 story building).
Description: The fire station is located on Mulholland Drive, which is placed along a
ridge on the Santa Monica Mountains. To the north, you can overlook the San Fernando
Valley, and the Sta. Monica Mountains continue to the south. The location is surrounded
by vegetation and heavy brush. There are some recent landslides west of the station,
about 1 mile away (on some steep cliffs). Exposed surface beneath the fire station looks
like soil, likely residual soil.
An outcrop on 13030 Mulholland, shows highly jointed, yellow-brown, poorly
cemented sandstone. Major joint sets dips north (about 70º). In the same road cut there is
also a medium to coarse grained sandstone, bedded. Sedimentation plains dip ≈ 70º N,
crumbles easily with hand pressure. Another exposure further west shows a deeper soil
horizon, also outcrop of sandstone, less weathered and more cemented than the previous
one.
425
USC 57: Canyon country – W. Lost Canyon
Date: June 2, 1998
Time: 2:00 p.m.
Instrument: Bolted to concrete floor on a one floor building (Main office, closet next to
the lounge)
Description: The instrument is located in a valley next to the Santa Clara River.
426
USC 65 - Glendora - N. Oakbank
Date: June 3, 1998
Time: 4:00 p.m.
Instrument: Could not access the church. The instrument is located in a church in N.
Oakbank.
Description:
Very close to the foothills of the San Gabriel Mountains. Site is located in the point
where the slope starts to increase, probably in the edge of recent alluvial fans. No
inferences can be made with respect to the soil depth.
427
USC 34 - LA Fletcher Dr.
Date: June 3, 1998
Time: 1:00 p.m.
Instrument: Located inside the fire station (LA Fire Station 52). Could not access
instrument in site visit, building was closed.
Description: The site is located in a very broad valley, possibly a deep site.
428
CDMG 24271 - Lake Hughes #1 - Fire Station #78
Date: June 2, 1998
Time: 10:00 a.m.
Instrument: Located in the fire station building (1 story garage), bolted on concrete.
Description: On edge of hills overlooking the rift valley formed by the San Andreas
fault, in the north side of the fault. The fire station is located in the mouth of a small
canyon; this may explain why the soil cover is so thick in this station.
429
Leona Valley sites (1-6)
CDMG 24305
CDMG 24306
CDMG 24307
CDMG 24308
CDMG 24055
CDMG 24309
Date: June 2, 1998
Time: 12:00 noon
Instrument: All of the instruments located directly on soil. All but #5 have metal
instrument shelters with an antenna. #5 has an older instrument shelter.
Description:
#1 Located in a large hill in the middle of a (cherry?) orchard. There is a rock outcrop
next to the stations (moderately weathered crystalline rock). According to the geologic
map, the rock is likely to be Pelona Schist. From this observation, I would classify the
site as B.
# 2 Located in a small, broad valley, low relief. To the north there is a hill where #3 is
located at about 50 m., to the south is the hill where #1 is located. The bottom of the hill
is at about 100 m. from the station. There is a small creek next to the station, exposes
about 1 m of gravel and clay. It is impossible to infer the depth of soil from just a cursory
observation, but given proximity of hills and older deposits, I would classify this site as
C2 or C3.
#3 Located on the top of a hill with moderate slope (slope angle ≈ 20º). Definitely not
metamorphic bedrock as indicated in CSMIP. Most likely a C1 site, probably a B site.
#4 Located on the slope of a small hill (about 5 -10 m. high). On the hill there is highly
weathered sedimentary rock exposed; crumbles with slight pressure from the hand. You
can find quartz pebbles around the station. Inferring from rock behind the instrument, the
site is on highly weathered sedimentary rock, likely a C1 site.
#5 On a flat region near a small hill with mild slope. This site is located near a hill, but
closer to the larger valley where the SAF is located. This site may be a C2 or a C3 site.
#6 Located on a hill on the north side of the road (N2). The road has been recently
moved to allow for the construction of a spillway in the creek running parallel to the San
Andreas fault (SAF) (Amargosa creek). The station is located apparently to the north of
the main trace of the SAF. to the north of the station there is a small valley and on the
other side of the valley there are larger mountains dividing the rift valley from the
Antelope valley. Soil around the station is gravelly, angular. A canyon on a creek about
300 m. east of the station exposes sedimentary rock. Due to this observation and the
geologic map, I would classify this as a C1 site.
430
CDMG 24396 - Malibu, Point Dume School
Date: June 4, 1998
Time: 10:00 a.m.
Instrument: Located in the Point Dume community center office (copy room)
Description: Instrument located at the foot of a mild sloping hill (~70 m. from the base
of the hill). A shallow soil cover may exist, probably formed from materials transported
from the hill. According to the map, the site is in Tertiary Marine deposits.
I would classify the site C1 because in general the tertiary deposits in this zone are
weathered. Note that USC 51 is close by and on the same formation; however it is on a
hill at a very different elevation than this site.
431
CDMG 14404 - Rancho Palos Verdes - Hawthorne Blvd.
Date: June 3, 1998
Time: 6:30 p.m.
Instrument: Located in a small maintenance building (1 story). It could not be accessed.
The site, formerly a Loyola Marymount University building, now belongs to the Salvation
Army.
Description:
Very weathered sandstone, light tan (looks the same formation as for CDMG 14159 Site).
Building located against a hill on what appears to be a cut section.
432
USC 44 - Rancho Palos Verdes - 30511 Lucania Dr.
Date: June 3, 1998
Time: 8:00 p.m.
Instrument: Located in Mira Catalina School.
Description: Site located in the side of a hill facing south. No rock exposures where
found close to the site, but the geology is likely to be similar to the CDMG 14159 Site.
433
CDMG 14405 - Rolling Hills Estates - Rancho Vista School
Date: June 3, 1998
Time: 7:00 p.m.
Instrument: Located in Rancho Vista School, the school was closed at the time of the
visit and access to the exact location of the instrument was not possible.
Description: Site is on an esplanade in a hillside facing towards the LA Valley (N).
434
CDMG 14159 - San Pedro - Palos Verdes
Date: June 3, 1998
Time: 6:00 p.m.
Instrument: Located in a fire station on 25th Street in front of Whites Park. The
instruments in a deposit in the east side of the building (1 story building). The instrument
is not the same one that recorded the Northridge earthquake, the previous one was located
in the other side of the wall from its current location.
Description: The fire station is located in top of a hill overlooking the ocean (at least 80
m from the top of the slope). In Whites Park, south of the site there are some impressive
rock exposures:
Sandstone, varies in grain size and color across the exposure, weathered to very
weathered at spots. The cut in the slope where rock is exposed is about 40 - 50 m. high.
If station were located here, it would be a C1 site, but probably bordering on a B site.
Given that the cut has been exposed, weathering in the cut may be due to its exposure,
and may not reflect true depth of weathering; therefore the site may also be a B site.
Small slide W of the station (about 500 m) on the side of the road that goes down
to Whites Park exposes sandstone, horizontal bedding, lightly tan, fine grained, very
weathered. Also layers of hard shale, dark tan, very weathered.
Recommended Classification from Site Visit: C?1
435
CDMG 24644 - Sandberg - Bald Mountain
Date: June 2, 1998
Time: 8:00 a.m.
Instrument: Instrument Shelter, placed on ground surface.
Description: Site located on the top of a mountain. Low vegetation cover. The
surrounding peaks all are dome-shaped; no abrupt peaks are seen. A recently scraped
road next to the instrument exposes soil (clay?) with considerable amount of coarse sand-
size particles of granitic origin; some larger (1in. - 2 in.) granitic pebbles also found. No
outcrops of intact rock visible in the near vicinity.
A small slide (10 m. wide) caused by road work about 0.5 miles east of the station
shows what seems to be residual soil at least 3 m deep. Grain size of mother rock
(granite, pink) increases downward. Some highly weathered granite blocks seen also at
about 1m. deep.
436
USC 77 - Santa Fe Spring - 11500 E. Joslin
Date: June 3, 1998
Time: 5:00 p.m.
Instrument: Located in Lake View School
Description: The site was visited because it had the particularity that it was the only C2
site in the LA basin, however, nothing particular in the topography surrounding the site
was noticed. From the site visit alone, it is recommended to change the classification to a
D site, given that the C2 classification is inferred only from the SCEC shear wave
velocity database (see Boore and Brown 1998).
437
USGS 5081
Date: June 3, 1998
Time: 2:00 p.m.
Instrument: Instrument located in the NW side of the station, in a small wash room (1
story building) in a fire station. (FS is in the intersection of Topanga Canyon Rd. with
Fernwood-Pacific Dr.)
Description:
Station located in the wall of a steep canyon (W side of canyon). There is a landslide in
the road in front of the station on Topanga Canyon Rd. A driller on the site indicated to
me that they found superficial colluvium over sandstone (soft, not well cemented).
Strangely, they found a 10 ft. layer of river sand interbedded with the Sandstone.
Recommended Classification from Site Visit: C2, although it may be also C1. Station
is likely located in landslide deposits.
438
CDMG 24047 - Vasquez Rock Park
Date: June 2, 1998
Time: 3:00 p.m.
Instrument: Not located
Description: The instrument could not be exactly located due to the high vegetation.
The site where the instrument should be located is near a creek bed just north of Vasquez
Rock Park.
439
USC 51 - Malibu - St. Adams Episcopal Church
Date: June 3, 1998
Time: 11:00 a.m.
Instrument: Located in a small deposit adjacent to the men's room in the church.
Description:
Instrument is located in a hill on the N side of Hwy. 1. The hill has relatively steep graded
slopes. The instrument is about 30 m from the edge of the slope.
440
APPENDIX C
Equations to Obtain Combined Spectral Acceleration Ratios for the

Northridge and Loma Prieta Earthquakes
441
Formula for Spectral Amplification Factors from Site j to Site i
The amplification factors from Site j to Site i (Fi/j) for a given reference peak
ground acceleration (PGAref) and a given spectral period are given by the following
formula:
( ) [ ] [ ]
ln Fi / j = w i / j, N a i / j,N + b i / j,N ln( R N + c N ) + w i / j,L a i / j, L + b i / j,L ln( R L + c L ) (C1)
where the subscripts N and L denote coefficients for the Northridge and Loma Prieta
Earthquakes respectively, ai/j and bi/j are given by Equation 4 (rewritten here for
convenience), and are listed in Table C1;
ai/j = a(Site i) - a(Site j) (4a)
bi/j = b(Site i) - b(Site j) (4b)
a, b, and c are period-dependent coefficients given in Table 3 (rewritten as Table C4 for
convenience) for each earthquake; R is the distance corresponding to the reference PGA
(PGAref), and is given by:
æ ln( PGAref ) − a ö
ç
ç
R=e −c
è b
(C2)
442
where a, b, and c are the coefficients given in Table 3 (Table C4) corresponding to the
peak ground acceleration (PGA) and wi/j are weights for each earthquakes. If a simple
geometric mean of both earthquakes is used, the weights are equal to 0.5 for each
earthquake. If the variance-weighted geometric mean is used, the weights are obtained
by:
(VAR ) −1
w i / j, N =
i / j, N
(VAR ) + (VAR )
−1 −1
(C3)
i / j, N i / j, L
for the Northridge Earthquake and:
(VAR ) −1
w i / j, L =
i / j, L
(VAR ) + (VAR )
−1 −1
(C4a)
i / j, N i / j, L
for the Loma Prieta Earthquake. The variance of the sample mean, VARi/j, is obtained for
each earthquake using the following formula:
σj
2
σ
2
VAR i / j = i + (C4b)
Ni N j
where σ and N are the standard deviation and number of sites corresponding to each site
condition, spectral period, and earthquake. The standard deviations are given in Table 3
443
(Table C4) and the number of sites is a function of period and is given in Table C2. The
resulting weights for the variance weighted scheme (wi/j) are given in Table C3.
Standard Deviations
The standard deviations associated with each site condition are obtained using a similar
weighting scheme as the amplification factors. For Site i, the standard deviations are
given by:
( )
σ i2 = (w ′i , N ) σ i2, N + (w ′i ,L ) σ i2,L ( ) (C5)
where σi is the standard deviation and w'i the weight for site condition i for each
earthquake. The weights w'i are equal to 0.5 if a simple geometric mean of both
earthquakes is used. If the variance-weighted geometric mean is used, the weights are
obtained by:
(VAR ) −1
w i,N =
VAR ( i ), N
(VAR ) + (VAR )
−1 −1
(C6a)
VAR ( i ), N VAR ( i ), L
for the Northridge Earthquake and:
(VAR ) −1
w i / j, L =
VAR ( i ), L
(VAR ) + (VAR )
−1 −1
(C6b)
VAR ( i ), N VAR ( i ), L
444
for the Loma Prieta Earthquake. The variance of the sample variance, VARVAR(i) is
estimated for each earthquake by:
2(N i − 1) 4
VAR VAR (i ) = σi (C7)
N i2
where σ and N are the standard deviation and number of sites corresponding to each site
condition, spectral period, and earthquake. The resulting weights, w'i, are given in Table
C5.
445
Table C1. Coefficients for determining spectral amplification ratios (Equation C1).
T ac/b bc/b ad/b bd/b ad/c bd/c ac/b bc/b ad/b bd/b ad/c bd/c
PGA 0.0000 0.1215 0.3198 0.0592 0.3198 -0.0623 0.0993 0.0452 -0.1503 0.1922 -0.2496 0.1470
0.055 0.0000 0.0922 0.1404 0.0728 0.1404 -0.0195 -0.3276 0.1334 -0.5213 0.2640 -0.1937 0.1306
0.06 0.0000 0.0914 0.1449 0.0709 0.1449 -0.0205 -0.3414 0.1374 -0.5322 0.2673 -0.1907 0.1299
0.07 0.0000 0.0893 0.1597 0.0653 0.1597 -0.0240 -0.3589 0.1426 -0.5456 0.2717 -0.1867 0.1290
0.08 0.0000 0.0875 0.1801 0.0585 0.1801 -0.0290 -0.3859 0.1515 -0.5655 0.2790 -0.1796 0.1275
0.09 0.0000 0.0873 0.2017 0.0527 0.2017 -0.0346 -0.4068 0.1595 -0.5793 0.2856 -0.1724 0.1261
0.1 0.0000 0.0896 0.2193 0.0497 0.2193 -0.0398 -0.4141 0.1637 -0.5821 0.2891 -0.1680 0.1254
0.11 0.0000 0.0917 0.2257 0.0497 0.2257 -0.0421 -0.4150 0.1665 -0.5788 0.2913 -0.1638 0.1248
0.12 0.0000 0.0946 0.2306 0.0507 0.2306 -0.0440 -0.4124 0.1673 -0.5740 0.2918 -0.1616 0.1245
0.13 0.0000 0.0982 0.2343 0.0525 0.2343 -0.0456 -0.4100 0.1680 -0.5695 0.2922 -0.1595 0.1242
0.14 0.0000 0.1022 0.2373 0.0552 0.2373 -0.0470 -0.4002 0.1682 -0.5545 0.2917 -0.1543 0.1235
0.15 0.0000 0.1066 0.2401 0.0584 0.2401 -0.0482 -0.3874 0.1676 -0.5346 0.2898 -0.1472 0.1222
0.16 0.0000 0.1112 0.2430 0.0621 0.2430 -0.0491 -0.3740 0.1665 -0.5110 0.2866 -0.1370 0.1201
0.17 0.0000 0.1158 0.2463 0.0661 0.2463 -0.0497 -0.3679 0.1659 -0.4975 0.2843 -0.1297 0.1184
0.18 0.0000 0.1179 0.2482 0.0682 0.2482 -0.0498 -0.3621 0.1654 -0.4848 0.2822 -0.1227 0.1168
0.19 0.0000 0.1200 0.2501 0.0701 0.2501 -0.0498 -0.3537 0.1648 -0.4572 0.2768 -0.1035 0.1119
0.2 0.0000 0.1237 0.2544 0.0742 0.2544 -0.0496 -0.3500 0.1650 -0.4293 0.2705 -0.0793 0.1055
0.22 0.0000 0.1267 0.2593 0.0780 0.2593 -0.0487 -0.3512 0.1661 -0.4018 0.2637 -0.0506 0.0976
0.24 0.0000 0.1288 0.2649 0.0815 0.2649 -0.0474 -0.3541 0.1671 -0.3881 0.2600 -0.0340 0.0929
0.26 0.0000 0.1294 0.2683 0.0829 0.2683 -0.0465 -0.3568 0.1679 -0.3755 0.2566 -0.0187 0.0887
0.28 0.0000 0.1300 0.2714 0.0843 0.2714 -0.0457 -0.3650 0.1700 -0.3508 0.2494 0.0142 0.0793
0.3 0.0000 0.1301 0.2796 0.0862 0.2796 -0.0438 -0.3736 0.1719 -0.3279 0.2423 0.0457 0.0704
0.32 0.0000 0.1291 0.2901 0.0869 0.2901 -0.0422 -0.3769 0.1723 -0.3170 0.2389 0.0599 0.0665
0.34 0.0000 0.1272 0.3039 0.0861 0.3039 -0.0410 -0.3799 0.1728 -0.3067 0.2357 0.0732 0.0629
0.36 0.0000 0.1244 0.3218 0.0836 0.3218 -0.0408 -0.3804 0.1724 -0.2964 0.2325 0.0840 0.0601
0.4 0.0000 0.1209 0.3444 0.0792 0.3444 -0.0417 -0.3737 0.1688 -0.2669 0.2238 0.1068 0.0550
0.44 0.0000 0.1169 0.3718 0.0730 0.3718 -0.0439 -0.3561 0.1626 -0.2461 0.2183 0.1101 0.0557
0.5 0.0000 0.1081 0.4385 0.0562 0.4385 -0.0519 -0.3135 0.1491 -0.2137 0.2110 0.0998 0.0619
0.55 0.0000 0.1038 0.4753 0.0467 0.4753 -0.0571 -0.2834 0.1399 -0.1935 0.2067 0.0899 0.0668
0.6 0.0000 0.0998 0.5120 0.0373 0.5120 -0.0625 -0.2578 0.1324 -0.1770 0.2034 0.0808 0.0710
0.667 0.0000 0.0962 0.5466 0.0288 0.5466 -0.0674 -0.2267 0.1232 -0.1571 0.1994 0.0697 0.0762
0.7 0.0000 0.0933 0.5774 0.0217 0.5774 -0.0715 -0.1980 0.1149 -0.1380 0.1955 0.0599 0.0806
0.75 0.0000 0.0910 0.6030 0.0167 0.6030 -0.0743 -0.1569 0.1030 -0.1109 0.1900 0.0461 0.0870
0.8 0.0000 0.0895 0.6227 0.0139 0.6227 -0.0756 -0.1148 0.0910 -0.0811 0.1839 0.0338 0.0928
0.85 0.0000 0.0888 0.6365 0.0135 0.6365 -0.0753 -0.0753 0.0798 -0.0531 0.1781 0.0222 0.0983
0.9 0.0000 0.0889 0.6448 0.0154 0.6448 -0.0735 0.0161 0.0540 0.0173 0.1630 0.0011 0.1090
0.95 0.0000 0.0898 0.6486 0.0195 0.6486 -0.0704 0.1146 0.0266 0.0999 0.1448 -0.0146 0.1182
1.0 0.0000 0.0914 0.6491 0.0251 0.6491 -0.0663 0.2165 -0.0017 0.1932 0.1237 -0.0233 0.1253
1.1 0.0000 0.0936 0.6475 0.0320 0.6475 -0.0616 0.3180 -0.0297 0.2941 0.1002 -0.0239 0.1298
1.2 0.0000 0.0961 0.6449 0.0395 0.6449 -0.0565 0.4150 -0.0564 0.3984 0.0753 -0.0167 0.1317
1.3 0.0000 0.0987 0.6423 0.0472 0.6423 -0.0515 0.5039 -0.0809 0.5014 0.0503 -0.0025 0.1311
1.4 0.0000 0.1013 0.6402 0.0547 0.6402 -0.0466 0.5815 -0.1022 0.5984 0.0263 0.0169 0.1285
1.5 0.0000 0.1036 0.6389 0.0614 0.6389 -0.0422 0.6458 -0.1199 0.6853 0.0044 0.0395 0.1243
1.7 0.0000 0.1072 0.6389 0.0721 0.6389 -0.0350 0.6961 -0.1338 0.7592 -0.0145 0.0631 0.1193
2.0 0.0000 0.1090 0.6411 0.0787 0.6411 -0.0302 0.7329 -0.1441 0.8187 -0.0301 0.0859 0.1140
2.2 0.0000 0.1094 0.6424 0.0806 0.6424 -0.0287 0.7459 -0.1477 0.8423 -0.0363 0.0964 0.1114
2.6 0.0000 0.1096 0.6434 0.0818 0.6434 -0.0277 0.7634 -0.1526 0.8753 -0.0451 0.1119 0.1075
3.0 0.0000 0.1096 0.6441 0.0825 0.6441 -0.0272 0.7733 -0.1554 0.8958 -0.0506 0.1224 0.1048
446
Table C2. Number of sites for each earthquake as a function of site condition and
spectral period. The number of sites for periods lower than one second is equal to the
number of sites for peak ground acceleration.

T B C D B C D
PGA 11 70 59 18 26 18
1.0 11 70 59 18 26 18
1.1 11 70 58 18 26 18
1.2 11 70 58 18 26 18
1.3 11 70 58 18 26 18
1.4 11 69 58 18 26 18
1.5 11 69 58 18 26 18
1.6 11 69 58 18 26 18
1.7 11 67 58 18 26 18
1.8 11 67 58 18 26 18
1.9 11 67 58 18 26 18
2.0 11 67 58 18 26 18
2.2 11 65 57 18 26 18
2.4 11 65 57 18 26 18
2.6 11 65 57 18 26 18
2.8 10 48 41 18 26 18
3.0 10 47 41 18 26 18
447
Table C3. Weights used to combine spectral amplification ratios for the Northridge and
Loma Prieta Earthquakes. Weights are inversely proportional to the variance of the
sample mean.
T b/c b/d c/d b/c b/d c/d
PGA 0.56 0.64 0.70 0.44 0.36 0.30
0.055 0.45 0.53 0.75 0.55 0.47 0.25
0.06 0.46 0.53 0.75 0.54 0.47 0.25
0.07 0.46 0.53 0.75 0.54 0.47 0.25
0.08 0.48 0.54 0.75 0.52 0.46 0.25
0.09 0.50 0.57 0.75 0.50 0.43 0.25
0.1 0.54 0.60 0.75 0.46 0.40 0.25
0.11 0.56 0.61 0.75 0.44 0.39 0.25
0.12 0.57 0.63 0.74 0.43 0.37 0.26
0.13 0.58 0.64 0.74 0.42 0.36 0.26
0.14 0.58 0.64 0.73 0.42 0.36 0.27
0.15 0.58 0.63 0.73 0.42 0.37 0.27
0.16 0.57 0.62 0.72 0.43 0.38 0.28
0.17 0.56 0.61 0.71 0.44 0.39 0.29
0.18 0.56 0.60 0.70 0.44 0.40 0.30
0.19 0.55 0.60 0.69 0.45 0.40 0.31
0.2 0.54 0.58 0.68 0.46 0.42 0.32
0.22 0.53 0.58 0.67 0.47 0.42 0.33
0.24 0.52 0.57 0.66 0.48 0.43 0.34
0.26 0.52 0.57 0.66 0.48 0.43 0.34
0.28 0.52 0.57 0.66 0.48 0.43 0.34
0.3 0.52 0.57 0.66 0.48 0.43 0.34
0.32 0.52 0.57 0.67 0.48 0.43 0.33
0.34 0.51 0.56 0.67 0.49 0.44 0.33
0.36 0.51 0.56 0.67 0.49 0.44 0.33
0.4 0.53 0.58 0.69 0.47 0.42 0.31
0.44 0.54 0.59 0.71 0.46 0.41 0.29
0.5 0.55 0.61 0.73 0.45 0.39 0.27
0.55 0.56 0.63 0.74 0.44 0.37 0.26
0.6 0.57 0.64 0.74 0.43 0.36 0.26
0.667 0.57 0.65 0.75 0.43 0.35 0.25
0.7 0.58 0.65 0.76 0.42 0.35 0.24
0.75 0.58 0.65 0.77 0.42 0.35 0.23
0.8 0.57 0.65 0.77 0.43 0.35 0.23
0.85 0.57 0.65 0.78 0.43 0.35 0.22
0.9 0.56 0.65 0.79 0.44 0.35 0.21
0.95 0.56 0.65 0.80 0.44 0.35 0.20
1.0 0.56 0.65 0.80 0.44 0.35 0.20
1.1 0.55 0.64 0.81 0.45 0.36 0.19
1.2 0.55 0.64 0.81 0.45 0.36 0.19
1.3 0.55 0.63 0.81 0.45 0.37 0.19
1.4 0.55 0.63 0.82 0.45 0.37 0.18
1.5 0.55 0.63 0.82 0.45 0.37 0.18
1.7 0.55 0.63 0.82 0.45 0.37 0.18
2.0 0.55 0.62 0.82 0.45 0.38 0.18
2.2 0.55 0.62 0.82 0.45 0.38 0.18
2.6 0.55 0.62 0.83 0.45 0.38 0.17
3.0 0.52 0.59 0.80 0.48 0.41 0.20
448
Table C4a. Regression coefficients and Standard Error for spectral acceleration values at
5% damping for the Northridge Earthquake.
PGA 2.3718 -1.2753 6.3883 0.3209 2.3718 -1.1538 6.3883 0.4686 2.6916 -1.2161 6.3883 0.3559
0.055 3.5192 -1.4829 10.2486 0.4343 3.5192 -1.3869 10.2486 0.4661 3.5126 -1.3703 10.2486 0.3560
0.06 3.7423 -1.5138 11.8103 0.4343 3.7423 -1.4266 11.8103 0.4655 3.7970 -1.4257 11.8103 0.3654
0.07 4.3982 -1.6291 14.5768 0.4310 4.3982 -1.5480 14.5768 0.4636 4.4475 -1.5472 14.5768 0.3705
0.08 4.8097 -1.7006 16.9734 0.4180 4.8097 -1.6152 16.9734 0.4619 4.9774 -1.6422 16.9734 0.3754
0.09 4.9993 -1.7175 18.0000 0.3935 4.9993 -1.6366 18.0000 0.4617 5.2637 -1.6826 18.0000 0.3779
0.1 4.9768 -1.6855 18.0000 0.3615 4.9768 -1.6089 18.0000 0.4642 5.3000 -1.6679 18.0000 0.3774
0.11 4.9365 -1.6614 18.0000 0.3457 4.9365 -1.5844 18.0000 0.4667 5.2529 -1.6439 18.0000 0.3766
0.12 4.8748 -1.6330 18.0000 0.3322 4.8748 -1.5530 18.0000 0.4703 5.1563 -1.6072 18.0000 0.3759
0.13 4.7753 -1.5991 18.0000 0.3226 4.7753 -1.5140 18.0000 0.4750 5.0044 -1.5586 18.0000 0.3758
0.14 4.6161 -1.5564 17.3303 0.3179 4.6161 -1.4646 17.3303 0.4808 4.7947 -1.4991 17.3303 0.3766
0.15 4.3937 -1.5041 16.0757 0.3182 4.3937 -1.4037 16.0757 0.4877 4.5454 -1.4330 16.0757 0.3786
0.16 4.1376 -1.4471 14.9021 0.3232 4.1376 -1.3364 14.9021 0.4952 4.2958 -1.3685 14.9021 0.3820
0.17 3.8807 -1.3907 13.7997 0.3315 3.8807 -1.2694 13.7997 0.5030 4.0778 -1.3133 13.7997 0.3865
0.18 3.7555 -1.3635 12.7603 0.3368 3.7555 -1.2373 12.7603 0.5069 3.9820 -1.2900 12.7603 0.3893
0.19 3.6370 -1.3378 11.7771 0.3418 3.6370 -1.2069 11.7771 0.5105 3.8913 -1.2680 11.7771 0.3918
0.2 3.4048 -1.2891 10.8444 0.3531 3.4048 -1.1508 10.8444 0.5174 3.7044 -1.2249 10.8444 0.3974
0.22 3.1681 -1.2413 9.1112 0.3646 3.1681 -1.0982 9.1112 0.5234 3.4809 -1.1745 9.1112 0.4026
0.24 2.9146 -1.1904 7.5290 0.3759 2.9146 -1.0449 7.5290 0.5285 3.2196 -1.1160 7.5290 0.4071
0.26 2.7904 -1.1657 6.6312 0.3818 2.7904 -1.0198 6.6312 0.5308 3.0913 -1.0874 6.6312 0.4089
0.28 2.6754 -1.1429 5.8000 0.3872 2.6754 -0.9965 5.8000 0.5330 2.9725 -1.0610 5.8000 0.4106
0.3 2.5178 -1.1149 4.9000 0.3983 2.5178 -0.9682 4.9000 0.5372 2.8087 -1.0250 4.9000 0.4129
0.32 2.4644 -1.1117 4.4939 0.4087 2.4644 -0.9657 4.4939 0.5415 2.7420 -1.0110 4.4939 0.4141
0.34 2.4645 -1.1197 4.4254 0.4176 2.4645 -0.9768 4.4254 0.5463 2.7212 -1.0067 4.4254 0.4145
0.36 2.4594 -1.1242 4.3606 0.4242 2.4594 -0.9870 4.3606 0.5515 2.6916 -0.9999 4.3606 0.4142
0.4 2.4375 -1.1239 4.2415 0.4276 2.4375 -0.9935 4.2415 0.5570 2.6466 -0.9915 4.2415 0.4133
0.44 2.4279 -1.1279 4.1337 0.4277 2.4279 -1.0049 4.1337 0.5627 2.6269 -0.9946 4.1337 0.4119
0.5 2.4692 -1.1545 3.9890 0.4198 2.4692 -1.0526 3.9890 0.5739 2.7651 -1.0629 3.9890 0.4066
0.55 2.4447 -1.1582 3.8812 0.4140 2.4447 -1.0682 3.8812 0.5792 2.8613 -1.1091 3.8812 0.4023
0.6 2.3687 -1.1540 3.7828 0.4090 2.3687 -1.0710 3.7828 0.5843 2.9263 -1.1469 3.7828 0.3968
0.667 2.2699 -1.1513 3.6630 0.4060 2.2699 -1.0675 3.6630 0.5892 2.9650 -1.1752 3.6630 0.3901
0.7 2.1804 -1.1550 3.6084 0.4059 2.1804 -1.0660 3.6084 0.5937 2.9956 -1.1995 3.6084 0.3826
0.75 2.1276 -1.1664 3.5303 0.4090 2.1276 -1.0746 3.5303 0.5977 3.0096 -1.2199 3.5303 0.3750
0.8 2.1239 -1.1848 3.4573 0.4151 2.1239 -1.0966 3.4573 0.6009 2.9754 -1.2294 3.4573 0.3680
0.85 2.1516 -1.2064 3.3887 0.4235 2.1516 -1.1267 3.3887 0.6030 2.8866 -1.2261 3.3887 0.3621
0.9 2.1703 -1.2244 3.4413 0.4332 2.1703 -1.1539 3.4413 0.6041 2.7784 -1.2185 3.4413 0.3579
0.95 2.1451 -1.2353 3.2629 0.4435 2.1451 -1.1701 3.2629 0.6041 2.6965 -1.2187 3.2629 0.3556
1.0 2.0734 -1.2443 3.2048 0.4538 2.0734 -1.1775 3.2048 0.6033 2.6601 -1.2333 3.2048 0.3551
1.1 1.9888 -1.2635 3.0970 0.4637 1.9888 -1.1873 3.0970 0.6017 2.6461 -1.2583 3.0970 0.3563
1.2 1.9252 -1.2983 2.9986 0.4726 1.9252 -1.2071 2.9986 0.5995 2.6099 -1.2804 2.9986 0.3587
1.3 1.8811 -1.3390 2.9080 0.4799 1.8811 -1.2317 2.9080 0.5962 2.5295 -1.2884 2.9080 0.3618
1.4 1.8327 -1.3706 2.8242 0.4799 1.8327 -1.2510 2.8242 0.5909 2.4272 -1.2846 2.8242 0.3649
1.5 1.7582 -1.3853 2.7461 0.4799 1.7582 -1.2588 2.7461 0.5850 2.3331 -1.2785 2.7461 0.3664
1.7 1.5420 -1.3800 2.6045 0.4799 1.5420 -1.2565 2.6045 0.5800 2.1862 -1.2817 2.6045 0.3811
2.0 1.3896 -1.3970 2.4206 0.4799 1.3896 -1.2933 2.4206 0.5700 2.0500 -1.3154 2.4206 0.4130
2.2 1.2440 -1.3983 2.3128 0.4799 1.2440 -1.3004 2.3128 0.5600 1.8906 -1.3182 2.3128 0.4244
2.6 0.9829 -1.3739 2.1238 0.4799 0.9829 -1.2719 2.1238 0.5400 1.6293 -1.2941 2.1238 0.4145
3.0 0.6859 -1.3338 2.0000 0.4799 0.6859 -1.2207 2.0000 0.5200 1.3413 -1.2536 2.0000 0.3877
449
Table C4b. Regression coefficients and Standard Error for spectral acceleration values at
5% damping for the Loma Prieta Earthquake.
PGA 0.7219 -0.7954 1.0000 0.4713 0.8212 -0.7502 1.0000 0.3111 0.5716 -0.6032 1.0000 0.3896
0.055 1.6308 -0.9794 1.0000 0.4566 1.4230 -0.8769 1.0000 0.3708 1.3201 -0.7767 1.0000 0.4334
0.06 1.8207 -1.0119 1.0000 0.4561 1.4804 -0.8841 1.0000 0.3747 1.2568 -0.7489 1.0000 0.4338
0.07 1.9001 -1.0181 1.0000 0.4554 1.4819 -0.8734 1.0000 0.3798 1.2413 -0.7315 1.0000 0.4340
0.08 2.0559 -1.0383 1.0000 0.4538 1.5348 -0.8701 1.0000 0.3886 1.3041 -0.7271 1.0000 0.4331
0.09 2.1619 -1.0489 1.0000 0.4518 1.5875 -0.8642 1.0000 0.3973 1.4037 -0.7300 1.0000 0.4303
0.1 2.2305 -1.0551 1.0000 0.4500 1.6419 -0.8595 1.0000 0.4027 1.5122 -0.7400 1.0000 0.4269
0.11 2.2946 -1.0607 1.0000 0.4481 1.7031 -0.8527 1.0000 0.4074 1.6341 -0.7524 1.0000 0.4220
0.12 2.3215 -1.0625 1.0000 0.4472 1.7361 -0.8492 1.0000 0.4091 1.6890 -0.7575 1.0000 0.4187
0.13 2.3462 -1.0642 1.0000 0.4464 1.7665 -0.8461 1.0000 0.4108 1.7395 -0.7622 1.0000 0.4157
0.14 2.3659 -1.0613 1.0000 0.4451 1.8339 -0.8450 1.0000 0.4124 1.7916 -0.7621 1.0000 0.4084
0.15 2.3410 -1.0484 1.0000 0.4448 1.9079 -0.8523 1.0000 0.4120 1.7724 -0.7458 1.0000 0.4004
0.16 2.2804 -1.0268 1.0000 0.4460 1.9696 -0.8621 1.0000 0.4095 1.7156 -0.7191 1.0000 0.3924
0.17 2.2370 -1.0125 1.0000 0.4476 1.9792 -0.8631 1.0000 0.4071 1.6926 -0.7066 1.0000 0.3888
0.18 2.1960 -0.9991 1.0000 0.4491 1.9882 -0.8640 1.0000 0.4049 1.6710 -0.6949 1.0000 0.3853
0.19 2.0939 -0.9675 1.0000 0.4545 1.9513 -0.8531 1.0000 0.3989 1.6405 -0.6754 1.0000 0.3797
0.2 1.9861 -0.9352 1.0000 0.4626 1.8633 -0.8291 1.0000 0.3923 1.5961 -0.6551 1.0000 0.3761
0.22 1.8879 -0.9051 1.0000 0.4731 1.7414 -0.7944 1.0000 0.3861 1.5361 -0.6348 1.0000 0.3748
0.24 1.8523 -0.8933 1.0000 0.4797 1.6772 -0.7749 1.0000 0.3837 1.5154 -0.6295 1.0000 0.3754
0.26 1.8196 -0.8825 1.0000 0.4858 1.6182 -0.7570 1.0000 0.3815 1.4963 -0.6245 1.0000 0.3759
0.28 1.8136 -0.8775 1.0000 0.5001 1.5268 -0.7272 1.0000 0.3796 1.5140 -0.6328 1.0000 0.3796
0.3 1.8860 -0.8959 1.0000 0.5149 1.4800 -0.7104 1.0000 0.3812 1.5933 -0.6598 1.0000 0.3859
0.32 1.9446 -0.9120 1.0000 0.5223 1.4703 -0.7068 1.0000 0.3841 1.6498 -0.6788 1.0000 0.3906
0.34 1.9996 -0.9271 1.0000 0.5292 1.4613 -0.7033 1.0000 0.3868 1.7028 -0.6968 1.0000 0.3950
0.36 2.0373 -0.9373 1.0000 0.5358 1.4510 -0.7011 1.0000 0.3916 1.7453 -0.7128 1.0000 0.4012
0.4 2.0412 -0.9393 1.0000 0.5524 1.3972 -0.6925 1.0000 0.4092 1.7829 -0.7364 1.0000 0.4219
0.44 1.8966 -0.9057 1.0000 0.5600 1.3081 -0.6785 1.0000 0.4251 1.6708 -0.7148 1.0000 0.4396
0.5 1.5766 -0.8357 1.0000 0.5658 1.0905 -0.6402 1.0000 0.4486 1.3791 -0.6481 1.0000 0.4659
0.55 1.3683 -0.7909 1.0000 0.5678 0.9405 -0.6134 1.0000 0.4616 1.1859 -0.6031 1.0000 0.4808
0.6 1.2193 -0.7593 1.0000 0.5685 0.8299 -0.5944 1.0000 0.4699 1.0459 -0.5707 1.0000 0.4906
0.667 1.0380 -0.7209 1.0000 0.5694 0.6953 -0.5713 1.0000 0.4799 0.8757 -0.5314 1.0000 0.5025
0.7 0.9158 -0.6959 1.0000 0.5700 0.5954 -0.5543 1.0000 0.4867 0.7392 -0.4998 1.0000 0.5112
0.75 0.7412 -0.6602 1.0000 0.5708 0.4527 -0.5302 1.0000 0.4965 0.5444 -0.4547 1.0000 0.5235
0.8 0.6212 -0.6371 1.0000 0.5719 0.3418 -0.5106 1.0000 0.5038 0.3623 -0.4116 1.0000 0.5335
0.85 0.5083 -0.6155 1.0000 0.5728 0.2376 -0.4923 1.0000 0.5106 0.1913 -0.3712 1.0000 0.5428
0.9 0.2964 -0.5761 1.0000 0.5760 0.0693 -0.4630 1.0000 0.5215 -0.1385 -0.2932 1.0000 0.5598
0.95 0.0614 -0.5335 1.0000 0.5803 -0.0415 -0.4494 1.0000 0.5296 -0.3583 -0.2461 1.0000 0.5739
1.0 -0.1915 -0.4913 1.0000 0.5854 -0.0967 -0.4555 1.0000 0.5354 -0.4193 -0.2456 1.0000 0.5852
1.1 -0.4301 -0.4563 1.0000 0.5904 -0.1041 -0.4806 1.0000 0.5401 -0.3485 -0.2856 1.0000 0.5936
1.2 -0.6336 -0.4304 1.0000 0.5941 -0.0738 -0.5215 1.0000 0.5450 -0.2165 -0.3463 1.0000 0.5996
1.3 -0.8156 -0.4103 1.0000 0.5953 -0.0320 -0.5691 1.0000 0.5511 -0.0920 -0.4091 1.0000 0.6035
1.4 -1.0118 -0.3912 1.0000 0.5931 -0.0357 -0.6071 1.0000 0.5593 -0.0276 -0.4612 1.0000 0.6063
1.5 -1.2503 -0.3703 1.0000 0.5874 -0.1493 -0.6191 1.0000 0.5697 -0.0722 -0.4911 1.0000 0.6087
1.7 -1.5259 -0.3501 1.0000 0.5785 -0.3975 -0.6017 1.0000 0.5819 -0.2535 -0.4925 1.0000 0.6117
2.0 -1.7950 -0.3397 1.0000 0.5674 -0.7453 -0.5663 1.0000 0.5950 -0.5395 -0.4756 1.0000 0.6160
2.2 -1.9108 -0.3426 1.0000 0.5611 -0.9419 -0.5467 1.0000 0.6018 -0.7079 -0.4662 1.0000 0.6187
2.6 -2.0796 -0.3504 1.0000 0.5508 -1.2418 -0.5188 1.0000 0.6119 -0.9767 -0.4513 1.0000 0.6233
3.0 -2.1924 -0.3596 1.0000 0.5428 -1.4567 -0.5011 1.0000 0.6189 -1.1824 -0.4400 1.0000 0.6268
3.4 -2.3459 -0.3686 1.0000 0.5302 -1.7104 -0.4873 1.0000 0.6263 -1.5020 -0.4122 1.0000 0.6310
4.0 -2.4736 -0.3683 1.0000 0.5170 -1.8745 -0.4834 1.0000 0.6284 -1.7876 -0.3769 1.0000 0.6319
450
Table C5. Weights used to combine standard deviations for the Northridge and Loma
Prieta Earthquakes. Weights are inversely proportional to the variance of the sample
variance.
T B C D B C D
PGA 0.75 0.34 0.82 0.25 0.66 0.18
0.055 0.44 0.51 0.87 0.56 0.49 0.13
0.06 0.44 0.52 0.86 0.56 0.48 0.14
0.07 0.44 0.54 0.86 0.56 0.46 0.14
0.08 0.47 0.57 0.85 0.53 0.43 0.15
0.09 0.52 0.59 0.84 0.48 0.41 0.16
0.1 0.60 0.60 0.84 0.40 0.40 0.16
0.11 0.64 0.60 0.83 0.36 0.40 0.17
0.12 0.68 0.60 0.83 0.32 0.40 0.17
0.13 0.70 0.59 0.83 0.30 0.41 0.17
0.14 0.71 0.59 0.81 0.29 0.41 0.19
0.15 0.71 0.57 0.80 0.29 0.43 0.20
0.16 0.70 0.55 0.78 0.30 0.45 0.22
0.17 0.68 0.53 0.76 0.32 0.47 0.24
0.18 0.67 0.52 0.75 0.33 0.48 0.25
0.19 0.66 0.49 0.74 0.34 0.51 0.26
0.2 0.65 0.46 0.72 0.35 0.54 0.28
0.22 0.64 0.44 0.70 0.36 0.56 0.30
0.24 0.63 0.42 0.69 0.37 0.58 0.31
0.26 0.62 0.41 0.69 0.38 0.59 0.31
0.28 0.64 0.40 0.70 0.36 0.60 0.30
0.3 0.64 0.40 0.71 0.36 0.60 0.29
0.32 0.63 0.40 0.71 0.37 0.60 0.29
0.34 0.62 0.40 0.72 0.38 0.60 0.28
0.36 0.62 0.40 0.73 0.38 0.60 0.27
0.4 0.64 0.43 0.77 0.36 0.57 0.23
0.44 0.65 0.46 0.80 0.35 0.54 0.20
0.5 0.68 0.50 0.84 0.32 0.50 0.16
0.55 0.69 0.51 0.87 0.31 0.49 0.13
0.6 0.70 0.52 0.88 0.30 0.48 0.12
0.667 0.71 0.54 0.90 0.29 0.46 0.10
0.7 0.71 0.54 0.91 0.29 0.46 0.09
0.75 0.71 0.56 0.92 0.29 0.44 0.08
0.8 0.70 0.56 0.93 0.30 0.44 0.07
0.85 0.68 0.57 0.94 0.32 0.43 0.06
0.9 0.66 0.59 0.95 0.34 0.41 0.05
0.95 0.65 0.61 0.96 0.35 0.39 0.04
1.0 0.64 0.62 0.96 0.36 0.38 0.04
1.1 0.63 0.63 0.96 0.37 0.37 0.04
1.2 0.61 0.64 0.96 0.39 0.36 0.04
1.3 0.60 0.66 0.96 0.40 0.34 0.04
1.4 0.60 0.68 0.96 0.40 0.32 0.04
1.5 0.59 0.70 0.96 0.41 0.30 0.04
1.7 0.57 0.72 0.95 0.43 0.28 0.05
2.0 0.55 0.75 0.94 0.45 0.25 0.06
2.2 0.54 0.77 0.93 0.46 0.23 0.07
2.6 0.52 0.80 0.94 0.48 0.20 0.06
3.0 0.49 0.78 0.94 0.51 0.22 0.06
451
View publication stats

Near-Fault Seismic Site Response

Hochgeladen von

Dokumentinformationen

Copyright

Verfügbare Formate

Dieses Dokument teilen

Dokument teilen oder einbetten

Freigabeoptionen

Stufen Sie dieses Dokument als nützlich ein?

Sind diese Inhalte unangemessen?

Copyright:

Verfügbare Formate

Near-Fault Seismic Site Response

Hochgeladen von

Copyright:

Verfügbare Formate

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

Near-Fault Seismic Site Response

Article · January 2000

Resilient and Sustainable Buildings: Performance-Based Engineering View project

Bayesian Methods for Performance Based Engineering View project

The user has requested enhancement of the downloaded file.

B.S. (Washington State University) 1994

A dissertation submitted in partial satisfaction of the

requirements for the degree of

UNIVERSITY OF CALIFORNIA, BERKELEY

Professor Jonathan D. Bray, Chair

University of California, Berkeley

NEAR-FAULT SEISMIC SITE RESPONSE

Adrian Rodriguez Marek

Doctor of Philosophy in Engineering-Civil Engineering

University of California, Berkeley

Professor Jonathan D. Bray, Chair

The understanding of ground motions in the near-fault region is important for

Understanding the effect of local site conditions is important for a complete

characterization of near-fault ground motions. A geotechnically-based site classification

Northridge earthquakes. Regression analyses resulted in estimates of ground motion with

pulse-like near-fault forward-directivity ground motions. Thus, a separate database of

near-fault forward-directivity ground motions is studied. Each record in the database is

conclusions regarding site response based exclusively on empirical analyses. Numerical

response. A nonlinear site response methodology using a bounding surface plasticity

TABLE OF CONTENTS ............................................................................................... ii

LIST OF TABLES ........................................................................................................ vi

LIST OF FIGURES ....................................................................................................... ix

CHAPTER 1 - INTRODUCTION .............................................................................. 1

CHAPTER 2 - LITERATURE REVIEW.................................................................. 7

CHAPTER 3 - CHARACTERIZATION OF SITE RESPONSE GENERAL SITE

CHAPTER 4 - EMPIRICAL CHARACTERIZATION OF NEAR-FAULT

CHAPTER 5 - SITE RESPONSE ANALYSIS METHODOLOGY ................... 199

CHAPTER 6 ....... - SITE RESPONSE ANALYSIS OF NEAR-FAULT GROUND

CHAPTER 7 - CONCLUSION............................................................................... 370

REFERENCES ......................................................................................................... 381

APPENDIX A - LIST OF GROUND MOTION SITES WITH CORRESPONDING

APPENDIX B - SITE VISITS TO SELECTED GROUND MOTION SITES......... 419

APPENDIX C - EQUATIONS TO OBTAIN COMBINED SPECTRAL

intellectual experience of my life. My deepest thanks go to the faculty in the Civil

his support during my first year in the doctoral program.

providing a linear-elastic brick element and helping me in the finite element

his willingness to come to my help anytime it was needed.

My greatest thanks go to her.

A number of major metropolitan centers lie in close proximity to major faults.

large population centers motivates research seeking a better understanding of ground

motions close to faults.

effects of forward-directivity. Forward-directivity conditions produce ground motions

many efforts have been devoted to the characterization of forward-directivity motions

intermediate to long distances from the zones of energy release.

Near-fault ground motions, and in particular forward-directivity motions, have

specific characteristics that differentiate these motions from those recorded at

site conditions on near-fault ground motions. In addition, a new site classification

Loma Prieta and 1994 Northridge earthquakes.

The relevance and importance of correctly characterizing seismic site response

earthquake originating on nearby faults. A better understanding of the mechanisms that

understanding, in turn, will permit the development of adequate code-based regulations

that ensure the safety of engineered structures.

The assessment of seismic demand on a structure within the framework of

A better understanding of site amplification and, in particular, site amplification in

increased the available database of strong motions recordings by an order of magnitude,

of paramount importance in the evaluation of the current database of strong motions

qualitative and a quantitative evaluation of seismic site response to near-fault ground

can be summarized as follows: