Sie sind auf Seite 1von 12

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 105, NO.

B6, PAGES 13,631-13,642,JUNE 10, 2000

Static stresstransfer and earthquaketriggering:


No lower threshold in sight?
Alon Ziv and Allan M. Rubin
Departmentof Geosciences,
PrincetonUniversity,Princeton,New Jersey

Abstract. The main objective of this study is to see if a lower threshold for
earthquaketriggeringexists.Resolvingthis issueis importantfor the understanding
of earthquakemechanicsand for the purposeof hazard analysis.We computethe
cumulative static stresschangesimposedon 63 M _>4.5 earthquakesin central
Californiabetween1969 and 1998,by addingthe stresschangesimposedby all
previousM _>4.5 earthquakesas a function of time prior to the events. We find
that 85% of the cumulativestresschanges at the time of ruptureare positivefor
stresschangemagnitudesof >10 kPa (>0.1 bar), and 70% are positivefor stress
changes of <10 kPa (< 0.1 bar) as well as <1 kPa (<0.01 bar). A comparison
betweentheseresultsand thoseobtainedfor syntheticcatalogs,in whichthe timing
or focalmechanisms of the earthquakeswererandomized,showsthat this degreeof
triggeringis very unlikelyto be foundin a randomcatalog. Thus we concludethat
no lowerthresholdfor earthquaketriggeringin central Californiahas been found.
We showthat the temporal distributionof stresschangesthat discouragefailure
is consistentwith the theoreticalpredictionthat the time delay increaseswith the
magnitudeof the stresschange.

1. Introduction Neglect of very small stresschangescan be rational-


ized by several arguments. The absenceof a clear cor-
The effectof staticstresschangesonthe temporaland relationbetweentidal stresses and seismicity[Emter,
spatial distribution of earthquakeshas receivedmuch 1997]mayraisethe expectationthat staticstresschanges
attentionin the literature[Harris,1998,and references of the samemagnitudemay not be significantlycorre-
therein].A stresschange mayadvance or delaythetime lated with earthquake occurrenceas well. Coseismic
of a future rupture. Accordingto the Coulombfailure stressdrops are typically between 3 and 10 MPa. A
criterion,the time advanceor delay equalsthe ratio of stresschangethat is only a small Ëaction of that seems
the Coulombstresschangeto the long-termtectonic unimportant. Uncertainties in earthquake location and
loadingrate resolvedon the ruptureplane. If, for ex- source parameters result in uncertainties in the com-
ample,the background stressingrate is 20 kPa/yr (a puted stresschange. For very small stresschangesthese
reasonableestimate for many faults in central Califor- uncertaintiesmay be comparableto or larger than the
nia [Simpson
andReasenberg,
1994]),a 10kPa stress magnitudeof the stresschangeitself. In addition, since
changewouldadvancethe time to future failureby 6 the aftershock rate decreases with distance from a main-
monthsif that changein stressbringsthe fault closer shock,the effectof absolutestresschangessmallerthan
to failure,or it wouldpostponeit by the sameamount about 10 kPa may become statistically undetectable
if the stresschangemovesthe fault awayfrom failure. [Reasenberg and Simpson,1992;Simpsonand Reasen-
According to therate-and-state
model[Dieterich, 1994] berg,1994;Hardebeck et al., 1998].
the amount of time advanceor delay, in addition to Several studies of stress transfer between mainshocks
beinga functionof the stressstep and the long-term and aftershocks found that the correlation between stress
loadingrate, alsodependsuponother fault constitutive change and seismicity rate change disappears below
parametersand on whenin the seismiccyclethat stress somesmall stress,typically about 10 kPa (e.g., 1992
stepis applied,i.e. the stateof the faultsurface[Harris LomaPrieta [Reasenberg andSimpson,1992];1992Lan-
andSimpson,1998].In boththe Coulombfailuremodel ders [Hardebeck et al., 1998];1995 Kobe [Todaet al.,
and the rate-and-statemodel,there is no physicalrea- 1998];and 1987SuperstitionHills [AndersonandJohn-
son why a lower threshold should exist.
son,1999]).Whetherthis is a physicallowerthreshold
for triggering or merely a lower threshold for detection
Copyright 2000 by the American GeophysicalUnion. is as yet an unresolved
question[Harris, 1998;Harde-
becket al., 1998]. The main conclusionof this study
Paper number 2000JB900081. is that in central California static stresschangesmuch
0148-0227
/ 00/ 2000JB900081$09.00 smallerthan 10 kPa have a noticeabletriggeringeffect.
13,631
13,632 ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES

237 ø 238 ø 239 ø 240 ø closelymatchesthe orientationand geologicoffsetof


the nearest active fault; for those earthquakes with re-
39 ø ß ., .,.......
........... ...... ....:::..' 39 ø ported surfacerupture, we choosethe nodal plane most
ß
"' "•:•
.......
:'-•""::•....i::.•i'"•.":.•.
.... ..0:•,•::::
?:•"•::•'•..
....:..%.,•.
•'•' •
. ....
•-:•
•t•:"•
.•........
:-•::•...-•.

closelymatchingthe reported plane. Through this pro-
cesswe exclude11 earthquakesfor which neither of the
38 ø focal mechanismnodal planes seemconsistentwith the
surroundingfaults. In the appendixwe presentthe re-
sults of Monte Carlo simulations which indicate that
the main conclusionof this study would not changeif
37 ø those 11 earthquakeswere included. For the remaining
37
* 60 eventswe assumesquareruptures and compute rup-
ture dimensionand slip basedon the scalingof slip u
with sourcedimension
L [Scholzet al., 1986],
36 ø 36 ø
u - 6- 10-SL, (1)

35 ø
237 238 239 ø 240 ø
35 and the moment-magnituderelationshipof Hanks and
Kanamori[1979],
logm(Nm) - 1.5M+ 9 - 1.5(GL2u)
+9 (2)
Figure 1. Map showingthe studied area and the 64 where m is seismicmoment, M is magnitude,and G is
M _>4.5 earthquakes between 1969 and 1998 that are the shear modulus taken to be equal to 30 GPa.
includedin this work (data from NCEDC). Solidlines Using a three-dimensionalelastic half-spacedisloca-
indicateactivefaults[Jennings,1992],anddottedlines tion program[Erickson,1986],we computethe change
indicateblind faults[WorkingGroupon NorthernCal- in normalstressresolvedon the fault planeAa (positive
ifornia EarthquakePotential, 1996]. Two eventsnear in tension)and the changein shearstressin the direc-
the Great Valley fault with uncertain fault plane are
highlightedin gray. SAF, San Andreasfault; C, Calav-
tion of slipA•-slip,inducedby previous
earthquakes
on
eras fault; H, Hayward fault; and GVF, Great Valley the earthquake in question. In those calculations,the
fault. shear modulus and Poisson's ratio were chosen to be
equal to 30 GPa and 0.25, respectively.The changein
Coulombstress(ACS) that the jth earthquakein the
2. Method cataloginduceson the hypocenterof the/th earthquake
We use a subset of the Northern California Earth- is givenby [Reasenbergand Simpson,1992]
quake Data Center (NCEDC) fault solutioncatalog, ACSij-- ATi•
ip•-•t'A(•ij (j < i), (3)
which includes earthquakes with magnitude 4.5 and
higher occurring between January 1969 and January where•u'is the effectivecoefficientof friction, account-
1998 and located within the regionenclosedby the rect- ing for the effect of changesin fluid pressure. In the
angle in Figure 1. The 75 earthquakessatisfyingthese remainingof this paper we assume•u• -- 0.6 and discuss
criteria are listed in Table 1. The major faults in this the implications of different values in the next section.
area are the San Andreas, the Calaveras, and the Hay- The cumulativeCoulombstresschange(ZACS) on the
ward, all of which are right-lateral strike-slipfaults, and ith earthquake as a function of the time prior to that
the Great Valley fault, which is a low-angleblind thrust earthquake,At, is obtainedby adding the effectsof the
[Steinand EkstrSm,1992]. previousearthquakesaccordingto
We modeled earthquakesas rectangular dislocation
planes in an elastic half-space. For the four largest EACSi(At) - • ACSij (ti- tj) _>At, (4)
J
earthquakeswe define the sourceparameters based on
the available geodetic, strong motion, or teleseismic whereti and tj are the time of the earthquake
in ques-
studies(M-5.6 1979CoyoteLake[Du andAydin,1993]; tion and the times of previousearthquakes,respectively.
M=6.5 1983Coalinga[HartzellandHeaton,1992];M=6 We chooseto studythe cumulativestresschangeas a
1984MorganHill [HartzellandHeaton,1986];and the functionof time insteadof simplyat the time of the rup-
M=7 1989 Loma Prieta [Wald et al., 1991]). We do ture because,as we will show later, this approach pro-
not attempt to model complex slip distributions in de- vides additional insight into the processof earthquake
tail. For example, the Loma Prieta earthquake is mod- triggering.Becausefor a givenearthquakeY•ACS(At)
eled as two dislocationsonly. Although slip distribu- is the sum of all stresschangesfrom the first day in the
tions are greatly simplified, seismicmoment and aver- cataloguntil time At beforethe rupture of that earth-
age slip match those proposedin the referenceslisted quake, the sameAt for earthquakesthat occurredearly
above. For the rest of the catalog, we selectas a rup- in the catalog implies a summationover a shorter time
ture plane the focal mechanismnodal plane which most period than that for an earthquakethat occurredlate in
ZIV AND RUBIN: TRIGGERING AT SMALLSTRESSES 13,633

Table 1. Seventy-Five
M >_4.5 Earthquakes
1969-1998
Date Latitude, Longitude, Depth Magnitude Notes
øN øW km

Oct. 27, 1969 36.77 121.37 10.5 4.6


Aug.4, 1970 36.73 122.06 9.8 4.7
March9, 1971 36.79 122.13 5.9 4.6
April16,1971 36.80 122.14 ,0.7 4.5
Feb.27, 1972 36.54 121.08 11.0 4.7
Sept.4, 1972 36.64 121.25 5.3 4.7
Oct. 3, 1972 36.80 121.52 6.6 4.8
Nov. 12, 1973 37.24 121.97 13.5 4.5
Nov. 28, 1974 36.92 121.47 6.0 5.0
Aug.3, 1975 36.45 120.45 19.0 4.9
Aug.15,1975 36.46 120.42 11.5 4.6
Sept.13, 1975 36.00 120.54 11.6 4.8
Jan. 14, 1976 36.08 120.23 10.5 4.9
May8, 1979 37.29 121.66 7.4 4.6
CL
Aug.6, 1979 37.10 121.51 8.9 5.6
Jan. 24, 1980 37.83 121.77 14.5 5.4
Jan. 27, 1980 37.75 121.70 14.7 5.1
April13,1980 36.78 121.50 8.8 4.7
Jan.7, 1981 36.87 121.62 9.7 4.6
Jan. 15, 1981 37.37 121.73 9.4 4.6
Aug.11,1982 36.65 121.27 11.0 4.5
Oct. 25, 1982 36.32 120.51 12.1 5.3
CO
May2, 1983 36.23 120.32 9.8 6.5
as
May3, 1983 36.28 120.37 11.1 4.7
May3, 1983 36.24 120.30 8.8 4.7
as
July9, 1983 36.25 120.40 8.5 4.8
July22, 1983 36.24 120.41 8.0 5.2
July22, 1983 36.22 120.41 8.4 4.8
July25, 1983 36.22 120.40 8.6 5.0
Aug.29,!983 35.84 121.34 5.9 5.3
Sept.9, 1983 36.21 120.26 7.0 5.4
Sept.11,1983 36.24 120.39 9.7 4.7
Jan. 23 1984 36.36 121.89 10.0 4.9
Jan. 23 1984 36.36 121.89 8.7 4.5
MH
April24 1984 37.31 121.68 8.6 6.0
Aug.4, 1985 36.14 120.16 13.4 4.6
Aug.4, 1985 36.15 120.15 16.0 5.5
Nov. 24, 1985 36.03 120.88 10.4 4.6
Nov. 28, 1985 36.58 121.03 10.5 4.6
Jan. 14, 1986 36.58 121.18 7.1 4.8
Jan. 26, 1986 36.80 121.28 8.8 5.5
March31, 1986 37.48 121.68 9.0 5.7
May31, 1986 36.64 121.25 6.0 4.7
Feb. 14, 1987 36.17 120.34 15.4 5.2
Feb.20, 1988 36.79 121.31 9.7 5.0
June13, 1988 37.39 121.74 9.7 5.0
June27, 1988 37.13 121.90 13.4 4.7
July26,1988 36.58 121.15 4.1 4.5
Nov. 10, 1988 37.37 121.73 8.7 4.8
April3, 1989 37.43 121.77 11.8 4.5
Aug.8, 1989 37.14 121.93 13.9 5.4
Aug.8, 1989 37.16 121.95 15.1 4.5

the catalog.Notethat the cumulative Coulomb stress 3. Results


change EACS canbeeitherpositive or negative.
Posi- 3.1. Triggering Due to Cumulative Stress
tiveEAC$ promotes rupture,whereas negativeEAC$ Changes
inhibitsit. Throughout this paper,earthquakesthat
receivea positiveEAC$(At = 0) are referredto as In Figure2, the numberof positiveEACS values
triggered earthquakes,
regardlessof the magnitudeof is compared to that of the negativeEACS valuesas
r,xcs(,xt = o). a functionof time prior to rupture. The numberof
13,634 ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES

Table 1. (continued)
Date Latitude, Longitude, Depth Magnitude Notes
øN øW km

Oct. 18, 1989 37.04 121.88 16.8 7.0 LP


Oct. 18, 1989 37.05 121.78 4.1 4.8 as

Oct. 18, 1989 37.19 122.05 15.4 5.1 as

Oct. 18, 1989 37.03 121.73 1.8 4.5 11c

Oct. 19, 1989 36.93 121.24 12.2 4.5 11c

Oct. 19, 1989 36.97 121.84 11.5 4.6 as

Oct. 21, 1989 37.05 121.86 13.4 4.5 as

Oct. 21, 1989 37.06 121.88 14.9 4.7 11c

Oct. 25, 1989 37.06 121.81 10.4 4.7 11c

Nov. 2, 1989 37.06 121.80 9.5 4.7 11c

April 18, 1990 36.93 121.66 5.7 4.9 as

April 18, 1990 36.93 121.66 5.8 5.4 as

April 18, 1990 36.96 121.68 6.4 5.1 as

Feb. 4, 1991 36.03 121.55 5.3 4.7 IIC

March 24, 1991 36.95 121.71 15.7 4.2 11c

Sept. 17, 1991 35.82 121.33 6.0 5.2


Jan. 16, 1993 37.02 121.46 7.9 4.8
Aug. 11, 1993 37.31 121.68 9.3 4.6
Nov. 14, 1993 35.95 120.50 11.6 4.6
Dec. 20, 1994 35.92 120.47 8.8 4.7
April 23, 1995 36.60 121.20 8.0 4.7
May 21, 1996 37.36 121.72 8.1 4.5
Aug. 12, 1998 36.75 121.46 9.2 5.3

CL, CoyoteLake; CO, Coalinga;MH, MorganHill; LP, Loma Prieta;


nc, not considered;as, within 30 km and 6 monthsof a M _>6.5.

positive EACS values is shown in the upper half of of [EACS[ in Figure 2b between1 kPa and 10 kPa.
eachdiagram,and the numberof negativeEACS values However,this would result in a significantreductionin
is shown in the lower half. In addition, to summarize the total number of •AC$ valuesin that diagram and
these diagrams, we plot for each diagram the ratio of weaken the statistical validity of the result. The fact
positive EACS values to the total number of EACS thattheabruptriseofthert+ot
curvetoward0.7shortly
values as a function of time. In the remainder of this beforerupture is precededby a longtime intervalduring
paper, these summary curves are referred to as rtot which the curve is relatively stable providesconfidence
curves," and the fractions of positive EACS at At - 0 that this result is statistically significant. The results
are referredto as .,-tot(O)
ratios." Becausethe number of Monte Carlo simulations presentedin the appendix
indicate
of EACS valuesdecreaseswith increasingtime before that theshapeof thert+otcurves in Figures2b
rupture,abruptjumpsin the rt+ot
curvesa longtime and 2c would remain almost unchangedif the 11 events
beforerupture are statistically insignificantand should for which focal plane solutions were inconsistentwith
be ignored. The distributions of stressesgreater than the surroundingfaults were includedin the calculation.
10 kPa, smaller than 10 kPa, and smaller than I kPa In Figure 3 we explorethe effectof differentvaluesof
are shown in Figures 2a through 2c. In each of the it' onthert+ot(O)
ratios.Thisreveals
thatrt+ot(O)
ratios
three diagrams there is an increasein the fraction of increasewith increasingvaluesof it' for 0 < it' < 0.6
positiveEACS valuesbeforethe rupture. For IEACSI and becomealmost insensitiveto it' for it' > 0.6. We
greaterthan 10 kPa, the rt+ot(O)
ratio is about0.85. note that this result is at odds with previous studies
We point out that this ratio can be raised to 0.95 by on stresstransfer in California that find a generaltrend
choosingthe other nodal planesfor two eventsfor which for the fraction of encouragedearthquakesto increase
this choicewas least certain (the gray/white circlesin with decreasing/•';in thosestudiesthe bestresultswere
Figure 1). This is becausethe EACS(At = 0) valuesof obtainedfor 0 <_ it' <_0.2 [Simpsonand Reasenberg,
thoseeventschangefrom large negative(<-10 kPa) to 1994; Harris et al., 1995; Harris and Simpson, 1996;
Deng and Sykes,1997].
a smallnegativeand to a largepositive(> 10 kPa) if the
In summary,in Figures2b and 2c an increaseis found
other nodal plane is chosen.For IEACSI smallerthan
10 kPaandsmallerthan i kPa,the rt+ot(O)
ratiosare in the fraction of triggered eventsshortly beforerup-
about 0.7. Note that the data in Figure 2c are contained ture. In the next section we show that this result is
in Figure 2b, and thereforethe two are not independent. significantin the sensethat it is very unlikely to be
This could have been avoided had we binned the range found in a random catalog.
ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES 13,635

TIME BEFORE RUPTURE [days] with respectto time meansthat the map in Figure 1 is
not changed,but the sequencein which the earthquakes
rupture is different, whereasrandomizationwith respect
3O ! I I I I I I I ! I • I I I i i 1.0
to focal mechanismsmeansthat the black/white cir-
20
MACS(at)
>10kPa 0.9 cles are switchingposition. The main purposesof this
0.8 exercise
are (1) to seeif the rt+ot(O)
ratios(Figure2)
10
0.7 are significantlylarge and (2) to see if the geometry
• - 0.6 of faulting is suchthat the high degreeof triggeringis
.. .,,

...........
'........""•
....•'":"x O• +' independentof the sequence of earthquakes(if, for ex-
........•..:
..... .•..• -• •:•........ ::..... g•
ample, all earthquakesare coplanarand nonoverlapping
ß

..........
.........: ..

.............
' - 0.4
10 and havethe samefocal mechanism,earthquakeswill al-
0.3
ways increasethe Coulombfailure stresson subsequent
0.2 earthquakes).
20 A
EACS(A10kPa 0.1 Observed rt+ot
curves
(heavysolidlines)arecompared
30 I t I I I I'I I I t I I t l'"t'"i I I I , 0.0 withrt+ot
curvesof50catalogs
randomized withrespect
1.0 to times (Figure 4) and 50 catalogsrandomizedwith
0.9
respectto focal mechanisms (Figure 5). Here the main
0.8
ideais to seehowmanyof the synthetic
rt+ot(O)
ratios
exceedthe observedratio. These diagramsshow that
0.7
rt+ot
(0) ratiosaresignificantly
reduced
owingto random-
• 10 0.6 izing the times and the focal mechanisms.
• 0 The decreasein triggering is strongerfor catalogsin
• 10 0.4 whichfocalmechanisms are randomized(Figure5) than
r• 20 0.3 for catalogsin whichtimes are randomized(Figure 4).
0.2
The fraction of positive MACS valuesin catalogswith
randomizedfocal mechanisms is on average50%. That
0.1
is, if faulting geometrywas random, earthquakeswould
50 0.0 have had equal chancesof being delayedor triggered.
On average,the fraction of positiveMACS valuesin cat-

20! 0<EA•(At)
<1kPaf0.9 alogswith randomizedtimesis about 65% for ]EACSI
valuesgreaterthan 10 kPa, about60% for ]EACSI val-
ues smallerthan 10 kPa, and about 50% for IEACSI
10 ;:;•:.•:i%.:::.:.?
;:.' ::.:•:
..........
f:'•
•:•'•;::•:•'•:::"'--
•:.•.•
..::;:•.77•
. .-:?•
j 0.7
?.:•?:"
0.6 of a few
valuessmallerthan I kPa. The catalog we use consists
mainshocks that are followed and sometimes
":•:•m•:•::::'•
..' .........• '•:•"''>..,•'•-:&
...............
::•,•:?•::::•:•.
'•-•-•-.• '.'•..:.•
.......•:•.:• '.•.'-'5.
'•:.•:.
'•;.'......
......

•'•
::.:. • precededby periodsof increasedseismicactivity (see
Table 1), i.e., the catalogis clusteredin time and space.

'•:•'?':•='""='•?'
':•=
.

0.2
c o.1
• ' t t t i :i 1 I I t t t I i: I t I t I I 0.0
IzACSl> 10 kea
[] IZaCSl< lO kPa
0.9 - ¸ IEACSl
< 1kPa
T•E BEFO• .R•• [days]

Figure 2. Distributionof EACS values(•u'= 0.6) asa I).8-


functionof the time prior to the earthquake(left axis):
(a) IaC$1 > 10 kPa, (b) Iac$1 < 10 kPa, and (c)
lEACS1 < 1 kPa. Alsoplottedis the ratio betweenthe 0.7
number of positive MACS valuesand the total number
of MACS valuesas a functionof time (right axis). In
thetext,thesecurves
arereferred
to asrt+ot
curves. 0.6

0..5 I

3.2. Comparison With Randomized Catalogs 0.0 0.2 0.4 0.6 0.8 1.0
We computed
rt+ot
curves
for synthetic
catalogs
pro-
ducedby shufflingthe times in the original catalog and
for synthetic catalogsproducedby shufflingthe focal Figure 3. Effectof different
values
of•u'onthert+ot(O)
mechanismsin the original catalog. Randomization ratios.
1.0 Figure5. Comparison between
observedr,+ot
curves
0.9 -
(heavy
solidlines)andthose
calculated
for50synthetic
catalogsthat wereproduced by randomizing thefocal
0.8 mechanisms (lightlines)for lEACS[values(a) >10
kPa, (b) <10 kPa, and(c) < I kPa. Insetsprovide
0.7 enlargedviewsof the resultsat thevicinityof theob-
servedr,+ot(O)
ratios.
0.6

0.5 1.0

0.4 0.9 -

03 0.8

0.2 0.7

0.1 0.6

0.0 0.5 -
0.4

03

0.2,

0.1

0.6 0.0

+' 0.5

0.4- 0.9
'":"',
"[ IZzXcs,(zXt)l
o.8 ................
< •okPa

' ,'" ',.',, ,


0.7 '...,",,
./ ,.-",'x />t'.. :•;.,
',.........
.,' ?}•:'?:,,%,
,,.,,,:,½;',,
i",i
:/•.,'16•---;•;-'•>'.•;,•,,;e;-.
......
, ,',•.,,'•,,.*•
.- ..-..
L--' ";:::•t:'.•--' •;•.:•:•.•!.:,.;f
.........
'w•.w .•--..,,:½,'.,,,-•
•,,:• ,',,,

.10•0 0 '•":'..;.v,_,"
".• •$.,.,•,:
•,',:
.....:••'.'-'"';'
".".• • '-';•
"•.•,."'
'. '••.... ':

I I I I I I I I I I I I I I I I I I I I I + ,•.•
.:•i:.•I
t;•.•!,•.•
iizACS(At)
I

0.9
o.s -;"/?,:,,:,;C
...........
/ i.,.:
!•?v-'>',•.
:.,•_:.½
•,'• ,•::/',.
.......
//;i:::u
'/ :' ",'•i,X---<
:-;.-,"',.
'.:':.-•--:;.J(,•'; "•",,
"'•-•',•
..... ".',]:":,

,,,,.,.,,,. 0.0 I• II';I I I' I I I I I I f I I I i I I I I


0.6
, lI,,.,,
.......
i,, ,:. ., .,,.........,,.,,,,.-,,:
,,.,,............ ,......... 0.9o.8
•!l!l/i
i
0.8- '%?x,./"'•
.:,,::>.'•:v,'.:':•.- ;;7,".•:,.,,""
.....• ........
i,]'•.-Y'•iv'-
.......
..... 'v"\ •"•'
.....',:, .......
1
i ... I--
0.7 •,,.._.•.p..,,,,,•
:•::•.:,<,,•
;• ",•:'...t'\"
[('"'"B,.'.'rY'--
"Y•':•i?';;
"),':'
•":•":
....
4"-',,:,,:,:,,
.....•;,,½•'e.
'"':
'........
,,'........
*:',?'•...'
':•
0.6 o.6 i.Y':
½'•,',,•,rP,,:
[•"',: •-,'•'
I':,"•"'
,"i ', 7 <,
•_. ,,....,.,;,:• . .;
! I I ! I I I I I
0.5: i•,:
,:. '-.",
:i-:•.ii"!i,.'1."'."
',: "- •,""

TIME BEFORF, RUPTURE [days] 0.3 i:,7?i:,•jl•i'½i"i?i•i"F'":"""':-f••:.'•


"•'•"
"'"'•••
'"'?'"
'"
'-"'''• .....
t•'...•,
, ":i g-,A.IU•,K,.•--'.--..•
.. :•.:,t.:
•.,:::i. ,.,.:•/,,__':_•,t...,
,,•,,f
œ"/ ¾ \ ,.,,.•,,.•,.__.,
/v / ,.,,,-v_/

Figure 4. Comparison between observedr,+ot


curves 0.1- (
(heavysolidlines)andthosecalculated
for50synthetic
catalogsthat wereproducedby randomizing
the times I t I I I I I I I
(lightlines)for [EAC$I values(a) >10 kPa, (b) <10
kPa, and (c) <1 kPa. Insetsprovideenlarged
viewsof
theresults
at thevicinityoftheobserved
r,+ot
(0) ratios. TIME BEFORE RUPTURE [days]
ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES 13,637

Following randomization, the synthetic catalog is still with large EACS are locatedcloseto a previousearth-
clusteredin time and space,but those two are uncorre- quakeand are thereforevery sensitiveto changesin the
lated and the pattern of foreshocksand aftershocksthat sourceparameters of that earthquake, whereasthose
characterizesthe original catalogis lost. It thus follows with small Y•ACS are located further away from preced-
that much of the triggering is due to interaction with ing earthquakesand are thus lessaffectedby changing
nearby events that occurred shortly before rupture. In source parameters.
addition, in the original catalog strike-slipearthquakes The sign of Y•ACS of a particular earthquakein a
tend to occur next to strike-slip earthquakesand thrust particular time window is well resolvedif the standard
earthquakesnext to thrust earthquakes. In catalogsin deviations of all EACSMc values is considerably less
which the focal mechanismswere shuffled,earthquakes than Y•ACS, that is, for points falling well below the
with very different focal mechanismsmay be located lines in Figures6a-6b. Overall, most of the standard
next to each other. It is therefore clear why randomiz- deviationsfall below the line. To quantify the distribu-
ing the focal mechanismsresultsin lesstriggering than tion of the standard deviations with respect to the line
randomizing the times. as a function of EACS, we define the parameter R as
To testthe nullhypothesisthat theobservedrt+ot(O)
ratiosareno greaterthanthe syntheticrt+ot(O)
ratios,
wecomputed rt+ot(O)
ratiosfor 1000synthetic
catalogs R-log(STD(EACSMc))
IZCSl ' (5)
with randomized times and compared them with the
whereSTD(EACSMc) is the standarddeviationof the
observedratios. Testing the null hypothesisagainstcat-
Monte Carlo Coulomb stresschanges. The parameter
alogswith randomized times is more strict than testing
R is the distance along the vertical between a point in
againstcatalogswith randomizedfocal mechanismsfor
the diagram and the line and is negative if the stan-
the reasonsmentioned in the precedingparagraph. In dard deviation is smaller than EACS and positive if it
this test the null hypothesiscan be rejectedat a 95%
is larger than EACS. The mediansof all R parameters
levelofconfidence
if theobserved
rt+ot
(0) ratioisgreater within bins of logIr,/xcsI that are one unit wide are
than that of 95% of the syntheticcatalogs. We found
plotted in Figures 6c-6d as a function of EACS for 1
that the null hypothesiscan be rejected at the 99%
and 2 standard deviations. These plots show that it is
confidencelevelfor IrA½$1 > 10 kPa and at the 95% generallypossibleto distinguishpositivestresschanges
confidence levelfor IEACSI < 10 kPa and IEACSI < 1 from negative ones even for very small stresses. In
kPa. We concludethat cumulative static stresschanges
fact, it is easier to resolvethe sign of EACS for small
much smaller than 10 kPa do have a significanttrigger-
IZCsI, becausethe fractionaluncertaintyin EACS
ing effect.
due to uncertainty in both location and focal mecha-
nism is larger for nearbyevents(with large IEACSI)
4. Discussion than for distant events. Slip distributions are not per-
turbed in the Monte Carlo simulations. Perturbing the
4.1. Can Very Small Stress Changes Be slip distributionswould mainly affect stresseson nearby
Resolved?
events. It is thus expected that incorporating the slip
Our stress calculations have uncertainties that stem distributions into the Monte Carlo simulations would

from uncertaintiesin the catalog. To distinguisha pos- further increasethe uncertaintyof large Ir,x½sI val-
itive Y•ACS from a negative one, those uncertainties ues but would hardly affect the uncertainty of the small
must be smaller than the Y•ACS itself. It is essential Ir,/XCSl values.
to seeif and how the uncertaintiesare dependentupon
Y•ACS and whether it is possibleto distinguishpositive 4.2. Effect of Stress Change on Time to
stresschangesfrom negativeonesfor IEACSI smaller Rupture
than 10 kPa. Accordingto the Coulombfailure hypothesis,follow-
To addressthose questions,we performed 500 Monte ing a Coulomb stressdecreaseon a given fault plane,
Carlo simulationsin which the strikes,dips, rakes,and the time of the next rupture on that plane is postponed
hypocentersof all the earthquakeswere perturbed ac- at least until the Coulomb stress is recovered as a result
cording to the published ranges of uncertainty in the of otherearthquakesand/or long-termtectonicloading.
N CEDC catalog. In every Monte Carlo simulationwe Thus,aswaspointedout by Harris andSimpson[1998],
computed the cumulative Coulomb stressEACSMc at knowledgeof the time and magnitudeof the stressde-
the epicentersof 63 earthquakesfor 70 time windows. In creasecan be used to test the Coulomb failure hypoth-
Figures 6a and 6b, the standard deviationsof the Monte esis. As a simple example, consideran event for which
Carlo Coulombfailure stressesare plotted againstthe the most recentstresschangeis negativeand whichthus
cumulativeCoulombstressof the unperturbedcatalog can be used to test the Coulomb failure hypothesis. In
for all negativeand positiveY•ACS values,respectively. that casethe Coulombhypothesiswould be rejectedif
These plots show that the standard deviation is posi- the magnitude of the stressdecreasewas larger than
tively correlatedto EACS. This is becauseearthquakes the product of the long-term loading rate and the time
13,638 ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES

-ZACS [kPa] ZACS [kPa]

10-410-310-z10-1100 101 10z 103 10-410-310-2104 100 10• 102 103

10
3 I I I I I I•,• I I I I I I I , 10
3
101 101•
1oo - 1oo]
10'1•*'* ****•:**
• - 10'1•
lO-,
10'4 , •: • • • I I - 10
lO-,*
'4

C D

,o
-1 , , , , , , • , , , , , , •-1
10-410-310-z104
100 10• 10z 103 10-410-310-z104100 10• 10z 103

-ZACS [kPa] ZACS [kPa]

Figure6. (a,b)Log-log
diagrams
showing
thestandard
deviations
of500Monte
CarloEACSMc
against[EACS[ oftheunperturbedcatalog
for(a)negative
and(b)positive
EACSvalues.(c,
d) PlotsofthemedianofR, log[STD(EACSMc)/[•ACS[],against
•ACS oftheunperturbed
catalog
for(c)negative
and(d)positive
•ACS values.
ThemedianofR iscomputed
forbinsof
log[•ACS[ thatareoneunitwide,with50%overlap
between
twoadjacentbins.

intervalbetween
that stress
changeandthe earthquakeas a functionof the stressrangeand the time interval.
in question.However, because
in equation(4) stress In theseplots, 30 regionsare definedby 5 time bins
changes are summedfrom the first day in the cata- and6 stressbins.The regionsareshaded accordingto
log until At beforerupture,the valueof EAC$ at the the numberof earthquakes
with negativeEAC$- val-
timeof the mostrecentstresschange
reflects
not only ueswithinthat region.The distributionof all negative
that stressreduction,but insteadthe sumof all stress EAC$- valuesis shownin Figure 7a. Uncertainties
changesup to and includingthat moment. Therefore of EAC$- valueswerecalculated in the samewayas
theCoulomb stress
change asdefined in equation (4) is in section4.1. In Figure 7b, EAC$- valuesthat are
inadequatefor the purposeof this test of the Coulomb smaller thanthecorresponding $TD(EAC$•c ) values
failurehypothesis. Herewe computea differentcumu- are excluded.In addition,in generating that figurewe
lativestress change by addingthe stress
changes from eliminated the two earthquakes nearthe GreatValley
the time of the earthquakein questionbackto a time thrustwith uncertainfault planesreferredto in section
At beforethe earthquake accordingto 3.1 (thegray/whitecirclesin Figure1). Forlong-term
stressing
ratesof 20 and40 kPa/yr, weplotasa function
•/kCS•-(/kt) - E/kCSij (ti- tj) _•At, (6) of EAC$- theminimumtimeperiodthat according to
J
the Coulomb failurehypothesismustelapsebetween the
wherethe "minussign"superscript indicatesthat stress application
ofthestress change andtheearthquake (the
changes
areaddedbackward
in time,andti andtj are dotted lines). The lack of data belowthe dotted lines
definedasin equation (4). In Figures7aand7bweplot is not only consistentwith the predictionthat the time
the distributionof the negativeEACS- earthquakesdelayis proportionalto the magnitudeof the stressde-
ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES 13,639

11

11
.......
:.•&• :::•

:::•'• ....

ia:i•::...a:::....5
.,..
lO lO

-zaCS- g•Pa] -ESCS- [kPa]

Figure 7. Plot of the distributionof negativeY•ACS- earthquakesas a functionof time interval


and stressrange, where Y•ACS- indicates that the summation is done from the time of the
earthquaketo a time At before. Each region is shadedaccordingto the number of earthquakes
with a negativeEACS- valuewithin that region. (a) Distributionsof all negativeEACS- values.
(b) The samedistributionafter the removalof the very uncertainEACS- values.Dotted lines
indicatea reasonable
rangefor long-termstressing
rateson faultsin centralCalifornia[Simpson
and Reasenber•7,
1994].

I I I I I I I t I I i I t t t I I I I
crease, but is also consistent with the Coulomb failure
hypothesis.Clearly, the amount of data in Figure 7b is EACS(At) > 10 kPa
very small, and more data are needed before a definite
conclusion can be made.

4.3. Lower Threshold in Studies of Aftershock


Sequences .A T.ACS(At) <-10 kPa
I t i I I I t I i t I I I I t t I i t

The catalog we use includestwo mainshocks(M >_ ! I I I I I I 1 I 1 I i t I I t t I I I

6.5) that were followedby periodsof increasedseismic


activity. The distributionof IY•ACSI valuesof earth-
quakes that occurred within 6 months after a main-
shock and within 30 km from a mainshockepicenter is
1.0

0
0<2ACS(At)
<10
kPai
'," % •_I•; :-.'..... ;•.."" --' ' ":

shownin Figure 8 (the 11 eventsthat satisfythesecri- I0

teria are labelled "as"in Table 1). This showsthat the B 0 > EACS(At) >-10 kPa
I t t . t ,, I I
earthquakepopulationthat contributesto the rising of
thert+ot
curveshortlybeforerupturefor IEACSI< 10
kPa (Figure 2b) differssubstantiallyfrom the popula- TIME BEFO• RUPTURE [days]
tionthat contributes
to the risingof the rt+ot
curvefor
IEACSI > 10 kPa (Figures2a and 8a) and that an in-
Figure 8. Distribution of EACS values(p' - 0.6)
crease in the fraction of positive EACS smaller than as a function of the time prior to the earthquake for
10 kPa could not have been observed had we restricted
earthquakes that occurred within 6 months and 30 km
our study to the aftershock sequencesin our data set. of a M > 6.5: (a)IEACSI > 10 kPa and (b)lEACS] <
Also notice that in our data set, earthquakeswith cu- 10 kPa.
13,640 ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES

mulative stresschangesmaller than and larger than 10 do havea significanttriggeringeffect. Randomizingthe


kPa are equallyabundant(i.e., Figure2). catalogwith respectto either time or focal mechanism
Previous workerswho studied much larger data sets resultsin a significantreductionin the fractionof trig-
of aftershock sequencesreported that the correlation geredearthquakes.We concludethat if faultinggeome-
betweenstresschangeand seismicityrate changedisap- try were random, earthquakeswould haveequal chances
pearsbelow30-10 kPa (e.g.,1992LomaPrieta [Reasen- of being delayedor triggered. Monte Carlo simulations
bergand Simpson,1992]; 1992 Landers[Hardebeck et indicate that we are able to distinguishpositive stress
al., 1998];1995Kobe [Todaet al., 1998];and 1987Su- changesfrom negativeones,evenif EACS is very small.
perstitionHills [Anderson andJohnson, 1999]).Whether We show that the time delay due to a stresschangein-
this reflectsa physicallowerthresholdfor triggeringor creaseswith the magnitudeof that stresschange.
merely a lower threshold for detection due to the fact
that aftershock rate decreases with distance from the
Appendix
mainshock,and thus becomesstatistically undetectable
for very small stresschanges,is as yet an unresolved As is mentioned in section 2, 11 events with focal
mechanisms dissimilar to the focal mechanisms of the
question[Harris, 1998;Hardebeck
et al., 1998].Studies
of aftershocksequences computethe stresschangedue nearby active faults were not included in our calcula-
to the mainshockand disregardinteraction betweenaf- tion. The largest of these earthquakeswas M - 5.5,
tershocks. However, the effect of a small but nearby the secondlargest was M = 5.4, and the remaining 9
event may be more significantthan the effectof a large
and distant event. Thus accountingonly for the effect
1.0
of the mainshockmay not be appropriate for the cal- ..?=.=.=.,.•.•.
:n=•.=• •.•?•=• ................

culation of the stresschangeson aftershocksthat are 0'9 -


0 25 50 75 I00
located far from the mainshock.In addition, although 0.8 -
% probability
the number of eventsin studiesof aftershocksequences 0.7-
is very large, the fraction of stresschangessmallerthan 0.6-
10 kPa is very small. Thus owingto the scarcityof small
stresschangesin aftershocksequencesand the neglect -

of interaction with nearby aftershocks,aftershockdata 0.4-

setsmay not be ideal for the searchfor a lowerthreshold 0,3-


in triggering. 0.2-
o

0.I
4.4. Comment on Tidal Stresses
0.0
Whether tidal stressestrigger earthquakesis highly
controversial[Emter, 1997]. If tidal stressesdo trig-
ger earthquakes,the amount of suchtriggeringis very
small [Vidale et al., 1998a,b]. Why then do static •ol 0 25 50 [
stress changesthat are of the order of tidal stresses o.o-• %probability F
0.7 •
have a greater triggering effect? A major difference
betweentidal stressesand static stresschangesgener-
ated by coseismicslip is that they act over very dif-
ferent timescales. While tidal stressesare periodic, 0.4
static stresschangesremain until the time of the next 0,3
earthquake, which may either intensify or weakenthem. 0.2
Rate-and-statemodels[Dieterich,1994;Gomberg et al.,
1998]predictthat the longerthe durationoverwhicha
stresschangeacts, the greater is its effect on the time 0.0

of the future earthquake. Thus in order for tidal stress


and static stresschangeto have an equal effect,the am- TIME BEFORE RUPTU• [days]
plitude of the tidal stressmust be larger than that of
the static stressstep. Figure 9. Probability distribution maps of Monte
Carlort+otcurvesfor [EACSI < 10 kPa, computed
5. Conclusions from 300 simulationsfor (a) the 64 earthquakesof this
study due to adding the effect of the 11 earthquakes
We use a three-dimensionalelastic half-spacemodel
for which neither of the nodal planes was consistent
to computecumulativestatic Coulombstresschangeon with the nearbyfault and (b) all 75 earthquakesin the
the hypocentersof 63 M >_4.5 earthquakesin central N CEDC catalog meeting the original selectioncriteria
California between 1969 and 1998. We find that cumu- in terms of origin time, epicenter,and magnitude. The
lative static stresschangesmuch smaller than 10 kPa solidcurveis theobserved
rt+ot
curveasin Figure2b
ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES 13,641

were M ( 5. Here we examine the extent to which the Deng, J., and L. R. Sykes,Stressevolutionin southernCali-
r•+ot
curve
for[•AC$ I • 10kPa(i.e.,Figure2b)maybe fornia and triggering of moderate-, small-, and micro-size
affectedif thoseearthquakesare includedin the model. earthquakes, J. Geophys.Res., 102, 24,411-24,435, 1997.
Dieterich, J., A constitutive law for rate of earthquake pro-
To addressthis question, we ran Monte Carlo simula- duction and its application to earthquake clustering, J.
tions.In eachsimulation wecomputed ther•+otcurvefor Geophys.Res., 99, 2601-2618, 1994.
a catalog that combinedthe 64 earthquakesfor which Du, Y., and A. Aydin, Stresstransfer during three sequential
focal mechanisms were consistent with the surround- moderate earthquakesalong the central Calaverasfault,
ing activefaults with the remaining11 earthquakesfor California, J. Geophys.Res., 98, 9947-9962, 1993.
Emter, D., Tidal triggering of earthquakes and volcanic
which slip planes were randomly picked from the two events, in Tidal Phenomena, Lect. Notes in Earth Sci.,
possiblenodal planes. The results of 300 such simula- vol. 66, edited by H. Wilhelm, W. Ziirn, and H.-G. Wen-
tions are plotted in Figure 9 in the form of a probability zel, pp. 293-310, Springer-Verlag,New York, 1997.
distribution mapandiscompared withtheobserved Erickson, L., A three-dimensionaldislocation program with
curve in Figure 2b. In interpreting the result it is im- applicationsto faulting in the Earth, M.S. thesis,167 pp.,
Stanford Univ., Stanford, Calif., 1986.
portant to understandthat the fact that we are dealing Gomberg, J., N.M. Beeler, M. L. Blanpied, and P. Bodin,
with very small •ACS values((10 kPa) impliesthat Earthquake triggering by transient and static deforma-
these earthquakesare located severalrupture lengths tions, J. Geophys.Res., 103, 24,411-24,426, 1998.
fromany previousearthquake.In Figure94 weplot the Hanks, T. C., and H. Kanamori, A moment magnitudescale,
probability distribution ofther•t curveofthe63earth- J. Geophys.Res., 8•, 2348-2350, 1979.
Hardebeck, J. L., J. J. Nazareth, and E. Hauksson, The
quakes(•ACS is undefined for the first) dueto adding static stress change triggering model: Constraints from
the effectof the 11 uncertainearthquakes.The prob- two southernCalifornia aftershocksequences,J. Geophys.
ability distribution is unique for some time windows Res., 103, 24,427-24,437, 1998.
and bimodial for others. The bimodiality is entirely Harris, R. A., Introduction to specialsection: Stresstrig-
due to the two choicesof the nodal plane for a single gers, stressshadows,and implications for seismichazards,
J. Geophys.Res., 103, 24,347-24,358, 1998.
earthquake(the secondlargest). The uniqueprobabil- Harris, R. A., and R. W. Simpson, In the shadow of 1857,
ity distributionis due to the fact that stresschangeson the effectof the great Ft. Tejon earthquakeon subsequent
earthquakehypocentersare relatively insensitiveto the earthquakes in southern California, Geophys.Res. Left.,
choiceof nodal plane of a previousearthquakethat is 23, 229-232, 1996.
sufficientlydistant, i.e., at distancesthat are at least 3 Harris, R. A., and R. W. Simpson, Suppressionof large
times greater than the rupture length of the previous earthquakesby stressshadows: A comparisonof Coulomb
and rate-and-state failure, J. Geophys.Res., 103, 24,439-
event(seeequation(8.64) of Pollardand $egall[1987] 24,451, 1998.
for the two-dimensional case).In Figure 9b we includeHarris, R. A., R. W. Simpson,and P. A. Reasenberg,In-
all 74 earthquakes in the computation of the proba- fluenceof static stresschangeson earthquakelocationsin
bility distributionof the r•+ot curve.In this casethe southern California, Nature, 375, 221-224, 1995.
probability distribution is nonunique,becausechanges Hartzell, S. H., and T. H. Heaton, Rupture history of the
1984 Morgan Hill California earthquakefrom the inversion
in normal stress(but not in shearstress)on each of of strong motion records, Bull. Seismol. Soc. Am., 76,
the 11 hypocentersare sensitiveto the choiceof nodal 649-674, 1986.
plane of that earthquake. Both diagramsshowthat the Hartzell, S. H., and T. H. Heaton, Teleseismicmechanism
shape ofther•+ot curveisunlikely to bechanged signifi- of the May 2, 1983 Coalinga, California, earthquakefrom
cantly when the effect of the 11 excludedearthquakesis long-period p-waves, in The 1983 Coalinga, California
taken into account. Thus we conclude that the amount
Earthquakes, Spec. Publ. 66, edited by J. H. Bennett and
R. W. Sherburne,pp. 241-246, Calif. Dep. of Conserv.,
of triggeringfor I•AC$[ • 10 kPa is not an artifact of Div. of Mines and Geol., Sacramento, 1992.
disregardingsomeof the earthquakes. Jennings,C. W., Preliminary fault activity map of Califor-
nia, Open File Rep. 92-93, Calif. Dep. of Conserv., Div.
of Mines and Geol., Sacramento, 1992.
Acknowledgments. Data were providedby Northern Pollard, D. D., and P. Segall, Theoretical displacementsand
California Earthquake Data Center, Northern California stressesnear fractures in rock; with applicationsto faults,
SeismicNetwork, U.S. GeologicalSurvey,Menlo Park, and joints, veins, dikes, and solution surfaces,in Fracture Me-
BerkeleySeismologicalLaboratory,University of California, chanicsof Rock,edited by B. K. Atkinson, Academic,San
Berkeley. This paper benefited from discussionswith John Diego, Calif., 1987.
Suppe,Tony Dahlen, and Thomas Meier. Review by Ruth Reasenberg,P. A., and R. W. Simpson,Responseof regional
Harris, JeanneHardebeck,Thorsten Becker,and an anony- seismicity to the static stress change produced by the
mous reviewer greatly improved the presentation and the Loma Prieta earthquake, Science, 255, 1687-1690, 1992.
contentof the manuscript. This researchwas partly funded Scholz, C. H., C. Aviles, and S. Wesnousky,Scaling differ-
by USGS grant 11434-HQ-97-GR-03096. encesbetweenlarge intraplate and interplate earthquakes,
Bull. Seismol. Soc. Am., 76, 65-70, 1986.
References Simpson,R. W., and P. A. Reasenberg,Earthquake-induced
static stresschangeson central California faults, in The
Anderson, G., and H. Johnson, A new statistical test for Loma Prieta, California Earthquakeof October17, 1989-
static stresstriggering: Application to the 1987 Super- TectonicProcessesand Models, edited by R. W. Simpson,
stition Hills earthquakesequence,J. Geophys.Res., 10•, U.S. Geol. Surv. Prof. Pap., 1550-F, F55-F89, 1994.
20,153-20,168, 1999. Stein, R. S., and G. EkstrSm, Seismicityand geometryof a
13,642 ZIV AND RUBIN: TRIGGERING AT SMALL STRESSES

110-km-long blind thrust fault, 2, Synthetic of the 1982- Wald, D. J., D. V. Helmberger, and T. H. Heaton, Rapture
1985 California earthquake sequence,J. Geophys.Res., model of the 1989 Loma Prieta earthquake from the in-
97, 4865-4883, 1992. versionof strong-motionand broadbandteleseismicdata,
Toda, S., R. S. Stein, P. A. Reasenberg,J. H. Dieterich, and Bull. Seismol. Soc. Am., 81, 1540-1572, 1991.
A. Yosida, Stresstransferredby the 1995 M•o - 6.9 Kobe, Working Group on Northern California Earthquake Poten-
Japan, shock: Effect on aftershocksand future earthquake tial (WGNCEP), Databaseof potentialsourcesfor earth-
probabilities, J. Geophys.Res., 103, 24,543-24,565, 1998. quakes larger than magnitude 6 in northern California,
Wald, D. J., D. V. Helmberger,and T. H. Heaton, Rapture U.S. Geol. $urv. Open File Rep., 96-705, 40 pp., 1996.
model of the 1989 Loma Prieta earthquake from the in-
versionof strong-motionand broadbandteleseismicdata,
Bull. $eismol. $oc. Am., 81, 1540-1572, 1991.
Vidale, J. E., D.C. Agnew, M. J. S. Johnston,and D. H. Op-
penheimer, Absenceof earthquake correlation with Earth A.M. Rubin and A. Ziv, Department of Geosciences,
tides: An indication of high preseismicfault stressrate, Guyot Hall, Princeton University, Princeton, NJ 08544.
J. Geophys.Res., 103, 24,567-24,572, 1998a. (arubin@princeton.edu; alonziv@princeton.edu;)
Vidale, J. E., D.C. Agnew, D. H. Oppenheimer, D. Ro-
driques, and H. Houston, A weak correlation between
earthquakes and extensional normal stress and stressrate
from lunar tides (abstract), Eos Trans. AGU, 79, (45), (ReceivedSeptember3, 1999; revisedFebruary,182000;
Fall Meet. Suppl., F641, 1998b. acceptedMarch 8, 2000.)

Das könnte Ihnen auch gefallen