Seismicity Remotely Triggered by the Magnitude 7.3 Landers, California, Earthquake Author(S): D

Home , 1980 Eureka earthquake

Seismicity Remotely Triggered by the Magnitude 7.3 Landers, California, Earthquake Author(s): D. P. Hill, P. A. Reasenberg, A. Michael, W. J. Arabaz, G. Beroza, D. Brumbaugh, J. N. Brune, R. Castro, S. Davis, D. dePolo, W. L. Ellsworth, J. Gomberg, S. Harmsen, L. House, S. M. Jackson, M. J. S. Johnston, L. Jones, R. Keller, S. Malone, L. Munguia, S. Nava, J. C. Pechmann, A. Sanford, R. W. Simpson, R. B. Smith, M. Stark, M. Stickney, A. Vidal, S. Walter, V. Wong and J. Zollweg Source: Science, New Series, Vol. 260, No. 5114 (Jun. 11, 1993), pp. 1617-1623 Published by: American Association for the Advancement of Science Stable URL: http://www.jstor.org/stable/2881709 . Accessed: 28/10/2013 21:58

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

. JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact [email protected].

American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve and extend access to Science.

http://www.jstor.org

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions ...... ---.----..;- Rv'>'E S5'5.' EA ; a ar"T I_Cl E

few tens of kilometersor less of the induc- Seismicity Remotely Triggered by ing source (4). In marked contrast, the widespreadsurge in post-Landersseismic activity extendedover 1250 km (17 source the Magnitude 7.3 Landers, dimensions)from the mainshock. This abrupt,widespread, and unexpected California, Earthquake seismicityincrease challenges a long-stand- ing skepticismregarding the realityof trig- D. P. Hill,P. A. Reasenberg, A. Michael,W. J. Arabaz, geredseismic activity at greatdistances from G. Beroza, D. Brumbaugh,J. N. Brune, R. Castro, S. Davis, an earthquake.This skepticismhas its roots, D. dePolo, W. L. Ellsworth,J. Gomberg, S. Harmsen, L. House, in part, in the uncertainsignificance associated with an increasein earthquakeactivity S. M. Jackson, M. J. S. Johnston, L. Jones, R. Keller,S. Malone, at a solitary,remote site aftera largeearth- L. Munguia,S. Nava, J. C. Pechmann, A. Sanford, quake,and, in part,in the lackof a plausible R. W. Simpson, R. B. Smith, M. Stark, M. Stickney, A. Vidal, physical model for remote triggering.In S. Walter,V. Wong, J. Zollweg particular,Earth models based on linear elasticity, which have been successfulin The magnitude 7.3 Landers earthquake of 28 June 1992 triggered a remarkably sudden explaininga vast rangeof seismologicalphe- and widespread increase in earthquake activity across much of the western United States. nomena, seem incapableof accountingfor The triggered earthquakes, which occurred at distances up to 1250 kilometers (17 source triggeredseismicity beyond a few source dimensions) from the Landers mainshock, were confined to areas of persistent seismicity dimensionsof an earthquakerupture (5). and strike-slip to normal faulting. Many of the triggered areas also are sites of geothermal The simultaneousincrease in seismicactiv- and recent volcanic activity. Static stress changes calculated for elastic models of the ity at many remote sites following the earthquake appear to be too small to have caused the triggering. The most promising Landersearthquake thus forces consider- explanations involve nonlinear interactions between large dynamic strains accompanying ation of an expandedrange of models that seismic waves from the mainshock and crustal fluids (perhaps including crustal magma). includenonlinear interactions. In this article we document the spatial distributionand temporalevolution of the seismic activity that followed the Landers At 1157 UT (04:57 PDT) on 28 June gered by a nearby source is well known. mainshock and the tectonic settings in 1992, southern California was rocked by Aftershocksof large and moderateearth- which the activityoccurred. We arguethat the magnitude (M) 7.3 Landers earth- quakescommonly occur at distancesof one this activity is not explained by random quake, the largest earthquake to strike the or two sourcedimensions from a mainshock coincidence and commenton evidence for region in 40 years (1). The earthquake (3). Seismicityinduced by humanactivities remote triggeringby other major (M 2 7) resulted from a northward propagating rup- (which include the undergrounddetona- earthquakes.We conclude by exploring ture producing up to 6 m of right-lateral, tion of nuclearexplosions, filling or empty- some possiblephysical processes that might strike-slip displacement on a series of faults ing of water reservoirs,injecting and ex- explain remote triggering. extending over 70 km north-northwest into tractingfluids in deep boreholes,and min- Distribution and nature of triggered the Mojave desert from the epicenter 5 km ing) is usuallyconfined to an areawithin a seismicity. Recognition of remotely trig- southwest of the town of Landers (Fig. 1). Within minutes after the Landers mainshock, earthquake activity abruptly in- Fig. 1. Map showing seis- creased at widely scattered sites across the mograph stations in the western United States (2). >. ,7t .< . .* 8t >+' < '. l . _>. _ * Grat western United States and f ~PlainsPlaiDs That earthquake activity can be trig- K.W t: *WAs *Mexico used in this study .- J *. MT (2). The Landers main- D. P. Hill, P. A. Reasenberg, A. Michael, W. L. Ell- shock ruptureis shown by sworth, M. J. S. Johnston, R. W. Simpson, and S. / ; * t vx [*-* ,N heavy solid line. Major Walter are at the U.S. Geological Survey, Menlo Park, CPSP t +w * / ; , physiographic provinces CA 94025. W. J. Arabaz, S. Nava, J. C. Pechmann, *tt --J___ * \*_/* b .2* 0 are outlined by dashed and R. B. Smith are at the University of Utah, Salt Lake /cot; ^ so . So. X * 4 , lines. Abbreviation: CP- City, UT 84112. G. Beroza is at Stanford University, ,L'*_*I ( R \ Columbia-- o .-: Plateau- Stanford, CA 94305. D. Brumbaugh is at Northern !SRP, | ..... ?et Arizona University, Flagstaff, AZ 86011. J. N. Brune | a .CA1 NV WY_ Snake River Plain. and D. dePolo are at the University of Nevada, Reno, NV 89557. R. Castro, L. Munguia, A. Vidal, and V. Wong are at CICESE, Ensenada, Mexico. S. Davis is at 'Great Basin the University of Texas, Austin, TX 78712. J. Gomberg and S. Harmsen are at the U.S. Geological Survey, Golden, CO. L. House is at Los Alamos National Laboratories, Los Alamos, NM 87545. S. M. Jackson is at Idaho National Engineering Laboratory, Idaho Falls, Colorado ID 83401. L. Jones is at the U.S. Geological Survey, Pateau Pasadena, CA. R. Keller is at the University of Texas at El Paso, El Paso, TX 79968. S. Malone is at the University of Washington, Seattle, WA 98195. A. San- ford is at the New Mexico Institute of Mining and Technology, Socorro, NM 87801. M. Stark is at Unocal Corporation, Santa Rosa, CA. M. Stickney is at the * .'.Southern. Basin Montana Bureau of Mines and Geology, Butte, MT 500 km and Range 59701. J. Zollweg is at Boise State University, Boise, - - NM ID 83725.

SCIENCE * VOL. 260 * 11 JUNE 1993 1617

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions gered seismicitydepends critically on the geothermalactivity or youngvolcanism are lion yearold) basalticvents (9). Only a few distributionof seismic networks. Because concentratedjust east of the California- of the many seismicallyactive geothermal the networks operating in the western Nevada borderfrom the easternside of the and young volcanic systemsin the region, United Statesand northernMexico (Fig. 1) White Mountains to near Lake Tahoe. however,responded with triggeredactivity. do not provide spatiallyuniform coverage Triggeredsites in southernNevada (Little The seismogenicBrawley and CerroPrieto and their detection thresholdsvary, recog- Skull Mountain)and westernUtah (Cedar geothermalareas in the ImperialValley- nition of seismicitytriggered by the Landers City) are not near geothermalsystems but Salton Trough south of Landers(distance event is probablyincomplete. For example, are within 20 km of Pleistocene (<1 mil- fromLanders epicenter A\ = 150 to 250 km; a sequenceof M < 2 earthquakestriggered in southeasternOregon or southernArizo- na would not be detected by any of the Fig. 2. Map showing - / operating networks. Furthermore,before earthquake activitydetect- NV the late 1970smany of the presentnetworks ed by the combined seis- 'MSN / , were sparse or nonexistent. Had the mic networkin the 10 days C Landersearthquake occurred before then, immediately after the ? a, j* I, Landers earthquake. Ma- (,L,' i /o I. the chances are poor that triggeredactivity jor physiographic provinc- CM& \ (/a ) would have been recognized. es are outlined by dashed 0 $,u roatBasin The most dramaticincrease in earth- lines. Faultswith quaterna- f{( , quake activity occurredin the Landersaf- ry movement are shown X) tershockzone within 100 km of the rupture by solid lines. Shaded el- (Fig. 2), which we do not considerfurther lipse extending approxi- (6). We also do not considerpossible trig- mately 100 km beyond the \ L gered activity in southern Californiabe- Landers rupture repre -L , tween the Garlock fault and the interna- sents the aftershock zone. M Abbreviations: AZ, Arizo- Magnitudes M Colorad; which a earth- tional border,for complete na; B, Burney;C, Coso Hot 1.0+ \. ( Plateau quake catalog is not yet availablefor the Springs; CA, California; o 2.0+ periodof interest.There are, however,two Z- CC, Cedar City, Utah; CM, 3.0 + candidatesfor triggeredactivity in southern 1992 Cape Mendocinoo California:A clusterbeneath Pasadena (PA earthquake epicenter; 0 L 0? in Fig. 2) 60 km west of the aftershock CNSB, central Nevada 0 5.0+ ellipsethat includeda M = 3.9 event on 29 seismic belt; D, Death Val- Southern Basin June and a lineation of earthquakesalong ley; E, Excelsior Moun- and Range and east of the White Wolf fault (sourceof tains; G, Geysers; GF, 300 km MEXIN = Garlock fault; HFZ, Hurri- the M 7.5 Kern County earthquakeof cane fault zone; I, Indian 1952; WW and K in Fig. 2). Wells Valley; IV, Imperial Valley; K, 1952 Kern County earthquake epicenter; LP, 1989 Loma Prieta North of the Garlockfault, post-Landers earthquake epicenter; L, Lassen Peak; LV, Long Valley caldera; LT, Lake Tahoe; M, Mono Basin; ML, earthquakeactivity was concentratedwith- Medicine Lake caldera; MS, MountShasta; NV, Nevada; OV, Owens Valley; P, Parkfield;PA, Pasadena; in a belt of persistentseismicity and Ho- SAF, San Andreas fault zone; SCR, Southern Cascade Range; SM, LittleSkull Mountain;UT, Utah; W, locene faulting that extends north-north- White Mountains;WF, Wasatch fault zone; WW,White Wolf fault. west fromthe southernmargin of the Great Basin (7) as far north as the southern 3. Cascadevolcanoes in northernCalifornia. Coso - IndianWells (165 to 205) 13 Fig. Cumulativenumber Within this region, triggeredactivity was Con0173 of locatable earthquakes in selected zones, begin- primarilyconcentrated along the boundary LittleSkull Mtn,NV (240) B ning 1 January 1992. 538 zone between the SierraNevada and Great Numbers in parentheses Basin (SNGBZ), although some activity White Mtns, CA (380 to 420) are distances (in kilome- occurredin more isolated sites, each also 0 ters) from Landers earth- with a historyof persistentseismic activity Long Valley(415) quake. Total number of (Figs. 2 to 7). Particularlynoteworthy as- v0_ _ - I 1952 earthquakes in each zone is shown at right. Vertical pects of the triggeredsites include: (i) all t Mono Basin (450) sites of (recognized)remote triggeringare 54 lines mark times of the 25 northof the Landersmainshock and (ii) all C , April 1992 Petrolia (Cape = and sites show strike-slip to normal faulting CedarCity, UT (490) : 41 Mendocino)(M 7.1) E Landers (M = 7.3) earth- (implyinga horizontalorientation for the : Western Nevada (450 to 650) 1843 quakes. least principalstress). Many sites of persis- 1 1 tent seismicity north of Landersdid not Lassen (840) 3 respondwith triggeredseismicity, however. 3~~~~~~~~~~~~~~~~~3 Notably,active sectionsof the San Andreas Burney(900) 1 fault systemin central and northernCali- the Intermountainseismic belt in fornia, ID centraland northernUtah, and the central Cascade, (1100) 51 Nevadabelt (8) showedno response.Many Yellowstone (1250) sites of remotelytriggered activity are also Y 102 closely associated with areas of geothermal activity or young volcanism (last 1 million 0 50 100 150 .200 years). Sites of triggered activity lacking Days since 1 January1992

1618 SCIENCE * VOL. 260 * 11 JUNE 1993

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions RESEARCHARTICLE

IV in Fig. 2) areparticularly notable for the the Landersmainshock (even earlierevents Valleycaldera has been the most active area absenceof triggeredactivity (10). could have been obscuredby seismicwaves within the SNGBZ duringthe last decade Both the maximummagnitude and the generatedby the intense Landersaftershock andhas hadfrequent earthquake swarms and total seismicmoment of the remotelytrig- activity). It was the largestearthquake to over 0.5 m of grounduplift. It is the site of gered earthquakesdecreased with distance occur in this section of the southernGreat the most recent volcanism within the fromthe Landersepicenter (Table 1 andFig. Basinsince 1868. SNGBZ;the mostrecent eruptions were just 4). The largesttriggered earthquake was a M Triggeredactivity along the easternmar- 500 to 600 yearsago along the Inyo-Mono = 5.6 event beneathLittle Skull Mountain gin of the SierraNevada and adjacentsec- cratersvolcanic chain (12). The triggered in southernNevada (A = 240 km) on 29 tionsof westernNevada was concentrated in seismicitywithin the SNGBZ, which in- June at 1014 UT located approximately20 scattered clusters from the Indian Wells cludednumerous events between M = 3 and km east of a group of basalticvents that Valley-Cosoarea north of the Garlockfault 4, coincidesclosely with the locations and erupted sometime between 0.02 and 1.0 (A = 165 to 205 km) to nearLake Tahoe (A depths of earthquakeslocated there during millionyears ago (Ma) (9). This earthquake, = 500 km). The SNGBZ is markedby the past several decades. Specific areas of which had an oblique normal mechanism numerousQuaternary basaltic and rhyolitic triggeredactivity within this regioninclude with a northwest-southeastT-axis (exten- volcanic centersand geothermalareas and (see Fig. 2): (i) a 60-km-long,north-north- sion direction),was precededby at least 19 has been a persistentsource of moderate-to- west-trendinglineation along the baseof the smallerevents within 100 km of LittleSkull largeearthquakes throughout historic time, SierraNevada from IndianWells Valley to Mountainbeginning as soon as 91 min after which datesfrom the mid-1800s( 1). Long the vicinity of Coso Hot Springs; (ii) a diffuse cluster cutting westwardfrom the California-Nevadaborder near 37?N across Fig. 4. Cumulative seis- Coso - Indian Wells (165 to 205) the northernend of Death Valley through mic moment in selected 7.2 x 102 zones, beginning 1 Janu- the InyoMountains into OwensValley; (iii) a dense cluster concentratedin the south ary 1992. Numbersin pa- Little Skull Mtn, NV (240) 2.6 x 1024 rentheses are distances 2.6_x_1024 moatof LongValley caldera, including three (in kilometers) from White Mtns, CA (380 to 420) clustersalong the California-Nevadaborder, Landers earthquake. To- 6.2 x 1021 two east of Long Valley along the east tal seismic moment (in Long Valley (415) 1.2 x 1022 marginof the White Mountains,and one dyne-cm) for each zone is c north of Long Valley in the Mono Basin; shown at right. Vertical E 0 x (iv) a clusterelongated to the northeastin lines mark times of the Mono Basin (450) 2.6 1 the ExcelsiorMountains (E) in westernNe- Petrolia Mendo- .2 -_ (Cape vada; and (v) a diffuselineation along the cino) (M = 7.1) and , Cedar City, UT (490) 2.6 x 1022 Landers (M = 7.3) earth- I quakes. > Western Nevada (450 to 650) 2.6 x 1023 80 A Geysers 60 E Lassen (840) earthquakes 6.4 x 1020 ~;40 E 20 Burney(900) 1.4 x 1022 *, -0-4 -0.2 0.0 0.2 0.4

Cascade, ID (1100) 6.2 x 1019 71000 B LongValley 600 earthquakes Yellowstone (1250) 1.6 x 1022 200 0 - 0 50 100 150 200 -20 0 20 40 Days since 1 January 1992 0.2 C LongValley volumetricstrain 0 01

Unfiltered Long Valley Caldera (MTCM) .2 0.0 I IIOi"Al -~~~~~~~~~~1Ih v _ -. MF$A . -20 0 20 40 s Days relativeto LandersEarthquake 1.3 x 1 5 to30Hzbandpass Fig. 6. Response of seismicity and strain to the Geysers (GGPM) Landers earthquake. (A) Cumulative number of 1JUA ~~~~~~~The earthquakes detected at the Geysers geothermal area in a 24-hour period surrounding the

3.12x S 1 b Landers earthquake. (B) Cumulative number of earthquakes detected at Long Valley caldera between 1 June and 20 August 1992. (C) Parkfield (PMMM) Dilatational strain recorded at Devils Postpile, located approximately 10 km west of the triggered seismicity in Long Valley caldera, during 2.36x 'uwLA the same interval. Increasing values corre- 1 Minute spond to increasing compression. Tidal strains and a secular dilatational strain rate of -0.007 Fig. 5. Unfiltered and filtered vertical-component, ground-velocity seismograms from low-gain microstrain per day have been removed, but a Calnet stations at Long Valley caldera, the Geysers geothermal area, and Parkfield. Number next to residual tidal strain signal remains. The tran- filtered trace indicates its magnification relative to corresponding unfiltered trace. S indicates arrival sient dynamic Landers signal (6.4 microstrain) of Landers S wave and 1 indicates first identified triggered event at each site. is not shown on this record.

SCIENCE * VOL. 260 * 11 JUNE 1993 1619

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions easternmargin of the SierraNevada between Fig. 7. Map showing stat- 125- 120- 115- 110- Bridgeportand LakeTahoe. ic mean stress change 45- + The triggeredswarm near CedarCity in and peak dynamicstress + + in bars for the Landers . southwesternUtah (A = 490 km) included Yel morethan 60 earthquakes(at least seven of mainshockat mid-crustal CascadeID o depth together with the which were felt locally) within the Hurri- first 10 days of post- cane fault zone, the principal zone of Landers seismicity. The boundingfaults between the ColoradoPla- static mean stress teau and the eastem Great Basin. The change was calculated 1 bar * earthquakeswere all within 15 to 20 km of for an elastic half-space 0.8. /0.7 Quaternarybasalt flows with dates of 1.0 to using the three-dimen- . 1.2 Ma (9). No other sites along the seis- sional boundary element 4 micallyactive, easternmargin of the Great subroutines of Okada (34) and a model for the Basin [the Intermountainseismic belt (13)] Landers based showedevidence of triggeredactivity. rupture (.)2br on Caltech Terrascope 0 Beyond600 km, isolatedclusters of less data and surface slip ob- L:2) *' .5 intense activity occurred in the Geysers servations (1). T indi- \ geothermalarea (A = 740 km), the south- cates typical range of tid- ern Cascaderange (A = 840 to 900 km), al stresses. Shading \ 1 8 YellowstoneNational Park (A = 1250 km), shows the expansional . l 2) and western Idaho (A = 1100 km). The quadrants; the compres- 35 / to )'*J\ triggeredactivity at the Geysersconsisted of sional quadrantscan be to2)/ 2.5A 7 9 a surgeof M < 1 earthquakesthat began30 filled in by 900 rotation of 2ookm/ . s arrivalof the LandersS wave and the shaded pattern and a 0 2 2 bars , afterthe sign change. (Maximum 2 bar decayedto backgroundlevels 3 hourslater static shear stress (Fig. 6A). This is the only triggeredactivity changes resolved onto reLognizedwithin the northern San An- vertical planes based on dreasfault system.In the southemCascade the same model are list- 125- 120 T 115- Range, earthquake activity increased at ed in Table 1.) Numbers __ _ LassenPeak, Medicine Lake caldera, and indicate peak dynamic 0.0001 0.001 0.01 0.1 1 bars near Burney(an areacovered by Quaterna- mean stress from bore- ry lava flows). Triggeredseismicity near hole dilatometer data (tri- Cascade,Idaho, was locatedin the western angles) and peak dynamic shear stress from regional digital seismic stations (diamonds). Value at Parkfield (1.8) is mean of six dilatometer observations. Parenthetic numbers are peak dynamic part of the Idaho batholith (see Fig. 7) stresses for the M = 7.1 Petrolia (Cape Mendocino) earthquake, shown for comparison. Contours some 15 km west of the nearesthot spring. suggest general pattern of peak dynamic stresses from the Landers mainshock without distinguish- The most distant candidate for activity ing between mean and shear stresses. triggeredby the Landersmainshock occurredin a smallcluster 15 km northwestof Yellowstone caldera within Yellowstone the mainshock and its large aftershocks It is possiblethat the increasedseismic- National Park (14) (see Fig. 7). tended to obscure the onset time of the ity rate at some sites is merelycoincidental Temporalpatterns in triggeredactivity. triggeredactivity. Our approach to this (having no causal link) with the Landers Establishingthe onset time of triggered problemtakes advantage of the observation earthquake. Although the post-Landers seismicity is important both for judging that seismicwaves from nearby earthquakes surgein seismicityis unusualat each site, statistically whether triggeringhas taken arerelatively rich in high-frequencyenergy, all of the sites are seismicallyactive and place (rulingout randomcoincidence) and whereas waves from the Landers main- have had seismicitysurges (swarms) in the for constrainingphysical models for remote shock, having traveleda long distance, are past (see, for example,the swarmon day35 triggering.The apparentonset (time of the relativelydepleted in high-frequencyener- at LassenPeak in Fig. 3). Consideringthe firstdetected earthquake after Landers) var- gy as a result of intrinsic attenuationand relative infrequency of these seismicity ied from30 s afterpassage of the LandersS scattering.Applying a 5- to 30-Hz band- surges,however, noncausalcoincidence is wave to 33 hours after the Landersmain- pass filter to seismogramseffectively elimi- unlikelyfor the post-Landersactivity (17). shock (Fig. 3 and Table 1). In general,the nates the mainshockcoda wavesat distanc- Although the maximummagnitude of recognizedonset time dependson the local es beyond 500 km and enhances the local triggeredearthquakes at each site general- rate of seismicityand the sensitivityof the earthquakes(Fig. 5). The resultshows that ly decreasedwith distance from Landers, seismicnetwork, both of which varyregion- in LongValley caldera(A = 415 km) and the durationof the triggeredactivity at a ally. Using the earthquakecatalog data, we the Geysers (A = 740 km) the triggered given site showed no clear correlation tested the hypothesis that the observed activitybegan 30 to 40 s afterthe arrivalof with distance (Figs. 3 and 6 and Table 1). onset of activity at each site is consistent the S wave from the Landersearthquake At Long Valley and the Geysers, where with an instantaneousincrease in local and duringthe passageof the large-ampli- seismicity is normally high, the triggered seismicityrate at the time of the Landers tude Love and Rayleighsurface wave trains seismicity occurred as smooth transient mainshock(15). We were unable to reject (16). No evidence for such early activity surgesin seismicitywith durationsof a few the hypothesisat all but one of the sites, a was found at Parkfield,the White Moun- daysand a few hours, respectively(Fig. 6), resultconsistent with a causallink between tains, Mono Basin, westem Nevada, Las- whereasat Mono Basin and Burney, trig- the Landersearthquake and the remote sen, Cedar City, Bumey, or Yellowstone. gered activity showed little sign of dimin- earthquake activity. Inadequatestation coverage or severesignal ishing after 3 weeks or more (Fig. 3). The The strong direct waves and the reflect- clipping made this analysis impossibleat increase in number of events after the ed and scattered coda waves produced by other sites (Table 1). Landersearthquake varied widely from site

1620 SCIENCE * VOL. 260 * 11 JUNE 1993

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions mRESEARCH ARTICLE

Table 1. Summary of earthquake data and modeling results for selected regions. The epicentral strains of Rayleigh waves from distant distances expressed in source lengths reflect the 74 km of surface faulting observed in the earthquakesis well documented(21). Data mainshock. Static stress changes reflect maximum shear stress change on optimally oriented fromwater wells at YuccaMountain (25 km planes. Daily tidal stress variation is approximately 2 x 10-2 bar. The triggered seismicity in each region is characterized, using data in regional earthquake catalogs, by the maximum magnitude northwest of Little Skull Mountain) and earthquake during the first 7 days after the Landers mainshock, and the numbers of events during LongValley calderashow pronouncedtran- the 7 days before (Nb, 21 to 28 June) and after (Na, 28 June to 5 July) the Landers mainshock. sient fluctuations (amplitudes from less Timing of the triggered activity is characterized by the interval T, between the Landers mainshock than 1 cm to several meters) associated and the first located post-Landers earthquake, and by the mean interevent time Tmof the first ten with seismicwaves from the Landersmain- local earthquakes in the post-Landers activity. shockbut no clearevidence for staticoffsets in local watertable levels (22). Approximate Static stress change Remote triggeringby other large (M > distance Max. N N T Region N 6.5) earthquakes?It might at first appear mg b h) (r from the paucityof examplesin the litera- RS(km) Source (bar) FractionMa (i) lengths (a) of tidal turethat the triggeringof remoteseismicity by large earthquakesis exceptional. The Coso-lndian 165-205 2.2-2.8 2 x 10-2 to 1 to 3 4.4 4 44 8.6 4.6 activity triggeredby the Landers earth- Wells 6 x 10-2 quake, however, consists largely of small Little Skull 240 3.2 2.8 x 10-2 1.4 5.6 10 59 1.5 0.9 Mt (M ? 3) events in relativelyremote areas, Death Valley 300 4.1 8.0 x 10-3 0.4 3.6 6 11 5.3 12.4 and, as noted above, the ability to recog- White Mts* 380-420 5.1-5.7 2.3 x 10-3 to 0.13 3.7 0 27 11.6 12.0 nize reliably this level of seismic activity 3.0 x 10-3 dates only from the late 1970s with the Parkfield* 410 5.5 6.9 x 10-3 0.35 8 11 deploymentof dense, telemeteredseismic Long Valley* 415 5.6 2.8 x 10-3 0.14 3.4 38 340 0.15 0.052 networks and computer-based,real-time Mono Basin* 450 6.1 1.9 x 10-3 0.10 3.1 3 12 19.1 12.0 Cedar City* 490 6.6 3.6 x 10-3 0.18 4.1 0 39 0.65 1.3 data processingsystems (23). Indeed, to Western NV* 450-650 6.1-8.8 0.6 x 10-3 to 0.03 to 4.0 58 504 0.15 0.10 searchfor evidence of reniote triggeringin 2.0 x 10-3 0.1 the past, we must consider progressively Geysers*t 740 10.0 8.7 x 10-4 0.04 1.6 70 60 0.05 0.01 larger(and thus less common)earthquakes, Lassen* 840 10.9 3.9 x 10-4 0.02 2.8 0 14 0.2 5.4 which in turnpresses the issue of statistical Burney* 900 12.4 2.9 x 10-4 0.015 2.8 1 9 23.0 21.0 significance. Cascade, ID 1100 14.9 2.5 x 10-4 0.01 1.7 0 38 33.3 1.2 = Yellowstone* 1250 16.9 2.3 x 10-4 0.01 2.1 0 16 1.8 2.1 Four other M 7 earthquakeshave occurredin the westernUnited States since *Denotes regionwhere high-pass filteringof seismogramswas used to search for early events in the Landers 1980: the M = 7.4 Eureka, California, coda. tAt the Geysers, N, and N, reflectCalnet data; T, and T_ reflectUnocal data. earthquake(located 50 km west of Cape Mendocino)of 8 November1980; the M = to site and was apparentlyunrelated to - kO, where k is the bulk modulus. The 6.9 Borah Peak, Idaho, earthquakeof 28 distance from Landers. seismometersprovide data on particle ve- October 1983; the M 6.9 LomaPrieta, Supportingobservations. Observations locities, which are proportionalto dynamic California,earthquake of 18 October1989; of both permanent and dynamic strain stresses (20). Peak particle velocities and and the M = 7.1 Petrolia, California, changes at sites of remotelytriggered seis- dilatationalstrains (and thus peak dynamic earthquakeof 25 April 1992. However, micity help constrain possible triggering stresses)occur within the S wave coda at with the possibleexception of minorswarm processes.Of all the sites showingtriggered distancesbeyond about 300 km and include activity at the Geysers geothermal field activity,however, only LongValley caldera early partsof the fundamentalmode Ray- coincident with the 1989 LomaPrieta and has a continuous instrumentalrecord of leigh and Love wave trains,with dominant 1992 Petroliaearthquakes (distances of 220 deformation.Data from daily measurements periodsof 5 to 20 s. and 230 km from the Geysers, respective- of a two-colorgeodimeter network showed The distributionof peak dynamicstress- ly), none of these M - 7 earthquakes no strain changes above the resolutionof es from the Landersmainshock (Fig. 7) appearsto have triggeredremote seismicity about 0.3 microstrain (18). Continuous shows a strong directivityeffect associated (24). The 1980 Eureka earthquakewas data from a 200-m-deepborehole dilatom- with the northward propagationof the essentially the same size as the Landers eter 4 km west of the calderashowed an mainshockrupture. Peak dynamicstresses earthquake(both with seismicmoments Mo instantaneouscompressional strain step of north-northwestof the mainshock were 1 x 1027 dyne-cm). The other three about3 x 10' betweenthe LandersP and roughlytwice those at comparabledistances events were smallerby factors of 2 to 3, S wavesfollowed by a slowercompressional to the west and over three times those to with seismic momentsranging from Mo = pulse that built to about 2 x 10-7 during the south. Forexample, at A = 410 km the 3.0 x 1026 (LomaPrieta) toMO = 4.5 x the 5 days afterthe mainshock.The strain peak dynamic stress was 1 to 2 bars at 1026 dyne-cm (Petrolia).Neither the 1980 pulse then decayed to backgroundduring Parkfieldbut 3 to 4 bars in Long Valley Eurekanor the 1992 Petrolia earthquakes the next few weeks (Fig. 6). Dilatometers caldera. Peak dynamicstresses in the San near Cape Mendocino triggeredseismicity alongthe San Andreasfault (solidtriangles FranciscoBay areaproduced by the Landers in the vicinity of the southern Cascade in Fig. 7) showed only an instantaneous mainshock (1.2 to 1.5 bars) were roughly volcanoes (distances 200 to 250 km), strainstep. twice those produced by the M = 7.1 whereas the Landers earthquake, which Informationon dynamicstresses comes Petrolia (Cape Mendocino) earthquake occurred just 64 days after the Petrolia fromon-scale records of the Landersmain- (0.5 to 0.6 bars), even thoughthe distance event, did triggeractivity in the southern shock recordedon both broad-banddigital fromthe Bayarea to Landersis nearlytwice Cascades(A = 840 to 900 km). The Eureka seismometersand dilatometers [Table 1; that to Cape Mendocino. and Petrolia rupturespropagated to the (19)]. The dilatometersprovide a direct Well-aquifer systems can behave as southwest and west, respectively (away measurementof dilatationalstrain, 0, and band-limitedvolume strain meters, and the from the continenftal United States), the associateddynamic mean stressesare o responseof such systemsto the dilatational whereas the Landers rupture propagated to

SCIENCE * VOL. 260 * 11 JUNE 1993 1621

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions the north and in the directionof the trig- the Landersmainshock at distancesbeyond relativelyshallow depths and (ii) deforma- gered activity. In any case, it appearsthat about250 km arguesagainst their efficacy as tion in the overlyingbrittle (seismogenic) remotetriggering requires more than simply a triggeringmechanism for this instance of domaintends to be dominatedby strike-slip the occurrenceof a M 2 7.3 (or MO2 1 x remoteseismicity (28). to normalfaulting (a horizontalleast prin- 1027 dyne-cm) earthquake. The relativelylarge, northward-directed cipal stress)such that fluid-filledcracks are Remote triggeringmay have occurred dynamicstresses associated with the shear- vertical. This combination is favorable for afterthe 1906 San FranciscoMs = 8?/4(M wave coda and the fundamentalmode Love the upwardflow of high-pressurepore fluids = 7.7) earthquake.Eight felt earthquakes and Rayleighwaves (Fig. 7) admit several from the plastic domain into the brittle occurredduring the first2 dayswithin 700 possiblemechanisms for the triggeringpro- crust. km of San Francisco;most notablewas a M cess. In principle, an S wave (or Love Morespeculatively, the largedilatation- = 6.2 event in the ImperialValley 11 hours wave) polarizedin the plane of maximum al strains associatedwith Rayleigh waves afterthe mainshock.Similarly, seven M = tectonic shear stress could triggerslip on interactingwith magmabodies in the upper 2 to 3 events near Riverside, California favorablyoriented faults that were close to crustmay acceleratethe exsolutionof vol- (distance100 km), occurredwithin the first the failure threshold. S waves or Love atile componentsand temporarilyincrease 6 hours after the M = 7.5 1952 Kern waves propagatingthrough a locally heter- pressurewithin the magma body or pore County earthquake.If the ImperialValley ogeneous stress field will generateparticle pressuresin overlyingrock as a resultof an and Riverside events were triggered,the accelerationsin the direction of propaga- increasedflux of volatilesout of the magma triggeringis not very remote, as they lie tion and transient stressesnormal to the body (32). Alternatively, crustal magma within 2 and 2.5 rupturelengths from their shear plane (29). The combinationwould bodies that are predominatelycrystalline respectivemainshock sources. facilitateslip on an optimallyoriented fault with only a small melt fractionbehave as Two additional candidatesfor remote by temporarilyreducing the normal stress solids ratherthan as liquidsat small strain triggeringinvolve an abruptseismicity in- acting across the fault plane. Nonlinear levels (that is, they transmitshear waves). creasein Rabaulcaldera (Papua New Guin- constitutivelaws for fault friction (30) ad- If the largedynamic strains associated with ea) on 10 May 1985 followinga M. = 7.2 mit the possibility that the frictional the surface waves caused such a magma earthquakein New Britain 180 km away strengthof faultscan be loweredas the fault chamberto liquifypartially, it wouldrelease (25) and a swarmof some 30 locally felt planesare worked by the dynamicstrains of differentialstress. The resultingload trans- earthquakeson Kyushu,Japan, that began the passing wave field, thereby triggering fer to the surroundingcrust could trigger about 16 min after the great (Ms = 8) slip on faultsnear the failurethreshold. earthquakesin much the same way that Nankaidoearthquake of 21 December1946 The interactionof the dilatationalcom- stressredistribution in an earthquaketrig- at a distanceof 450 km (26). In both cases, ponent of Rayleighwaves with fluidsin the gers aftershocks. the putative triggeredactivity occurredin crust may contributeto remote triggering. Possibly, several processescontributed close associationwith sites of young volca- Peak dynamic stressesassociated with the to the observedtriggering, with the domi- nism and geothermalactivity. Rayleigh and Love wave at mid-crustal nant processat a given site determinedby Triggering mechanisms. Competitive depths were on the orderof a few bars at the local crustalenvironment and location modelsfor the remotetriggering process fall distancesof at least 500 km north of the (both distanceand azimuth)with respectto into two broadclasses: One involving the mainshockepicenter (Fig. 7 and Table 1) the mainshocksource. The close associa- static stresschanges in the crust produced (20). Water levels in wells tappinguncon- tion of most sites of triggeredactivity with by the dislocation along the Landersrup- fined aquiferscan fluctuategreatly during recent volcanism and geothermalsystems, ture surfaceand the other involving the the passageof Rayleighwaves from a large, forexample, suggests that the interactionof dynamicstresses associated with the propa- distantearthquake (21). Where crustalflu- the dilatationalcomponents of the strain gating seismic waves generatedby abrupt ids are confined, a passingcrustal Rayleigh field from the Landers earthquakewith slip along the rupture surface. In both, wave will alternatelyelevate and depress geothermalfluids or crustalmagma bodies remote triggeringinvolves brittle slip on local porepressure over periodsof 5 to 20 s. may be an importanttriggering process. At local, favorablyoriented faults inducedby Seismicitymay be directlytriggered by the Long Valley caldera, the close temporal an incrementalchange in the local stress compressionalphase of the passingRayleigh associationbetween seismicity rate and the field sufficient to overcome frictional wave as elevated pore pressuresreduce the increasingpart of the transient compres- strength or an incremental reduction in effective strength of local faults. Indirect sionalstrain pulse (Fig. 6) stronglypoints to effectivefrictional strength. triggeringmay occurif the dynamicstresses an increasein fluid pressuresomewhere in Static stress changes decrease rapidly and pore pressuretransients rupture fluid the uppercrust driving the triggeredseis- with distance (as r-3 , comparedwith r-2 seals. Previouslyisolated fluids would then micity. and r-3/2 for dynamic stresses associated flowinto adjacentvolumes with lowerpres- The observationthat the triggeredactiv- with body and surfacewaves, respectively) sure, thereby increasingthe pore pressure ity persistedhours to 1 week or more after for a dislocation in an elastic half space. and decreasingeffective strength in these seismicwaves from the Landersevent sub- Maximumstatic shearstress changes calcu- volumes (30a). In this case, the onset of sided (and in many cases may not have lated for the Landersearthquake, for exam- triggeredseismicity would follow passageof begun until after the seismic waves had ple, fallbelow daily tidal stress changes (27) the Rayleighwave with a delaygoverned by subsided)emphasizes that the triggeredac- at distancesbeyond about 250 to 300 km local permeability,ambient pore pressures, tivity was not drivensolely by the dynamic (Table 1 and Fig. 7). The dilatational and shearstresses. stresses.Whatever the triggeringprocesses, componentof the static strainchange cal- Hot (and weaker) areas of the crust the resultswere a cascadingfailure sequence culatedwith this model for the vicinity of associated with young magmatic systems (earthquakeswarm) in crustalvolumes al- LongValley caldera (A = 415 km) agreesin and geothermalareas may be particularly readyloaded to a critical stressstate (33). both sense and magnitudewith the -3 x susceptibleto the latter processbecause (i) The form and durationof individualtrig- 10-9 compressionalstrain step detectedby rocks in the plastic domain (temperatures geredsequences are probablyinfluenced by the borehole dilatometer in the caldera. above 3500 to 4000C:), which tend to have the same factors (the degree and character- The small size of both theoretical and ob- low permeabilities and pore pressures that istic dimension of fracturing, pore pressure, served static stress (or strain) changes for are near lithostatic values (31 ), may exist at permeability, and so forth) that influence

1622 SCIENCE * VOL. 260 * 11 JUNE 1993

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions X RESEARCHARTICLE the evolution of an aftershocksequence. gradients and the southern Nevada transverse polarization directions, respectively, ui(x,t) is par- The observationsof remotetriggering of seismic belt [G. P. Eaton, R. R. Wahl, H. J. ticle displacement, ,t is the shear modulus, and 1B Prostka, D. R. Mabey, M. D. Kleinkopf, Geol. Soc. is shear velocity. Taking values appropriate for seismicityafter the Landersearthquake fo- Am. Mem. 152, 51 (1978), R. S. Smith and M. L. mid-crustal depths (Poisson ratio = 0.25, .t = 3.3 cus attention on the natureof earthquake- Sbar, Geol. Soc. Am. Bull. 85, 1205 (1974).] x 1011 dyne cm-2, p = 2.7 gm cm-3, and 1B= 3.5 fault interactionand the mechanicsof the 8. R. E. Wallace, J. Geophys. Res. 89, 5763 (1984). km s-I gives ITI : cu(x,t) bars, where c= 1 (bar 9. Maps showing the distribution of geothermal ar- s cm-1). Note that the relation between dynamic seismogenic crust. Although static-elastic eas and late Cenozoic volcanic centers are found stress and particle velocity does not depend on dislocationmodels in homogeneousmedia in L. J. P. Muffler, Ed., U.S. Geol. Surv. Circ. 790 frequency for harmonic plane waves [J. N. Brune, may account for the occurrenceof after- (1979) and R. G. Luedke and R. L. Smith, U.S. J. Geophys. Res. 75, 4997 (1970).] Geol. Surv. Misc. Inv. Ser. MAP 1-1091B (1978) 21. R. C. Vorhis, in The Great Alaskan Earthquake of shocks within one to two rupturelengths and Map l-1091C (1981), respectively; S. Hecker, 1964: Hydrology (Publ. 1603, National Academy from a fault, the predicted static stress Utah Geol. Surv. Bull., in press; B. D. Turin, D. of Sciences, Washington, DC, 1968). L.-B. Liu, E. changesat greaterdistances seem to be too Champion, R. J. Fleck, Science 253, 654 (1991). Roeloffs, X.-Y. Zheng, J. Geophys. Res. 94, 9453 small to explain remote triggering.The 10. A small swarm occurred in the Brawley seismic (1989). zone beginning a week after the Landers main- 22. M. L. Sorey, personal communication; G. M. temporalform and spatial distributionof shock. This swarm appears in Fig. 2 as a cluster O'Brian and P. Tucci, Eos 73, 157 (1992). the remote triggeringpoint, instead, to a of epicenters in the Imperial Valley (IV). 23. E. R. Engdahl and W. A. Rinehart, in Neotectonics class of explanations involving critically 11. J. D. VanWormer and A. S. Ryall, Bull. Seismol. of North America, D. B. Slemmons, E. R. Engdahl, Soc. Am. 70, 1557 (1980); A. M. Rogers, S. C. M. D. Zoback, D. D. Blackwell, Eds. (Geological loadedfaults in a heterogeneouscrust, stat- Harmsen, E. J. Corbett, D. M. dePolo, K. F. Society of America, Denver, 1991). ic strain amplificationwithin weak bound- Priestley in Neotectonics of North America, D. B. 24. The interval between eruptions of the Old Faith- aries (fault zones) between crustalblocks, Slemmons, E. R. Engdahl, M. D. Zoback, D. D. ful geyser in Calistoga, California (30 km south- Blackwell Eds. (Geological Society of America, southwest of the Geysers geothermal field), also andnonlinear interactions between dynam- Denver, 1991), pp. 153-184. This zone is also seems to be sensitive to large (M 2 5.5), region- ic stresses in seismic waves and crustal known as the eastern California-central Nevada al earthquakes [P. G. Silver and N. J. Valette- fluids. Although no single model seems seismic belt [D. P. Hill, R. E. Wallace, R. S. Silver, Science 257 1363 (1992)]. The Calistoga of all Cockerham Earthquake Predict. Res. 3, 571 geyser, however, failed to respond to the capable explaining the triggeredac- (1985)] and the eastern California shear zone [R. Landers mainshock (P. G. Silver, personal com- tivity following the Landersearthquake, K. Dokka and C. J. Travis, Tectonics 9, 311 munication). this generalclass of explanationsis appeal- (1990)]. 25. J. Mori et al., in Volcanic Hazards: Assessment ing because it involves geologicallymore 12. J. B. Rundle and D. P. Hill, Annu. Rev. Earth and Monitoring, J. H. Latter, Ed. (Springer-Verlag, Planet. Sci. 16, 251 (1988). R. A. Bailey and D. P. Berlin, 1989), pp. 429-462. realisticcrustal models than does the clas- Hill, Geosci. Can. 17, 175 (1990). 26. K. Abe, personal communication. sical, homogeneousview. 13. R. B. Smith and W. J. Arabaz, in Neotectonics of 27. Peak stresses due to the solid Earth tides typically North America, D. B. Slemmons, E. R. Engdahl, M. range from 0.01 to 0.03 bar (strains in the range REFERENCESAND NOTES D. Zoback, D. D. Blackwell, Eds. (Geological 0.02 to 0.06 microstrain). Peak tidal stresses the Society of America, Denver, 1991), pp. 185-222. day of the Landers mainshock were about 0.02 1. H. Kanamori, H.-K. Thio, D. Dreager, E. H. Hauks- 14. In addition, less certain observations include a bar (0.04 microstrain). son, Geophys. Res. Lett., in press. In this article swarm of M 2 1.7 events near Wallace, Idaho (A 28. The possibility remains, however, that static or = we use M for moment-magnitude [T. Hanks and 1400 km), beginning 4 days after the Landers quasi-static stress changes may result in delayed H. Kanamori, J. Geophys. Res. 84, 2348 (1979)], earthquake [P. Swanson, personal communica- triggering of earthquakes over a time scale of = M for local network magnitudes based on either tion] and three M 2 to 3 events in southeastern months to years as co-seismic stresses in the = amplitude or coda duration measurements, and Oregon near Crump's Hot Springs (A 940 km). lower crust and asthenosphere relax and transfer 15. We defined the onset delay at each site as the load to the overlying seismogenic crust, which M, for surface wave magnitude. T, 2. Earthquake data used in this study were provided time interval between the Landers mainshock and may, in turn, concentrate stress in relatively weak by regional seismograph networks operated by the first detected post-Landers earthquake. We faults. The ability of static strain changes within the U.S. Geological Survey, the California Institute represented the post-Landers seismicity at each weak zones to trigger local seismicity may be of Technology, the University of Nevada, the site by a Poisson process and estimated its rate, further enhanced if the weak zones are relatively University of Utah, the University of Washington, X, from the first ten events located at that site. impermeable such that the strain changes drive Boise State University, the University of Texas at Interevent times in a Poisson process are expect- large changes in local pore pressure [J. R. Rice El Paso, New Mexico Tech, the Idaho National ed to exceed 3/X 5 percent of the time. T, was and J.-C. Gu, Pure Appl. Geophys. 121, 187 Engineering Laboratory, Los Alamos National less than 3/X at all but one of the sites tested: the (1983); S. C. Jaum6 and L. R. Sykes, Science 258, Laboratory, the Montana Bureau of Mines and swarm near Cascade, Idaho, was the exception 1325 (1992); M. L. Blanpied, D. A. Lockner, J. D. Geology, the Centro de Investigaci6n Cientifica y (see Table 1). Byerlee, Nature 358, 574 (1992). de Educaci6n Superior de Ensenada, B.C., Mex- 16. Before transforming the seismograms to the fre- 29. C. Kisslinger and J. .T.Cherry, Eos 51, 353 (1970). ico, the Unocal Corporation, the U.S. Bureau of quency domain, we applied a 20% cosine taper in 30. P. Okubo and J. H. Dieterich, in Earthquake Mines, and the Montana Bureau of Mines and the time domain. In the frequency domain cosine Source Mechanics, S. Das, J. Boatwright, C. Geology. tapers were applied between 4 and 5 Hz and 30 Scholz, Eds. (American Geophysical Union, 3. The source dimension of an earthquake is usually and 35 Hz. Between 5 and 30 Hz the filterwas flat. Washington, DC, 1986), pp. 37-48. defined as the maximum linear measure of fault 17. We have not attempted to test the hypothesis that 30a.J. Byerlee, Geology 21, 303 (1993). surface that slipped. The source dimension of the a regional strain event acted as a common trigger 31. R. 0. Fournier, Geophys. Res. Lett. 18, 955 Landers earthquake is approximately 70 km. For to both the Landers mainshock and the remote (1991). comparison, the source dimension of a M 3 seismicity. Two observations stand against this 32. R. I. Tilling, J. M. Rhodes, J. W. Sparks, J. P. earthquake (generally the smallest earthquake to hypothesis, however: (i) The Parkfield and Long Lockwood, P. W. Lipman, Science 235, 196 produce locally felt shaking) is a few hundred Valley borehole dilatometers showed no evidence (1987); D. L. Sahagiam and A. A. Proussevitch, meters, while that of a great (M 2 8) earthquake is for a regional strain event before the Landers Eos 73, 627 (1992). a few hundred kilometers. mainshock and (ii) in no case did triggered seis- 33. In the terminology of nonlinear dynamics, the sites 4. C. Kisslinger, Eng. Geol. 10, 85 (1976). micity begin before the shear wave arrival from that responded to the Landers -earthquake with 5. R. A. Harris and R. W. Simpson, Nature 360, 251 the Landers mainshock. triggered seismicity had reached a state of self- (1992); P. A. Reasenberg and R. W. Simpson, 18. J. Langbein, J. Geophys Res. 94, 9453 (1989). organized criticality sometime before the Landers Science 255, 1687 (1992); R. S. Stein and M. 19. The Landers mainshock did not trigger strong- earthquake, with a correlation length spanning Lisowski, J. Geophys. Rev. 88, 6477 (1983); K. W. motion seismometers at distances beyond about much of the Great Basin [see, for example, J. B. Hudnut, L. Seeber, J. Pacheco, Geophys. Res. 200 km [A. Shakal, CSMIP Strong-Motion Records Rundle, in Chaotic Processes in the Geological Lett. 16, 199 (1989); R. S. Stein, G. C. P. King, J. from the Landers, Califomia Earthquake of 28 Sciences, D. A. Yuen, Ed. (Springer-Verlag, Ber- Lin, Science 258, 1328 (1992). June 1992 (Rep. OSMS 92-09, California Depart- lin, 1992), pp. 293-303]. 6. K. Sieh et a., Science 260, 171 (1993). ment of Conservation, CDMG, Sacramento, 34. Y. Okada, Bull. Seismol. Soc. Am. 82, 1018 7. We take the southern margin of the Great Basin as 1992)]. (1992). defined by the belt of pronounced geophysical 20. The relation between mean stress, a, and dilata- 35. We thank J. Savage, P. Spudich, T. Hanks, R. anomalies that cut across the southern tip of tional strain is ff = kO,where a = 1/3(or1 ?+ a22 + Aster, P. Silver, and an anonymous referee for Nevada the between 36.50 and 37?N (the dashed J33). The stress tensor for a plane shear wave has their helpful comments in review. line in Figs. 1 and 2). These coincidental anoma- the form T = -ua(x,t) (,u,) [bk + kb], where k lies include pronounced gravity and topographic and b are unit vectors in the propagationand 8 February1993; accepted 3 May1993

SCIENCE * VOL. 260 * 11 JUNE 1993 1623

This content downloaded from 128.95.104.66 on Mon, 28 Oct 2013 21:58:06 PM All use subject to JSTOR Terms and Conditions