The Magnitude 6.7 Northridge, , of 17 January 1994 Author(s): Scientists of the U.S. Geological Survey and the Earth Source: Science, New Series, Vol. 266, No. 5184 (Oct. 21, 1994), pp. 389-397 Published by: American Association for the Advancement of Science Stable URL: http://www.jstor.org/stable/2885318 Accessed: 26/04/2010 18:04

Your use of the JSTOR archive indicates your acceptance of JSTOR's Terms and Conditions of Use, available at http://www.jstor.org/page/info/about/policies/terms.jsp. JSTOR's Terms and Conditions of Use provides, in part, that unless you have obtained prior permission, you may not download an entire issue of a journal or multiple copies of articles, and you may use content in the JSTOR archive only for your personal, non-commercial use.

Please contact the publisher regarding any further use of this work. Publisher contact information may be obtained at http://www.jstor.org/action/showPublisher?publisherCode=aaas.

Each copy of any part of a JSTOR transmission must contain the same copyright notice that appears on the screen or printed page of such transmission.

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 m ARTICLE The Magnitude 6.7 Northridge, California, Earthquake of 17 January 1994 Scientists of the U.S. Geological Survey and the Southern CaliforniaEarthquake Center

The most costly American earthquake since 1906 struck Los Angeles on 17 January 1994. The magnitude 6.7 Northridge earthquake resulted from more than 3 meters of reverse slip on a 1 5-kilometer-long south-dipping thrust fault that raised the Santa Susana mountains by as much as 70 centimeters. The fault appears to be truncated by the fault that broke in the 1971 San Fernando earthquake at a depth of 8 kilometers. Of these two events, the Northridge earthquake caused many times more damage, primarilybecause its causative fault is directly under the city. Many types of structures were damaged, but the fracture of welds in steel-frame buildings was the greatest surprise. The Northridge earthquake emphasizes the hazard posed to Los Angeles by concealed thrust faults and the potential for strong ground shaking in moderate .

On 17 January1994, at 4:30 a.m. Pacific lost in longer commutes,and as even un- structural damage observed, and discuss standardtime (12:30 UT), the first earth- damagedresidential buildings lose renters the implications of this event for earth- quakesince 1933 to strikedirectly under an in the epicentral region. In the midst of quake hazardassessment and mitigation. urban area of the United States occurred these losses, the gains made through the in a northern suburbof Los Angeles, Cal- earthquakehazard mitigation efforts of the Tectonic Setting ifomia. This magnitude (Mw) 6.7 (1) past two decadeswere obvious.Retrofits of earthquake resulted from thrust faulting masonrybuildings helped reduce the loss The causativefault of the Northridgeearth- on a fault dipping down to the south- of life, hospitals suffered less structural quakeis partof a broadsystem of thrustfaults southwest beneath the northem San Fer- damage than in the 1971 San Fernando that resultfrom the 160-kmleft step in the nando Valley (Figs. 1 and 2). It produced earthquake,and the emergency response Pacific-NorthAmerican plate boundaryat the strongestground motions ever instru- was exemplary. The Northridge earth- the Big Bend of the San Andreasfault (Fig. mentally recordedin an urban setting in quake proved that preparing for earth- 2). The northwestwardmotion of the Pacific North America and the greatestfinancial quakescan greatlyreduce the damagethey plate along the San Andreasfault requires lossesfrom a naturaldisaster in the United do. In this article, we examine the geolog- compressionof the crustaround this easterly States since 1906. Although the ic setting and the shaking and ground bend (4). More than 10 mm/yearof this Northridgeearthquake was the same size deformationproduced by the Northridge compressionis accommodatedon east-strik- as the nearby 1971 San Femando earth- event, analyze how they relate to the ing thrustfaults and folds of the Transverse quake (Mw 6.7) (Figs. 1 and 2), it was much more damaging,in part because of 120? 118 117 its location directly beneath the San Fer- 35~ A" , nando Valley and its closer proximity to communities in the Los Angeles basin. The Northridge earthquakedisrupted N- the lives and livelihoods of many of the residentsof southernCalifornia. Casualties included33 dead as a direct result of the earthquake,more than 7000 injuriestreated at hospitals,and over 20,000 homeless (2). Financiallosses have been estimatedat $13 billion to $20 billion (3). Sections of three majorfreeways were closed, including the ! Sant - Bernardino busiest highway in the country, Interstate 3419117 , 10. The lossescontinue to growas damaged businessdistricts lose customers,as time is Angeles 34',-e Los\,r ' Contributing scientists are from the Universityof South- ern California,United States Geological Survey, Califomia 0'0 Institute of Technology, Jet Propulsion Laboratory, Uni- versity of Californiaat San Diego, Califomia Division of Mines and Geology, and Columbia University.Direct con- tributions were made by L. Jones, K. Aki, D. Boore, M. Celebi, A. Donnellan, J. Hall, R. Harris, E. Hauksson, T. km f. N Heaton, S. Hough, K. Hudnut, K. Hutton, M. Johnston, L------30 W. Joyner, H. Kanamori,G. Marshall,A. Michael, J. Mori, M. Murray, D. Ponti, P. Reasenberg, D. Schwartz, L. Seeber, A. Shakal, R. Simpson, H. Thio, J. Tinsley, M. Fig. 1. Digitalshaded reliefmap of southern Californiatopography produced from United States Todorovska, M. Trifunac, D. Wald, and M. L. Zoback. GeologicalSurvey (USGS) topographic data files, showing faults and M ?: 5.0 earthquakessince 1932. Communications should be directed to L. Jones, U.S. The earthquakesare shown by lowerhemisphere focal mechanisms with size proportionalto magnitude, Geological Survey, 525 South Wilson Avenue, Pasadena, CA 91106, USA, or K. Aki, Department of Geology, Uni- and compressionalquadrants are shaded ifthe mechanismis knownand indicatedby open circlesif itis versity of Southem Califomia, Los Angeles, CA 90089, not known.The mechanismof the 1994 Northridgeearthquake is shown in red, and an outlineof its USA. aftershockzone is in yellow.SS, Santa Susana;SF, San Femando.

SCIENCE * VOL.266 * 21 OCTOBER1994 389 Rangesand the Los Angeles basin that we partsof Califomia,closing at 7 to 10 mm/ 1) in two temporaland spatialclusters. The refer to as the Big Bend Compressional year (9, 10), lies between these two faults. firstwas from 1920 to 1942 along the south- Zone. The zone includes many north-dip- The causativefault of the Northridgeearth- em Los Angeles basin, whereasthe second, ping and south-dippingsubparallel faults, quake does not extend to the surfaceand from 1970 to the present, is concentrated only some of which come to the surface was not mappedbefore the event. along the northemedge of the Los Angeles (5). The interaction of these faults is not The dense system of exposed and con- basin(14) and involvesboth concealed(15) yet well understood,but no one fault dom- cealedthrust faults along the northemflank and exposed thrustfaults as well as strike- inates the deformation. of the Los Angelesbasin, coupled with high slip faults in the CompressionalZone. The The 1994 Northridge earthquakeoc- geodeticrates of compression,imply that the 1971 San Fernandoearthquake (Mw 6.7) curredat the intersectionof severalmapped northemLos Angeles region faces one of the (16) was located just northeast of the faults(Fig. 2) and south of the westem end greatestseismic hazards in southemCalifor- Northridge earthquake,also on a west- of the longest exposed thrust fault in the nia ( 11). A reportin revisionat the time of northwest-strikingplane but dippingdown Big Bend CompressionalZone, the Cu- the Northridgeearthquake put the northem to the north, as part of the SierraMadre camonga-SierraMadre fault. This faultdips San FemandoValley in the top one-sixthof system that raises the San Gabriel moun- north under the San Gabriel Mountains, the areasin southem Califomiain termsof tains (Figs. 1 and 2). That earthquake(17, the steepest mountainsof the belt (Figs. 1 seismicpotential (12). This systemof faults 18) is thoughtto have been boundedon the and 2), slipsat 3 mm/year(6), andproduced underliesthe heavilyurbanized sedimentary west by a northeast-strikingleft-lateral tear the 1971 San Femando earthquake(7). basins,so that althougheach fault in this fault, called the "Chatsworthtrend." Other Fifteen kilometers west of the causative zone moves more slowly and in smaller moderateearthquakes in the most recent fault of the Northridgemainshock, the Si- earthquakesthan is expectedof the strike- temporalcluster occurred on the SierraMa- erraMadre fault systemsplits into two sur- slip events on the , in dre (19), ElysianPark (20), and Raymond ficial faults, the north-dippingSan Cay- aggregate,the seismic risk (13) from these faultsystems (21). The northernflank of the etano fault and the south-dipping Oak faultsmay be greater. Los Angeles basin has also sustaineda high Ridge fault (8). The Ventura basin, the Since 1920, 18 moderate(M 4.8 to 6.7) degreeof backgroundmicroseismicity in the world'sthickest section of Pleistocenesed- mainshock-aftershocksequences have oc- past decade (22). Focal mechanismsin the iments and one of the fastest-deforming curredin the greaterLos Angeles area(Fig. immediate vicinity of Northridge show

*60LX \ \ .1983,M.6.3 SIERRA BA ASIN 360 V| ie ar93 RANGE % k

Pt. C -~~~~CNTALG9o 11-7.3

340~~~~ | 19;46,M946,Ms60~~~ Borb~~~~~~~~arrio tn F L

0 SonLuisPlai50 100Km a 44~~~~~~~M. r Obispo ] 1952f z I S

2400220 is01i87,M i987M' 199 iM60i1 M=7.33Ma 1947, *19St6M6.o

222? $21? 120? i Fgo C'%rz?*6.4 SQ Zon

faultrupture is shownin red. Large arrows indicate the sense and magnitudeof platemotion. Some faultnames are abbreviated as follows:ADF, Anacapa Dume fault;BAF, Banning fault; OF, Cucamonga fault; MOF, Mission Creek fault; ORF, Oak Ridge fault; PVF, Palos Verdesfault; RHF, Raymond Hill fault; SDT, San DiegoTrough-Soledad fault; SFF, San Fernandofault; SMF, Santa Monica fault; SRF, Sierra Madre fault; SSF, SantaSusana fault; and SOF,San Cayetanofault.

390 SCIENCE * VOL.266 * 21 OCTOBER1994 m ARTICLE thrustearthquakes in the north-dippingSan surfaceof the earth.The 1971 San Fernando 1981 to 1993 that wereabove M 1.7, which Fernandorupture zone, in its westwardex- earthquakeoccurred on a fault parallelto is typical of the dispersedbackground seis- tension,and near the Northridgemainshock the Northridgefault but dippingin the op- micity in the Los Angeles region.Ten days fault plane. Strike-slip events continued posite direction, down to the north. The before the event, a swarmof small earth- alongthe Chatsworthtrend (23) duringthis 1994 fault dips up to the north towardthe quakes (includingfour M 3.0 to 3.7) oc- period. 1971 plane (Fig. 3). cuffed along the coast west of Los Angeles. Models of both body wave and surface The alignment of the epicenters suggests The EarthquakeSource wave data give a seismicmoment of (1.2 ? that they were on a shallow south-dipping 0.2) x 1019N-m, and all the geodeticmod- fault, parallel to the Northridgefault but Mainshocksource parameters. The North- els suggestsimilar moments. All these mod- offset at least 25 km to the south. Because ridgeearthquake originated at 340 12.53'N, els imply that the slip was greaterthan is they occurredon a separatefault, and M 2 3 1180 32.44'W, about30 km west-northwest normallyseen for thrustfaults with a rup- events occur in the Los Angeles basin 5 to of downtownLos Angeles at a focaldepth of ture surfaceof less than 250 km2 (24). The 10 timeseach year,we see no directrelation 19 km. The first motion focal mechanism maximum slip in the Northridge earth- between these events and the Northridge showed almost pure thrust motion on a quake exceeded 3 m and was 5 to 10 km earthquake. plane strikingN70?W to N80?W and dip- northwest and updip of the hypocenter. Aftershocks.3000 aftershocksof M > 1.5 ping 350 to 450 to the south-southwest(Fig. The passageof the rupturefront through were recordedin the first 3 weeks of the 1). Models of long-periodwaveforms and this region of high slip produceda distinct sequence(Fig. 3) (25) in threeclusters. One geodeticoffsets suggest similar fault orienta- pulse of energy about 2 s after the start of cluster,associated with the mainshockrup- tions, althoughseveral of the longerperiod the earthquake,leading to early media re- ture, extends about 15 km west-northwest solutionsyield a more northerlystrike. The ports that the Northridgeearthquake was fromthe mainshockepicenter and about 15 earthquakebegan at the southeastemcomer actuallytwo events. Becausethe amountof km to the north-northeast.It defines a 350 of this plane and rupturedup to the north- slip, and thus the seismic moment release, to 450 dippingplane froma depth of 19 km west for about 15 km. This pattem suggests was large for the size of the fault, and be- to about8 km (Fig.3), the southemhalf of that rupture began on a plane striking cause the durationof the event dependson which is under the San Femando Valley. N75?W and bent to the north as it propa- the size of the fault, the seismicmoment of The largest(M 5.9) aftershockoccurred 1 gatedto the west.We have no teleseismicor the earthquakewas released in a shorter min afterthe mainshock,along the eastern geodetic evidence of slip above a depth of time than is usualfor an event of this size. edge of the dippingplane. A second cluster about 8 km. Shakingfrom the earthquake Foreshocks.The Northridgeearthquake of shalloweraftershocks above the dipping might have been more severe in a limited had no immediateforeshocks. The after- planeapparently reflects diffuse deformation area if the rupturehad come closer to the shockzone averaged22 eventsper year from of the overlyinganticline. A thirdcluster, a 10-km-longgroup in the northwestcorner of Fig. 3. (A) Map of 1500 A a a' the aftershockzone, has an approximately verticaldistribution beneath the Santa Su- epicenters of the 1994 0-. 4aLJ1l.I_LtJ..L t_.tiL?_IA t4 JlLAItL,hLI Northridge earthquake and g: ?'flt sanaMountains, appearing to be on second- aryfaults that did not rupturein the main- shocks recorded between . a shock.This clusterincludes the second larg- 17 January 1994 and 10 E . no est aftershock(M 5.6), which occurred11 February 1994, with at least - a hoursafter the mainshockat a depth of 11 30 arrivaltimes) and of the Q. km, with a thrust-faultingfocal mechanism 1971 San Femando earth- n -20 similarto that of the mainshock.This zone, quake and its aftershocks * 1994 Northridge a all aftershocksnorthwest of the M (blue) (1971-1980) with 1971 San Fernando including faults from (49). Earth- 4 5.6 aftershock,developed only after the M quakes of M ? 5.5 are -30 5.6 event occurred.The full extent of the shown by stars. The two 0 10 20 30 40 50 aftershockzone was definedby the 500 af- largest aftershocks are at Distance (km) tershocksin the first 24 hours of activity the eastem and westem B (26), and the 6300 aftershocksof the subse- edges of the aftershock quent 6 months (18 Januaryto 16 July) zone. (B)The hypocenters in \ occurredwithin the same region. the box in (A)projected onto Althoughthe Chatsworthtrend bounded a line trending N30?E. do the 1971 aftershockzone to the west (17), we do not see a similarstructure in the 1994 aftershocks. Rather, the probable 1994 2 *0 ? l ../ mainshockrupture plane extends about 15 km and is bisectedby the Chatsworthtrend. Northridgeaftershocks occurred within the Northridgehanging wall near the Chat- sworthtrend; however, unlike the situation in 1971, few of the 1994 aftershockshad 200m strike-slipmechanisms and the few that did, did not align. It thus appearsthat the 1994 1 0t_- mainshock broke across the Chatsworth trendand produced slip on both sidesof that tearfault without reactivating it. Mostof the 1971 aftershocksdefining the Chatsworth 50' 40' 30' 20' 10' trendwere shallow (<10 km), so this struc-

SCIENCE * VOL.266 * 21 OCTOBER1994 391 ture might exist only in the hangingwall of show that the region was lifted up as much lisecondsbefore the event. Some minorre- the 1994 earthquake. as 70 cm and displaced horizontally as laxation occurredin the few minutes after By fitting a decay rate equation to the much as 21 cm (Fig. 4). The earthquake the main rupture. aftershockdata from the first 12 weeks,we raised the Santa Susana Mountains and Groundshaking. The Northridgeearth- can estimatethe numberof aftershocksto the northern San Fernando Valley. Be- quake produced strong ground motions expect in the future (27). The Northridge sides the groundmotion directly attribut- acrossa largepart of the Los Angeles met- aftershocksequence has an overallproduc- able to slip on the fault, seven stations to ropolitanarea (Fig. 4). Morestrong motion tivitythat is greaterthan averagebut is dying the north and west of the ruptureshow recordswere obtainedwithin 25 km of the off slightlymore quickly than is averagefor several centimeters of westward motion sourcethan had been recordedever before Califomiaaftershock sequences. The proba- that cannot be modeledby the mainshock for a single event. More than 200 free-field bility of an additionalaftershock above M 5 or the majoraftershocks. recordingswere made that showed acceler- between 1 August 1994 and 31 July 1995 is Continuous high-precisionstrain mea- ations above 0.01g (28). 45%. surementswere made in boreholesat dis- On average,the peakhorizontal acceler- tances of 74 and 196 km from the North- ationsrecorded in this earthquakewere larg- Earthquake Effects ridge earthquake.The coseismic,peak dy- er for its magnitudethan those recordedfor namicstrains observed in the boreholesdur- other reverse-faultingevents (Fig. 5). How- Crustaldeformation. Elastic strain released by ing the passageof the S wavesexceeded 10 ever, how the accelerationdiminishes with the Northridgeearthquake measurably de- microstrain.Net offsetswere 21 nanostrain distance from the source in this type of formedthe crustover 5000 km2surrounding of extension and 5 nanostrainof compres- earthquakehad not been well established the epicenter.Observations of the displace- sion,which is consistentwith the momentof becauseof the paucityof data availablebe- ments of 60 surveystations determined by the earthquake.No systematicchange in fore now, especially at very short dis- meansof Global PositioningSystem (GPS) strain above the backgroundnoise of 0.1 tances.Although early reports suggested that satellites before and after the earthquake nanostrainoccurred during the hoursto mil- high vertical accelerationsmay have con-

30' 30 | s | | s | | |I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....I.[X1t...1tv*|{ ......

20' 1E_ _> c...... ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~....

207t . Grouqidds op1 *wAg!l z:lb ' .b_W! .

_i. LA ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ b

_I |- T Xs s s *. *'-* g | | f | g f | | w | f | | f w . 10'l~~~~~~~~~~~~~~~~~~~~~~~ 10'

34. .I T 50' 40 301 20' 10 118 Fig. 4. Digital shaded relief map of southern Califomia topography produced from USGS topographic data files showing data from the 1994 Northridge earthquake, including zones of ground deformation (red dots), contours of uplift(blue lines) and honzontal displacements (blue arrows) inferredfrom a model (white box) fit to the geodetically recorded deformations, most of the liquefaction sites (orange stars; some may not have been included in the early reporting), and measurements of peak ground accelerations (red to pink triangles).

392 SCIENCE * VOL.266 * 21 OCTOBER1994 ARTICLE

tributed to the extensive damage caused by amplitude pulse that is indicative of source tectonic deformation. These zones were con- the Northridge event (an idea adopted by directivity. For many larger structures, peak centrated near the epicenter, in Granada many who confused the vertical motion of ground velocity is a better measure of dam- Hills just east of the inferred rupture surface, the fault block with vertical shaking at a age potential than is peak ground accelera- and along the north flank of the Santa Su- site), the ratio of peak vertical to peak hor- tion. sana Mountains (Fig. 4). izontal acceleration in this earthquake is not As in other earthquakes, soft soils may Most of the surface displacements ob- anomalous. The vertical and horizontal have produced larger ground motions locally served in natural ground in these areas are ground accelerations and velocities were (34). Several of the larger peak accelerations extensional and cumulative displacements large, but the average peak accelerations are were located south of the epicenter, where across zones of fractures rarely exceed a few no more than one standard deviation above propagation and site effects were probably tens of centimeters. Where these fractures the mean of the data from other earthquakes. more important than source radiation alone. cross streets and sidewalks, they are refracted The high seismic moment release for a small The short duration of the peak accelerations into complex series of pavement cracks and area of fault, and thus a short duration, could at these sites leads to lower peak velocities buckles, spalled and extended curbs, and have led to higher peak accelerations (29). as compared with the northern sites. Far- tented sidewalk slabs; many pavement cracks The systematic variation in overall acceler- ther south in the northern Los Angeles resulted from decoupling of the pavement ation between earthquakes has been recog- basin, the generation of surface waves from the ground below. Most of the surface nized for some time (30). along the edge of the basin (35) may have displacements appear to be shallow and dis- The most important factor controlling played a role in the high accelerations and play no afterslip. They have orientations and the amount of strong shaking in one event is extensive damage caused- in Santa displacements consistent with liquefaction- the distance of the site from the fault plane. Monica, Hollywood, and south-central induced lateral spreading (36) and compac- Sites closest to the Northridge earthquake Los Angeles. When data on site condi- tion of loosely consolidated sands within are those north of the epicenter, because the tions have been collected, this earthquake artificial fill and buried stream deposits. For plane is shallower to the north. Besides dis- will provide an opportunity to learn more these reasons, we attribute most of the de- tance to the source, a number of other fac- about the effect of site conditions on formation to ground failure from strong tors contributed to the variability of ground ground motion. shaking. However, some of the features are motions that is apparent in Figs. 4 and 5. Ground rupture.Surface deformation from associated with mapped faults and fold axes Directivity (31) probably increased the earthquakes can be produced by ground rup- and may involve bedrock; thus, some fea- ground motions at sites to the north of the ture on the causative fault (primary fault- tures may be a response to folding during epicenter as the rupture propagated toward ing), displacement on nearby faults (sympa- coseismic uplift of the mountains and the them. In the region 10 to 15 km north- thetic faulting), bedding-plane slip and ex- northern San Fernando Valley above the northeast of the epicenter, where we would tension to accommodate coseismic folding, concealed causative fault. expect the combined effects of radiation pat- and shaking-induced ground failures such as Although displacements on the second- tern (32) and directivity to be maximized for liquefaction and landsliding. In contrast to ary ground ruptures are small, the linear ex- this fault geometry, the recorded ground ve- the 15 km of well-defined surface faulting in tent of these zones is comparable to what locities are among the largest ever recorded. the 1971 San Fernando earthquake (7), the might be expected for a surface-faulting The recorded peak horizontal ground veloc- Northridge earthquake produced no clear earthquake of similar magnitude. They also ity at a free-field site near the county hospi- evidence of primary surface rupture. It did, caused substantial damage in densely devel- tal in Sylmar (15 km north-northeast of the however, cause permanent ground deforma- oped areas (37). The ability to recognize epicenter) was about 130 cm/s; the peak tion over a wide area. Along with obvious zones of secondary ground deformation be- velocity was over 170 cm/s at the Los Ange- liquefaction and slope-failure features, broad fore the next large earthquake represents a les Department of Water and Power Rinaldi zones of secondary surface ruptures were ob- major but not insurmountable challenge. receiving station several kilometers south of served that are not readily attributable to One key may be that much deformation has the hospital (33). The ground velocities in common modes of shaking-induced ground occurred repeatedly in the same areas (38), this region are dominated by a single large- failure and may in part be a response to and in many cases, these zones have subtle topographic expression. Secondary fractures

Fig. 5. The larger peak of the two ._ .______._._ are also commonly aligned along fault zones horizontal components of accelera- or axial surfaces of folds and preferentially tion in the Northridge earthquake 0 occur in regions underlain by soft or steeply (filled circles) (28) compared with 1 dipping sediment. the median and +1 cf curves given o -. Liquefaction.Liquefaction produced sand by the equations of (30) for a M X" blows and other evidence of permanent 6.7 earthquake and site class B a ** ground deformation in Holocene alluvial de- (generally, very stiff soil or soft rock), X posits and filled land at several sites within which we believe represent typical Xe 48 km of the This defor- Northridge sites. The acceleration a : epicenter (Fig. 4). data are plotted against the closest < 0.1 - mation damaged pipelines, water supply horizontal distance to the rupture C channels, filtration facilities, parking lots, defined by GPS data. We plotted s- 0~~~~~~~~~~~~ residential and commercial buildings, storm only records whose motion was ? drains, and flood control debris basins. How- judged to be unaffected by struc- ever, the Northridge earthquake caused tures they were on or near. The o * 1994 Northridge, M =6.7 * - - much less ground failure due to liquefaction open circles show values from the BJF, reverse slip, site class B than many other earthquakes of its size. The 1971 San Fernando earthquake. 0010BJF ...... E 1 a em o 1971 San Fernando, M = 6.6 near-sufface deposits in the epicentral region are mainly cohesive clay and clayey silt. 1.0 10 100 Such sediment is not generally susceptible to Shortest horizontal distance liquefaction, which might explain the rela- to map view of rupture surface (km) tively sparse incidence of observed liquefac-

SCIENCE * VOL. 266 * 21 OCTOBER 1994 393 tion-relateddamage in this area.A clusterof very strongshaking and were mostly unaf- Nonductilereinforced concrete buildings sites 10 to 15 km northeastof the epicenter fected, as was the new Los Angeles subway built before the mid 1970s, when lessons (Fig. 4) experiencedliquefaction both in system. from the 1971 San Femando earthquake 1994 and in the 1971 San Fernandoearth- Many URMs crackedand partsof their were incorporatedinto the codes, behaved quake,but displacementsat the groundsur- walls fell outward,but few life-threatening poorly,with partialcollapses of a multistory face weresmaller in 1994. It is still unclear collapsesoccurred. Many of these buildings medicalclinic (Fig. 6) and of a mall, both whetherthe smallerground displacements are residentialbut not one life was lost, high-occupancystructures during business resultedfrom compaction of soils causedby althoughfew URMs exist in the immediate hours. In addition, a hotel, condominium shaking,a loweredground-water table that epicentralregion. The San FemandoValley tower,hospital, and office buildingwere se- increasedthe resistanceto liquefaction,the had few buildings in 1933 when damage verelydamaged. Modem reinforced concrete shorterduration of shakingin 1994 as com- fromthe Long Beach earthquakeinspired a structuresfared fairly well, with the excep- paredwith 1971 (39), or to engineeredcoun- change in the Califomiabuilding code. An tion of pre-cast concrete parking garages tenneasurestaken at some sites to mitigate URM retrofitprogram instituted by the City (Fig. 6), six of which partiallycollapsed. the liquefactionhazard. of Los Angeles in the early 1980s has Inadequateconnections and poorlateral de- Regionalliquefaction hazard maps of the strengthened6000 buildingsand undoubt- formationcapability of componentsintend- LosAngeles region are based on the assump- edlyhelped to preventloss of life duringthe ed to carryonly verticalload mayhave con- tion that highly liquefiable,loose, clay-free, Northridgeearthquake. Damage to URMs tributedto these collapses. sandy,alluvial fan depositsor narrowchan- occurredboth in LosAngeles, where most of Wood-framebuildings, both old andnew, nel depositsof formerstreams could experi- thesestructures have been retrofitted,and in showed deficiencies.Inadequate bracing in ence liquefactionwhen persistentshallow adjacentcities withoutretrofit programs, so parkingareas in the groundstory of multi- groundwater is present.However, in many the benefitsof retrofitcan be studied. story residentialstructures caused collapses areas,including the westernSan Fernando Valley,these deposits are not mappablefrom surfaceexposure. Thus, regionsof persistent shallowground water (<3 m) were mapped to highlightareas where additional site-spe- cific studiesmight be advisableto determine if susceptibledeposits were present (40). Ad- ditional studies of the permanentground deformationdescribed above are needed to determine if liquefaction-inducedground failure, settlement, or seismicallyinduced compaction of small bodies of dry, loose sedimentcould explain its occurrence.For muchof the epicentralarea, if ground-water levels are maintainedat or below present levels, the risk of liquefactionin buried channeldeposits confined by nonsusceptible sedimentfor a comparablysized earthquake is probablylow.

Damage to Structures Structural damage from the Northridge earthquakewas extensive but not devastat- ing. About 3000 buildingswere deemed un- safe by buildinginspectors, which is only a -. smallfraction of the buildingsin the region - of strong shaking, and many of those are repairable.Substantial failures were seen in unreinforced masonry buildings (URMs, which areprimarily older brick buildings), in nonductile reinforced concrete (pre-1971 multistorycommercial buildings), and in wood-frame(primarily two- to four-story -~N apartments)and steel-frame(modern engi- neered buildings)buildings. Nonstructural losses, includingdamage to the contents of Fig. 6. Illustrationsof structural damage from the Northridge earthquake. (A) This medical clinic in the buildings,were major, probablyexceeding northem San Femando Valley is an older, nonductile concrete building. Its second story completely the total cost of the structuraldamage. Sub- collapsed as did both ends of the building in the stones above. (B) The collapse of this parking garage stantial damagealso occurredto bridges,a probably originated in the interior,and the post-tensioning cables in the slabs pulled down surrounding majordam, electric power facilities, and wa- areas. (C) A close-up photo of part of a beam-to-column joint in a steel building that suffered a weld fracture between the lower flange of the beam and the column. The bolts shown are replacements of ter and gas pipelines.Restoration of utilities original ones that sheared off when the lower flange weld broke. (D) This failed column supported a was rapidbecause of the use of redundant section of Interstate 10 and is another example of the poor behavior of nonductile concrete. The bridge and backupsystems. The high-risesin down- deck fell onto some masonry walls of a leased storage space. (Photos courtesy of MarvinHailing [(A) and town LosAngeles were outside the regionof (D)]and John Hall [(B)and (C)])

394 SCIENCE * VOL.266 * 21 OCTOBER1994 a ARTICLE

of the ground story at several apartment tion records at the sites of the base-isolated diction of the ground motions (given an complexes. Reliance on brittle materials buildings showed only limited long-period earthquake of some size at some distance, such as stucco for lateral strength proved motions, so this successful experience does how will the ground move under a building), unwise as these materials broke down under not prove them fail-safe. Other sites with and building response (how do buildings be- cyclic loading. The modem trend toward much larger long-period ground motions, have when subjected to those ground mo- fewer, heavier shear walls in wood-frame especially the county hospital in Sylmar, tions). Some of the results from the construction created large overturning forc- would have provided a more demanding Northridge earthquake confirmed research es and caused base anchors to fail. However, test of a base-isolated structure. over the past decade by the National Earth- postearthquake reconnaissance of damaged The Northridge earthquake struck during quake Hazards Reduction Program. Thrust wooden buildings revealed that many were California's ongoing seismic retrofit program faults concealed below Los Angeles present a not constructed according to approved for bridges that began in earnest after the threat to the region that approaches that plans, which suggests that a lack of proper 1989 Loma Prieta earthquake. Altogether, posed by the San Andreas fault. When inspection was a major reason for much about 1600 state and county bridges were earthquakes occur directly beneath a city, it damage. subjected to shaking exceeding 0.25g. Free- will be subjected to ground motions with The most disturbing structural behavior, way bridges in California are typically com- peak accelerations approaching the force of with potentially great economic implica- posed of reinforced concrete box girders sup- gravity, exceeding the amounts of shaking tions, was the brittle fracture of welded con- ported on reinforced concrete columns. Sev- anticipated by building codes in some re- nections in steel-frame buildings, most often en such bridges collapsed. Five of these were spects. Some of the results, especially con- the beam-to-column connections (Fig. 6) of pre-1971 nonductile design and had been cerning the effects of earthquakes, reinforced that give a building its lateral earthquake scheduled for retrofit, and the other two findings from previous earthquakes. Ground resistance. The steel used in buildings is date to the mid-1970s and were of better failure induced by shaking can be as exten- commonly believed to possess excellent duc- design. One of the collapses was of a high sive as that caused by direct faulting, and the tility, but apparently the welding procedures bridge; excessive sway pulled the expansion system of concealed faults under Los Angeles currently in use do not achieve this desired joints apart, causing decks to fall. Inade- is more complex than previously thought, behavior. Laboratory shaking tests some- quately reinforced columns (Fig. 6) caused dipping both to the north and south. The times show poor behavior but not to the the other collapses; the columns of the two severe damage done to two- to four-story extent seen in this earthquake. Although the more recent bridges were still substandard residential buildings built over parking ga- problem seems serious, none of the 100 or so even though these two bridges had not been rages has important implications for regional buildings identified as having connection placed on the retrofit list. Several older housing needs because of the pervasive exis- fractures collapsed or even developed a seri- bridges that had had their columns retrofit- tence of such structures. Some results were a ous lean. Some damaged buildings are yet to ted by steel jackets performed well but did surprise. The widespread fracturing of welds be identified, because the cracks are hidden not experience the very strong shaking. in steel-frame buildings was unexpected be- behind fireproofing and architectural finish- Of the more than 100 dams located with- cause of the ductility of steel. Understanding ing materials. Many aspects of the problem, in 80 km of the epicenter, only Pacoima the cause and correcting the problem will be including proper repair strategies, remain to Dam, an 111-m-high arch dam (located 18 essential for continued construction in be resolved (41). km from the epicenter but only 10 km from earthquake-prone regions. Damage to furnishings, ceilings, glass, the fault plane), suffered notable damage. Because no one fault dominates the Big piping, and equipment was extensive. Al- Recorded accelerations as high as 2g on the Bend Compressional Zone, hazard mitiga- though hospitals are designed for higher canyon walls triggered numerous rock falls. tion efforts that focus on avoiding one or a seismic forces than are ordinary buildings, A 5-cm-wide crack opened at the left abut- few fault structures are not appropriate. several were forced to close temporarily ment of the dam because of movement of the With scores of faults, each moving no more solely because of nonstructural damage. adjacent rock mass, and cracks in the upper frequently than once or twice a millennium, This included the county hospital in Syl- part of the dam were a testament to the the approach to mitigation must be region- mar, an extremely strong post-1971 struc- strong shaking. The water level during the ally based. The most active fault need not, ture with steel shear walls, where a peak earthquake was low, and because Pacoima and probably will not, be the fault that horizontal acceleration of 2.3g was recorded Dam is operated for flood control, high water ruptures in our lifetime. For instance, the on the roof in response to 0.9g shaking at is infrequent. After an embankment dam Northridge earthquake raised the northern the building's base. Schools suffered much liquefied and was nearly overtopped during San Fernando Valley by several tens of nonstructural damage, and falling lights the 1971 San Fernando earthquake, the Cal- centimeters, yet other earthquakes must would have claimed lives had schools been ifornia dam regulatory agency has overseen raise the adjacent mountains at a greater in session. Water damage from broken pip- seismic improvements at 27 dams in the rate than the valley was raised by this earth- ing required massive cleanup efforts in region, including the installation of rock an- quake for the valley to be a valley. Earth- many buildings. Malfunctions of backup chors on the left abutment of Pacoima Dam quake hazard mitigation in Los Angeles power systems affected hospitals, telephone that helped to limit the rock movement must take all these faults into account. service, and emergency response operations. there during the Northridge earthquake. Concealed faults beneath Los Angeles Two major base-isolated buildings were have been recognized from detailed maps of shaken by moderate ground accelerations, Implications for Earthquake Risk overlying geologic structures (42) and dis- reaching 0.5g horizontal at one site, and tributions of microseismicity (22). Geodetic performed well without damage. In this de- Detailed studies of major earthquakes data have been used to infer rates of motion sign, buildings are supported on rubber through the earth sciences and engineering that together with seismological or geolog- pads, which provide flexibility to isolate provide us with the knowledge to mitigate ical evidence of the location of the faults against horizontal ground motions. A criti- future earthquake hazards. These studies in- can provide an estimate of the hazard from cal design objective is to avoid excessive clude characterization of the earthquake the faults. Such studies have found several pad displacement, and this can be difficult source (using geologic, geodetic, and seismo- probable concealed thrust systems in the if the ground motion contains a strong logic data to estimate how large an earth- Los Angeles basin south of the San Fer- long-period component. Strong ground mo- quake can occur with what probability), pre- nando Valley. However, neither of the con-

SCIENCE * VOL. 266 * 21 OCTOBER 1994 395 cealed faults in the 1987 Whittier Narrows nearby faults, including the San Andreas sures among the public, such as securing or 1994 Northridge earthquakes was recog- fault. However, calculations suggest that if computers, water pipes, ceiling tiles, and nized before the events. Our inability to the frictional strength of the San Andreas bookcases, could save billions of dollars in recognize the causative structures before the fault is low (47), then stress changes on the future earthquakes. Nontraditional design earthquake is a testament to the inadequa- fault caused by Northridge are small. The technologies such as base isolation show cies of present geologic data and the non- most the next large earthquake on the San promise in reducing property losses and uniqueness of structural models. An impor- Andreas fault might be delayed or advanced maintaining functionality after an event, tant question is whether more detailed is 2 to 3 years, a calculation based on com- and their development should continue. analyses of the San Fernando Valley before parison of the magnitude of Northridge-in- The widespread ground failure caused by the Northridge earthquake would have al- duced stress changes with the normal tecton- the Northridge earthquake was similar to lowed us to recognize that concealed fault ic loading rate for the San Andreas fault. If the ground failures caused by the 1989 Loma as active before a major event. Analysis of the nucleation point of the next southern Prieta earthquake, although in neither case the secondary zones of deformation pro- San Andreas event is not in the region near did failure show any direct connection to duced in the Northridge earthquake may the Northridge earthquake, then the stress the causative fault. These types of deforma- provide important constraints on models of change caused by Northridge is not likely to tion are, in part, related to near-surface geo- concealed thrust faults. Recognition and have much effect at all on the timing of that logic conditions that can be identified and analysis of such features in other regions event, although post-Northridge aseismic af- mapped for all urban areas of California. potentially could provide a method to assess terslip or relaxation in the region might con- Such hazard identifications should become the activity and recurrence intervals of siderably alter the stress distribution over part of future land-use planning practices. earthquakes on concealed faults. time. Almost 100 faults in the Los Angeles The full extent of the urban corridor from The damage to modem buildings in the metropolitan area have been identified as San Bernardino through Los Angeles and Northridge earthquake will trigger an inves- capable of damaging, M 2 6 earthquakes northwest to Santa Barbara is at risk from tigation into the adequacy of the current (48), and more are probably still unmapped, both the thrust faults and the San Andreas building code in California (the Uniform but only a few of these (and we do not know fault, and the two risks are comparable. The Building Code). The code is based on an which) will produce events in our lifetimes. earthquake history of the San Andreas fault earthquake thought to be the largest one Mitigation strategies for this heavily popu- suggests that the part of the San Andreas with a reasonable chance of occurring. The lated metropolitan area should focus on rec- fault nearest to Los Angeles, the Mojave ground motion in the Northridge earthquake ognizing that large earthquakes in the urban segment, produces great earthquakes on av- exceeded the design code, especially at high- areas are not rare events, on predicting the erage every 131 years (43). The individual er frequencies. Such ground motions should effects of these earthquakes, and on design- concealed and surficial thrust faults in the be regarded as the norm near a large thrust ing buildings and response strategies that Big Bend Compressional Zone move more earthquake. Moreover, this event highlight- adequately account for these effects. slowly, the necessary geological data are dif- ed some of the poorly understood aspects of ficult to obtain in an urban setting, and the the earthquake proc-ess, such as how an REFERENCES AND NOTES faults' earthquake history is in most cases earthquake starts, how it stops, what controls 1. Magnitude calculated from seismib moment (Mw) [T. unknown. However, geologic and geodetic the dynamic and static stress drops of an Hanks and H. Kanamori, J. Geophys. Res. 84, 2348 estimates of slip on all the faults in Los earthquake, and the effect of these variables (1979)] is considered more representative of the size Angeles suggest that earthquakes as large as on ground motions. Furthermore, because of an earthquake and is 6.7 for Northridge. The local Richter magnitude, which saturates and thus under- Northridge must occur on average every 40 earthquakes larger-than Northridge will oc- estimates the magnitude above M 6, was ML 6.4, years somewhere in the Los Angeles region cur, other deficiencies in the code may exist, determined from TERRAscope data by H. Kanamori. (14). such as insufficient consideration of long- The National Earthquake Information Center assigned the earthquake a 20-s surface-wave magnitude (Mt) The rate of earthquakes recorded in Los period ground motion, near-fault ground of 6.8. Angeles since 1800 cannot account for this motions, and duration effects. A lack of 2. J. Hall, Ed., Northridge Earthquake January 17, accumulation of slip. Three possible expla- knowledge about some aspects of the be- 1994: Preliminary Reconnaissance Report (Earth- quake Engineering Research Institute, Oakland, CA, nations for this discrepancy are: (i) that the havior of structures also leads to inadequa- 1994). The dead included 20 who died from struc- rate of the past 200 years is anomalously low, cies in the code; the fracture of steel welds tural failures of their buildings, including 16 at the and in the future, moderate earthquakes will is one example. However, conclusions Northridge Meadows apartment complex and 13 be more (ii) who died from nonstructural causes such as fires, common, that Los Angeles is about the code made from damage at electrocution, and falls. Over 30 fatalities from heart accumulating slip to be released in an earth- Northridge must wait until the degree to attacks have also been attributed to the earthquake. quake of M. 7.5 or larger, and (ifi) that which the code was enforced in the dam- 3. Personal communication, California Governor's Of- significant deformation is occurring aseismi- aged buildings can be determined. fice of Emergency Services. 4. The rescaled NUVEL-1 rate for relative motion be- cally (44, 45). The rate of M 2 5 earth- Because the goal of a building code is to tween the Pacific and North American plates at this quakes in southern California has doubled protect life by preventing collapse, damage, latitude is 44.5 mm/year in the N38?W direction [poles since 1986, and the increase in Los Angeles sometimes unrepairable, is to be expected from C. Demets, R. G. Gordon, S. Stein, Geophys. J. Int. 101, 425 (1990)]. About 8 mm/year of this relative has been even greater (46). The origins of from strong shaking even in new structures. motion occurs along the Eastern California Shear this increase are unclear, so until we have However, the present rate of Califomian Zone extending into the Basin and Range [J. C. Sav- evidence that the rate has changed again, seismicity suggests that earthquakes similar age, M. Lisowski, W. H. Prescott, Geophys. Res. Lett. the rate of the decade- one : 17, 2113 (1990)1. The remaining plate motion can be past M 5 to Northridge, with ground motions at the partitioned into a San Andreas fault-parallel compo- earthquake (excluding aftershocks) per year extreme of the present code, will occur nent of 32.2 mm/year (with the use of the average in the Los Angeles area-should be consid- several times in a structure's lifetime. N66?W trend of the San Andreas fault directly adja- ered the best estimate of the seismic hazard cent to Northridge) and a fault-normal component of Therefore, the focus of current design prac- 17.1 mm/year in a N24?E direction. in the next few years. At this higher rate, the tice, which is solely on avoiding collapse, 5. The understanding of this system is rapidly evolving as slip accumulated since 1800 will be relieved should be reconsidered. Engineers need to new results are obtained, so no one paper provides a within 2 to 3 decades. devise methods for limiting damage that definitive overview. The surface geology is described in (48) and the seismotectonics in (22). Slip on the Northridge fault plane un- can be offered as design options. Encourag- 6. R. J. Crook, C. R. Allen, B. Kamb, C. M. Payne, R. J. doubtedly changed the static stresses on ing simple cost-effective mitigation mea- Proctor, U.S. Geol. Sunv. Prof. Pap. 1339,27 (1987).

396 SCIENCE * VOL. 266 * 21 OCTOBER 1994 &ARTICLE

7. U.S.G.S. staff, ibid. 733, 55 (1971). factors may be important. did not experience unusually great damage, the large 8. R. S. Yeats, J. Geophys. Res. 88, 569 (1983). 27. P. Reasenberg and L. Jones [Science 243, 1173 peak acceleration must have been caused by very 9. A. Donnellan, B. H. Hager, R. W. King, Nature 366, (1989)] showed that the rate at which aftershocks localized amplification. 333 (1993). occur after mainshocks can be described by A(t ,M) 35. T. C. Hanks, Bull. Seismol. Soc. Am. 65,193 (1975); 10. R. S. Yeats, J. Geophys. Res. 93,12137 (1988). = 1 o[a + b(Mm - M)] (t + c)-P, where A is the rate, t is L. A. Drake, ibid. 70, 571 (1980); H.-L. Liu and T. H. 11. S. Ward, Bull. Seismol. Soc. Am. 84, 1293, (1994); S. time, M is the magnitude of an aftershock, Mm is the Heaton, ibid. 74, 1951 (1984); J. E. Vidale and D. V. G. Wesnousky, J. Geophys. Res. 91, 12587 (1986). magnitude of the mainshock, and a, b, c, and p are Heimberger, ibid. 78, 122 (1988). 12. The Northern San Fernando Valley was in the top constants (a is the overall productivity of the se- 36. Many extensional fractures and grabens parallel con- one-sixth in terms of the potential for moment release quence, b is the frequency of magnitudes, p is the rate tours, exhibiting displacements consistent with move- from earthquakes of M 2 6.5 [D. Jackson, Seismol. at which the aftershocks decay with time, and 6 is the ment down a gradual slope. However, unequivocal Res. Lett. 65, A5 (abstr.) (1994)]. The region was as- time delay until the sequence starts to decay). Once evidence for liquefaction, such as the venting of sand signed a high value because of its high geodetic rates these four parameters are determined for a sequence, and water to the surface, was not observed in asso- of closure. the rate of aftershocks is fully described, and the ciation with these ground fractures except for a few 13. Seismic risk is defined as the hazard times the expo- probability of future aftershocks can be determined. isolated occurrences in Potrero Canyon, along the sure, where hazard is the potential for damaging The parameters for Northridge are almost exactly av- north flank of the Santa Susana Mountains. Studies earthquake shaking, and exposure is the number of erage for California aftershock sequences except for are in progress to assess whether much of the defor- potentially damagable structures in that region. the high overall productivity, a. For Northridge, a is mation could have resulted from liquefaction at depth 14. E. Hauksson, in Engineering Geology Practice in -1.3, b is 0.90, c is 0.09 days, and p is 1.2. The that did not expel sand. Southern California (Southern California Section, As- Californian average is a, -1.67; b, 0.91; c, 0.05 days; 37. Ground fractures were the cause of numerous sociation Engineering Geology, Belmont, CA, 1992), and p, 1.08. For the 1971 San Fernando earthquake, breaks in water and gas lines, including the rupture p. 167. a is -2.2, b is 1.08, and p is 1.2. For the 1933 Long and explosion of a 22-inch gas main that destroyed 15. Concealed faults (sometimes called blind faults) are Beach earthquake, a is -1.0, b is 1.0, and p is 1.3. five homes. Ground failure was also wholly or partial- faults that do not crop out at the earth's surface but 28. A. Shakal et al., California Strong Motion Instrumen- ly to blame for extensive foundation damage to hun- instead are capped by folds that accommodate the tation Program, Office of Strong Motion Studies Re- dreds of homes in the San Fernando Valley. fault's slip in nonbrittle deformation. The slip rates on port 94-07 (1994); R. Porcella, E. Etheredge, R. Ma- 38. Several trenches excavated across normal faults these faults can be inferred from the growth rates of ley, A. Acosta, U.S. Geological Survey Open-File Re- along the south margin of Potrero Canyon reveal the overlying folds [R. S. Stein and G. C. P. King, port-94-141 (1994); M. D. Trifunac, M. I. Todorovska, evidence for two earlier displacement events that Science 224, 869 (1984)] (18). S. S. Ivanovic, Int. J. Soil Dyn. Earthquake Eng., in were equal in size to the -28-cm displacement pro- by the Northridge earthquake. Whether 16. This moment magnitude is based on T. Heaton's press. duced there moment of 1.3 x 1019 N-m [Bull. Seismol. Soc. Am. 29. T. C. Hanks and R. A. McGuire, Bull. Seismol. Soc. tectonic or induced by shaking, these observations this least two earth- 72, 2037 (1982)]. The local magnitude of the 1971 Am. 71, 2071 (1981); D. M. Boore, ibid. 73, 1865 indicate that area experienced at in the Holocene with ground motions event was 6.4, the same as for Northridge. (1983); ibil. 76, 43 (1986). These studies model the quakes late similar to that of the Northridge earthquake. Prelim- 17. T. C. Hanks, J. Geophys. Res. 79, 1215 (1974). earthquake source in terms of moment and a term inary radiocarbon dates suggest that the penultimate 18. J. H. Whitcomb, U.S. Geol. Surv. Prof. Pap. 733, 30 called the stress parameter, which controls the dy- event occurred within the past 1200 years. (1973). namic ground motion amplitudes. A model of 39. The duration of rupture in the 1971 earthquake was 19. The 1991 Sierra Madre earthquake was M 5.8 [E. Northridge data with a frequency-dependent attenu- 10 to 12 s [T. Heaton, ibid. 72, 2037 (1982)] as Hauksson, Bull. Seismol. Soc. Am. 84, 1058 (1994)]. ation Q (Q = 61 x w(Hz), from an earlier analysis of the compared with 6 to 8 s for the Northridge earth- 20. "Elysian Park fault system" and "Santa Monica Anti- San Fernando earthquake [A. Papageorgiou and K. quake. Each shaking cycle increases the compac- clinorium" have both been used to describe some or Aki, ibid. 73, 953 (1983)]} compared the observed tion of soils, so that duration is an important factor all of the concealed faults along the north-to-north- peak horizontal accelerations at rock sites with those controlling liquefaction [T. L. Youd, J. Soil Mech. eastern flank of the Los Angeles basin. As defined in predicted for a single a-squared model with fmax = Found. Eng. 98, 709 (1972)]. (22), the Elysian Park system has included the 1987 10 Hz with the use of the site amplification factor 40. J. C. Tinsley, T. L. Youd, D. M. Perkins, A. T. F. Chen, Whittier Narrows (M 5.9) earthquake [E. Hauksson estimated by the coda method [F. Su et al., ibid. 82, U.S. Geol. Sun'. Prof. Pap. 1360, 263 (1985). L. 580 (1992); B. H. Chin and K. Aki, ibid. 81, 1859 and M. Jones, J. Geophys. Pes. 94, 9569 (1989)] 41. Earthquake damage is cumulative, with each cycle (1991)]. The 'stress parameter" of the best-fitting (45) and the 1979 (M 5.2) and 1989 (M 5.0) Malibu of loading increasing the likelihood of damage and model was about 150 bars, which is more than twice earthquakes (E. Hauksson and G. M. Saldivar, ibid., p. collapse. Failure to repair damage from this earth- the representative value for the western United States 9591). quake will increase the chance of life-threatening [G. Atkinson arad D. Boore, Earthquake Spectra 6, 15 21. The earthquake on the system was damage in the next event. (1 the 1988 (M 5.0) Pasadena earthquake [L. M. Jones, 990)]. 42. R. S. Yeats, in preparation; T. L. Davis and J. Nam- 30. D. MiV.Boore, W. B. Joyner, T. E. Fumal, U.S. Geo- K. E. Sieh, E. Hauksson, L. K. Hutton, Bull. Seismol. son, in preparation; T. L. Davis, J. Namson, R. F. rogical Survey Open-File Report 93-509 (1993); W. Soc. Am. 80, 474 (1990)]. Yerkes, J. Geophys. Res. 94, 9644 (1989). B. Joyner and D. M. Boore, Bull. SeismoL Soc. Am. 22. E. Hauksson, J. Geophys. Res. 95,15365 (1990). 43. K. E. Sieh, M. Stuiver, D. Brillinger, J. Geophys. Res. 71, 2011 (1981); D. R. Brillinger and H. K. Preisler, 23. These results are from a study of small earthquakes 94, 603 (1989). The last earthquake on this segment ibid. 74, 1441 (1984); ibid. 75, 611 (1985). in the region from 1980 to 1992 [L. Seeber, Seismol. was in 1857. Res. Lett. 65, Al0 (abstr.) (1994)]. 31. The velocity of rupture propagation is close to the 44. R. S. Stein and R. S. Yeats, Sci. Am. 260, 48 (June 24. Slip on a concealed fault plane must be inferred from velocity of shear-wave propagation. As a conse- 1989). indirect measurements. Data from many sources, quence, if rupture propagates toward a station, radi- 45. J. Lin and R. S. Stein, J. Geophys. Pes. 94, 9614 ation from a relatively long portion of the ar- including seismological body and surface waves, lo- rupture (1989). cal strong ground motions, displacements mea- rives at the station in a relatively short time as com- 46. Since 1900, moderate (M > 4.8) earthquakes in the sured by the GPS, and levelirrg lines can be inverted pared with stations at other azimuths, hence the Los Angeles region have occurred in two clusters, to estimate the distribution of slip. The details of the ground motions are larger. A station located in the with 5 events between 1920 and 1942 and 10 events distribution vary depending on the data sets and opposite direction will record a more elongated wave so far since 1970. No moderate events were recorded model assumptions. The results in the text are the train with less severe ground motion. between 1942 and 1970 (14). Previous changes in robust features common to all of the models so far. 32. The shear motion across a fault leads to a nonho- the rate of moderate earthquakes in California have in 25. The aftershock,data were collected and analyzed by mogerious distribution of energy radiation from an some cases been followed by major earthquakes the Southern California Seismographic Network op- earthquake. Santa Monica was in the direction of [such as the 1906 (M 8) San Francisco, 1983 (M 6.5) erated by the California Institute of Technology and maximum S wave radiation, whereas Pasadena was Coalinga, and 1989 (M 7.1) Loma Prieta events]. In the U.S. Geological Survey. at a minimum in both P and S waves. other cases, such as the Long Valley area from 1978 26. Unlike the situation after the larger 1992 Landers 33. Velocities are determined from integration of acceler- to 1986 and the Banning region of southern California earthquake [D. P. Hill et al., Science 260, 1617 ation records. A glitch in the film record at the Rinaldi from 1935 to 1948, no event within the cluster was (1993)], no widespread increase in seismicity was ob- station, interpreted as a stall in the advancing mech- significantly larger than the others. In the earlier Los served at regional distances after the Northridge anism, makes the temporal integration uncertain. The Angeles cluster, the largest event (M 6.4, at Long mainshock. However, within the coda of the value of 170 cm/s assumes that the film stalled for Beach) was in the middle of the group. The mecha- Northridge mainshock, a few small (M < 1) events 0.05 s. Assuming no stall gives a minimum estimate of nisms controlling rate changes or their relation to the were observed at Medicine Lake, east of the Long 130 cm/s. Any longer delay would imply an even high- largest earthquakes of a region are not understood. Valley Caldera, and in The Geysers geothermal area. er velocity. 47. M. D. Zoback et al., Science 238, 1105 (1987). The Geysers geothermal field has shown increased 34. The most extreme local geologic site effect in the 48. J. I. Ziony and R. F. Yerkes, U.S. Geol. Surv. Prof. seismicity after 6 regional events but not after 22 oth- Northridge earthquake was at a site in Tarzana with a Pap. 1360, 43 (1985). ers, depending on the distance to the mainshock and peak horizontal acceleration of 1.82g. In the 1987 49. C. W. Jennings, Fault Map of California with Volca- the mainshock magnitude Dr[S. Davis, Eos 74, 317 Whittier Narrows earthquake, the same site had a noes, Thermal Springs and Thermal Wells (Scale (abstr.) (fall 1993)]. The Geysers region has a high, peak horizontalacceleration of 0.62g, which is more 1:750.000), California Division of Mines Geology, fairlyconstant degree of seismicity, which suggests than 10 times greater than the geometric mean of Geologic Data Map No. 1 (1975). that it is in a metastable condition and thus is prone to seven sites at similar distances. In other events, in- 50. We thank W. Ellsworthand R. Page for insighifuland triggeringby small perturbationssuch as the dynamic cluding a WhittierNarrows aftershock, the peak hori- constructive reviews. Sponsored by NSF through shaking of regional events. Northridge should have zontal acceleration at this site was not especially large. the Southern CaliforniaEarthquake Center, the U.S. been too small to have triggered a swarm, so other Because the nearby nursery and homes in the area Geological Survey, and NASA.

SCIENCE * VOL. 266 * 21 OCTOBER 1994 397