Empirical Analysis of Strong Ground Motion from the 1992 Landers, California, Earthquake

Transcription

1 Bulletin f the Seismlgical Sciety f America, Vl. 84, N. 3, pp , June 1994 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake by K. W. Campbell and Y. Bzrgnia Abstract A ttal f 167 grund-level accelergrams recrded during the 28 June 1992 Ms 7.6 (Mw 7.3) Landers, Califrnia, earthquake were used t study the dependence f peak hrizntal acceleratin and 5%-damped hrizntal respnse spectra n distance, azimuth, and site gelgy. A cmparisn f these recrdings with grund-shaking amplitudes predicted by cntemprary attenuatin relatinships indicated that (1) relatinships develped prir t the Landers earthquake did a reasnably gd jb at predicting the Landers grund mtins within abut 60 km f the fault, but underpredicted the grund mtins at further distances by as much as a factr f 2 t 3; and (2) relatinships develped after the Landers earthquake did a reasnably gd jb at predicting the Landers grund mtins within the distance ranges fr which they were applicable. An extraplatin f near-surce attenuatin rates based n the newer relatinships was fund t significantly underpredict the Landers grund mtins at distances beynd abut 60 km, indicating a need t develp attenuatin relatinships that can accunt fr differing rates f attenuatin at near-surce and far-surce distances. Peak acceleratins and respnse spectra were fund t be significantly affected by surce radiatin pattern, surce directivity, and site gelgy, with higher grund mtins bserved in the directin f rupture prpagatin and lwer grund mtins bserved n rck beynd distances f abut 40 km. Regressin analyses f the Landers recrdings indicated an unusually lw rate f attenuatin, cntrlled primarily by recrdings btained at distances greater than abut 60 km, which can be attributed t ne r mre f the fllwing factrs: (1) a change in gemetric attenuatin rate as a result f the dminance f surface waves; (2) amplificatin in the Ls Angeles Basin, the San Femand Valley, the San Bernardin Valley, and the Cachella Valley because f the effects f deep sediments and basin respnse; and (3) the arrival f critical reflectins ff the base f the crust (MOHO). The fault-nrmal cmpnents f respnse spectra at sites lcated clsest t the fault were fund t be systematically larger than the fault-parallel cmpnents at perids f 1 sec and greater. This effect was largest at statins lcated nrth f the epicenter in the directin f rupture prpagatin and indicates a need t accunt fr this effect in the design f lng-perid structures. Intrductin The Landers earthquake (Ms 7.6, Mw 7.3) is the largest earthquake t strike Califrnia since the 1952 (Ms 7.8, Mw 7.5) Kern Cunty earthquake. It is the first earthquake f magnitude 7 r greater t ccur in Suthern Califrnia since a large array f strngmtin accelergraphs was installed beginning in the 1970s. As a result, this earthquake prvides the first cmprehensive set f strng-mtin recrdings frm an earthquake appraching the magnitude f that ex- pected t ccur n the San Andreas fault near Ls Angeles. The imprtance f the Landers recrdings is heightened by the cnclusin f the Ad Hc Wrking Grup n the Prbabilities f Future Large Earthquakes in Suthern Califrnia (1992) that the chances f a large earthquake (M > 7) ccurring within 100 km f the Landers rupture is 2 t 5% in the first year, and pssibly as high as 50% in the next 5 yr, as a result f the increased 573

2 574 K.W. Campbell and Y. Bzrgnia seismic activity in the regin since 1985 and the ccurrence f the Landers earthquake sequence. Accrding t this Ad Hc Grup, the mst likely lcatin fr such an event is either the Mjave shear zne (the suthern part f which ruptured during the Landers earthquake); the San Bernardin Muntains and Cachella Valley segments f)he San Andreas fault, a cmbinatin f the San Bernardin Muntains segment with either the Cachella Valley segment, the Mjave segment, r bth, r the nrthern San Jacint fault. The latter tw scenaris wuld have a far greater impact n the San Bernardin and Ls Angeles areas than the Landers earthquake. In this article, we present a preliminary empirical analysis f strng-mtin data frm the Landers earthquake in rder t study the attenuatin characteristics, gelgical effects, and azimuthal dependence f the peak hrizntal acceleratins and hrizntal respnse spectra recrded during this earthquake. Of particular interest was the ptential impact this earthquake might have n the empirical predictin f grund mtins cmmnly used fr design. Strng-Mtin Data The data base used fr this study cnsisted f 167 grund-level accelergrams recrded in the free field r in buildings less than three stries high by strng-mtin netwrks perated by the Califrnia Divisin f Mines and Gelgy (CDMG), the U.S. Gelgical Survey (USGS), and the Suthern Califrnia Edisn Cmpany (SCE) (Table 1). Only a small number f recrdings had been digitized and prcessed as f the writing f this article. As a result, the data base fr respnse spectral rdinates was cmpsed f nly 24 recrdings: 22 frm CDMG and 2 frm SCE (Table 2). Peak acceleratins were taken directly frm Shakal et al. (1992), Etheredge et al. (1993), and Hawkins et al. (1994). We btained respnse spectra fr the CDMG recrdings frm R. B. Darragh f the Califrnia Strng-Mtin Instrumentatin Prgram (CSMIP). We calculated respnse spectra fr Table 1 Distributin f Recrdings by Organizatin and Site Cnditins Number f Recrdings Organizatin Alluvium Shallw Sil Sft Rck Hard Rck Ttal Peak Hrizntal Acceleratin USGS CDMG SCE Respnse Spectra USGS CDMG SCE the SCE recrdings frm baseline-crrected accelergrams riginally btained frm H. G. Hawkins and D. K. Ostrm (written cmm., 1993). Figure 1 shws the gegraphic distributin f the recrdings used in this study. The distributin f recrdings by rganizatin and gelgical site cnditins is given in Table 1. By far the majrity f the recrdings were btained n Alluvium, defined as Quaternary depsits with at least 10 m f sil verlying rck. The next-largest numbers f recrdings were btained n Hard Rck (primarily crystalline rck and Cretaceus and lder sedimentary depsits) and Sft Rck (primarily Tertiary-age sedimentary depsits). The smallest number f recrdings was btained n Shallw Sil (Quaternary depsits <10 m deep). There were n recrdings frm sft-sil sites, similar t the Bay Mud depsits lcated alng the margins f San Francisc Bay that were respnsible fr the highest grund mtins recrded during the 1989 Lma Prieta earthquake. Gelgical site cnditins were derived primarily frm a data base f strng-mtin recrdings cmpiled by the first authr. These site cnditins were fund t be cnsistent with thse reprted by Shakal et al. (1992) fr the CDMG recrding statins. Etheredge et al. (1993) did nt reprt site cnditins fr the USGS recrding statins, s it was nt pssible t perfrm an independent check f the site cnditins derived fr these sites. Site cnditins fr the SCE recrding statins were taken frm Hawkins et al. (1994). In rder t study the attenuatin characteristics f the Landers grund mtins, a definitin f surce-tsite distance was needed. Fr purpses f this study, we selected the clsest distance t seismgenic rupture as riginally defined by Campbell (1987) as this distance measure. The seismgenic rupture zne fr the Landers earthquake was assumed t cincide in plan view with the nrthern three-furths f the bserved surface rupture, extending alng the Jhnsn Valley, Hmestead Valley, and Camp Rck-Emersn faults sme 60 km frm a pint 4 km nrth f the epicenter t the terminus f surface rupture. This cincides with the zne f significant fault slip at depth as determined frm an analysis f brad-band TERRAscpe recrdings by Kanamri et al. (1992). Kanamri et al. (1992) identified tw majr subevents that ccurred abut 10 sec apart. The first subevent was n the Jhnsn Valley and Hmestead Valley faults and had a maximum slip at depth f abut 4.5 m. The secnd and larger subevent was n the Camp Rck- Emersn fault and had a maximum slip at depth f abut 7.5 m. The prtin f the fault suth f the epicenter, which was characterized by bth intermittent surface rupture and an alignment f aftershcks, generated very little seismic energy accrding t Kanamri et al. and was nt cnsidered t be seismgenic in this study. The depth t the tp f the seismgenic rupture zne was assumed t be 3 kin, cnsistent with the average depth

3 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 575" Table 2 Respnse Spectra Statin Site* Recrding Statin Organizatin Number Class BARSTOW--Vineyard & H Street CDMG A DESERT HOT SPRINGS--Fire Statin CDMG A INGLEWOOD--Unin Oil Yard CDMG A JOSHUA TREE--Fire Statin CDMG A PALM SPRINGS--Airprt CDMG A YERMO--Fire Statin CDMG A SILENT VALLEY--Pppet Flat CDMG HR TWENTYNINE PALMS--Park Maintenance Building CDMG HR AMBOY CDMG SS BAKER--Fire Statin CDMG A HEMET--Stetsn Avenue Fire Statin CDMG A INDIO--Cachella Canal CDMG A FEATHERLY PARK--Park Maintenance Building CDMG A POMONA--4th & Lcust CDMG A PUERTA LA CRUZ--USFS Strage Building CDMG SS RIVERSIDE--Airprt CDMG A TARZANA--Cedar Hill Nursery CDMG SS BORON--Fire Statin CDMG A DOWNEY--Cunty Maintenance Building CDMG A FORT IRWIN CDMG A LOS ANGELES--116th Street Schl CDMG ~ A LOS ANGELES--Obregn Park CDMG A LUCERNE VALLEY SCE -- HR COOLWATER SCE -- A *A = Alluvium; HR = Hard Rck; SS = Shallw Sil. 1 I17 FORT IRWIN 115~50 ' RECORDING STATIONS ALLUVIUM ROCK , BARSTOW YERMO COOtWATER I Krn I 3500 ' l LUCERNE VALLEY-'~~I~'~ ~ J. JOSHUA TREE TWENTYNINE PALMS DESERT HOT SPRINGS PALM SPRING~ 117 leilp ~ INDIO ' vc 44~X Figure l. Map shwing the gegraphic distributin f recrdings used in the analysis. JVF = Jhnsn Valley Fault; HVF = Hmestead Valley fault; CRF -- Camp Rck-Emersn fault.

4 576 K.W. Campbell and Y. Bzrgnia f seismgenic faulting bserved fr similar faults by Marne and Schlz (1988). Attenuatin Characteristics In rder t study the general attenuatin characteristics f the Landers strng-mtin data, we develped fur representative attenuatin relatinships frm recrdings btained n Alluvium: ne relatinship fr peak grund acceleratin (PGA) and three relatinships fr 5%- damped pseud-abslute acceleratin (PSA) at natural perids f 0.3, 1, and 3 sec. Althugh many different functinal frms fr these relatinships were cnsidered (e.g., Campbell, 1985; Jyner and Bre, 1988), the nly frm fund t be statistically significant at the 90% cnfidence level was that given by the equatin In Y =/31 +/32 In Rs + e, (1) where Y is the gemetric mean f the tw hrizntal cmpnents f PGA r PSA in units f gravity (g), Rs is the clsest distance t seismgenic rupture n the fault in kilmeters, and E is a randm errr term with a mean f zer and a standard deviatin f O-l,r, the standard errr f regressin. The regressin cefficients /31, /32, and O'ln Y are summarized in Table 3. The crrespnding attenuatin relatinships are pltted in Figure 2. The Landers attenuatin relatinships are interesting frm tw standpints. First, the data d nt statistically require an anelastic attenuatin term [the cefficient f Rs, nt shwn in equatin (2)] as expected n the basis f simple wave-prpagatin thery (e.g., Bre, 1983). Secnd, even thugh an anelastic attenuatin term was nt included in the relatin, the gemetric attenuatin term [the cefficient f In Rs in equatin (2)] was fund t be less than the value 1 expected fr bdy waves (e.g., Bre, 1983). Similar results were btained by Bre et al. (1993) in their analysis f recrdings frm the westem United States, which included distant recrdings frm bth the 1989 Lma Prieta and 1992 Landers earthquakes. Pssible reasns fr this disparity are discussed later in the article. Cmparisn with Cntemprary Attenuatin Relatinships The Landers attenuatin relatinships were cmpared t tw sets f cntemprary attenuatin relatinships: ne set that excluded recrdings frm the 1989 Lma Prieta and 1992 Landers earthquakes (Sadigh et al., 1986, 1993; Jyner and Bre, 1988; Campbell, 1993), and anther set that included recrdings frm these as well as ther recent earthquakes (Bre et al., 1993; Campbell and Bzrgnia, 1994). These attenuatin relatinships are representative f thse cmmnly used in the seismic evaluatin and design f structures in the western United States. Campbell and Bzrgnia (1994) did nt develp attenuatin relatinships fr respnse spectra, s their relatinship culd nly be cmpared with values f PGA. As a matter f cnvenience, the attenuatin relatinships used fr the cmparisns are identified thrughut the remainder f the article by the initials f the last names f the authrs fllwed by the last tw digits f the year in which they were published. This crrespnds t the fllwing naming cnventin: SEY86: JB88: BJF93: C93: SCACP93: CB94: Sadigh et al. (1986) Jyner and Bre (1988) Bre et al. (1993) Campbell (1993) Sadigh et al. (1993) Campbell and Bzrgnia (1994) All f the selected attenuatin relatinships used widely differing measures f surce-t-site distance. C93 and CB94 defined their distance measure as the clsest distance t seismgenic rupture, identical t that used in this study. JB88 and BJF93 defined their distance measure as the clsest distance t the surface prjectin f rupture n the fault. SEY86 and SCACP93 defined their distance measure as the clsest distance t the rupture surface. Since rupture during the Landers earthquake ccurred n a nearly vertical fault plane, the SEY86, JB88, BJF93, and SCACP93 relatinships were easily adjusted fr use with the distance measure adpted in this study by means f the simple transfrmatin R' = (R~ - d2s) 1/2, (2) Table 3 Regressin Results: in Y = /3~ + /32 In Rs Parameter, Y Perid (sec) /3t /32 q n y r 2 Value* (%) PGA PSA PSA PSA *The percentage f the variance explained by the mdel.

5 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 577 where Rs is the clsest distance t seismgenic rupture n the fault, ds = 3 km is the assumed depth t the tp f the seismgenic rupture zne, and R' is used in place f the distance measures R (SEY86 and SCACP93), r0 (JB88), and d (BJF93). SEY86, SCACP93, and CB94 included style f faulting as a parameter in their relatinships. This parameter is 0 fr strike-slip faults and >1 fr reverse, reverse-blique, and thrust faults. Cnsistent with the strike-slip mechanism f the Landers earthquake, this parameter was set equal t zer. C93 incrprated depth t basement rck in his relatinship t characterize amplificatin assciated with the lng-perid respnse f sedimentary depsits. Althugh this depth is site depen- dent, it was assigned a cnstant value f 4 km fr purpses f cmparisn, t be generally cnsistent with the depth f the sediments in the Ls Angeles Basin. All f the attenuatin relatinships used different methds fr characterizing lcal site cnditins. We picked the site classificatin that came the clsest t thse defined in this study fr use in the cmparisns. We cnsidered Site Classes A, B, and C f BJF93 t crrespnd rughly t ur Hard Rck, Sft Rck, and Alluvium categries, respectively. We cnsidered the Sil classificatin f SEY86 and JB88 t be cnsistent with ur Alluvium categry. Finally, we cnsidered the Rck classificatins f JB88, C93, and SCACP93 t be cnsistent with ur Sft Rck and Hard Rck categries. (a) PGA (SOIL) (b) PGA ('ROCK),f ~ ~\',.q~ -, ALLUVIUM \~ \. (c) rsa (0.3 s).... IDQ al SOFF ROCK \~? I I I I,1111 I I * 11111' ~ ~'t t I I J I II (d) PSA (1.0 s) ~ I,, A _ N D E ~ -- C93 \\\ "~\ (e) PSA (3.0 s) C93 \\ ', ALLUVIUM \ \ ", i i i itltft i ~ i i111,, x,\,,,,,, ,000 Distance t Seismgenic Rupture (km) 0.1 : lo 100 1,000 Distance t Seismgenic Rupture (kin) Figure 2. Cmparisn f the Landers strng-mtin recrdings and attenuatin relatinships with attenuatin relatinships develped by Sadigh et al. (1986) (SEY86), Jyner and Bre (1988) (JB88), and Campbell (1993) [C93]. All f these relatinships exclude recrdings frm the Lma Prieta and Landers earthquakes. PGA = peak hrizntal acceleratin; PSA = 5%-damped pseud-abslute acceleratin.

6 K.W. Campbell and Y. Bzrgnia Cmparisns with the lder attenuatin relatinships are shwn in Figure 2. Since there were nly three rck sites with values f PSA, n plts are shwn fr these recrdings. This cmparisn indicates that all f the attenuatin relatinships d a reasnably gd jb at predicting grund mtins frm the Landers earthquake fr distances less than abut 60 km. The predictins fr Alluvium are all within abut ne standard errr f the Landers estimates (Table 3). Hwever, all f these attenuatin relatinships underpredict the Landers grund mtins at lnger distances by as much as a factr f 2 t 3, much larger than the standard deviatins f either predictin. Similar behavir was bserved fr the Lma Prieta recrdings (e.g., Campbell, 1991). It shuld be nted that the authrs f the JB88 relatinship have since recmmended that their relatinships nt be used t predict grund mtins beynd 100 km (W. B. Jyner, persnal cmm., 1993). Cmparisns with the newer attenuatin relatinships are shwn in Figure 3. Since there were nly three rck sites with values f PSA, n plts are shwn fr these recrdings. This cmparisn indicates that all f the attenuatin relatinships d a reasnably gd jb at predicting grund mtins frm the Landers earthquake ver the range f distances fr which the authrs cn- sider them valid (i.e., 100 km fr BJF93 and 60 km fr CB94). The predictins fr Alluvium are all within abut ne standard errr f the Landers estimates (Table 3). Nte that an extraplatin f the CB94 relatinship beynd 60 km wuld significantly underpredict the Landers acceleratins. This disparity between the attenuatin characteristics f near-surce and far-surce recrdings was als bserved fr the 1987 Whittier Narrws earthquake (Campbell, 1988) and the 1989 Lma Prieta earthquake (Campbell, 1991) and is discussed later in the article. Nte als that an extraplatin f the BJF93 relatinship beynd 100 km wuld verpredict the Landers PGA recrdings n Sft Rck by abut 50%, but wuld accurately predict the recrdings n Hard Rck. Analysis f Residuals An analysis f residuals was used t better understand the bserved behavir f the Landers recrdings and t further explre the dependence f these recrdings n surce-t-site distance, gelgical site cnditins, and azimuth. Fr this purpse, azimuth was defined as the angle between a vectr drawn frm the epicenter t the site and a similar vectr drawn t true nrth, measured clckwise frm nrth. Residuals were (a) PGA (SOIL) (b) PGA (ROCK)....'~.~ g~d LAND... O l~rf93 ALLUVIUM, NO O n i ~ I,IIH i w, JJIIH J (c) PSA (0,3 s) - CB91(SR),,~..tr. c~ )... m~3 t~) ^ ~... ~-~3 (A) ~ ~ t Slzr ROCK ItARD ROUK \ \ \ i f liill i (d) PSA (10 s) ~.~.~ ~, tip f~ gh 0.1 O.Ol LANDERS.... BdlF93 t~ ALLUVIUM l ,t~ 1 Distance t Seismgenic Rupture (km)... a,lrt~$ ALLIIVH~ i i i iiiiii t i i... t,,, t,, ,000 Distance t Seismgenic Rupture (km) Figure 3. Cmparisn f the Landers strng-mtin recrdings and attenuatin relatinships with attenuatin relatinships develped by Bre et al. (1993) (BJF93) and Campbell and Bzrgnia (1994) [CB94]. Bth f these relatinships include recrdings frm the Lma Prieta and Landers earthquakes. PGA peak hrizntal acceleratin; PSA = 5%-damped pseud-abslute acceleratin; SR = Sft Rck; HR = Hard Rck; A = Site Class A; B = Site Class B.

7 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 579 cmputed with respect t equatin (1) and nrmalized with respect t the standard errr f regressin by the equatin ei = (ln Y i - ln Y)/qny, (3) where the index i represents the ith recrding, ei is the nrmalized residual, In Yi is the bserved value, In Y is the predicted value, and -],r is the standard errr f regressin. Nte that psitive residuals crrespnd t bserved grund-shaking amplitudes that are larger than thse predicted (relatively high grund mtins), and negative values crrespnd t thse that are smaller than thse predicted (relatively lw grund mtins). As we will shw in the fllwing sectins, the residuals were fund t be strngly biased with respect t bth site cnditins and azimuth. Furthermre, as demnstrated in Figure 4, the distributin f recrdings with respect t distance was fund t be highly crrelated with azimuth. In rder t separate these varius effects, the recrdings were segregated by distance, site cnditins, and azimuth befre being cmpared. The cmparisns were dne visually and als were dne statistically by means f a frmal hypthesis test (t test). Only thse grups having at least five recrdings were used in the tests. The hypthesis that the means r the variances f any tw sets f residuals are the same was rejected if the level f significance (the p value) was fund t be 0.1 r less (i.e., the cnfidence level was 90% r greater). These tests were cnducted nly fr PGA, since there were t few respnse spectra with which t perfrm reliable statistical tests. The results f the hypthesis tests are described belw and summarized in Tables 4 thrugh 6. In these tables, RESID represents the mean value, and e represents the standard deviatin f the nrmalized residuals, ei. Distance Bias Figure 5 shws plts f the nrmalized residuals as a functin f distance fr the fur grund-mtin parameters f interest in this article. Because f the large number f recrdings, plts fr PGA have been separated by gelgical site cnditins. All f the plts fr Alluvium shw relatively cnstant residuals ver all distances, demnstrating an verall lack f bias in the regressin results. The rck sites, hwever, demnstrate a bias twed smaller grund mtins (i.e., smaller residuals) at all distances. T test whether there was a distance bias in the residuals, we cmpared the PGA residuals fr Alluvium and Hard Rck at azimuths ranging frm 240 t 300 fr tw distance ranges, Rs <= 115 km andrs > 115 km, ~ lz E 4 m 20 ~ ~ (a) PGA (SOIL) ALLUVIUM SHALLOW SOIL _~'~ O E 00~_ O ~ = c~ mml... lih... I,H... I... I... i... I... (c) PSA i ALLUVIUM SHAI.LOW SOIL B ~ _ HARD ROCK ~- (b) PGA (ROCK) SOFT ROCK e % -- HARD ROCK r~ L L 0000 "'C~ - O ~ 2 I... I... I... h,,,,,,,i.,,,.,,,[...,,,h,, Azimuth (deg). IW O [4 "G 20 ~ 0 - Q... I... I... I... I... I... I... h,,m,,, % S 360 Azimuth (dt~) Figure 4. Distributin f Landers strng-mtin recrdings with respect t clsest distance t seismgenic rupture and surce-t-statin azimuth. Azimuth is measured clckwise with respect t the epicenter f the earthquake. PGA = peak hrizntal acceleratin; PSA = 5%-damped pseud-abslute acceleratin.

8 580 K.W. Campbell and Y. Bzrgnia fr which sufficient data were available. The results f the hypthesis test are given in Table 4. This test indicates that there is n significant bias with respect t distance fr either the Alluvium r Hard-Rck recrdings. Based n these results, we did nt segregate the recrdings by distance fr the remainder f the tests. Azimuthal Bias Plts f the nrmalized residuals as a functin f azimuth are shwn in Figure 6. As befre, the plts fr PGA are separated by gelgical site cnditins. The plts fr PGA and fr the 0.3-sec rdinate f PSA shw a tendency fr relatively high grund mtins at sites lcated nrthwest (azimuths f 300 t 015 ) and sutheast (azimuths f 115 t 175 ) f the epicenter, and relatively small grund mtins at sites lcated suthwest f the epicenter (azimuths f 175 t 240 ). This difference appears t vanish at perids f 1 and 3 sec as the lngperid respnse f the deep sedimentary basins lcated tward the suth and west begins t becme as imprtant as the lng-perid effects f surce radiatin and directivity that impact sites lcated n relatively shallw sediments tward the nrth. T test whether the bserved bias with azimuth was significant, we cmpared the residuals fr Alluvium and Hard Rck at fur azimuths, 115 t 175 (sutheast), 175 t 240 (suth t suthwest), 240 t 300 (west), and 300 t 015 (nrthwest t nrth), fr which sufficient data were available. The results f the hypthesis tests are summarized in Table 5. These tests indicate that the bserved differences in the mean residuals are significant fr cmparisns amng all azimuths except the nrthwest and sutheast, which exhibited similarly high grund mtins. Fr nne f the grups were differences in the variances f the residuals fund t be significant. Gelgic Bias Figures 5 and 6 and Tables 4 and 5 indicate that rck sites appear t have had relatively smaller grund mtins than Alluvium. T test whether this bias was significant, we cmpared the residuals fr Alluvium, Sft Rck, and Hard Rck at tw azimuths, 175 t 240 and 240 t 300, fr which sufficient data were available. The results f the hypthesis tests are given in Table 6. In all cases, the mean residual f the Alluvium recrdings was fund t be significantly higher than the mean Table 4 Hypthesis Test fr Distance Bias; Azimuth: 240 t 300 (West) Dis~nce (km) Difference in Means Difference in Variances Number f Recrdings RESID e R~st p value tr~ p value Alluvium ~115 >115 Hard Rck _-<115 > Table 5 Hypthesis Test fr Azimuthal Bias Azimuth (deg) Difference in Means Difference in Variances Number f Recrdings RESID e RES1D p value ~r e p value Alluvium Hard Rck

9 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 581 Table 6 Hypthesis Test fr Gelgic Bias Difference in Means Difference in Variances Site Number f Cnditins Recrdings RESID e RF'sl p value 'e p value Azimuth: 175 t 240 (Suthwest) Alluvium Hard Rck Azimuth: 240 t 300 (West) Alluvium Sft Rck Hard Rck Sft Rck Hard Rck Table 7 Summary f Site Gelgical Effects (Peak Grund Acceleratin) Relative Site Factr Site Gelgy Landers Lma Pricta BJF93 CB94" Alluvium Sft Rck ,77 Hard Rck ,60 *Evaluated at a distance f 60 kin. differences bserved at 3 sec fr the Pppet Flat and Twentynine Palms recrdings (apprximately a factr f 2 instead f the factr f 3 t 5 bserved at the shrter perids) appear t be due t a lw signal-t-nise rati at these statins. The relatively large 3-sec value f PSA at Lucerne Valley, as cmpared t the Alluvial sites lcated at similar azimuths, is easily explained by nearsurce directivity and wave-passage effects, which are knwn t have a prfund influence n lng-perid grund mtins (e.g., Smerville and Graves, 1993). Table 8 Summary f Azimuthal Effects (Peak Grund Acceleratin) Azimuth Directin frm Relative Azimuthal (deg) Epicenter Factr Landers Suthwest West Nrthwest Sutheast 1.59 Lma Prieta Nrtheast Nrthwest 1.68 residuals f the rck recrdings. In cntrast, the mean residuals f the Sft-Rck and Hard-Rck recrdings were nt fund t be significantly different. The nly significant difference in the variances f the residuals was fund between the Alluvium and Hard-Rck recrdings at azimuths f 240 t 300, principally because f the unusually large variance assciated with the Hard-Rck recrdings at distances beynd abut 100 km. A/thugh based n nly three recrdings, even larger differences were bserved between Alluvium and Hard- Rck spectra. Spectral amplitudes at similar azimuths (Fig. 6) were fund t be cnsistently 3 t 5 times lwer at Hard-Rck sites fr perids f 0.3 and 1 sec. The smaller Lng-Perid Effects Recent studies f selected earthquakes have shwn that lng-perid amplitudes f near-surce hrizntal respnse spectra fr cmpnents riented nrmal t the causative fault are larger than thse riented parallel t the fault (e.g., Smerville and Graves, 1993; Sadigh et al., 1993). This effect has been fund t be especially significant fr strike-slip faults. This behavir is examined fr the Landers earthquake by carefully analyzing the spectral ratis f the fault-nrmal t fault-parallel cmpnents fr nine grund-respnse accelergrams lcated within abut 60 km f the fault rupture. These recrding statins are all apprximately aligned alng the strike f ne f the tw principal rupture directins and, therefre, were psitined alng a maximum in the SH radiatin pattern f at least ne f the tw majr subevents. We excluded recrdings at further distances after a preliminary analysis did nt shw significant and cnsistent differences between the fault-nrmal and faultparallel cmpnents. We excluded the Twentynine Palms Hard-Rck recrding because f its apparent lw signalt-nise rati at lng perids. The hrizntal cmpnents f each selected accelergram were rtated t fault-nrmal and fault-parallel rientatins by means f an rthgnal crdinate transfrmatin. Since faulting during the Landers earthquake was cmplex and multidirectinal, tw predminant fault

10 582 K.W. Campbell and Y. Bzrgnia strikes were used t perfrm this rtatin: (1) 349 (measured clckwise frm nrth), the apprximate strike f the Jhnsn Valley and Hmestead Valley segments f the fault rupture and the lcatin f the first and smaller subevent (e.g., Kanamri et al., 1992), and (2) 315, the apprximate strike f the Camp Rck-Emersn segment f the fault rupture and the lcatin f the secnd and larger subevent (e.g., Kanamri et al., 1992). Rtated accelergrams and crrespnding respnse spectra were cmputed fr bth fault strikes. Hwever, an inspectin f the spectra indicated that the largest difference between the fault-nrmal and fault-parallel cmpnents was assciated with that fault strike with which the statin was mst clsely aligned, regardless f which rupture segment was clsest t the recrding statin. Based n this bservatin, the cmpnents rtated with respect t the clsest-aligned fault strike were used t cmpute the spectral ratis discussed belw. The Lucerne Valley statin was lcated nly a few kilmeters frm surface rupture n the Camp Rck- Emersn fault. Hwever, it was subject t directivity assciated with rupture n the Jhnsn Valley and Hmestead Valley faults. Our first thught was that this accelergram shuld be rtated with respect t the clsest fault. Hwever, upn inspectin f the rtated spectra, it became very bvius that the largest difference between the fault-nrmal and fault-parallel cmpnents f this recrding was assciated with rupture n the Jhnsn Valley-Hmestead Valley faults. This demnstrates the imprtant, and in this case cntrlling, effect 5 (a) PGA (SOIL) ~- ALLUVIUM... 4 ~_. SHALLOW SOIL " ~ 2 ~-" i--'o... Ir.-...~..... ta ~ : t~..0...cK- -._-.01Katr... " '""~... a a'~''''~b... ~.-~...=... %a6 E ~... ~... t i~"... (b) PGA (ROCK) SOFT ROCK... HARD ROCK ~... "~'"'l" %... ~ , 0?....? 4... "--... ~:~,~e ~... "~... ~5"... ~ --.q...e.?..~_... ~"C~ :... " J... "0" i.... I,,,, I,,,, n, r ~,),,,, I ~, T, I A,,, 5 4 m ~3 2 4, "4 E-2 t_ ~ m =. -4 U $ I ALLUVIUM (e) PSA (0.3 s) i SHALLOW SOIL HARD ROCK a.~..~......" ~ga...;... a~... i J i n n OF I I I I I I t I I,, J, ALLUVIUM S ~ SOIL (e) PSA (3.0 s) HARD ROCK ~b- -//~... ",~ IL "1~... $;1... O... t~... "O... Q" = I I I I I I I [ ' ' u t I t i I I OO Distance t Seismgenic Rupture (kin) (d) rsa (LO s) ALLUVIUM... SHALLOW SOIL ~ R O C K... II....O... g---t~... la 0 D ~... ~ " "...., 0 $ S0 200 Distance t Seismgenic Rupture (kin) Figure 5. Nrmalized residuals f the Landers strng-mtin recrdings pltted as a functin f clsest distance t seismgenic rupture. PGA = peak hrizntal acceleratin; PSA = 5%-damped pseud-abslute acceleratin.

11 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 583 that surce directivity can have n the lng-perid amplitude f strng grund mtin. The respnse spectra used t cmpute the spectral ratis are shwn in Figure 7. Fr cnvenience, nly the fault-nrmal (the largest) cmpnents are pltted. This figure clearly shws that the spectra fr the recrding statins lcated nrth f the epicenter (Lucerne Valley, Clwater, Barstw, Yerm, and Frt Irwin) had relatively larger lng-perid amplitudes than recrding statins lcated t the suth (Jshua Tree, Desert Ht Springs, Palm Springs Airprt, and Indi-Cachella Canal). These differences in spectral shapes are cnsistent with the expected increase in lng-perid energy at the nrthern statins as a result f surce directivity (e.g., Smerville and Graves, 1993). Figure 8 shws the rati f the fault-nrmal t faultparallel cmpnents f the respnse spectra fr the nrthern and suthern statins. This figure indicates that the fault-nrmal cmpnents have generally larger spectral amplitudes than the fault-parallel cmpnents, particularly fr perids lnger than abut 1 sec. We cmpare the average ratis fr the nrthern and suthern statins t the rati bserved at Lucerne Valley in Figure 8c. By far the largest rati is assciated with the Lucerne Valley recrding, fllwed by the ratis at the nrthern statins and the suthern statins, respectively. Each f these ratis generally increases with perid, reaching values f abut 5 at Lucerne Valley, 3 t 4 at the nrthern statins, and 1.5 t 2 at the suthern statins at lng perids. The larger spectral ratis at Lucerne Valley and the ther nrthern m ~0 m -~ t- O m E m "4 m E (a) PGA (SOIL) I ALLUVIUM... SHALLOW SOIL a... ~ %..~..~..._._.~..~... ~... I... I... I... [... I,,... I,,,... I... (b) PGA (ROCK) i... iiiiii... i/:/iiiiiiii:: i;ii].'.. iiiiii/ii;?iiiii... ~... ; } ~. ~... ~----;;---~r.~ ii... I... i... i....i... I... I... (c) PSA (0.3 s) (d) PSA (1.0 s) ALLUVIUM... SHALLOW SOIL HARD ROCK...~... ~.~ ~...~. D... =...~... ~.... ~... -g ~..._.g.... ~... i... I... I... I...,...?... I... I... (e) rsa (3.0 s)... SHALLOW SOIL... HARD ROCK ~...~... g... ~... _~... ~... ;q... n ]...m..... ~... la.... ~... g-..-" I... I... I... J... I... i... I Azimuth (deg) -.. ~ ALLUVIUM SItALLOW SOIL HARD ROCK... ~ ~... _~... L...L...~.... ~.... ~ m ;,... I... I... I... I...,,,1,...,h.... h,, Azimuth (deg) Figure 6. Nrmalized residuals f the Landers strng-mtin recrdings pltted as a functin f surce-t-statin azimuth. Azimuth is measured clckwise with respect t the epicenter f the earthquake. PGA = peak hrizntal acceleratin; PSA = 5%-damped pseud-abslute acceleratin.

12 584 K.W. Campbell and Y. Bzrgnia 1,000 (a) NORTHERN STATIONS (b) SOUTHERN STATIONS 10 i'j, ' %\., i ~",. k,/ 7-t L-, I i..'7",.,.;f - - Cimter, B~'stw.... Y~m... Frt Ir~t... ~tc~lm~ / ".~.z~ //".:~ llle~ert H~ Slri~... r~ s~,~... J e ~ Tree... lllill t I I t l l i l l I i i i, till, [ I I II II I I I I I Ilip i i i I i i i i i p l i,, it Perid (s~e) Perid (see) Figure 7. Fault-nrmal cmpnents f 5%-damped pseud-relative velcity (PSV) fr statins lcated t the nrth (a) and suth (b) f the epicenter. All f statins are lcated within abut 60 km f the seismgenic rupture zne and are generally aligned alng the strike f either the Jhnsn Valley-Hmestead Valley r Camp Rck-Emersn Segments f the surface rupture. The recrdings were rtated with respect t the strike f the fault segment with which they were mst clsely aligned. 10 (a) NORTHERN STATIONS (b) SOUTHERN STATIONS e~ m c~ s- S ~ Y ~! " 7 \, I... \I j - - Cwlnm- V.... yllllm... yal't ~ -- D,~nit,t smlp... ~srt~... J~lmm Tr~ i r,,,,1,,, i i 1lilt1 i p iinln (c) MEAN.l [ [ I I l i i L I I I I I ;, l l l I I I [ l l l l Perid f~., i e~ t_. t, - ~d" E Z Nratem 0.1 f r i l l l l l l I I I ritlip t [, i i t i , Perid (see) Figure 8. Ratis f the fault-nrmal t fault-parallel cmpnents f the 5%- damped pseud-relative velcity respnse spectra shwn in Figure 7.

13 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 585 statins are cnsistent with the effects f surce directivity as suggested by Smerville and Graves (1993). Discussin The gd agreement between the bserved rate f attenuatin f PGA and that predicted by the CB94 nearsurce attenuatin relatinship, which includes near-surce recrdings frm bth the Landers and Lma Prieta earthquakes (Fig. 3), suggests that the lw rate f attenuatin bserved during the Landers earthquake is cntrlled primarily by the mre distant recrdings (Rs > 60 km). This is cnsistent with similar results btained fr the Lma Prieta earthquake by Campbell (1991). There are several factrs that culd be respnsible fr this lw rate f attenuatin at the larger distances: (1) the increased dminance f surface waves, which exhibit smaller gemetric attenuatin than the bdy waves that predminate at clser distances; (2) the arrival f critical reflectins ff the base f the crust (MOHO), which can substantially add t the energy cntent f the recrdings beynd sme critical distance (e.g., Smerville and Yshimura, 1990); and (3) amplificatin in the Ls Angeles Basin, the San Fernand Valley, the San Bernardin Valley, and the Cachella Valley, pssibly because f the effects f deep sediments and basin respnse (e.g., Campbell, 1993; Smerville and Graves, 1993). Whatever the cause, these results indicate that it is imprtant t develp attenuatin relatinships that can accunt fr differing rates f attenuatin at near-surce and far-surce distances in rder t accurately predict grund mtins in bth distance ranges. This is particularly imprtant if ne wants t avid verpredicting grund mtins at near-t-intermediate distances. Table 8 gives the average PGA rati nrmalized with respect t the suthwest directin fr sites lcated west, nrthwest, and sutheast f the epicenter. These results indicate that, n average, PGA was abut 60% higher in the nrthwest and sutheast directins and abut 30% higher tward the west. The lwer values f PGA tward the suthwest are cnsistent with the cmbined effects f defcusing resulting frm unilateral rupture prpagatin tward the nrth and nrthwest and the apprximate alignment f this azimuth alng a minimum in the SH radiatin pattern f the tw subevents. The higher PGA tward the nrthwest is cnsistent with directivity resulting frm rupture prpagatin tward the nrth and the nrthwest and the alignment f this azimuth alng a maximum in the SH radiatin pattern, as bserved by Kanamri et al. (1992) and dcumented in articles included in this vlume. The intermediate grund mtins tward the west are cnsistent with this azimuth being aligned alng a maximum in the SH radiatin pattern, but nt subject t directivity. The relatively high PGA tward the sutheast is smewhat f an anmaly. Althugh this azimuth is generally aligned alng a maxi- mum in the SH radiatin pattern f the tw subevents, the grund mtins alng this azimuth were subject t defcusing as a result f the nrthward prpagatin f rupture. They were, hwever, subject t amplificatin because f the effects f deep sediments and basin respnse in the suthern Cachella Valley, which may have mitigated t sme extent the effects f defcusing. As indicated in Table 8, the maximum azimuthal effect bserved during the Landers earthquake is similar in magnitude t that bserved during the 1989 Lma Prieta earthquake (Campbell, 1991). The azimuthal effects f directivity are mst clearly bserved in the lng-perid characteristics f the respnse spectra recrded at nrthern and suthern azimuths. The relatively larger amplitudes f the respnse spectra bserved at the nrthern statins at perids exceeding abut 1 sec (Fig. 7a) are in stark cntrast t the relatively lwer amplitudes bserved at the suthern statins (Fig. 7b). Anther imprtant azimuthal effect bserved in the Landers recrdings was the larger amplitude f the fault-nrmal cmpnent as cmpared t the fault-parallel cmpnent f the respnse spectra (up t a factr f 1.5 t 2 at statins lcated suth f the epicenter and 3 t 4 at statins lcated nrth f the epicenter) fr perids exceeding abut 1 sec and statins generally aligned alng the strike f ne f the tw principal rupture directins. These effects have nly recently begun t be addressed in the develpment f strng-mtin attenuatin relatinships (e.g., Silva and Abrahamsn, 1993; Sadigh et al., 1993). The larger lng-perid grund mtins bserved n the fault-nrmal cmpnents f the near-surce respnse spectra, especially fr statins lcated nrth and nrthwest f the epicenter, are cnsistent with previusly bserved near-fault directivity effects presented by Smerville and Graves (1993) and Sadigh et al. (1993). Frm an engineering standpint, these results are imprtant fr structures with lng natural perids, such as base-islated structures, high-rise buildings, and lng-span bridges. Such structures, if lcated near a fault, culd be subjected t significantly higher lng-perid amplitudes in the directin nrmal t the fault. If the structure were unfavrably riented with respect t this cmpnent, the cnsequences culd be significant. The average azimuthal effects presented in Table 8 are generally cnsistent with the reginal distributin f Mdified Mercalli Intensity (MMI) (e.g., Brewer, 1993; Tppzada, 1993). Fr example, the gegraphic distributin f bth PGA and MMI indicates relatively high levels f shaking (r relatively lw rates f attenuatin) tward the sutheast, west, nrthwest, and nrth, and relatively lw levels f shaking tward the suth and suthwest. The higher levels f shaking are dminated by sites lcated in the San Bernardin Valley, the Ls Angeles Basin, the San Fernand Valley, the Mjave Desert, and the suthern Cachella Valley. The lwer

14 586 K.W. Campbell and Y. Bzrgnia levels f shaking are dminated by sites lcated in the nrthern Cachella Valley, the San Jacint Valley, and the Peninsular Ranges (e.g., the Brreg Valley). Table 7 gives the average PGA rati nrmalized with respect t Alluvium fr a site n Sft Rck r Hard Rck. These results indicate that sites n Sft Rck had acceleratins that were apprximately 20% lwer and that sites n Hard Rck had acceleratins that were apprximately 30% lwer than thse n Alluvium. These resuits cmpare favrably t similar results btained frm an analysis f residuals frm the Lma Prieta earthquake (Campbell, 1991) and frm regressin analyses n suites f earthquakes cnducted by BJF93 and CB94 (Table 7). Fr this cmparisn, Site Classes A and B f BJF93 were cnsidered t be generally cnsistent with the Hard Rck and Sft Rck categries used in this study. Althugh based n nly three Hard Rck recrdings, much larger differences in spectral amplitudes (up t a factr f 3 t 5) were bserved between recrdings n Alluvium and Hard Rck, cnsistent with predictins by cntemprary spectral attenuatin relatinships (e.g., BJF93; C93). The gelgical effects bserved during the Landers and Lma Prieta earthquakes are dminated by sites lcated mre than 40 km frm the rupture. This is als true f the regressin results f BJF93. Therefre, the results given in Table 7 are apprpriate fr lw t mderate strains. These factrs are expected t be smaller at clser distances and larger magnitudes as the grundmtin amplitudes, and therefre the strains, becme higher, and the seismic respnse f the sils becmes nnlinear. These effects are evident in the regressin resuits fc93, CB94, SEY86, and SCACP93, but are nt incrprated in the relatinships f JB88 and BJF93. Cnclusins Based n the analysis f the Landers strng-mtin recrdings presented abve, we ffer the fllwing cnclusins: 1. Recent attenuatin relatinships develped befre the Landers and Lma Prieta earthquakes d a reasnably gd jb at predicting grund-shaking amplitudes and attenuatin rates fr the Landers earthquake at distances within abut 60 km f the fault, but underpredict the grund mtins at lnger distances by as much as a factr f 2 t 3. Attenuatin relatinships develped after the Landers earthquake d a reasnably gd jb at predicting grund mtins frm this earthquake ver the distance ranges fr which they are cnsidered valid. Hwever, an extraplatin f the near-surce attenuatin relatinship f Campbell and Bzrgnia (1994), which includes recrdings frm bth the Landers and Lma Prieta earthquakes, significantly underpredicts peak acceleratins bserved at distances greater than abut 60 kin, indicating that there is a need t develp attenuatin relatinships that can accunt fr ptential differences in near-surce and far-surce attenuatin rates. 2. A regressin analysis f PGA and PSA recrded during the Landers earthquake indicates that the attenuatin f these parameters can be mdeled with a simple equatin that requires n anelastic attenuatin and a gemetric attenuatin rate that ranges between r - 87 at a perid f 0.3 sec t r - '54 at a perid f 3 sec. This unusually lw rate f attenuatin, which is cntrlled primarily by recrdings btained at distances greater than abut 60 km frm the fault, is prbably caused by ne r mre f the fllwing factrs: (1) a change in gemetric attenuatin rate as surface waves begin t dminate the grund mtins; (2) the amplificatin f grund mtins in the Ls Angeles Basin, the San Fernand Valley, the San Bernardin Valley, and the Cachella Valley, as a result f the effects f deep sediments and basin respnse; and (3) the arrival f critical reflectins ff the base f the crust (MOHO). 3. Recrding statins lcated nrth, nrthwest, and sutheast f the epicenter were fund t have acceleratins apprximately 60% higher and thse lcated west f the epicenter were fund t have acceleratins apprximately 30% higher than thse lcated suth and suthwest f the epicenter. These differences, significant at the 90% cnfidence level, are generally cnsistent with the effects f SH radiatin pattern and surce directivity and, tgether with similar bservatins during the 1989 Lma Prieta earthquake, suggest that such azimuthal effects shuld be cnsidered in the empirical predictin f strng grund shaking: 4. Peak acceleratins n Hard Rck were fund t be apprximately 30% lwer and thse n Sft Rck sites were fund t be apprximately 20% lwer than peak acceleratins n Alluvium at distances greater than abut 40 km frm the fault. These differences, significant at the 90% cnfidence level, are similar t thse bserved at similar distances during the 1989 Lma Prieta earthquake as well as thse predicted by attenuatin relatinships develped by Bre et al. (1993) and Campbell and Bzrgnia (1994). Althugh based n nly three recrdings, respnse spectral amplitudes at perids f 0.3 sec and lnger n Hard Rck were fund t be as much as a factr f 3 t 5 times lwer than n Alluvium, similar t predictins frm the spectral attenuatin relatinships f Bre et al. (1993) and Campbell (1993). 5. Respnse-spectral amplitudes, especially at perids f abut 1 sec and lnger, were fund t be relatively higher at recrding statins lcated nrth f the epicenter than at statins lcated t the suth, cnsistent with directivity effects assciated with the unilateral prpagatin f rupture tward the nrth. This effect

15 Empirical Analysis f Strng Grund Mtin frm the 1992 Landers, Califrnia, Earthquake 587 has imprtant implicatins regarding the design f lng-perid structures. 6. The fault-nrmal cmpnents f respnse spectra at perids f abut 1 sec and lnger at statins lcated within abut 60 km f the fault which were aligned apprximately alng the strike f ne f the principal rupture directins were fund t be as much as 5 times larger at the Lucerne Valley statin lcated adjacent t the Camp Rck-Emersn fault, as much as 3 t 4 times higher at statins lcated nrth and nrthwest f the epicenter, and as much as 1.5 t 2 times larger at statins lcated suth and sutheast f the epicenter as cmpared t the fault-parallel cmpnent, cnsistent with the cmbined effects f SH radiatin pattern and surce directivity suggested by Smerville and Graves (1993). This effect has imprtant implicatins cncerning the design f lng-perid structures whse rientatin is aligned unfavrably with respect t this cmpnent. Acknwledgments We wuld like t thank Gene Hawkins and Dennis Ostrm fr prviding us with digital recrdings and site infrmatin fr the tw Suthern Califrnia Edisn recrding statins. We wuld als like t thank Ed Etheredge, wh prvided us with a preliminary cpy f the strng-mtin reprt summarizing acceleratin data btained frm accelergraphs perated by the U.S. Gelgical Survey. We wuld like t acknwledge Bb Darragh and Tny Shakal f the Califrnia Strng Mtin Instrumentatin Prgram (CSMIP) fr prviding us with digitized and prcessed strng-mtin recrdings frm accelergraphs perated by the Califrnia Divisin f Mines and Gelgy. This article was reviewed by Chris Cramer, whse cmments significantly imprved the manuscript. References Ad Hc Wrking Grup n the Prbabilities f Future Large Earthquakes in Suthern Califrnia (1992). Future seismic hazards in Suthern Califrnia, phase I: implicatins f the 1992 Landers earthquake sequence, Spnsred by the Natinal Earthquake Predictin Evaluatin Cuncil, the Califrnia Earthquake Predictin Evaluatin Cuncil, and the Suthern Califrnia Earthquake Center, Published by the Califrnia Divisin f Mines and Gelgy, Sacrament, Califrnia. Bre, D. M. (1983). Stchastic simulatin f high-frequency grund mtins based n seismlgical mdels f the radiated spectra, Bull. Seism. Sac. Am. 73, Bre, D. M., W. B. Jyner, and T. E. Fumal (1993). Estimatin f respnse spectra and peak acceleratins frm western Nrth American earthquakes: an interim reprt, U.S. Gel. Surv. Open- File Rept Brewer, L. R. (1993). The preliminary damage and intensity survey fr the Landers, Califrnia, earthquake f June 28, 1992, U.S. Gel. Surv. Open-File Rept Campbell, K. W. (1985). Strng mtin attenuatin relatins: a tenyear perspective, Earthquake Spectra 1, Campbell, K. W. (1987). Predicting strng grund mtin in Utah, in Assessment f Reginal Earthquake Hazards and Risk alng the Wasatch Frnt, Utah, P. L. Gri and W. W. Hays (Editrs), U.S. Gel. Surv. Open-File Rept , L1-L90. Campbell, K. W. (1988). The Whittier Narrws, Califrnia, earthquake f Octber 1, preliminary analysis f peak hrizntal acceleratin, Earthquake Spectra 4, Campbell, K. W. (1991). An empirical analysis f peak hrizntal acceleratin fr the Lma Prieta, Califrnia, earthquake f 18 Octber 1989, Bull. Seism. Sc. Am. 81, Campbell, K. W. (1993). Empirical predictin f near-surce grund mtin frm large earthquakes, in Prc., Internatinal Wrkshp n Earthquake Hazard and Large Dams in the Himalayas. The Indian Natinal Trust fr Art and Cultural Heritage (IN- TACH), January 1993, New Delhi, India (in press). Campbell, K. W. and Y. Bzrgnia (1994). Near-surce attenuatin f peak hrizntal acceleratin frm wrldwide accelergrams recrded frm 1957 t 1993, in Prc., Fifth U.S. Natinal Cnference n Earthquake Engineering, July 1994, Chicag, Illinis (in press). Etheredge, E., R. Maley, R. Prcella, and J. Switzer (1993). Strngmtin accelergraph recrds frm the M = 7.5 Landers, Califrnia, earthquake f June 28, 1992, U.S. Gel. Surv. Open- File Rept Hawkins, H. G., D. K. Ostrm, and T. A. Kelly (1994). Cmparisn f strng mtin recrds, Landers earthquake 28 June 1992, near Lucerne Valley, Califrnia, Bull. Seism. Sc. Am. (in press). Jyner, W. B. and D. M. Bre (1988). Measurement, characterizatin, and predictin f strng grund mtin, in Prc., Cnference n Earthquake Engineering and Sil Dynamics ll--recent Advances in Grund-Mtin Evaluatin, J. L. Vn Thun (Editr), American Sciety f Civil Engineers, New Yrk, Getech. Spec. Pub. N. 20, Kanamri, H., H. K. Thi, D. Dreger, E. Haukssn, and T. Heatn (1992). Initial investigatin f the Landers, Califrnia, earthquake f 28 June 1992 using TERRAscpe, Gephys. Res. Lett. 19, Marne, C. and C. H. Schlz (1988). The depth f seismic faulting and the upper transitin frm stable t unstable slip regimes, Gephys. Res. Lett. 15, Sadigh, K., J. Egan, and R. Yungs (1986). Specificatin f grund mtin fr seismic design f lng perid structures, Earthquake Ntes 57, 13 (relatinships given in Jyner and Bre, 1988). Sadigh, K., C.-Y. Chang, N. A. Abrahamsn, S. J. Chiu, and M. S. Pwer (1993). Specificatins f lng-perid grund mtins: updated attenuatin relatinships fr rck site cnditins and adjustment factrs fr near-fault effects, in Prc., Seminar n Seismic Islatin, Passive Energy Dissipatin, and Active Cntrl, Applied Technlgy Cuncil, Pal Alt, Califrnia, Rept. N. ATC 17-1, Shakal, A., M. Huang, T. Ca, R. Sherburne, R. Sydnr, P. Fung, P. Malhtra, C. Cramer, F. Su, R. Darragh, and J. Wample (1992). CSMIP strng-mtin recrds frm the Landers, Califrnia earthquake f 28 June 1992, Calif. Div. Mines Gel., Office f Strng Mtin Studies, Sacrament, Califrnia, Rept. N. OSMS Silva, W. and N. A. Abrahamsn (1993). Attenuatin f lng perid strng grund mtins, in Seismic, Shck and Vibratin Islatin, Prc., 1993 Pressure Vessels and Piping Divisin, ASME Smerville, P. and J. Yshimura (1990). The influence f critical MOHO reflectins n strng grund mtins recrded in San Francisc and Oakland during the 1989 Lma Prieta earthquake, Gephys. Res. Lett. 17, Smerville, P. and R. Graves (1993). Cnditins that give rise t unusually large lng perid grund mtins, in Prc., Seminar