JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 106, NO. B1, PAGES 771-785, JANUARY 10, 2001
Sourceparameters for the 1952 Kern County earthquake, California: A joint inversionof leveling and triangulation observations
Gerald W. Bawden u.s. GeologicalSurvey, Menlo Park,California
Abstract.Coseismic leveling and triangulation observations are used to determinethe faulting geometryand slip distribution of theJuly 21, 1952,Mw 7.3 KernCounty earthquake on theWhite Wolf fault. A singularvalue decomposition inversion is usedto assessthe ability of thegeodetic networkto resolveslip alonga multisegmentfault andshows that the networkis sufficientto resolveslip along the surface rupture to a depthof 10km. Below10 km, thenetwork can only resolvedip slipnear the fault ends. The preferred source model is a two-segmentright-stepping faultwith a strikeof 51ø anda dipof 75ø SW. The epicentralpatch has deep (6-27 km) left- lateraloblique slip, while the northeastern patch has shallow (1-12.5 km) reverseslip. Thereis nearlyuniform reverse slip (epicentral,1.6 m; northeast,1.9 m), with 3.6 m of left-lateralstrike sliplimited to theepicentral patch. The seismic moment is Mo = 9.2 _+0.5 x1019N m (Mw=7.2). The signal-to-noiseratio of theleveling and triangulation data is reducedby 96% and49%, respectively.The slipdistribution from the preferred model matches regional geomorphic featuresand may provide a drivingmechanism for regionalshortening across the Comanche thrustand structural continuity with the Scodieseismic lineament to the northeast.
1. Introduction slip distributionalong the White Wolf fault. Finally, I compare the resultswith previousstudies and discussthe implicationsin a The Kern County earthquake was one of the largest regionalcontext. earthquakesin California during the twentiethcentury (M,• 7.7, Mw 7.3 [Richter, 1955;Ben-Menahern, 1978]), secondonly to the 2. Geodetic Data great 1906 San Franciscoearthquake. The 1952 eventruptured 2.1. Triangulation 60 km of the White Wolf fault, northof the junctionof the San Andreasand Garlock faults, and near a restrainingbend in the This studyuses coseismic angle changes for the triangulation San Andreasfault (Figure 1). Even thoughthis earthquakewas array that spansthe White Wolf fault (Figure 2 and Table l a). one of the most well-studied events in southern California at the The U.S. Coast and Geodetic Survey (USCGS) collectedover time [Oakeshott, 1955], nearly 30 years elapseduntil the first four decades(1926-1972) of triangulationdata throughoutthe rigorousgeodetic source models were published[Dunbar et al., Big Bend region of the San Andreas fault [Thatcher, 1979; 1980; Stein and Thatcher, 1981]. Unfortunately,differences in Dunbar et al., 1980;Stein and Thatcher,1981' King and Savage, the geometryand slip distributionbetween these models provide 1984;Eberhart-Phillips et al., 1990; Snayet al., 1996;Bawden et conflictinginterpretations on the relationshipbetween the White al., 1997], includingtwo surveysthat bracketthe July 21, 1952 Wolf fault and a number of other active faults and structures mainshock. Two months prior to the earthquakethe USCGS throughoutthe region. completeda comprehensive,6-month-long survey of the trian- I presenta simple sourcemodel for the Kern County earth- gulation array that spansthe southernhalf of the White Wolf quake based on a comprehensiveanalysis of the coseismic fault; this array was again reoccupiedbeginning 2 months triangulationand levelingobservations. This studydiffers from (September1952 to January1953) after the mainshock.Both the previous coseismic analysis by including a larger set of preseismicand postseismicsurveys were conductedto first-order triangulationobservations and by directlyinverting the geodetic tolerances,where the standarderror, o, assignedto each meas- observationsrather than using forward modelingtechniques to urementwas basedon the consistencyof the anglesturned during match shearstrain estimatesand uplift patterns. Furthermore,I each setup and the consistencyamong different setupsat the examinewhere the geodeticobservations can resolve slip and, same station. The standarderror values usedin this study are more importantly,where the resolutionis poor and slip cannotbe <0.8 arc sec. determined.I first constrainthe geometryof the fault modelby The 143 coseismic angle changes were calculated by combiningthe geodeticdata with aftershocklocations and then differencing654 preseismicand postseismic angles (Table lb and invert the observations,in a least squaressense, to estimatethe Figure 2). If repeatedmeasurements were made in either epoch (1951-1952 or 1952-1953), then the angle was determinedby averagingeach observation,weighted by its standarderror. To Thispaper is notsubject to U.S. copyfight.Published in 2001by the AmericanGeophysical Union. assessthe quality of the data, I evaluated the misclosuresfor monumentcombinations that formed a closedtriangle, a total of Papernumber 2000JB900315. 16 closedtriangle (48 angles)for bothpreseismic and postseismi-
771 772 BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL
Bakersfield
...... Ca/iente ', ! __L...... i--.:
:!:--:½• ......
",,.i. % • 7':"L'•: ! ;. ..
! ! ß--* I i .,.,,,,•,•..•.•...... •, ...... •......
ß ..•.• ..-•.:•,-: .
....•..•.•..•l% I IJ I II IIII ...... :•-:•,.. 110l• ...... :..:•:"•:.•:: ?'•"'::".•-•,•;;•!•.:(:::•';•...;.--•-•:'"::.'•...•:• -, -•;.:.-..--::':;::' .. Figure1. Topography,monument, and fault map of theWhite Wolf fault area. The 1952 surface ruptures areshown in boldblack lines, dashed where buried. Star, 1952 epicenter; triangles, triangulation stations; circles,leveling benchmarks.. The figure generated with GMT [Wesseland Smith, 1991].
c epochs. If the misclosurewas largerthan the sum of the errors whereN is the numberof observations,Oi is the ith observation, in the three angle measurements,then the misclosure was and o'• is the ith standarderror. The S/N for the triangulation distributed equally among the three measurements[Bomford, data is 3.33. About 27% of the angle changesare equal to or 1980; Hodgkinsonet al., 1996]. Such misclosureadjustments smallerthan the observationuncertainties. The low S/N is likely were made to 12 of the 654 observations. a functionof the geometryof the network:the majority of the The signal available for modeling is best describedby the triangulationobservations are within the upthrownblock and do signal-to-noiseratio (S/N). The signal representedhere is the not spanthe 1952 surfacerupture. Furthermore,the orientations coseismicangle change, and the standarddeviation of eachangle of many of the observedtriangles are not optimal for resolving changeis set to 1.18 arc sec. The standarddeviation is basedon deformationalong the fault. the weightedaverage of the triangleclosures [King and Thatcher, 1998]and is consistentwith first-order uncertainties (0.8 x/•) 2.2. Leveling . typically assignedto triangulationobservations [Gergen, 1975]. The leveling data used in this study were collected by the The signal-to-noiseratio can be expressedas USCGS [Whitten, 1955] and evaluatedby Stein and Thatcher [1981] for slope-dependentand misclosureerrors (Table 2a). S/N= 1 The majority of the monumentswere initially surveyedin 1926 N-1 and then resurveyedin 1953 following the mainshock. A small segmentof the southernleveling line was surveyedin 1947 and BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL 773
,,,,
Caliente 35' 15'
C-07
35 39
G-14 * • 35 ø 00'
LevelingEpochs 1952-53 Seismicity -•-1926-1953 • M> 7.0 + 1942-1953 • M6.0 • 1948-1953 e M 5.0 • M 4.0 0 10km • M 3.0 I -119' 00' -118' 45' -118' 30'
Figure 2. Geodeticnetwork and seismicity, 1952-1953. The numberedtriangulation stations (triangles) and levelingbenchmarks (circles) correspond to observationsin Tables1 and2. The 1952 surfaceruptures are shownin bold black lines. Star, 1952 epicenter.
1953 (Figure2). In the analysis,I usetwo primaryleveling lines 3. SingularValue Decomposition thatspan both the northernand southern end of the 1952surface rupture,as well as the spur line (1948-1953) to the top of I use single-valuedecomposition [Menke, 1989] to estimate WheelerRidge near the 1952 epicenter(Figure 2). Most of the slip alongthe White Wolf fault andto evaluatewhere fault slip is levelingmonuments used in the inversionsfor all threelines are well constrained(resolution) and to what limits the slip can be locatedon thehanging wall blockof theWhite Wolf fault (Figure determined(uncertainty). I solved = Am, where A is a partial 1) and are lesssusceptible to possiblenontectonic contamination derivative matrix that relates the observeddata, d (i = 1, m) to the from groundwaterand hydrocarbon-relatedsubsidence observed modelparameters that I seek,m (j = 1, m). The modelparameter northof theWhite Wolf fault [Lofgren,1975; Stein and Thatcher, mk is eitherthe dip-slipor strike-slipcomponent of displacement 1981]. I use the elevationchanges between successive bench- on a rectangularfault segment.The A matrixcan be decomposed markpairs as the levelingobservable, so that the observations are to A = UAVx, whereU is an M x J set of eigenvectorswhich independentof elevation changes at the endp_?ints (Table 2b). spanthe dataspace, V is a J x M matrixof eigenvectorsthat span The standarderror for each segmentis axil in millimeters, themodel parameter space, and A is a diagonalmatrix of singular where •x is a constantbased on the precisionof the leveling valuesordered by size [Menke, 1989]. The solutionvector m = surveyset to 2.0 mm andL is thedistance between monuments in V•,At,Up rangesfrom an a priorifixed model (p = 0) to the kilometers. Stein and Thatcher [1981] furnish a complete least squaressolution (p = M), where p are the number of descriptionof the levelinguncertainties. The root-mean-square singularvalues used (0 < p < M) [Kingand Thatcher,1998]. (rms)uncertainty of theleveling data is 4 mm with a S/N of 27.4. The model resolutionmatrix R quantifieshow well the slip The combinedrms S/N for both the triangulationand leveling componentsare resolved along the fault segments[Menke, 1989]. data is 12.2. Each row of the resolution matrix correspondsto one slip 774 BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL
Table la. Kern CountyTriangulation Stations improvementin the model fit, the rows of R with the largest diagonal elementsrepresent the slip componentson the fault Number Latitude,deg Longitude,deg Station segmentsthat are best resolved. The rows of R that are poorly resolved need to be reassessed, excluded from the inversion, 01 34.97693 -117.65200 ACROSS combined with neighboring segments, or constrained by 02 35.20144 -117.03239 ADOBE additionaldata. Once thesepoorly resolvedsegments have been 03 34.90006 -117.64121 BAJADA addressed,the final stepinvolves a least squaresinversion where 04 35.09420 - 117.58681 BED 05 35.16190 -117.30266 BLACK OAK eachof the modelparameters will be uniquelyresolved. 06 35.12565 -117.47012 BRITE The single-value decomposition method of treating the 07 35.07204 - 117.66608 CAMERON geodeticdata differs from previouscoseismic geodetic studies 08 35.14708 -117.54242 CHAPI [Dunbar et al., 1980; Stein and Thatcher, 1981] by directly 09 35.10517 -117.19454 COMMANCH modelingeach angle changeas a discreteobservation, instead of 10 35.16332 -117.42014 CUB 11 35.08576 - 117.49225 DEER forward modeling the derived componentsof shear strain by 12 35.08553 -116.06091 DESERT groupsof 3-4 monuments. I use direct observation:no points 13 35.03949 -117.67960 DOLEMITE need to be fixed or constrainedto calculatethe coseismicangle 14 35.03325 -117.51327 DOUBLE changes. Thus assumptions that apply to shear strain 15 35.27005 -118.57803 ELK calculations, such as uniform shear within a network, no network 16 35.07448 -117.13563 EL PASO 17 34.88021 -117.69918 FAULT dilation, and no network rotation, are avoided. Furthermore,by 18 35.09058 - 117.45730 FENCE modelingthe leveling data as relativeelevation changes between 19 35.13933 -117.35423 FLANNAGAN adjacent monuments, errors associatedwith datum offsets 20 34.96121 -117.71704 GOLD between the preseismic and postseismicsurveys are averted. 21 35.04803 - 117.23360 GORGE 22 35.11281 -117.27735 HORSETHIEF However, since the covariance between successive monument 23 35.08901 -117.33955 JACKS pairs is not utilized, then the S/N for the leveling data may be 24 35.12109 -117.44330 JAIL overestimated. Another difference between this study and 25 34.84961 - 117.63548 JOSHUA previousanalyses is that I include additionaltriangulation data 26 35.41431 -117.03411 KERN 27 35.09958 - 117.51794 KILN bothnorth and southof the surfacerupture. 28 34.95580 - 117.42666 LIEBRE 29 35.01060 -117.63385 LIMESTONE 30 34.79750 -117.69263 LITl•,270 4. Coseismic Fault Model 31 34.79750 -117.69263 LITI•E B 32 34.80829 - 117.63970 LOPE Determination of the coseismicfault model proceededin a 33 35.12904 -117.25615 MART multistepprocess. First, an overparameterizedfault model was 34 34.35224 -117.57044 MAY constructedto examinewhere slip could be resolvedboth along 35 35.12598 -117.63456 MONOLITH 36 34.85209 - 117.62957 MOVE the length of the fault and at depth. Second, the dip of the 37 34.74566 -117.67492 , NUMBER R overparameterizedfault model was varied over a wide rangeof 38 34.83214 - 117.68483 OLD RESE values to evaluate how well the network can resolve variations in 39 35.12074 -117.70544 PAJUELA dip. Third, basedon where the geodeticnetwork could resolve 40 35.07763 - 117.69385 PASS slip, the fault model was reparameterizedinto a four-segment 41 34.92017 -118.59458 PELATO 42 34.56097 - 117.64396 PELONA fault. Model parameters were then incrementally varied to 43 35.05808 -117.60712 QUARTZ determinethe fault geometry (endpointlocation, dips, and fault 44 34.98887 - 117.67360 QUICK depths). Fourth, the model resolution was then reevaluated, 45 35.13810 -117.32969 ROCK SPRINGS resulting in a further reparameterizationof the fault into two 46 34.80498 - 117.64746 SAND 47 34.69314 -117.43861 SAWMILL segments.I again useda grid searchon the two-segmentmodel 48 35.09913 - 117.39790 SCHOOL to determinethe geometryof the preferredmodel. Finally, the 49 35.08543 -117.64567 SHRUB 2 datawere invertedto determinethe coseismicslip distribution. 50 34.86327 -117.67350 SINGLE 2 51 34.98251 - 117.81116 SOLEDAD 4.1. Model Resolution 52 34.91282 -117.69998 STRAIGHT 53 35.13365 -117.59064 SUMMIT To evaluatethe resolutionof the triangulationand leveling 54 34.93639 -117.69052 T10N R14 data on the fault plane, a fault plane with a dip of 75ø, the 55 35.12379 -117.17744 TEJON228 56 34.94562 - 117.64670 WASH steepestdip suggestedfor the White Wolf fault, was divided into 57 35.01059 -118.98572 WHEEL250 30 patches(ten 6.6-km patchesalong strike and three 5-km patchesdown dip) (Figure 3a). This overparameterizedfault model was not used to estimate slip; rather, it was used to examine where on the modeled fault surface the data could componentfor a given fault segment. As p increasesfrom 0 to resolveslip and,more importantly, where the resolutionis poor. M, the fault segmentsthat are resolved best have maximum Figure 4 shows how the model fit varies as the number of valuesalong the diagonalof R, while unresolvedslip parameters singularvalues increases. The approachused here is to find the in otherrows in the matrix remainsmall. Whenp = M, R is the turning point in the trade-off curve of the model fit versusthe identity matrix and all of the parametersare uniquelyresolved, numberof singularvalues [e.g., Harris and Segall, 1987]. The the leastsquares inversion. Typically, asp increases,the model modelfit improvesrapidly for the first nine singularvalues and fit improvesto the point where additionaleigenvalues do not thenimproves more slowly. After 29 singularvalues, the model significantlyimprove the model fit. An F test is used here to fit levelsoff, suggestingthat thereis little gain with additional assesswhether the improvement in the model fit justifies parameters.Evaluating the resolution at the 29th singularvalue, I increasingthe number of singularvalues. Once there is little foundthat the data best resolves shallow (0-5 km deep)slip along Tablelb. The1951-1953 White Wolf FaultAngle Changes a 775
Triangle A0, Triangle A0,
A V B s A V B s
05 02 57 7.36 16 21 55 -7.84 16 02 57 4.88 22 21 45 0.93 27 04 53 0.54 33 21 22 0.87 49 04 27 0.70 23 22 21 -9.66 53 04 35 -1.39 45 22 21 -2.43 53 04 49 0.91 45 22 23 6.98 10 05 14 3.41 19 23 48 -1.42 10 05 19 2.41 22 23 19 -3.97 19 05 14 1.23 22 23 45 -0.84 55 05 02 -5.71 33 23 19 -2.93 57 05 55 -1.08 33 23 45 -1.18 08 06 11 1.76 33 23 48 -4.51 08 06 18 3.38 45 23 19 -3.08 08 06 27 3.47 06 24 11 -1.67 08 06 48 -1.80 06 24 18 1.08 11 06 18 1.65 06 24 48 -4.68 18 06 48 -5.18 11 24 18 2.75 27 06 11 -1.10 18 24 48 -5.82 48 06 24 1.94 05 26 57 4.12 04 08 27 1.68 06 27 08 -6.80 11 08 06 1.44 53 27 04 2.31 27 08 11 0.67 09 33 55 -3.74 53 08 06 2.41 16 33 09 -1.58 53 08 11 0.97 21 33 16 1.93 53 08 27 1.36 23 33 21 -8.87 21 09 16 4.04 45 33 21 -5.51 33 09 16 5.22 45 33 23 1.45 33 09 21 1.96 04 35 53 1.15 14 10 18 1.61 53 35 39 -1.72 18 10 48 -6.59 18 39 35 -0.85 19 10 05 1.01 49 39 35 0.87 39 10 05 -4.19 19 45 23 3.13 39 10 18 2.01 19 45 48 3.91 39 10 19 -5.54 21 45 22 1.64 39 10 48 -5.04 22 45 33 1.52 48 10 19 -0.70 23 45 21 -4.43 06 11 08 -4.49 48 45 23 -0.78 06 11 27 -1.55 06 48 18 2.19 08 11 27 2.48 10 48 24 0.66 18 11 06 -1.28 18 48 23 -4.6 18 11 08 -5.18 19 48 10 -7.30 18 11 24 -1.16 23 48 06 2.04 18 11 27 -2.83 23 48 19 7.64 05 14 19 0.53 23 48 24 1.54 18 14 10 3.76 23 48 45 4.78 19 14 18 -4.35 24 48 06 0.28 19 14 39 -4.95 45 48 19 2.85 57 14 19 6.24 04 49 53 0.66 02 16 55 0.77 53 49 35 0.95 09 16 33 -3.09 04 53 27 -1.17 21 16 57 7.81 08 53 35 0.42 33 16 21 -3.86 27 53 08 1.49 55 16 09 2.96 35 53 27 -1.91 55 16 33 0.70 35 53 49 -0.28 57 16 02 -5.42 49 53 04 0.40 57 16 09 -3.33 02 55 05 5.27 57 16 33 -4.21 05 55 21 -8.96 57 16 55 -5.53 05 55 33 -5.99 10 18 24 -1.54 05 55 57 1.72 14 18 19 0.78 16 55 57 2.28 19 18 39 -5.87 21 55 16 8.01 05 19 10 -4.73 33 55 21 -2.97 10 19 14 8.54 57 55 02 -6.70 10 19 18 9.08 02 57 16 0.65 10 19 23 3.21 02 57 55 -0.76 10 19 48 10.17 05 57 14 1.06 18 19 14 -0.62 05 57 16 1.25 23 19 45 -0.05 26 57 02 -0.66 45 19 05 1.46 26 57 05 -0.98 45 19 10 -3.16 55 57 05 -0.82 48 19 23 -6.96 55 21 09 7.70b aA,V,and B aretriangle vertices, where V isthe observation point; A0 is the angle change clockwise from AV to BV. bRejectedobservation. 776 BAWDEN: KERN COUNTY EARTHQUAKE SOURCEMODEL
Table 2a. KernCounty Leveling Stations Zoback, 1995], focal mechanismstudies find dips between 60 ø and 66ø [Gutenberg, 1955], and seismicprofiling yielded dips Station Latitude,deg Longitude,deg between 60 ø and 65 ø [Goodman and Malin, 1992]. To evaluate the ability of the networkto resolveslip at variousdips, I varied T 55 35.29028 - 118.62833 the dip of the 30-segmentresolution model in 10ø increments U 55 35.28138 -118.64833 between 25ø and 85ø . Changesin the dip of the fault only V 55 35.27361 - 118.64583 slightlychange the spatialresolution described above for the 75ø 1732 USGSa 35.27527 -118.63527 W 55 35.27138 -118.62416 dippingfault in Figure 3. Gently dippingfault planestended to Y 55 35.24639 - 118.58222 increasethe ability to resolvenearer-surface (0-5 km) dip slip in 2410 USGS 35.23889 - 118.57694 the centerof the fault, while decreasingthe dip-slipresolution at Z 55 35.22556 - 118.55806 depths greater than 5 km. For nearer-surface strike-slip 2719 USGS 35.21083 -118.55138 displacement,gently dipping faults have a modest boost in A 56 35.20472 -118.53417 3064 USGS 35.19694 - 118.53750 resolutionat either end of the fault plane. This patternis also B 56 35.19444 -118.52278 observed at moderate depths (5-10 km) and along the C 56 35.18472 -118.50889 southwestern15 km at greater depths. The geometryof the A 54 34.78417 -118.81556 geodeticnetwork can adequatelyresolve shallow strike slip and B 54 34.79583 - 118.85167 dip slip alongthe lengthof the fault andat a varietyof dips,with C 54 34.81027 - 118.88361 D 54 34.83500 - 118.86417 the exceptionof the last 6-km at either end of the fault. The F 54 34.84528 - 118.86972 networkcan resolvemoderate to deepdip slip below the leveling G 54 34.86777 - 118.88333 linesbut lacksresolution along the centralportion of thefault. H 54 34.88806 - 118.90667 K 54 34.92722 - 118.92583 4.2. Fault Geometry T 824 34.94111 - 118.93028 M 54 34.95639 -118.93556 N 54 34.98222 - 118.94278 To determinethe geometryand fault modelparameters to be S 604 34.99389 - 118.94694 usedin the preferredinversion, I useda grid searchmethod that E 608 35.02000 -118.95444 incrementallyvaried each parameterwhile minimizing the p 64 35.03472 - 118.95889 reducedchi square(Z 2) in a joint inversionof theleveling and R 824 35.04861 - 118.96361 triangulationobservations. Each iterationsolved for both the S 604 34.99389 - 118.94694 V 604 34.99444 - 118.99972 Q 55 35.32166 -118.70944 R 55 35.30305 - 118.67416 Table 2b. ElevationChanges S 55 35.29638 -118.66333
From To Elevation aUSGS,U.S. GeologicalSurvey. Change,cm
C-01 C-02 3.42 0.33 mostof the surfacetrace of the fault (Figures3b and c). Shallow C-02 C-03 1.06 0.33 C-03 C-04 2.75 0.33 dip slip can be resolvedalong both the epicentraland northeast C-04 C-05 - 1.25 0.33 endsof the fault, but is poorlyresolved in the centralportion of C-05 C-06 2.38 0.33 the fault. This is also true to a lesserextent for moderatedepths C-06 C-07 0.12 0.36 (5-10 km) along the fault plane. Deep dip slip (10-15 km) can C-07 C-08 4.47 0.33 only be resolvedon two segmentsnear the southwesternend of C-08 C-09 -0.43 0.56 C-09 C-10 -7.71 0.33 the fault (resolutionof >0.4). The ability of the data to resolve C-10 C-11 -14.01 0.33 dip slip is primarily controlledby the locationof the leveling C-11 C-12 -11.20 0.33 routes and, to a lesser extent, the triangulationstations in the C-12 C-13 2.89 0.38 centralregion of the fault. C-13 C-14 15.54 0.38 C-14 C-15 -13.02 0.38 Strike slip is best resolved for shallow (0.0-5.0 km) C-15 C-16 7.28 0.33 displacementsalong the central45 km of the fault (Figure3c), C-16 C-17 7.22 0.41 the result of the high density of triangulationstations in the G-01 G-02 1.61 0.40 hangingwall. The resolutionfor theepicentral end of thefault is G-02 G-03 2.59 0.40 the productof the fault-crossingtriangulation stations and the G-03 G-04 1.04 0.38 G-05 G-06 3.38 0.33 leveling line (Figure 3a). The lack of geodeticcontrol at the G-06 G-07 4.85 0.34 northeasternend of the fault tracelimits the ability to resolveslip G-07 G-08 5.00 0.38 in this region. The geometryof the networkalso allows minimal G-08 G-09 7.00 0.36 strike-slipresolution at moderate(5 to 10 km) anddeep (10 to 15 G-10 G-11 6.49 0.33 G-11 G-12 6.02 0.33 km) portionsof theWhite Wolf fault. G-12 G-13 5.87 0.34 Since the surface rupture was poorly expressedand the G-13 G-14 -1.38 0.33 aftershocklocations have large uncertainties,the dip of the G-14 G-15 -5.61 0.34 rupture plane is uncertain. Previous geodetic models of G-15 G-16 -9.85 0.33 coseismicslip have found a variety of dips for the rupture G-16 G-17 -15.56 0.33 G-17 G-18 -37.09 0.38 plane(s)that rangefrom 20ø to 75ø [Dunbaret al., 1980;Stein G- 14 W-01 3.82 0.46 and Thatcher,1981], aftershockand seismicitystudies find dips W-01 W-02 - 14.42 0.38 between50 ø and 75ø [Cisternas, 1963; Ross, 1986; Castillo and BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL 777
12
10
8
0 10 20 30 40 50 60 Number of singular values -119' Off -118' 30' Figure4. Modelfit (reducedZ2) versusthe number of singular values. The dashedline representsthe point at which there is SW NE little improvementin the modelfit with additionaleigenvalues.
stirke-and dip-slip components oneach fault segment. The Z 2 is computedas
whereN is thenumber of observations,N/is thenumber of free modelparameters, Oi is the ith observation,and Ci is the ith calculatedelevation change. I beganwith a four-segmentversion of the resolutionfault modeland changed the dip on eachfault plane in 5ø incrementsbetween 25 ø and 85ø for a total of 20,736
0 10 20 30 40 50 60 iterations.I foundthat the models with the lowestZ 2 haddips between 65 ø and 85ø. On the basis of this observation, I Distancealong fault plane constrainedthe dip of each fault segmentto 75ø , thereby SW NE loweringthe number of freemodel parameters. 0...... I usedthe aftershockdistribution and reducedchi squareto determinethe fault planedepths (depth is definedas the vertical elevationbelow the surface). The Kern Countyearthquake
. aftershocksand recent seismicity show a bimodaldistribution of õ earthquakedepths along strike, with deeperseismicity (<25 km) == 0.6 5-10 t near the epicentralregion to the southwestand shallowevents (<10 km) in the northeast[Gutenberg, 1955; Castillo and • 0.4 Zoback, 1995; Bawden et al., 1999]. Since the network can resolvedeeper slip (Figure 3) andthe seismicitysuggests deeper coseismicslip in the southwest,I initially set the downdip depth 0.2...... / :...... X...... /, / .... .% / / ...... •...?.\ ' of the epicentralpatch to 20 km and similarlyset the maximum depthat 10 km for the northeasternsegments. Incrementally adjustingboth the minimumand maximumdepths in 0.5-km 0.0 .... , .... , .... • ...... , .... , .... , ...... i"•..., ..... , .... ,. •*., .... , .... ,.. 0 10 20 • 40 50 60 steps for each fault segmentand evaluatingthe Z 2, the southwesternfault segments had the lowest Z 2, with upper depths Distan alongfault plane of >6.0 km andlower depths of >27 km. The Z2 continuedto Fibre 3. (a) Geome• of the resolutionmodel to the g•detic decreasewith bothdeeper minimum and maximum fault depths, network;½) dip-slip resolutionand (c) s•ke-slip resolutionas a but sincethe rate of improvedmodel fit waslower at increasing functionof depthalong the •ite Wolf faults. Solid lines •e for depthsand because of theinability to distinguishdeep slip, these dep•s of 0-5 •, dashedlines •e 5-10 •, •d dottedlines •e valueswere used in the inversion.Using a similarrationale, the 10-15 •. northeasternfault segments had low •2 withminimum depths of 778 BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL
,.,
• :• Faultplane
• ß.... t•. /
. : "' I•:';...... •...... Cal:ien• _.
•: J_ i:'":':"I
• e ' ======• • '•., • ':•.-'.-" ..---•-..-- • •
-110 ø 00' -118 ø 45" -118' 30'
b) A Alongfault A' c) B Faultnormal B' 0
•,10
• 20 ......
3O • • , • • • , 30 0 20 40 60 0 20 40 Distance(krn) Distance(km) Figure5. (a) Geometryof thepreferred fault model to thegeodetic network, (b) along-faultseismicity (1980-1999),and (c) fault normalseismicity. Bold linesassociated with the fault modelare the surface projectof the fault plane. Star, 1952 epicenter; triangles, triangulation stations; circles, leveling benchmarks. near 1.0 km and a maximumdepth of 12.5 km. Thereforeboth 220 m) overa 0.10ø range(11 km). The strikeof eachsegment fault depthswere determinedby the geodeticdata, with the was constrainedat 51ø (the averagestrike of the surfacetrace of maximumdepth of theepicenteral patch partly constrained by the the White Wolf fault), and the segmentswere not allowed to seismicity. overlapor separatealong strike. The modelswith the lowest Z 2 To determinethe coordinatesof the fault endpoints,I valuesshifted the fault segmentsto the southwestalong the minimizedthe Z 2 forjoint inversions of thetriangulation and WhiteWolf fault. The placementof the northeasternfault patch levelingdata carrying out a grid searchof the faultposition, is stronglycontrolled by leveling data, with poor resolving varyingthe latitudeand longitude in 0.002ø increments(about capabilitynortheast of the levelingline. The dataare insensitive BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL 779
Table 3. Fault Model Parameters explains 96% of the leveling data, but only 49% of the triangulationdata. The reasonfor the modestreduction in the Parameters Southwest Northeast reducedchi squarefor the triangulationis unclear. The misfit is scatteredfairly evenly throughoutthe network except at the Strike,deg 51 51 eastern end where the data are satisfied at or near the data Dip, deg 75 75 uncertainty(Figure 8). One possibilityis that sincethere are only Length,km 29.7 23.6 16 closed triangles to assess the data quality, I have Upperdepth, km 6 1 underestimated the data uncertainties for the network. One Lowerdepth, km 27 12.5 LatitudeNpt a 35.132 35.265 possibleimplication of Z2 > 1 in the preferredmodel is that LongitudeNpt - 118.840 - 118.636 assigneduncertainties to the slip model are underestimated.I LatitudeSpt 34.970 35.132 account for this shortcoming by recomputing the slip LongitudeSpt -119.100 -118.840 uncertaintiesby multiplying the formal slip uncertaintyby the squareroot of thereduced Z2 value [Thatcher et al., 1997]. These aNpt,northern endpoint of fault;Spt, southern endpoint of fault. larger uncertaintiesare listed in Table 4 and are specificto the fault geometryof the preferredmodel. to extensionof the epicentralfault patchto the southwest.From Given that the leveling observationshave a high signal-to- this, the fault was resized and partitionedinto two patchesof noise ratio and that the locationsof the three leveling lines cross equal length and a new set of inversionswas performedto both of the fault ends, the leveling data alone were inverted to determinetheir position while minimizing the Z 2. Theendpoints estimateslip (Table 4). The distributionof the coseismicslip in were allowed to incrementally move, in 0.10ø and 0.5-kin the southwest is similar to the joint inversion at the 2(; increments(position and depth), but the strikewas constrained at confidencelevel with 3.8 m of left-lateral strike slip and 1.6 m of 51o and again the fault patcheswere not allowed to overlapor reverse slip. However, the models differ for the northeastern separate.The ge9metry of thepreferred fault model is shownin segment.The leveling-onlymodel produced4 times (0.9 m) the Figure5 andTable 3 andis in goodagreement with the present- amount of left-lateral strike slip than the combined day microseismicity.A comparisonof the Goodmanand Malin triangulation/levelinginversion, with only minor increasein the [1992] fault map, determinedby geologicalmapping, well log reverseslip (2.0 m) (Table 4). data, and deep crustal seismicprofiles, with the geodetically derivedfault model showsan excellentagreement of the fault geometries,including the right-steppingjog approximately midwayalong the fault (Figure6). Eocene-Piiocene Sedimentary Rooks 5. CoseismicSlip Distribution Basement Complex
The coseismic slip distribution was determined by first 5 km invertingthe triangulationdata to obtainthe coseismicstrike-slip displacementsand then fixing thesevalues while invertingthe leveling data for the dip-slip components.This approachwas necessarybecause the low S/N of the triangulationrelative to the leveling data placed an inordinate weight on leveling observationsduring the joint inversion. This resultedin slip models inconsistentwith observed coseismic offset patterns (Table 4). The preferredmodel, which is, of course,not the only possiblemodel, fits the levelingdata well alongall threeprofiles (Figure7). The offsetof the WheelerRidge spurline to the west from the Highway 99 line (Figures7a and b) providedneeded constraintson the geometry of the fault model, becauseminor changesin the geometrywould result in largereduced chi square values in these lines. Similarly, lateral variations in the monument spacing along the Caliente leveling line, which followed a winding road somewhatperpendicular to the surface rupture(Figure 1), resultedin an "apparentscatter" of the data (Figure7c, e.g., between10 and 25 kin) or unusualuplift patterns (Figure7d, e.g., between40 and 50 kin). This apparentscatter is in part due to the lateral vertical deformationgradient and was usefulin determiningthe fault geometryfor the northeasternfault segment. This preferredmodel places3.6 m of left-lateral strike slip and 1.6 m of reverseslip on the southwesternsegment (Table 119'00' 118'50' 4 andFigure 7). Left-lateralstrike slip decreasesin the northeast Figure 6. Detailedfault mapof the centralWhite Wolf fault. to 0.2 m, with1.7 m of reverseslip. TheZ2 value decreases from The bold faults and folds are active structures[Goodman and 12.16 to 3.48 (Table 4). Malin, 1992]. Barbsrepresent exposed thrusts (solid barbs) and Sincethe Z 2 valuesare greater than 1.0, the modeldoes not blind thrusts(open barbs). CPT, ComanchePoint thrust; Rdg., satisfyall of the datato withintheir uncertainties (a perfectfit Ridge;WWF, WhiteWolf fault. Modifiedfrom Goodman and correspondsto a reducedchi squareof 1.0). Thepreferred model Malin [ 1992]. 780 BAWDEN' KERN COUNTY EARTHQUAKE SOURCE MODEL
Table 4. Model Misfits
SouthwestSegment NortheastSegment
Model Signalto Misfitto PercentSignal Reverse Left-Lateral Reverse Left-Lateral Noise Noise Explaineda Slip,m StrikeSlip, m Slip,m StrikeSlip, m
Triangulation 3.33 2.83 49 ..- 3.56 ñ 0.28 ß.. 0.18ñ0.13 Leveling-dip slip only 27.39 7.15 93 1.64 ñ 0.03 ... 1.61 ñ 0.04 ... Leveling 27.39 5.37 97 1.60 ñ 0.03 3.81 ñ 0.35 2.02 ñ 0.11 0.88 ñ 0.25 JointTrig andLeveling 12.16 4.15 52 1.62 ñ 0.03 3.64 ñ 0.23 1.39 ñ 0.05 -0.31 ñ 0.11 Preferred Model 12.16 3.48 69 1.63 ñ 0.03 3.56 ñ 0.32 1.89 ñ 0.04 0.22 ñ 0.14
aReductionin the misfit-to-noiseratio.
Is therea significantdifference in the reductionof residuals combinedleveling and triangulationinversion provides a better betweenthe inversionof leveling only and the joint inversion? fit to the data,even though the modelreduced the triangulation To addressthis question,I comparedthe misfits of the two signalby only 49%. models with an F test: The geodeticmoment was calculatedfor the Kern County earthquakefrom the preferred model (Table 5) by using
4 Me = , i=1
where r is the residual(observed minus calculated);o is the data where!x = 3x10•ø N/m 2 is rigidity,and A, andS t arethe area and uncertainty;and v •, v2 are the numberof degreesof freedomfor slipestimated from fault segmenti, respectively.This yields Me models1 and 2, respectively. The joint inversionproduced F = = (9.2 + 0.5) x 10•9 N m (Mwof 7.22),in goodagreement with 2.56, which is significantat the 99% confidencelevel. Thusthe resultsfrom other seismic and geodetic studies (Table 5).
0.5 a) ßbj i I i I , I , I , 0.4 0.3 0.2
-0.1 0.1 o -o.1 DataModel -0.2 Gorman ß a -0.3 . Wheeler ß v Wheeler ...... I - i ß i ß i ß -0.4 ßI I I i I I ß i...... ß 40 aO 2O 10 5O 4O 30 2O 10 0 Distance along leveling line - km Distance along leveling line - km
0.2 I I I I I I I I 0.5 , I , I , I • I i ß d) 0.4 ...... E 0.1 &...... A 0.3
c: 0
ß • o.1 c: -0.1 ._o ._> 0
ß -0.2
ß
-0.2 ß -0.3 Data Model Caliente ß • ß Caliente , -0.4 i ß i i -• i - ß' i r . i I ß I• ßData i Modelß 4O 30 2O 10 5O 4O 30 2O 10 Distance along leveling line - km Distance along leveling line - km Figure7. Coseismicand modeled elevation changes for the (a, b) Highway99 andWheeler Ridge spur and (c,d) Calienteleveling lines. Elevation change is thecoseismic change in heightbetween adjacent leveling monuments.Relative uplift is the sumof the elevationchanges along leveling line startingat the southernmostleveling point and represents the coseismic uplift pattern along the leveling profile. BAWDEN' KERN COUNTY EARTHQUAKE SOURCE MODEL 781
Leveling residual < 10 rnrn residual<50 rnrn
35 ø 20' -
35 ø 00'-
Triangulation 0 residual < lo
• , .• residual< 50
' I -119 ø 00' -118 ø 30' Figure $. Residualtriangulation and leveling signal from the inversion. Circles, triangulationresiduals normalizedby their uncertaintyfor eachobservation. Each circle representsthe residualfor a given angle changemeasured at the monumentrepresented by the circle. Small circlesrepresent a goodmatch between the model andthe observation.Hexagons, residual elevation changes in millimeters.
6. Discussion SierraNevada basement rock and basinsediments (Figure 1) [e.g., Goodmanand Malin, 1992]. 6.1. Data Misfit Alternatively, the modeled misfit may be a product of Since none of the fault models that I tested completely nontectonicor secondarydeformation that locally disturbedthe replicatedthe signalto within the observationaluncertainty, some networkor, may reflect complexitiesassociated with oblique- aspectof the faulting behaviorremains unmodeled. There are a reverse fault earthquakes. The most notable source of numberof possibilitiesto explainthe model'sinability to explain nontectonicdeformation is from groundwater-relatedsubsidence the entiresignal. The preferredmodel may be an oversimplified in the southernSan Joaquinvalley [Lofgren, 1975; Stein and approximationof the true fault geometry.Unfortunately, the data Thatcher, 1981; Bawden et al., 1997]. To minimize this source available will not allow us to explore models with added of contamination,leveling monumentsselected were located in complexities.Another explanation for the unmodeledsignal may the hangingwall of the White Wolf fault, placedin bedrock,had lie with the modelingassumptions. I assumeuniform-slip faults shallowdepths to bedrock,or were surveyedclose in time to the in an isotropic, homogenous elastic media. Additional mainshock (Figures 1 and 2). Since the triangulation segmentationof the fault model with subsequentinversions observationswere taken within 6 months of the mainshock, then suggeststhat the largest of strike-slipdisplacement occurred in the horizontalcomponent of any signalother than the coseismic the central region of the fault. However, this model is not should be much lower than the uncertainties associated with the significant at the 99% confidencelevel owing to the added measurements. Triggered slip on smaller faults or local numberof free parametersfor the additionalfault planes. The monument instabilities, such as those induced from ground assumptionof an isotropicand homogenousmedium is also not shaking,would alsohave similar effect. Three aftershocksM•. > realisticalong the White Wolf fault. The geologyvaries both 6.1 located near the network (Figure 2) may have introduced along fault and crossfault with a complexmixture of southern local movementat a few of the monuments.In particular,station 782 BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL
57 on Wheeler Ridge was within 2 km of a M,. 6.4 aftershock (Figure 2).
6.2. ComparisonWith Other Studies
The geometryand the slip distributionfrom the preferred model differ significantlyfrom previousgeodetic analysis that usedforward modeling techniques (Figure 9 and Table 5). One of the mostapparent differences among the threegeodetic models is the spatialgeometry of the fault planes. The Dunbar et al. [1980] model is a straightfault that approximatesthe endpoints of the primary surfacerupture and is dividedinto two shallow and two deeppatches (Figure 9b and Table 5). The differences between the Dunbar et al. [1980] model and the model here in are that I find a right-steppingjog approximatelymidway along the fault and the fault is shorter. Additionally, the data are insufficientto resolve slip on multiple downdippatches, as well as along the entire length of the northeasternfault patches. The Stein and Thatcher [ 1981] three-segment,curvilinear fault model poorly matchesthe surfacerupture trace of the White Wolf fault and extendswell beyondthe resolvingcapabilities of the geodetic data to the northeast(Figure 9c). Inversionsthat only included triangulationdata tended to have a southwesternfault segment with an easterly orientation, similar to the work of Stein and Thatcher [1981]. However, these more easterly trending fault segmentsfailed to satisfythe levelingdata, in particular,the spur leveling line on Wheeler Ridge that was not used in their modeling(Figure 2). Another difference in the fault models is the dip of the northeasternfault segment. The Dunbar et al. [1980] model uses a constantdip of 65ø, while the Steinand Thatcher[1981] model Mojave has a decreasingdip from 75 ø in the epicenteralregion to 20ø at 35' 00' - the northeastern end (Table 5). The inversion found that a uniform dip of 75 ø to the southeastexplains the leveling data best. More gentlydipping (<60 ø ) fault planesfailed to matchthe elevation changes along the northern leveling line and were rejected. It is unclear why the dip for the northeasternfault ModeledSlip segmentvaries so much between the Stein and Thatcher [ 1981] 2.0m • modeland the preferredmodel. I omittedleveling data that may 0 t0 have been subjectedto subsidenceassociated with groundwater or hydrocarbonpumping north of the White Wolf fault and monumentswith questionablestability near Tehachapi (Figure 9). Stein & Thatcher The exclusionof thesedata provided a morerobust solution with fewer potential sourcesof nontectioniccontamination. These 35' 20' data omissionsmay be one of the reasonsthat the preferred model differs from the Stein and Thatcher [1981] model. Additionally, since the northern fault patch on their model extendswell beyond the resolutioncapabilities of the geodetic array (Figure 9), the slip and geometry of this fault patch are unreliable. Anotherdifference among the threegeodetic models is the slip partitioningalong the fault surfaces. I found a nearly uniform reverse slip along the fault, while other geodetic studies
Figure9. Comparisonof geodetic models from (a) thisstudy, Gorman (b) Dunbaret al. [1980],and (c) Steinand Thatcher [1981]. The Modete• Slip boldgray lines are the leveling segments used, and the thin black 2.0 m linesare the triangulation measurements used in eachstudy. The km fault coordinatesfor Figures 9b and 9c are approximate. T, 0 Tehachapi. BAWDEN' KERN COUNTY EARTHQUAKESOURCE MODEL 783
determinedgreater reverse slip in theepicentral region (Table 5). Left-lateralstrike-slip displacements for thisstudy were up to 1.5 m greaterthan the otherstudies in the southwestand were significantlylower along the northeastern segment. The reason for the disagreementamong the differentmodels is likelyfrom the geometryand placement of the fault models.Observed surfacedisplacements from the Kern Countyearthquake are small and often contradictory[Buwalda and St. Amand,1955] and so providelittle informationto discriminateamong the models.Additionally, the spatial distributions of aftershock focal mechanismsalso provide little insightwith a mixedpattern of reverse,strike-slip, and obliquemechanisms throughout the region[Dreger and Savage, 1999]. Wherethe geodetic studies agreeis that mostof the slip occurredbelow 5 km for the southwesternhalf of therupture and the upper and lower depth of the fault is shalloweralong the northeasternportion of the fault (Table 5). Thegeodetic moment that I calculatedis at thelower end of the rangeof both seismicand geodeticmoments previously determinedfor the Kern Countyearthquake (Table 5). Observed surface breaks from the mainshock extend 12 km northeast of the preferredfault model. Sincethe geodeticnetwork could not resolveslip beyond this fault patch, then the moment calculated in thisstudy can be takenas a lowerbound for the KernCounty earthquake.
6.3. Implications Surface displacementsand geometry associatedwith the preferredmodel are consistentwith the regional topographyand the currenttectonic structures along the White Wolf fault (Plate 1). The southwestern half of the fault has no discernible topographicrelief, with the exceptionof Wheeler Ridge at the southwesternend of the fault. Conversely,the northeasternhalf of the fault has elevatedtopography in the hangingwall block with elevationsas high as 2100 m, while the footwall block, for the most part, remainsat an elevationof 200 m (Figure 1). The regionwith the greatestcoseismic uplift and sharpestdeformation gradientagrees well with the present-daytopography (Plate 1). Even thoughthe area of model maximumuplift doesnot directly correspondto the highesttopography, it doesinclude Comanche Point, a siteof contemporaryfolding and thrustfaulting (Figure 6 and Plate 1) [Goodmanand Malin, 1992]. The preferred slip modelhas high left-lateralstrike slip in the southwest(3.6 m) and minimal strike slip in the northeast (0.2 m). If this slip distributionpattern were typical of earthquakesalong the White Wolf fault, then some structure would need to accommodate the differential strike slip between the two fault patches. The position and orientation of the Comanche thrust system is consistentwith the regional shorteningthat would be expected with the strikeslip differentialthat the preferredmodel produced. Aftershocksfor the Kern County earthquakeare compatiblewith northeastshortening across the Comanchethrust system [Dreger and Savage, 1999]. The steeplydipping fault patchalong the northeasternportion of the fault is consistentwith the present-dayseismicity and providesstructural continuity between the White Wolf fault and the Scodie lineament, a newly forming strike-slip fault that extendsnortheast from the White Wolf fault (Figure 1) [Bawden et al., 1999]. Aftershocks immediately following the Kern Countyearthquake provide little structuralcontrol on the dip of 784 BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL
Calculatedelevation change (m) 1952 Surface rupture Faults
- 35o10 ,
ß %
...
lO km
119ø00 ' 118ø30 ' Plate 1. Comparison of modeled elevation changesand regional topography. Contours along the northeasternportion of the fault have been truncated to avoid covering all of the topography. CPT, Comanche Point thrust.
the northernfault becausethey lack a consistentpattern in the southeastand a slip distribution pattern with nearly uniform focal mechanism distribution [Dreger and Savage, 1999]. reverse slip (1.6 and 1.9 m, southwestto northeast)along the However, recent seismicity studiesat the northern end of the length of the fault. Left-lateral strike slip was primarily limited White Wolf fault indicatepredominantly strike-slip mechanisms to the southwestern half of the fault with 3.6 m. A resolution on near-verticalplanes, which is consistentwith left-lateralshear analysisof the data showsthat the previousgeodetic studies may along the White Wolf fault and the Scodie lineament [Castillo have overestimated the resolution of the data by either and Zoback,1995; Bawdenet al., 1999]. The dip on the northern subdividingthe fault into multiple downdippatches or estimating end of the Scodielineament (750-80 ø) is in goodagreement with slip on fault patchesconstrained by sparsedata. the preferredcoseismic model (75ø) and further suggestingthat The coseismicslip model is supportedby severalgeomorphic these structures are related. and structural features, as well as a kinematic fault model of southern California. The deep-seated strike slip along the 7. Conclusions epicentralpatch correlateswith both convergenceand uplift in the Comanche Point region, an area of active folding and Observations of coseismic elevation and angle changes thrustingmidway along the White Wolf fault. Furthermore,the associatedwith the 1952 Kern County earthquakefavor a two- regionwith the highestcoseismic uplift and sharpestdeformation section right-steppingfault with deep (6-27 km) left-lateral gradient correspondswell to the present-daytopography. The obliqueslip along the southwesternpatch and shallow(1.0-12.5 geometryof the northernend of the fault providescontinuity with km) reverseslip along the northeasternpatch. The preferred the Scodie seismic lineament to the northeast. Plate kinematic source model has a strike of 051ø and a dip of 75ø to the models from geologic and very long baseline interferometry BAWDEN: KERN COUNTY EARTHQUAKE SOURCE MODEL 785 measurementsfor the greaterSan Andreas,Garlock, and White Hanks, T. C., J. A. Hileman, and W. Thatcher, Seismic moments of the Wolf fault region show plate velocities orthogonal to the larger earthquakesof the southernCalifornia region, Geol. Soc.Am. Bull., 86(8), 1131-1139, 1975. northeastern end of the fault, while the southwestern White Wolf Harris, R. A., and P. Segall, Detectionof a locked zone at depthon the and Pleitofaults have a high componentof left-lateralmovement Parkfield, California, segmentof the San Andreasfault, J. Geophys, [Saucier et al., 1993]. This is in good agreement with the Res., 92(8), 7945-7962, 1987. coseismic model. Hodgkinson,K. M., R. S. Stein, and G. Marshall, Geometryof the 1954 Fairview Peak-Dixie Valley earthquake sequence from a joint Acknowledgements. Discussionand reviews with Ross Stein and inversionof leveling and triangulationdata, J. Geophys.Res., 101(11), Wayne Thatcher provided valuable insightsthat greatly improvedthe 25,437-25,458, 1996. manuscript.I thankJim Savage,Will Prescott,and Chuck Wicks for their King, N. E., and J. C. Savage,Regional deformationnear Palmdale, reviews,comments, and discussion.Jeff Freymueller,Mark Murray, and California, 1973-1983,J. Geophys.Res., 89(4), 2471-2477, 1984. an anonymous reviewer provided thoughtful reviews that greatly King,N. E., andW. Thatcher,The coseismicslip distribution of the 1940 improved the manuscript. I used inversionand dislocationmodeling and 1979 Imperial Valley, California, earthquakes and their programsdeveloped by Kathleen Hodgkinsonand Shinji Toda. The implications,J. Geophys.Res., 103(8), 18,069-18,086,1998. earthquake locations were provided by the Southern California Lofgren,B. E., Land subsidencedue to ground-waterwithdrawal, Arvin- EarthquakeData Center and the SouthernCalifornia SeismicNetwork, a Maricopaarea, California, U.S. Geol. Surv.Prof. Pap., 437-D, 55 pp., joint projectof the U.S. GeologicalSurvey, Pasadena, and Seismological 1975. Laboratory,Pasadena. The GMT program[Wessel and Smith,1991] was Menke, W., GeophysicalData Analysis:Discrete Inverse Theory,289 usedto preparea numberof the figures. pp., Academic,San Diego, Calif.,1989. Oakeshott,G. B., The Kern Countyearthquakes in California'sgeologic history,in Earthquakesin Kern CountyCalifornia During 1952, Bull, References 171, edited by G. Oakeshott,pp. 15-22, Calif. Div.of Mines, San Francisco, 1955. Bawden,G. W., A. Donnellan,L. H. Kellogg,D. Dong, andJ. B. Rundle, Richter, C. F., Foreshocksand Aftershocks,in Earthquakes in Kern Geodetic measurements of horizontal strain near the White Wolf fault, CountyCalifornia During 1952, Bull. 171, editedby G. Oakeshott,pp. Kern County,California, 1926-1993, J. Geophys.Res., 102(3), 4957- 177-198, Calif. Div. of Mines, San Francisco, 1955. 4967, 1997. Ross, D.C., Basement-rock correlations across the White Wolf- Bawden, G. W., A. J. Michael, and L. H. Kellogg, Birth of a fault: Breckenridge-southernKern Canyon fault zone, southern Sierra Connecting the Kern County and Walker Pass, California, Nevada,California, U.S. Geol. Surv.Bull., B 1651, pp. 25, 1986. earthquakes,Geology, 27(7), 601-604, 1999. Saucier,F., E. Humphreys,D. E. Smith,and D. L. Turcotte,Horizontal Ben-Menahem, A., Source mechanism of the 1906 San Francisco crustal deformation in Southern California from joint models of earthquake,Phys. Earth Planet. Inter., •7(2), 163-181,1978. geologic and very long baseline interferometrymeasurements, in Bornford,G., Geodesy,452 pp.,Clarendon, Oxford, England, 1980. Contributionsof SpaceGeodesy to Geodynamics:Crustal Dynamics, Buwalda, J.P., and P. St. Areand, Geologicaleffects of the Arvin- Geodyn.Ser., vol. 23, edited by D. E. Smith, and D. L. Turcotte,pp. Tehachapiearthquake, in Earthquakesin Kern CountyCalifornia 139-176, AGU, WashingtonD.C., 1993. During 1952, Bull. 171, editedby G. Oakeshott,pp. 41-56, Calif. Div. Snay,R. A., M. W. Cline, C. R. Philipp,D. D. Jackson,Y. Feng,Z. K. of Mines, San Francisco, 1955. Shen, and M. Lisowski, Crustal velocity field near the big bend of Castillo, D. A., and M.D. Zoback, Systematicstress variations in the California'sSan Andreasfault, J. Geophys.Res., 101(2), 3173-3185, southern San Joaquin Valley and along the White Wolf Fault: 1996. Implications for the rupture mechanicsof the 1952 M, 7.8 Kern Stein, R. S., and W. Thatcher, Seismic and aseismic deformation County earthquakeand contemporaryseismicity, J. Geophys.Res., associatedwith the 1952 Kern County, California, earthquakeand 100(4), 6249-6264, 1995. relationshipto the Quaternaryhistory of the White Wolf fault, J. Cisternas,A., Precisiondetermination of focal depthsand epicentersof Geophys.Res., 86(6), 4913-4928, 1981. local shocksin California, Bull. Seismol.Soc. Am., 53(5), 1075-1083, Thatcher, W., Horizontal crustal deformation from historic geodetic 1963. measurementsin southernCalifornia, J. Geophys.Res., 84(B5), 2351- Dreger, D., and B. Savage, Aftershocksof the 1952 Kern County, 2370, 1979. California,earthquake sequence, Bull. SeismoLSoc. Am., 89(4), 1094- Thatcher, W. R., G. A. Marshall, and M. Lisowski, Resolution of fault 1108, 1999. slip alongthe 470-km-longrupture of the great 1906 San Francisco Dunbar,W. S., D. M. Boore,and W. Thatcher,Pre-, co-, and postseismic earthquakeand its implications,J. Geophys.Res., 102(3), 5353-5367, strain changes associatedwith the 1952 ML = 7.2 Kern County, 1997. California, earthquake,Bull. Seismol.Soc. Am., 70(5), 1893-1905, Wessel, P., and W. H. F. Smith, Free softwarehelps map and display 1980. data, Eos Trans., AGU, 72(41), 441,445-446, 1991. Eberhart-Phillips,D., M. Lisowski,and M.D. Zoback,Crustal strain near Whitten, C. A., Measurements of Earth movements in California, in the big bend of the San Andreasfault: Analysis of the Los Padres- Earthquakesin Kern County California During 1952, Bull. 171, Tehachapitrilateration networks,California, J. Geophys.Res., 95(2), edited by G. Oakeshott, pp. 75-80, Calif. Div. of Mines, San 1139-1153, 1990. Francisco, 1955. Gergen, J. G., The new adjustmentof the North American horizontal datum-theobservables, Am. Congr.Surv. Mapp. Bull., 51, 9, 1975. Goodman,E. D., and P. E. Malin, Evolutionof the southernSan Joaquin G. W. Bawden, U.S. Geological Survey, 345 Middlefield Road, MS Basin and mid-Tertiary "transitional"tectonics, central California, 977, Menlo Park, C 94025. ([email protected]) Tectonics,••(3), 478-498, 1992. Gutenberg,B., The first motionon longitudinaland transversewaves of the main shock and the directionof slip, in Earthquakesin Kern CountyCalifornia During 1952, Bull. 171, editedby G. Oakeshott,pp. (ReceivedDecember 16, 1999; revisedJune 13, 2000; 165-170, Calif. Div. of Mines, San Francisco,1955. acceptedAugust 18, 2000.)