Geodetic Determination of Relative Plate Motion in Central California
Transcription
Geodetic Determination of Relative Plate Motion in Central California
VOL. 78, NO. 5 IOURNAL OF GEOPHYSICAL RESEARCH FEBRUARY 10, 1973 Geodetic Determination of Relative Plate Motion in Central California J. C. SAVAGEA•V R. O. Bvxwoxm National Center ]or Earthquake Research U.S. GeologicalSurvey, Menlo Park, Cali]ornia 94025 Geodetic data along the San Andreas fault between Parkfield and San Francisco,California (latitudes 36øN and 38øN, respectively),have been re-examinedto estimate the current relative movement between the American and Pacific plates acrossthe San Andreas fault system. The average relative right lateral motion is estimated to be 32 ñ 5 mm/yr for the period 1907-1971.Between 36øN and 37øN it appears that most, if not all, of the plate motion is accommodatedby fault creep. Although strain is presumably accumulating north of 37øN (San FranciscoBay area), the geodeticevidencefor accumulationis not conclusive. We have re-examinedthe geodeticdata per- slip (called fault creep) occursat the surface; this behavior presumablyrepresentsan extenmotion betweenthe Pacificand Americanplates sion of the area of stable sliding on the fault along the 280-kin-long segment of the San surface upward through the entire crust, so Andreas fault that extends southeast from San that the motion measured acrossthe segment Francisco (Figure 1). The data consistof re- over a few years is similar to the motion in the peated surveys of triangulation networks over rigid-block model shown in the lower part of the period 1907-1962, repeated surveys over Figure 2. In other regions (e.g., the sectionof the period 1960-1970of the geodimeternetwork the San Andreas fault from just north of Holshown in Figure1, anddirectmeasurements of lister to well north of San Francisco)the upper fault creepmostly in the period 1968-1971.The portion of the fault surface appears to be objective is to determinethe relative motion of locked, and strain should accumulate in the geodetic stations (e.g., Santa Ana and Mr. vicinity of the fault (upper sketchesin FigToro in Figure 1) on oppositesides of, and at ure 2). In measuring relative motion of the some distance from, the faults in the San plates acrossa sectionwhere the fault is locked, Andreas system. it is of coursenecessaryto be sure that the Figure 2 showstwo simplemodelsof plate measurementsextend far enough beyond the motion alonga transformfault. The platesslip fault to spanthe zoneof strain accumulation. past one anotherin responseto shearstresses This distanceturns out to be surprisinglylarge. tinent to the determination of the relative that presumablyoriginate from the drag of A roughestimateof it can be obtainedfrom a mantle convectioncurrentsupon the bottom of dislocation model similar to that shown in the the lithosphereplates.Our presentunderstand- upper part of Figure 2. For simplicitywe asing of the San Andreas fault system suggests sume that slip on the fault is zero down to that below about 15-kin depth the platesslip depthD but is an amountb everywherebelow past one another fairly uniformly alonga steep that depth. Then, in the coordinatesystem contact; i.e., a conditionof continuousstable shownin Figure 2, the strike slip displacement sliding may very well prevail at depth, as is v and the tensor shear strain e• on the free evidenced by rather uniform long-term dis- surfaceare givenfrom the simplescrewdislocaplacementrates measuredat the surfaceand tion model[Weertmanand Weertman,1964] by by an apparent absenceof earthquakefoci at depthsgreaterthan about 15 km. Along certain segments of the fault (e.g.,Parkfieldto Hollister; e• - (bD/27r)(x •' -{- D2)-1 see Figure 1) continuousor quasi-continuous v- (b/•r) arctan (x/D) (1) Both quantitiesare plottedas functionsof x/D on the right-handsideof Figure2. Most of the Copyright (•) 1973 by the American Geophysical Union. 832 833 SAVAGE AND BURFORD: ]•ELATIVE PLATE MOT•O• 15 km, as is commonlyaccepted,measurements TAMALPAIS.• •-/•"-d "•/, must '•_ ('-)X•6.'• MTDIABLO SA. • %X• / FRANCISC, OI ""•\ '/ SIERRA.X} "-' MORENA• I•, o I "]•"• ' / "• LOMb'( 2 ¾0 "• -5_*-•4,•1o*.SANTA • ANA e•w•.x• . 3" 0 RO $6 øN X• KM , .... extend to 100 km on either side of the fault. It is this requirement that creates the principal obstacleto interpreting geodeticdata on plate movement: any result can be challengedon the basisthat measurementsmay not have extended sufficiently far from the fault. On the other hand, where the rigid-block type of motion obtains, there is no particular difficulty in measuringrelative block motion. 5•0 TRIANGULATION DATA Part of the primary schemeof triangulation in California spans the San Andreas fault system.Repeated surveysof this systemby the U.S. Coast and Geodetic Survey have furnished PARKFIELD ß Fig. 1. The San Andreas fault system between Parkfield and San Francisco. The principal faults, shown by heavy sinuous lines, are, from west to east, the Safi Andreas, Hayward, and Calaveras. The solid triangles represent stations in the primary triangulation scheme.The lines connecting triangles show the California geodimeter network. invaluable data on relative movement across the fault system. Whirten [1948] comparedthe geodeticpositionsof the stations shownin Figure 3 as determined from tke surveys of 1906, 1922, and 1946 to determine the relative move(The four lines farthest south are identified by ment of the stations. The geodetic positions numbers25, 28, 30, 32. The other lines are identiwere calculatedsubject to the assumptionthat fied in Figure 7.) Observationsof fault creep (in millimeters per year) are shownby numbers along the position and length of the line Mocho-Mt. the faults. Fault creep data marked by * are Diablo was the same for each survey. The from Nason [1971]; other data. are from R. O. results of Whitten's analysis are shown by Burford (unpublished data, 1971). arrows originating at each station in Figure 3. It appearsthat Santa Ana and Tamalpais have relative movement is concentrated close to the undergone no significant net displacement in fault (e.g., 50% of the relative motion occurs the period 1906-1946, whereas Mr. Toro, between x - --D and x -- +D, but, to inGavilan, Loma Prieta, and Sierra Morena have clude 90% of the relative motion, measurements must be made across a zone extending from all moved northwest approximately parallel to x -- --6.3D to x -- +6.3D. Thus if D is ,•bout the fault at average rates of 49, 48, 32, and 36 exy O t• -15 -3 0 +15 i•.•l•• X/O •$ X/D-• DISPLACEMENT STRAIN •• 'ß I•xy tt,tt tttJ y -$ 0 +3 X / O ---P Fig. 2. Simple models for fault motion showingthe distribution of shear strain and strike slip displacementon the surface.The upper sketchesrefer to a locked-fault model, in which slip on the fault occursonly at depths greater than D. The lower sketchesrefer to the rigidblock model, in which slip is approximately uniform with depth. 834 SAVAGE AND BURFORD: RELATIVE PLATE MOTION mm/yr, respectively.Becausethese results are contaminated by appreciable errors accumulated in carrying the triangulation network forward from the base Mocho-Mt. Diablo, it is probablethat the 17-mm/yr differencebetween the lowest and highestrates is not significant. The magnitudesof the accumulatederrors are suggestedby the apparent movementat Santa Ana (1 meter northwest in 1906-1922 and ! meter southeast in 1922-1946). Presumably Santa Ana remained fixed relative to Mocho- Mr. Diablo during the entire'period (all three stations are on the same plate well back from the fault), and the apparent motionsin 19061922 and 1922-1946 merely representaccumuFig. 3. Movement of stations in the primary lated survey error. Thus correctionsof similar scheme of triangulation from a comparisonof the magnitudeand direction (i.e., 1 meter southeast adjusted surveys of 1906, 1922, and 1946 [after for 1906-1922 and 1 meter northwest for 1922- 1946) shouldbe appliedto the displacements at Mr. Toro and Gavi!an. Such corrections would clearly make the motionsat thesestationsmore uniform for the tw(• periodsof time. The point we wish to make here is that significanterrors accumulate in triangulation adjustments,but the relative motion of nearby stations is preserved. The relative motion between nearby stations can frequentlybe calculateddirectly from the observedchangesin azimuth. For example,in cases where the azimuth between two stations is Whitten, 1948]. The arrows represent the apparent station displacementin the periods 1906-1922and 1922-1946. ment parallel to the strike of the San Andreas fault; for the other lines in Figure 4 we have assumedthat the relativedisplacement is parallel to the fault. As can be seen, the data do define reasonablylinear trends. The slopesof the trends have been calculated from a least- squares analysis of all data except the 1882 survey (the 1882 survey was excludedfrom the approximatelyperpendicularto a fault, the fit because the interval 1882-1906 would concomponentparallelto the fault of the displace- tain the sudden displacementsassociatedwith ment betweenthose two stationsis simply the the San Francisco earthquake of 1906); the productof the azimuthchangeand the distance slopes,along with the standard deviation indibetween stations. Such a calculation would catedby the least-squares analysis,are shownin apply, for example,to the line Santa Ana-Mt. Figure 4. Although it does not form part of the priToro in Figure 3. Alternatively, if one assumes that the relative motion is parallel to the fault strike, the magnitude of the displacementcan readily be calculatedfrom changesin azimuth even though that azimuth is not even approximately perpendicularto the fault strike. Whirten [1948] and Meade [1963] have applied these techniquesto the triangulation network in Figure 3. We have used their data for surveysin 1882, 1906, 1922, 1947, 1951, !957, and 1962 to show in Figure 4 the calculated relative displacement as a function of time. For the lines approximately perpendicularto the fault strike (Mr. Diablo-SierraMorena, Mocho-LomaPrieta, Santa Ana-Mt. Toro, and Santa Ana-Gavilan) the method gives the componentof displace- mary triangulation network, Farallon L.H., a lighthouseon a small island about 36 km southwest of the San Andreas fault (Figure 5), has been intersectedin surveysof 1855, 1882, 1906, 1922, 1948, and 1957. The accuracy of the azimuth to Farallon L.H. is, of course,somewhat lower than azimuths in the primary scheme. Nevertheless,the positionof Farallon L.H. (so far seaward from the fault) makes it desirable to attempt to ascertain its motion from the observedchangesin azimuth.Figure 5 showsthe two lines to the Farallon L.H., one from Tamalpais and one from Sonoma Mountain, which have been used to calculate the fight lateral relative motion of Farallon L.H. For the motion SAVAGEAND BURFORD' RELATIVE PLATE MOTION of FarallonL.H. relativeto Tamalpaiswe added the azimuth of Mr. Diablo from Tamalpais in the main-schemetriangulation to the observed (i.e., unadjusted) angle at Tamalpais from Mr. ROSS '•RMTN. •. % 835 ' \ ' \ , 38' 30 N- •.SO.OMA • "% P Diablo to Farallon L.H. to calculate the azimuth MTN , AMALPAIS - •MT•DIABLO f E 1900 1950 . . ! ! ! ! ! ! ! ! i • I MT. DIABLO - MT.TAMALPAIS • LrJ E ß -2'16 • mm/yr • •N , , •RIETA • ,• '• / / -I • 2t <[ 28 * I0 mm/yr • MORENA 0 ,••BLO-SIERRA •o 2 3,9'8 Fig. 5. Map showing the location of Farallon L.H. relative to the main scheme of triangulation. The faults (heavy lines) are, from left to right, the San Andreas fault, the Hayward-Rogers Creek-Healdsburg system, and the CalaverasGreen Valley system. mm/yr 0 '•1 MOCHO - SIERRA MORENA </: -2 of Farallon L.H. The results (Figure 6) suggest a right lateral motion of 8 ñ i mm/yr for Farallon L.H. relative to Tamalpais. That mo- • 2:F MOCHOLOMA •-,"J"-' ,,••,•--•,,, , PRIETA 30"$ mm/yr k- 0 , • .... : MOCHO- MT. TORO <[ 47" 17 mm/yr tion can also be calculated by using the angle from Farallon L.H. to Ross Mountain (seeFigure 5) measuredat Tamalpais. Because the line Tamalpais-RossMountain is parallel to the San Andreas fault, it is reasonableto assume that .j 2 .••..•_•. '( 26'6 • 2 mm/yr ANA - GAVII..AN I • 22 * 13mm/yr bJ ANA - MT.TORO > - 2.[ SANTA •..J ' TaGAVILA. -' MT. TORO ; ;';: I I _1, • ' LrJ_11 I I I 1900 I i I I I I -4"6 mm/yr ß 1950 Fig. 4. Relative right lateral displacement as a function of time for stations in the primary triangulation scheme (Figure 3) as calculated from observed changesi• azimuth [from Whirten, 1948; Meade, 1963]. The average rate of motion in millimeters per year as calculated by a leastsquares linear fit to the postearthquake (1906) data is shown beside each plot. The uncertainty quoted is one standard deviation, as obtained from the least-squaresanalysis. The pre-1906 observations (open circles) were not included in the least-squaresfit. its azimuth remains fixed. Thus the changein azimuth of Farallon L.H. from Tamalpais shouldbe equal to (but of opposite sign from) the change in the angle Farallon L.H. to RossMountainmeasuredat Tamalpais. That analysissuggestsa right lateral motion of Farallon L.H. relative to Tamalpais of 25 ñ 11 mm/yr, an estimate that, becauseof its large standarddeviation,doesnot differ significantly (i.e., within the 95% confidencelevel) from the earlier estimate of 8 ñ 1 mm/yr. The second estimate (25 ñ 11 mm/yr) does suggestvery strongly, however,that the standard deviation (ñ1 mm/yr) of the earlier estimate is much too low, a result that is not very surprising in view of the fact that the least-squaresfit was basedon only 3 data points. The right lateral motion of Farallon L.H. relative to SonomaMountain (Figure 5) can be estimatedfrom the changein the anglebetween Mr. Diablo and Farallon L.H. measured at Sonoma Mountain. Because the line Mt. Diablo- SonomaMountain is practically parallel to the San Andreas fault, its azimuth is not likely to changeappreciably.Thus the changein the 836 SAVAGEAND BURFORD:RELATIVEPLATE MOTION and Mt. Diablo, as is assumedin Whitten's adjustment.If it has not remainedfixed, its movement would introduce a bias tending to causethe right lateral movement of Farallon E i TAMALPAIS TOFARALLON L.H. 0 9•)6'' : : 19•)0 ' mm/yr 6d'l L.H. to be overestimated. Thus there is reason to believe that Whit•en's estimatemay be too high. A triangulationnetwork (SalinasRiver valley net,) extendingnorthwestfrom Parkfield along Z SONOMA MT.TOFARALLON L.H. 21" _$L ø •-•'• 12 1950 the San Andreas fault to within about 30 km of Hollister was established in 1944 and re- Fig. 6. Relative right lateral displacement as a function of time for Farallon L.I-I. relative to Tamalpais and Sonoma Mountain (Figure 5) as calculated from observedchangesin azimuth. The average rate of motion in millimeters per year as calculated by a least-squareslinear fit to the postearthquake (1906) data is shown beside each plot. The uncertainty quoted is one standard deviation. The pre-earthquakedata (open circles) were not included in the least-squaresfit. surveyedin 1963. The network coversa zone extendinglaterally from a few kilometersnortheast of the San Andreas fault to about 40 km southwestof the fault. Meade and Small [1966] compared geodetic positions calculated fromthe two adjustedsurveysto detect relative movement across the San Andreas fault. That com- parisonshowedclearlythat all the significant relative movement occurred within a few kilom- etersof the fault; there was no significantrelaangle Mt. Diablo to Farallon L.H. at Sonoma tive motion between stations on the same side Mountain is equal to the change in azimuth of the fault (i.e., on the samefault block). The of Farallon L.H. to Sonoma Mountain. The measurementsare quite consistentwith the changesin the azimuth so calculatedsuggesta rigid-blockmodelshownin the lower half of right lateral motion of Farallon L.H. relative Figure 2. Meade and Small [1966] also calcuto SonomaMountain of 21 ñ 12 mm/yr (Fig- lated the right lateral componentof motion ure 6). acrossthe fault from the changein azimuth of Dr. C. A. Whitten (quotedby Bolt [1970, p: six lines that crossedthe fault approximately 34] ) has investigatedthe movementof Farallon perpendicular to strike.They foundthe average L.H. relative to Mr. Diablo, SonomaMountain, rate of right lateral slip acrossthe fault was and RossMountain as indicatedby the surveys 32 ñ 1 mm/yr. of 1882, 1906, 1922, 1948, and 1957. Whirten A triangulationnetwork (the Monterey Bay adjustedthe resultssubjectto the assumption arc) with a high density of stationswas exthat Mt. Diablo, Sonoma Mountain, and Ross tended in 1930 from the coastline north of Mountain all remainedfixed, and he found that Mt. Toro to Santa Ana, crossingthe San AnFarallon L.H. moved N 14øW at the rate of 46 mm/yr. The indicatedmotionis quite closely parallel to the strike of the San Andreasfault (N 35øW) in this area. Whitten's value should be comparedwith our own estimateof 21 ñ 12 mm/yr for the right lateral motion of Farallon L.H: relative to Sonoma Mountain. The stan- dreasfault systemnear Hollister (seeFigure 1 for locations).The network was resurveyedin 1951 and 1962. This network has been discussed by Whirten[1960]andby SavageandBur[ord [1970]. Savageand Burfordcomparedthe observedanglesfor the three surveysand concludedthat within the accuracyof the surveys dard deviation of our own estimate (ñ12 there was no evidence for strain accumulation mm/yr) is sufiqcientlylarge that the estimate itself is probablynot inconsistentwith Whitten's value. However, we believethat Whitten's estimate is subjectto very appreciableuncertainty. It is very doubtfulthat RossMountain, scarcely in the blocks on either side of the fault system. The relative motion between blocks appeared to be accommodatedby slip on San Andreas and Calaveras faults or strain accumulation in that underwent 3.5 meters of slip in 1906, has very narrow zones centered on those faults. Savageand Burford estimatedthe combined relativefight lateral slip acrossthe SanAndreas remained and Calaveras faults to be 25 ñ 6 km from the section of the San Andreas fault fixed relative to Sonoma Mountain I mm/yr. SAVAGEAND BURFORD.' [RELATIVE [PLATE Mo•xoN all the 1951 data in Figure 4 are too low; in fact, the valuesare so low that they require left rather than right lateral motion acrossthe San Andreas fault system in the period 1947-1951. 122'00' + 38'00' 0 +M 2 BI.O •• 1,, ,• 1957- ? 1951 -1957 .KM, 3,0 An error in the 1951 azimuth of the line Mr. Diablo-Mocho would also explain this anomaly. Thus we feel that the large right lateral motions inferred from the 1951-1962 San FranciscoBay area triangulation data are probably the consequenceof horizontal refraction in the baseline MOGHO stations in the San Francisco Bay area from com- observations parison of the adjusted triangulation surveys of 1951, 1957,and 1962.In the calculationthe length tive movement and azimuth of the line Mt. Diablo-Mocho assumed to remain invariant in the interval 837 were 19'51- 1962. The data are from Pope et al. [1966]. average right lateral strain accumulationacro• the 70-km-widezone to be 0.04 ñ 0.08 • strain/yr which is equivalentto 3 ñ 5 mm/yr right lateral relative motionof the end points alternative to this explanationis to accept a highly irregular relabetween the Pacific and Ameri- can plates (left lateral slip in 1947-1951, very rapid right lateral slip in 1951-1957, and no slip in 1957-1962).We believethat an error in the azimuth Excludingslip on the faults,they estimatedthe of 1951. The observations in 1951 is much more likely. An error of this magnitudecan be tolerated in a long-termseriesof observationssuch as are shownin Figure 4; however,it will completely dominate the changesin a period as short as 1951-1957. of the arc. Thus, the analysis of Savage and GEODIMETER DATA Bur/ord[1970]leadsto an estimat•of 28 ñ 5 mm/yr right lateral motion for a point near Mt. Toro relative to a point near Santa Ana. Finally, a high-densitytriangulation network Figure 1 shows the northern part of the California geodimeternetwork, which extends was established in 1951 in the area between Mr. Diablo and Mocho on the east and Sierra establishedin 1959 and was surveyedapproximately a•nually until 1968 by the California alongthe SanAndreasfault. Th9 networkwas Morena and Tamalpaison the west (Figure 7). Division of Water This network was resurveyedin 1957 and 1962. Pope et al. [1966] have shownthat the surveys suggesta large fight lateral movement (80 cm for Sierra Morena relative to Mocho) distributed acrossthe network during the period 1951-1957 but negligible and rather random motion in the period 1957-1962.These motions are indicatedby arrowsin Figure 7. As Whirten [1959] pointed out, a closeexaminationof the displacementpattern for the period 1951-1957 suggeststhe possibilityof a clockwiserotation been surveyed by the California Division of Mines and Geology. Line length measurements are available for a period of 9 to 12 years for eachof the lines shownin Figure 1, and the data of the whole network The geodimeternetwork near Hollister contains a rigid configuration (Figure 8) that permits a unique solution (exclusiveof displacementsgeneratedby rigid-bodymotionsof the figure as a whole) for the relative motions relative to the fixed base line Mt. Diablo-Mocho. Such an apparent rotation would result if the observation from either end of the base line to the other end were de- flected by horizontal refraction. Dr. C. A. Whitten (verbal communication,1971) has informed us that up to 4 secondsof horizontal refraction has recently been observed from Mocho, and it doesseempossiblethat the 1951 surveymight have been similarly affected.This possibilityis supportedby the observationthat Resources. Since 1968 it has for most of these lines exhibit a trend in time. We have taken this trend as definedby a leastsquareslinear fit to representthe average rate of changeof line length (dL/dt), and that value, along with its standard deviation, will be used in this analysisto estimate fault slip and strain accumulation. of the five stations involved. Savage and Bur- [ord [1971] have alreadydiscussed this solution, and it will sufficehere 'to review it briefly. We chose to calculate motions relative to Browns, and for the moment we neglect the effects of possiblerigid rotationsof the figure by requir- 838 SAVAGEAND BURFORD.'RELATIVE PLATE MOTION ing that the azimuth of line 21 remainfixed. Then the observedrates of lengtheningof the lines involved determine the relative motions of the stationsas shownby the arrowsin Figure 8. The restriction to a fixed azimuth for line 21 canbe removedby addingany motiongenerated by a rigid rotation of the trilateration figure about Browns. Clearly such a rotation imposes a componentof motion perpendicularto the fault system at Fairview as well as at other stations.Becausethe length of line 21 has not changedsignificantly(dL/dt = 2 ñ I mm/yr) and the line itself is practically parallel to the San Andreas fault, we do not think it likely that station Fairview has moved significantly with respectto Browns.Thus we doubt that any significantrotation should be imposedon the trilateration figure, and consequentlywe believe that the annual rate vectors shown in Figure 8 are approximately correct. I•I•/YR 0 2O DISPLACEMENT 0 I Kid ß I 20 , I Fig. 8. Trilateration network near Hollister. The arrows represent the average rate of displacement of the stations relative to Browns for the period 1960-1970as inferred from the average rate of lengthening of the observed lines. tivity was basedon measurements at the end points and occasionally on a balloonmeasurement at the midpoint [Ho[mann, 1968]; later measurements(shownas solidpoints in the figshownin Figure 8 can be obtainedby making ure) weremadewith laserinstruments(geodolite the plausibleassumptionthat all motionsare or Model 8 Geodimeter), and the correctionfor parallel to the San Andreasfault system.Then atmosphericrefractivity was measuredsimulthe changesin the lengthsof the pairs of lines taneouslyfrom an aircraft flying along the line that link Fairview to Fremont dictate the relaof sight. The later measurementsare of course tive right lateral motion between those points. somewhatmore accurate, but a careful examiLines 17 and 23 require a relative motion of nation of the data indicatesno systematicerror 32 ñ 5 mm/yr, lines 18 and 22 require 27 _ 4 was introduced by changing the method of mm/yr, and lines 20 and 21 require 27 ñ 3 measurement.The linear least-squaresfit to the mm/yr. (The same assumption requires that data for each line is shown in Figure 10, and Sargent and Gilroy both move at the rate of the slope of that line (dL/dt) along with its standard deviation is shown beside the line in 15 ñ 4 mm/yr relative to Fairview.) Fault creep has been measuredat several localities Figure 9. alongthe Calaverasand San Andreasfaultsnear The configuration(•f the geodimeternet near Hollister (seeFigure 1). The averagevalue for Loma Prieta (Figure 11) is suchthat one can fault creep on the Calaveras in this sector is estimate the relative right lateral motion across seento be 15 ñ 2 mm/yr and on the San An- the San Andreas fault system,which at that dreas 10 ñ 2 mm/yr. Thus it would appear latitude consistsof three strands (the Calaveras, that all the relative movementdetectedby the the Hayward, and the San Andreasproper). geodimeternetwork can be accountedfor by Lines 6 and 13 in Figure 11 are nearly colinear and can be taken together to approximate a observedfault creep. Next let us considerthe geodimeternetwork singlelong line from Loma Prieta t(• Allison. north of Hollister (Figures I and 9). The ge- Then the triangle Loma Prieta-Allison-Mt. odimeter data for this area are generally less Hamilton forms a rigid configurationthat can satisfactory than data found elsewhere,and we be analyzedfor relativemotions(againexcludhave chosen to show the individual length ing possiblemotio.nsgeneratedby rigid rotameasurementsas a function of time (Figure 10), tions of the figure as a whole). We shall calso that the reader can judge for himself the culate the motions relative to Mr. Hamilton uncertaintiesin the analysis.Before 1969 mea- and, to eliminatethe contributionsfrom rigid surements were made with a Model 2A Geodimrotation .of the figure, hold the azimuth (•f line eter, and the correctionof atmosphericrefrac- 7 fixed. Then the rates of lengtheningof lines A similar solution of the trilateration network SAVAGEAND BURFORD: RELATIVE PLATE MOTION 839 quire that the motions of the stations relative to Mr. Hamilton be as they are shown by the dashedarrows and numbers in Figure 11. The indicated relative right lateral motion acrossthe fault systemin this area appearsto be 38 ñ 7 mm/yr (Eagle Rock relative to Mt. Hamilton). The configurationof the geodimeternetwork southeastof Hollister (lines 25, 28, 30, and J32 in Figure 1) doesnot permit a uniquesolution for the relative motion acrossthe fault system. However, it appears reasonableto assumethat the motionsare approximat.elyparallel to the fault trend, in which caseit is easy to infer the am.ount of relative right lateral motion required between stations at the ends of each line. The values for the individual lines are shown in Fig. 9. Geodimeter network in the vicinity of San Francisco Bay. The numbers beside the lines represent the average rate of lengthening (with standard deviation) of the line in millimeters per year. The Hollister trilateration network (Figure 8) is shown in the lower right. Table 1, and a meanvalue for the relativeright lateral motion appearsto be 32 ñ 3 mm/yr for the decade 1960-1970. Fault creep has been measured at several localities along this section of the fault (Figure 1). The measurementsgen- erally refer to the right lateral motion observed 6, 7, 10, and 13 determinethe motionsof Allison and Loma Prieta, as is shown by the arrows in Figure 11. The restrictionto a fixed azimuth for line 7 can be removed by adding any motions generated by a rigid-body rotation of the triangle Loma Prieta-Allison-Mt. Hamilton about Mt. Hamilton. Such rotations add a component of motionat Allisonthat is perpendicularto the trend of the San Andreas system. Inasmuch as the predominantmort.on at all stations should be at least approximatelyparallel to the San Andreasfault, the additionalcomponentat Allison producedby the acceptablerotationsmust be rather small. The rotational contribution across a zone about 100 meters wide centered on the active trace of the fault. The measure- ments in this section range from 18 to 36 mm/yr, with an averagevalue of about 25 ñ 6 mm/yr. There is reason to believe that such direct measurementsof fault creep may underestimate'the over-all relative motion within the fault zone [Savageand Bur/ord, 1971], and for this reasonthe higher valuesmay be more reliable; nevertheless,the averagemeasuredvalue for fault creep (25 ñ 6 mm/yr) is not significantly different from the average inferred from geodimetermeasurements(32 ñ 3 mm/yr). at Loma Prieta must be of almost equal magnitude to that at Allison (lines 7 and 10 are of similar CONCLUSIONS The measurements of relative motion across length) but will be almost parallel to the San Andreasfault (becauseline 10 is almostperpen- the San Andreasfault systembetweenParkfield dicular to the fault). Thus we have established that Loma Prieta must moveparallelto the fault trend at a rate of about 31 ---+6 mm/yr plus or minusa few millimetersper year associated with possiblerigid-body rotations of the trilateration figure. The latter contribution is small compared with the uncertainty in the former estimate and can be neglected.We can proceed further by assumingthat the other stations in Figure 11 also move parallel to the trend of the For conveniencethat segmentof the fault has been divided into three sections(northern, central, and southern), with the dividing points San Andreas fault. The observed values of dL/dt for the variouslines (Figure 9) then re- and San Francisco are summarized in Table 2. about 30 km northwest and 30 km southeast of Hollister. In quoting triangulation resultsfrom Figure 4 in Table 2, only lines approximately normal to the San Andreas fault system have been used, so as to obtain more reliable esti- matesof relativeright'lateralmotion. For the southernsectionthe evidenceis quite goodthat the measuredvalue (32 mm/yr) does representthe total relative right lateral motion 840 SAVAGE ANDBURFORD' RELATIVE PLATEMOTION LINE I DL/DT-+10* I mm/yr / LINE 2 0 o 0 -5 I LINE 5 I LINE 6 +20I .DL/DT+14 "2•m/yr IDL/DT-6"2 mm/yr LINE 4 DL/DT+2"I mm/yr o ..... I / DL/DT'+Sa'2 mm/yr +201 " LINE I LINE 9 I LINE 7 . IO DL/DT--5 •,$ mm/yr 0DL•o":• o15 o mm/yr -I0 o o o •) _.1 0,•,ø, ,, ,o,,6', , :: ,, ',/,,oS' ...... ' -I0 ?,',;oS LIN[ I$ DL/DT--I$ LINE II DL/DT--7 I LINE 12 o o •4 mm/yr •' 3 mm/yr DL/DT=+5 "2 mm/yr _,q,, T, +101 o o ' -I0 - 20 ;'o'q ...... LINE 15 DL/OT ß -8" 4 mm/yr / LINE 16 I LINE 14 I DL/DT+7"$mm/yr +11fDL/DT-+I •'3 mm/yr o ß o o, -IO ß • ..... i i i i +Iø t o• Fig. 10. Change of linelengthasa function of timefor geodimeter linesin theSanFran- ciscoBay area.The opendata pointsrepresent measurements by the CaliforniaDepartment of Water Resources, and the solid pointsrepresentmore recentmeasurements, in which atmospheric reftactivitywasdetermined simultaneously froman aircraft.The straightline in eachplot represents the least-squares linearfit to the data.The slopeof that lineis quoted with its standarddeviation as the averagerate of lengthenin• (dL/dt). angulationdata for the SalinasRiver valleynetwork, 1944-1963 [Meade and Sin.all, 1966], relativeplate motionhasbeenmeasured in the centralsectionis also quite good.The primary schemeof triangulation (Figures 3 and 4) show no evidence of relative motion within the shows that Mr. Toro and Gavilan both move block southwest of the fault (i.e., no measur- within the limits of significanceat the same rate relative to Santa Ana, as we shouldexpect acrossthe San Andreas fault system.The tri- ablestrain). Moreover,the relativeright lateral motionmeasuredby the geodimeternetworkis in reasonableagreementwith fault creep measured within 100 meters of the active fault trace. Thus the evidence certainly suggestsrigidblock motion of the type shown in the lower sketch of Figure 2. The evidencethat all the for the rigid-blockmodel.Similarlyno significant movement between Mt. Toro and Gavilan (bothon the samefault block)is demonstrated in Figure 4 (in fact, the indicatedmotion, thoughnot significantly differentfrom zero,is left lateral). However, becauseof the large SAVAGE AND ]•URFORD' RELATIVE PLATE MOTION 841 observedfault creepon the Calaveras(about 14 mm/yr) and the San Andreas (about 12 mm/yr) faults is adequateto explain the observedplate motion,as wasdiscussed previously [Savage and Bur/ord, 1971]. Finally, the various estimates of relative block movement for the centralsectionin Table 2 are quite consistent despitethe fact that they refer to measurements that span zonesof differentwidth acrossthe fault system.In the analysisof the northern section(Table 2), there is somewhatgreateruncertaintythat the geodeticcontroldoesextend far enougheitherto the eastor to the westto be sure that all the relative movement between plates has been included.The excellentagreeFig. 11. Station displacement rates relative to Mt. Hamilton for stations in the Bay area geodimeter network as inferred from the rate of changeof line lengths. In most casesthe information was inadequate to determine the motion without imposing the constraint that it was parallel to the San Andreas fault (N 40øW); for those cases displacement rate is indicated by a dashed rather than solid arrow. The rates of dis- placement (with standard deviations) in millimeters per year are shown beside the arrows. standard deviations these data do not exclude the possibilityof somerelative movementbetween stations on the same fault block (e.g., within the 95% confidence level Mt. Toro could move northwest at the rate of 8 mm/yr relative to Gavilan). Similarly the analysisof Savage and Bur/ord [1970] demonstrated no significantrelative motion in the fault blocks on either side of the fault system, but they could not exclude such movement below the level of significancein their analysis (about 7 mm/yr across 70 km at the 95% confidencelevel). The most convincingevidencethat the rigid-blockmodel applies is the fact that ment between the movement of Loma Prieta relative to Mocho (30 -- 3 mm/yr from Figure 4) and relativeto Mt. Hamilton (31 _ 6 mm/yr from Figure 11) suggeststhe absenceof significant relative motion between Mocho and Mt. Hamilton. The fact that all the stations on the block southwest of the San Andreas fault in Figure 11 moveat the samevelocityrelativeto Mt. Hamilton suggests that none of the relative motion extends appreciably west of the San Andreasfault. However,the standarddeviations associated with both observations are large enoughthat we cannotexcluderelative motion TABLE 2. Summary of Relative Motions of Blocks on Opposite Sides of the San Andreas Fault System between Parkfield and San Francisco Section of Fault Triangulation mm/yr Geodimeter Lines South of Period 32 ñ '1' 32 ñ 3 Central 26 22 ñ 6• ñ 13õ 28 ñ 4 58 ñ 7 Line Data Are Available Right dL/dt, mm/yr Northern Lateral Motion, mm/yr 28 30 32 1959 1959 1960 1959 to to to to 1971 1971 1971 1971 28 25 28 35 ñ ñ ñ ñ 1 2 3 3 35 28 29 37 ñ ñ ñ ñ 2 2 3 3 sll ñ 10ô 8** 59 ñ 21 ñ 12• *Salinas River Valley network, [Meade and Small, 1966]. 1944 to 1965 $Santa Ana-Gavilan, 1906 to 1962, Figure 4. õSanta Ana-Mt. Toro, 1906 to 1962, Figure 4. IIMontereyBay arc, 1950 to 1962 [Savageand Burford, 25 28 Hollister Relative for which Geodimeter Data, 1960 to 1970 mm/yr Southern 2s TABLE 1. Rate of Change of Line Length and Indicated Right Lateral Motion for Data, 1970]. ôMt. Diablo-Sierra Morena, 1906 to 1957, Figure 4. **Mocho-Sierra Morena, 1906 to 1957, Figure 4. ttSonoma Mountain-Farallon L.H., 1906 to 1957, Figure 6. 842 within SAVAGEAND BURFORD: RELATIVE PLATE MOTION the block east ,of the Calaveras fault or west of the San Andreasfault. Perhapsmore direct evidence on the lack of relative movement cepting a rate near 32 mm/yr. Moreover, the fault movement in the southern and central sec- tions seemsto be predominantly,if not entirely, within the western block is the absence of a of the rigid-block type (lower sketch in Figure significantchangein length for geodimeterline 2). Thus the problemof whether geodeticmea14 (Figure 9), which lies wholly within that surementsextend sufficientlyfar from the fault block. Less definite evidence for the lack of to include all significant movement is not inrelative movement within the eastern blocks is volved. After examiningall the uncertaintieswe affordedby the absenceof significantchangein have rather subjectively assigneda standard length of line 4 (Figure 9). The primary tri- deviation of ñ5 mm/yr to the estimatedrelative angulationdata for the northern sectionin Table plate motion, so that the final value we propose 2 extend only to Sierra Morena, which lies only is 32 ñ 5 mmfyr. about 3 km .southwest of the San Andreas fault. The motion of Tamalpais at the extremenorth The stationapparently was displaced1.7 meters end of the network may be anomalous.The art,o the northwest. relative to Mocho-Mt. Diablo rows in Figure 3 certainly suggestno net moveat the time of the 1906 San Franciscoearthquake ment of Tamalpais with respectto Mr. Diablo [Hay)•ord a•d Baldwin, 1908]. This of course and Mocho.However,whenthe repeatedobserindicatesthat the sameamount of relative right vations are plotted as a function of time (Figlateral motion must have accumulated to the ure 4) the uncertainty of the relative motion west of the station in an unknown interval of (--2 -- 16 mmfyr) is clearly demonstrated; time preceding the earthquake of 1906. Thus indeed, the data in that figure indicate that the primary triangulation data for the northern within the 95% confidencelimits the motion sectionin Table 2 may underestimatethe rela- couldbe identicalwith that at Loma Prieta (30 tive plate motion by an amount of the order of mmfyr right lateral), a station that occupiesa 17 mm/yr (based on the 100-year recurrence comparablepositionrelativeto the fault system. interval between great earthquakes). On the There doesappear to be a pattern of motion other hand, absenceof any significantshortening across both the San Andreas fault and the Calain geodimeterline 14 indicatesthat relative right veras-Hayward fault system in Figure 11. A lateral motion is not accumulatingwest of Sierra right lateral motion of about 19 _ 2 mm/yr Morena at present,and the supplementaltri- acrossthe Calaveras-Hayward system is sugangulationdata (last entry in Table 2), which gestedby the indicatedmovementsof Red Hill, extend far seawardto Farallon L.H., suggestan American,and Morgan relative to Mr. Hamileven lower rate of relative motion across the ton. Similarly a right lateral motion of about San Andreas system than do the primary tri7 ñ 2 mm/yr acrossthe San Andreas fault is angulation data. suggestedby the indicatedmovementsof Nike, Examination of Table 2 shows that, within Mindego, and Eagle Rock relative to Loma the limits of error in measurement,all the esti- Prieta and Shelford. The motion of the stations mates are consistentwith a relative right lateral onthe•southwest sideof the Calaveras-Hayward motion of about 32 mm/yr acrossthe San An- system(i.e., the line of stationsRed Hill, Amerdreas fault system.This agreement obtains de- ican, and Morgan) relative to those on the spite the fact that in the southern section all n,ortheast side of the San Andreas fault (i..e., the accommodationappearsto be concentrated stationsLoma Prieta and Shelford) can be estiwithin I km of the fault trace, in the central mated only from the data for the single consection it appears to be concentrated on two necting geodimeterline, line 13. The indicated faults (San Andreas and Calaveras) about 20 relative right lateral motion is 16 _ 5 mm/yr, a km apart, and in the northernsectionit appears rate that is surprisingly large in view of the t(• be spread over more than 40 km. It is true fact that line 13 doesnot crossone of the recogthat the agreementwouldbe more impressiveif nized throughgoingfaults. It is unfortunatethat the standard deviations of the individual meano other suitably oriented geodimeter line surementwere somewhatsmaller; nevertheless, crossesthe gap from the southwestside of the the remarkablecoincidenceof so many different Calaveras-Hayward fault system to the northmeasurementsis compelling evidence for ac- east side of the San Andreas fault; it would be SAVAGEAND BURFORD.'RELATIVE PLATE MOTION of great interestto have an independentcheck on the resultsindicatedby line 13. The data in Figure 6 on the motion of Farallon L.H. are roughlyconsistent with the picture .outlinedabove.The relative right lateral movement of 8 ----+1 mm/yr from 'Tamalpais to 843 The data for lines 1, 2, and 3 (Figure 9) do not appear to be consistentwith any simple interpretation. The change in line i certainly suggestsright lateral motion across the Calaveras fault, but the changein line 3 is inconsistentwith suchmotion. Line 2 appearsto be Farallon L.H. is quite consistentwith the ob- shorteningat the rate of 5 mm/yr, whereas served 7-mm/yr movement acrossthe San Ansimple right lateral motion modelswould sugdreasfault. The right lateral movement(21 _--+ gest that it shouldbe extendingat a very low 12 mm/yr) of Farallon L.H. relative to Sonoma rate. Perhaps an argument c.ould be advanced Mountainis reasonably well within the expected for omitting line 3 because of the scatter in range For a relative movement across the San line length measurementsshown in Figure 8. Andreas fault system of 32 ----+5 mm/yr. A (However, we find it hard to seehow the trend value somewhatlower than 32 mm/yr might could be reversed to a negative slope even by be expectedfor the line Sonoma Mountain to rejecting data.) But even if line 3 is omitted, it Farallon L.H. because it may not span the is difficult to explain the seeminglyinconsistent completezone of the San Andreas fault system behavior of lines I and 2. (Figure 5). Thus the data for the northern The geodimeterdata in Figure 9 do not really tier of stations (Mt. Diablo, Tamalpais,Sonoma fit either of the modelsof Figure 2. The rigidMountain, and Farallon L.H.) are consistent block model (lower sketchin Figure 2) doesnot with our over-all interpretation. apply becausethe measuredfault creep appears The sites of fault creep observationsin the inadequateto explain the observedmotion and also.becauseline 13 has lengthenedsignificantly fault system in central California are shown in Figure 1. Within the area shownin Figure 11, even though it does not cross a recognized fault creep has been observedon the Calaveras throughgoingfault. The locked-faultmodel (upand Hayward faults but not on the San Andreas per sketchin Figure 2) apparentlydoesnot apfault. (An offsetof about 6 mm in an alignment ply becauselines 3, 4, and 14 would be expected array establishedin 1968 on the San Andreas to shortenappreciably,whereasthey actually fault near the crossingwith geodimeterline 15, lengthen,althoughby an insignificantamount.It shownin Figure 9, has been observed,but it is is not completely clear that the discrepancies not clear that this motion is a real measure of are wholly due to the inadequacies of the fault creep.It is sh.own in Figure I as question- m.odels; it is possiblethat the difficulty is due able.) The observationsillustrated in Figure 1 to extraneoustrends in the data (e.g., incomsuggestthat about 10 mm/yr of fault creepoc- plete observationof fault creep,stationinstabilcurs alongthe Hayward-Calaverasfault system ity leading to erroneousvalues of dL/dT, or in the area wherelines6, 7, 9, and 10 crossthe apparent trends in Figure 10 introduced by system.The pattern of fault creepsuggests that random error). We think someof the difficulty slip is transferred from the Calaveras to the in interpreting the data may be associatedwith Hayward near the point where the latter the branching of the Hayward fault from the branchesoff from the former (i.e., just southof Calaveras fault in the vicinity of lines 9 and the crossingwith line 9). Becausethe transfer 10. It is likely that strain may be distributed in .of slip is not a simpleprocess,we are not at all a somewhatunusualmanner at such a Y juncsure that. the measurementsof fault creep re- tion. ported here representall the deformationwithin It is perhapsworth showingthat the geodimthe fault zone. We suspectthere may be con- eter data, in the San Francisco Bay area do centrated deformation within the thin wedge not fit a pattern of uniform strain accumulathat separatesthe Calaverasand Hayward faults tion.Figure 12 shows the• average strainrate in this area. However,the measuredvalues(10 (L -• dL/dt) as a function of azimuth for each mm/yr) for fault creep in this area are some- of the geodimeterlines.'The continuouscurve what less than the displacement across the representsthe uniform strain rate field, deterHayward-Calaveras system inferred from geo- mined by least squares,that best fits the obdimeter measurements(19 _ 2 mm/yr). served data. The principal strain rates for the 844 SAVAC, E AND BURFORD: RELATIVE PLATE MOTION • FIT FOR UNIFORM STRAIN 5 sured near Ross Mountain, of the surface de- NICEI• T earthquakeof 1906 to calculate'the depth to whichsuddenslip extendedat the time of the earthquake.He useda screwdislocati. on model LEASTSQUARE } / formation •36 N 14 ' •Js• •o' ,•s' % qeo. associated with the San Francisco that is just the complementof the one used in (1): constantslip to a depth D and no slip beneaththat depth.From the observed deforma13 9 Fig. 12. Average strain rates for lines in the Bay area geodimeternetwork as a function of azimuth (measured clockwise from north). The pointsare identifiedby line number,and the error bars representone standard deviation on either side of the point. If the strain field in the Bay area were uniform, a sinusoidalcurve would fit the points. The best uniform field fit as determinedby leastsquaresis shownby the heavyline, and the principal axes of the best-fit strain field are shown as an inset in the figure. tion Chinnew estimatedD to be not more than 6 km and perhapsas smallas 2 km. Again,because of the oversimplifiedmodel, this depth estimate must be underst.oodas representing some intermediate depth within the zone of transition from completeslip to no slip. Presumablythe regionthat slipsduring the earthquake is the regionthat did not slip continuously before the earthquake. Thus we would expectthat Chinnery's estimate of D (2-6 km) shbuldbe approximately the sameas our own estimate (9 km); this of course assumesthat the regionof fault creepdoesnot progressively approach closerto the surface until failure oc- field are alsoshownin the figure. It is clear that the field is not a reasonable fit to the observations. Meade [1971] has discussedmeasurements from 1906 (postearthquake) to 1969on a small triangulationfigure (8-km aperture) that spans the San Andreas fault near Ross Mountain (Figure 5). The data indicate approximately uniform right lateral shearstrain accumulation at the rate of 0.55 --+ 0.05 • strain/yr. (The value quoted is for the tensor shear, not the engineeringshear,asusedby Meade; the former is half of the latter.) There wasno indicationof fault c•reepalongthe trace of the 1906 rupture, which passesthrough the triangulation figure. The locked-fault model (upper sketch Figure 2) is presumedto apply to the San Andreasfault in this region,and we can use equation I to obtain an estimate of the depth D below which slip occurs.Using our estimateof relative plate motion (b - 32 mm/yr) and Meade's measurement (0.55 • strain/yr) for the peak strain (i.e., x = 0 in equation 1), we find D -- 9 km. Becausethe simplescrewdislocationmodel (no slip above depth D and constantslip below) is grossly oversimplified,the value of D quoted must be understoodas representingsomeintermediate depth within the zone •f transition from completeslip to no slip. Chinnery [1961] has usedthe values,mea- curs.Chinnery'scomparativelylow value of D wou|d indicatea proportionatelylower value of b (e.g., 16 mm/yr); such a low value would suggestthat part of the relative plate motion is being carried by the northern continuationsof the Hayward and Calaveras faults (Rodgers Creek-Healdsburg faults and GreenValley fault, respectively).However,in view of the tremendous oversimplificationsinv.olved in the screw dislocationmodels,it is possiblethat the differencebetweenthe two estimatesof D is not significant. Acknowledgments. The 1968 geodimeter data in Figure 10 were furnished through the courtesy of the California Division of Mines and Geology; the data for 1969-1971 were furnished through a cooperative effort of the California Division of Mines and Geology, the U.S. Geological Survey, and the University of California, Berkeley (through NSF grant GA 12205 to B. A. Bolt and F. H. Mofiqtt). All the triangulation measurements were made by the U•S. Coast and Geodetic Survey (now National Geodetic Survey). Publication has been authorized by the Director, U.S. Geological Survey. REFERENCES Bolt, B. A., Cause of earthquakes, in Earthquake Engineering, edited by R. L. Wiegel, pp. 21-45, Prentice Hall, Englewood Cliffs, N.J., 1970. Chinnery, M. A., Deformation of the ground SAVAGEAND BURFORD: RELATIVE PLATE MOTION around surface faults, Bull. Seismol. Soc. Amer., 51, 355-372, 1961. Hofmann, R. B., Geodimeter fault movement investigations in California, Calif. Dep. Water Resour. Bull., no. 116, part 6, 183 pp., 1968. Hayford, J. F., and A. L. Baldwin, Geodetic measurementsof earth movements, in The California Earthquake o[ April 18, 1906, vol. 1, pp. 114-145, Carnegie Institution of Washington, D.C., 1908. Meade, B. K., Horizontal crustal movements in the United States, 25 pp., U.S. Coast Geod. Surv., Washington, D.C., 1963. Meade, B. K., Horizontal movement along the San Andreas fault system,Roy. $oc. N. Z. Bull., 9, 175-179, 1971. Meade, B. K., and J. B. Small, Current and recent movement on the San Andreas fault, Calif. Div. Mines Geol. Bull., 190, 385-391, 1966. Nason, R. D., Investigations of fault creep slippage in Northern and Central California, Ph.D. thesis, 231 pp., Univ. of Calif., San Diego, 1971. Pope, A. J., J. L. Stearns, and C. A. Whitten, Surveys for crustal movement along the Hayward fault, Bull. Seismol. $oc. Amer. 56, 317323, 1966. 845 Savage, J. C., and R. O. Burford, Strain accumulation in California, Bull. Seismol. $oc. Amer., 60, 1877-1896,1970. Savage, J. C., and R. O. Burford, Discussion of paper by C. H. Scholz and T. J. Fitch, 'Strain accumulation along the San Andreas fault,' J. Geophys. Res., 76, 6469-6479, 1971. Weertman, J., and J. R. Weertman, Elementary Dislocation Theory, 213 pp., Macmillan, New York, 1964. Whitten, C. A., Horizontal earth movement, vicinity of San Francisco, California, Eos Trans. AGU, 29, 318-323, 1948. Whitten, C. A., Notes on remeasurement of triangulation net in the vicinity of San Francisco, Calif., Calq. Div. Mines Geol. Spec. Rep. 57, 56-57, 1959. Whitten, C. A., Horizontal movement in the earth's crust, J. Geophys. Res., 65, 2839-2844, 1960. (Received May 23, 1972; revised October 4, 1972.)