Heat Flow and Energetics of the San Andreas Fault Zone
Transcription
Heat Flow and Energetics of the San Andreas Fault Zone
JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 85, NO. Bll, PAGES 6185-6222, NOVEMBER 10, 1980 Heat Flow and Energeticsof the San AndreasFault Zone ARTHUR H. LACHENBRUCH AND J. H. SASS U.S. GeologicalSurvey,Menlo Park, California Approximately100heatflowmeasurements in the SanAndreasfault zoneindicate(1) thereis no evidencefor localfrictionalheatingof themainfaulttraceat anylatitudeovera 1000-kmlengthfromCape Mendocinoto SanBernardino, (2) averageheatflowis high(~2 HFU, ~80 mW m-2) throughout the 550-kmsegment of theCoastRanges thatencloses theSanAndreas faultzonein centralCalifornia'; this broadanomalyfalls off rapidly towardthe Great Valley to the east,and over a 200-km distancetoward the MendocinoTriple Junctionto thenorthwest.As othershavepointedout, a localconductiveheat flow anomalywouldbe detectable unlessthe frictionalresistance allocatedto heatproductionon the main tracewere•<100bars.Frictionalworkallocated to surface energyof newfractures is probablyunimportant,andhydrologic convection is not likelyto invalidatethe conduction assumption, sincethe heat dischargeby thermalspringsnearthefaultis negligible. Explanations for thelow dynamicfrictionfall into two intergradationalclasses:thosein which the fault is wdak all of the time and thosein which it is weak onlyduringearthquakes (possibly justlargeones).The firstclassincludes faultscontaining anomalously weak gougematerialsand faultscontainingmaterialswith normal frictionalpropertiesunder near-lithostaticsteadystatefluid pressures. In the secondclass,weakeningis causedby the event(for example,a thermallyinducedincrease in fluidpressure, dehydration of clayminerals, or acoustic fiuidization). In thisclass, unlikethefirst,theaverage s•rength andambienttectonic shearstress maybelarge,~ 1 kbar, but thestress allocated to elasticradiation(theapparentstress) mustbe of similarmagnitude, an apparent contradiction withseismic estimates. Unlessseismic radiationis underestimated for largeearthquakes, it is difficulttojustify averagetectonicstresses on the main traceof the SanAndreasfault in excessof ~200 bars.The development of the broadCoastRangeheatflowanomalysouthward from CapeMendocino suggests thatheatflowincreases by a factorof 2 within4 m.y.afterthepassage of theMendocinoTriple Junction.Thispassage leavesthe SanAndreastransformfault zonein itswake;thedepthof the anomaloussources cannotbe muchgreaterthanthe depthof the seismogenic layer.Someof the anomalous heat may be suppliedby conductionfrom the warmer mantle that must occur south of the Mendocino transform(wherethereis no subducting slab),andsomemightbe suppliedby shearheatingin the fault zone.With no contributionfrom shearheating,extrememantleupwellingwouldbe required,and asthenosphere conditions shouldexisttodayat depthsof only ~20 km in thenorthernmost CoastRanges. If there is an appreciablecontributionfrom shearheating,the heat flow constraintimpliesthat the seismogenic layer is partially decoupledat its baseand that the basaltractionis in the sensethat resists fightlateralmotionon the fault(s).As a resultof thesebasaltractions, the averagesheafingstress in the seismogenic layerwouldincrease withdistance fromthemainfault,andtheseismogenic layerwouldoffer substantial resistance to platemotioneventhoughresistance on the mainfault mightbe negligible. Thesespeculativemodelshave testableconsequences. INTRODUCTION hapsan orderof magnitude)we can scarcelyclaim to understandthe physicsof earthquakes or the resistance to platemoThe simplestmodelsof pure strikeslip upper-crustalearth- tions. It is seen that with seismologicestimates of two quakessuchas thoseon the San Andreasfault can be considindependent relationsamongthe stresses (for example,apparered in termsof three fundamentalstresses: an averagecom- ent stressand stressdrop), it is in principlepossibleto estimate ponentof shearingtractionin the directionof fault slip that the magnitudesof all three stressesfrom an estimate of the existsat the onsetof faulting,a corresponding componentthat magnitudeof any one of them. Measurementsof rock friction existsat the conclusionof faulting, and an averageresisting provideadditionalusefulconstraints. In the first part of this stress that results in the local conversion and loss of mechani- cal energy during an earthquake. Indirect estimatesbasedon seismological observationsprovide information on differences among these generalized stressesbut not on their absolute magnitudes. Similarly, geodetic observationsof deformation at the earth's surfaceprovide information on ratesof accumulation and releaseof strain, but they do not generallydistinguish betweenelastic and inelasticcomponents,and in any case,they provide no meansof extractingthe absolutevalue of elasticstrain necessaryto estimatethe magnitudesof the fundamentalstresses. Although seismologicand geodeticobservations have resulted in a substantial increase in the under- paper,weaddress theproblemof estimating themagnitude of the resistingstressfrom considerations of the thermalbudget of the fault zone. It waspointedout by Bruneet al. [1969]that for the longterm averagedisplacementrates that have been documented on the San Andreas fault, the frictional heat generation shouldresultin a conspicuous localheatflow anomalyif the averagedynamicfrictionalresistance in the seismogenic zone exceeded a hundred bars or so. Since no such heat flow anom- aly had been observed,it was concluded that this stresslimit applied. Applying seismicallyderived constraintson stress drop and apparentstress,they concludedthat the initial (tectonic) stresswas probablylimited to about 250 bars. Addi- standingof earthquakes,our ignoranceof the magnitudesof the stressesprecludesan understandingof their energetics. tional heat flow measurementsand analysis[Henyey and WasWith the presentuncertaintyin thesethree quantities(per- serberg, 1971; Lachenbruch and Sass, 1973] generally supportedthese conclusions.In a recent study of New ZeaThis paperis not subjectto U.S. copyright.Publishedin 1980by land's Alpine fault, however, Scholz et al. [1979] interpret lothe American GeophysicalUnion. cal thermal metamorphismand argon depletion as evidence Paper number 80B0929. 6185 6186 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE O=0 y • 1 dO = •. O(x), t>O thermal part of the resistanceon the assumption that the frictionally generatedheat is transmitted by conduction (an assumptionwe examine in a later section). If the sliding motion of two rigid blocks with uniform relative velocity2v is opposedby an averageresistingstressR, the rate of energydissipationper unit fault area can be written as 2vR-- D + e (1) where D is the average rate at which energy is consumedby creating new surface,and Q is the averagerate of mechanical heat generation.If R• and ß representthe portions of the total X2 resistingstress• allocated,respectively, to eachprocess, then D -- 2vR• (2) Q -- 2w (3) R = R• + •- (4) x Fig. 1. Conditionsfor the problemof frictionalheatingon a strike slip fault. that dynamic friction in the seismogeniczone of that fault averaged at least I kbar. Conclusionsfrom all of these studieswere basedupon heat conduction models, which are elaborated somewhat in the next sectionand then applied to the heat flow data presently available near the San Andreas fault. The new data generally confirm the conclusions of earlier studies, viz., that there is no detectable local heat flow anomaly at the San Andreas fault We shall first considerthe relative importanceof R•. The Role of SurfaceEnergy Consider a portionof a faultsurface of areaA thathasbeen active for a time t during which relative displacement2vt -- 2u has occurred.We supposethat this motion has resulted in the creation of newsurfaces of areaa andsp6cific surface energy •, i.e., trace. After consideringthe probability of heat removalby hydrologic processes,we conclude that it is unlikely and con- D= a• -At sequently,that the 100-barslimit on averagefrictional resistance probably applieswhile fault motion is in progress.We Then from (2), then considerthe heat flow constraint in terms of generalized a• results on rock friction and seismic stress differences and confirm that if the constraint on total seismic radiation R•= 2uA is taken at (5) face value, the fault must be surprisingly weak even when If the newly created surfacea is representedby a gougezone fault motion is not in progress.Following theseconsiderations of half-width B composedof cubeswith dimension b, then of heat flow near the fault trace, we examine the broad heat 6AB flow anomaly observed throughout the California Coast Rangesand its implicationsfor the evolution and mechanical behavior of the San Andreas RESISTING fault zone. STRESS AND EFFECTS OF STRIKE LONG-TERM a• b (6) and THERMAL SLIP FAULTING When slip occurs on a fault, elastic energy stored in the earth is released.Some may appear as kinetic energy of seismic radiation that leaves the fault zone, and the remainder is consumedlocally in overcomingthe resistanceoffered by the fault to progressof the slip motion. (In a recentmodel by Melosh[1979],resistanceto the movingfault surfaceis reducedby acousticoscillations.In our analysisof energetics,the acoustic energyabsorbedin the vicinityof the fault will contributeto the fault resistance, and any remainder,to the radiatedenergy.)Despiteits importanceto the understandingof earthquakesand crustal stress,the magnitude of this resistanceis unknown. Some of the work done against fault resistancemust be allocated to the productionof heat through varioussmall-scaleinelastic frictional processes;the rest could be consumed by other energy sinks such as metamorphic and chemical reactions or by the surfaceenergy associatedwith the creation of new fractures and flaws. Although the nonthermal portions are generally expected to be small, they probably should be identified and evaluated systematically.In this section, we shall attempt to evaluate only the effects of surface energy. We shall then discussthe theory necessaryto evaluate the R• -• 3• B b u (7) Accordingto a careful studyby Braceand Walsh[1962],/Jis in the range200-2000erg/cm2 for a wide varietyof common rock-formingminerals.Taking b •- 1 micron(10-4 cm) and/J '• l03 erg/cm2, (7)yields R• ,-3x107 B(c-•) =30 B[bars](8) u u Hence for thesevaluesthe resistingstressallocatedto fracture will not exceed30 barsunlessthe ratio of gougewidth to fault displacement exceeds -• 10ø. For the San Andreas fault, this ratio is probably•<10-2 (B •< 1 km, u •> 100km), whichseems to allow us to neglect R• with some room for error in the choices of b and/J. It is important,to distinguishbetween surfaceenergy and the 'fracture energy' used in models of macroscopicfracture; most of the fracture energy ultimately appears as heat. In studies reported by McGarr et al. [1979], specific 'crushing energies'of about 5 x 105 erg/cm2 were determinedfrom crushingtestson quartzite. They emphasize,however,that it LACHENBRUCH AND SASS.'ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6187 is not known how the work of crushing was partitioned among surface energy, heat, and elastic waves. It might be arguedthat over periodsof millionsof yearsthe gougemight be repeatedlyhealed and refractured.Although this could be important to the energeticsof individual events, it seemslikely that the healing processwould releasethe surface energy as heat, and no long-term cumulative storageas surfaceenergywould result.In what follows,we shall neglect the contribution of surfaceenergyand use the symbolsr and R interchangeably. Thermal Effects Equation (3) formsthe basisfor attemptsto estimateresisting stress(R • r) from measurementsof surface heat flow [Henyey, 1968; Brune et al., 1969; Henyey and Wasserburg, 1971;Lachenbruchand $ass, 1973;$cholz et al., 1979]. In this section,we presentresultsfor heat flow and fault temperature that will be useful backgroundfor the discussionsto follow. We considerthe thermal effectsof frictional heating by a vertical strikeslip fault in the planey = 0 in the half-spacex > 0 (Figure 1). A frictional heat sourceof strength Q(x) = 2w-(x) 3 2 1 14 -<-----y/x• 28 42 y(krn)-•->- 0(•x2/K)-' _ Ol ,••,c. 0• 02 x x¾ _b (9) 2• startsat time t = 0, when the fault originates.From that time 46• onward, the averageslip velocity 2v and averagedissipative O(øC) for• =1 HFU resistancer(x) are both assumedto be independentof time. Fig. 3. Surfaceheat flow q (part a) and fault plane temperature8 Consequently,the rate of heat generationdoesnot, on the average, change with time. The earth's surface x - 0 is main- (part c) for sourcestrengthdistributionshownin Figure 3b. Dimensional resultsare for x: = 14 km, K = 6 mcal/cm s øC, and X = 1.6 tained at zero temperature,the medium is homogeneouswith thermal conductivityK - 0.006 cal/cm s øC (•2.5 W m-' K-'), and thermal diffusivitya = 0.01 cm:/s. Thermal con- for surfaceheat flow and fault plane temperatureare given for duction is the only mode of heat transfer,and the fault is contwo cases:model 1, a linear increasein Q(x) from zero at the sidered to be infinitely extended in the direction persurfaceto Q* at x = x2, with zero sourcestrengthelsewhere pendicularto the x, y plane. In Appendix A, analytical results (seeFigure 2b), and model 2, a sourceof uniform strengthQc betweendepthsx, and x2, with zero strengthelsewhere(Figures3b and 4b). (For model2, the transienttemperaturedistribution throughoutthe entire medium is given, equation (A8).) Resultsfor thesemodelscan be superimposedto describethe effectsof any sourcedistribution Q(x) composedof step and linear functions;a simpleusefulexampleis model3, Figure 5b. In Figures 2-5, the resultsare normalizedby Q, the average rate of frictional heat production between the surfaceand x2, and q representsheat flow at the earth's surface. The well- O8 O6 a 04 known resultsfor steadystateheat flow (appendixA, (A22b) and (A23b)) are representedby the curveslabeled t = oo in Figures2a, 3a, 4a, and 5a. The other curvesin partsa of these figures represent transient states labeled in dimensionless 0 . o• •p% 14 28 quantitieson the left, and dimensionaloneson the right (using x2 = 14 km, a = 0.01 cm2/s).The time constantis 2• = (X2)2/4ot; it is about1.6m.y. for x2 -- 14km, or 0.8 m.y. for x2 42 y(km)-•>- O(Ox:/K)-' õ 0,2 • o• _o?__03 _Oio = ß t(my) 0•"" = ! 0o 230 46o 690 920 o(øc)forfi--1 HFU Fig. 2. Surfaceheat flow q (part a) and fault plane temperature (partc) for a linearincrease in sourcestrength to depthx2 (part b). Dimensional resultsare for x2 = 14 km, K = 6 mcal/cm s øC, and X (x:,):'/4a= 1.6m.y. 10 km. Such depthsare comparableto the thicknessof the seismogenic layer for continentalstrike slip faults.Within this layer we often have reasonable assurance from microseismicitythat a substantialamountof slip occursin a narrow zone [e.g., Eaton et al., 1970].Beneaththe seismogeniclayer, the mode of mechanicaldeformation and heat generationis more obscure,but fortunately,the effectof suchdeep sources on the local heat flow anomalyis relativelyunimportant.For example,increasingthe depthx2 of heat productionfrom 14 km to 28 km in the model of Figure 4b would increasethe maximum heat flow anomaly by only 15% after 10 m.y. of faulting (A23a), and near the fault the general form of the anomaly would be changedlittle. 6188 LACHENBRUCHAND SASS:ENERGETICSOF THE SAN ANDREAS FAULT ZONE what fraction of the relative plate motion can reasonablybe assignedto a slip zone whosewidth is small in relation to the thicknessof the seismogeniclayer. According to Atwater and Molnar [1973], the relative right lateral motion between the Pacific and North American platesaveraged4 cm?yr between 10 and 4.5 m.y. ago, and 5.5 cm/yr thereafter. Graham and Dickinson[1978] believethat most of the motion was probably concentratedon the San Andreas fault proper for the past 6 m.y. Analysis of geodeticdata by Thatcher[1979a] [seealso Savageand Burford, 1973]impliesthat for the last century,localized displacementon the creepingportion of the San Andreas fault in central California probably averaged3 cm?yr, and Sieh [1977] estimatesan average localized slip rate of o ! 3 2 0 14 28 42 about 31/2to 4 cm?yr over the past 3400 yr in the presently • y/x• y(km) locked portion of the San Andreas fault in the Carrizo Plain e(Ox•/K)-' at the southernend of the Coast Ranges.At the latitude of the 0 •,Oc 0 01 02 03 04 Salton trough in southern California, Savage et al. [1979] found that right lateral motion -,5 cm?yr was distributedover a 120-km zone including the San Andreas, San Jacinto, and x Elsinore faults during the past 5 yr, and Thatcher [1979b] found a similar broad region of strain accumulationin southb • 28 ern California over the portion of the San Andreasfault zone that boundsthe Mojave Block. These and other observations suggestthat the presentrate of plate motion is typical of the 3 3•2 e(oC) forõ=1HFU averageover the last 10 m.y., -,5 cm?yr,and that the local slip Fig. 4. Surfaceheat flow q (part a) and fault plane temperature 8 on the San Andreas fault presently might range downward (part c) for sourcestrengthdistributionshownin Figure 4b. Dimen- from perhaps4 cm?yr in parts of the Coast Rangesto subsional resultsare for x2 = 14 km, K = 6 mcal/cm s øC, and 3, = 1.6 stantially smaller valuesin southernCalifornia. m.y. The disparitybetweeninterplateslip and fault slip suggests that the problem of mechanicallygeneratedheat flow in the The temperaturerise on the fault plane is shownin parts c interplate shear zone has two parts: (1) that which can be of Figures2-5. On the upper scale,they are normalizedby the identified with frictional resistanceon a thin fault, and (2) that 'scale temperature,' Qx2/K, which is the one-dimensional associatedwith more broadly distributedinelasticdissipation steady temperature that would result at the base of the ø I s_eismogenic layerfroma horizontal planesource of strength Q there. On the bottom abscissalscalethe temperaturesare shownin degrees centigrade for 0 -- 1 HFU -- 41.8mW/m• (and X2 •---14 km, K -- 6 x 10-3 cal/øC cm s). A usefulcon- 08 version for frictional heat generationrate is I HFU--41.8 mW/m: • 1.32 kbar x I cm/yr(10i 06 Asa ruleofthumb, reasonable models forthevertical distil- • bution of frictional heating in the seismogenic zone yield a 04 maximu_m temperature riseof 30-40%of thescaletemperature(Qx2/K);suchtemperatures areapproached in twoor 0: three time constants,typically a few million years. v, Similarly, inspection ofpartsa ofFigures 2-5suggests that the maximumlong-termheat flow anomalyq(y -- O, t ----oo) 0 3 over a narrow strike slip fault is typically 60-80% of the aver- agerate of frictionalheatingQ for reasonable verticaldistributions Q(x) in the seismogenic zone. Thesevalues,too, are generally approachedin a few million years;they fall off by an order of magnitude at horizontal distancesof about the thicknessof the seismogeniclayer. If appreciablefrictional heatingextendsto within a few kilometersof the surface(Figures 2 and 3), a substantialfraction of the anomaly will develop in a few hundredthousandyears,and the anomalywill be sharply peaked. For wider fault zones,the peaks of the heat flow curves would be subdued somewhat, but this effect 2 1 < 0 y/x• 0[.,.,•! o,q• 28 42 y(km) • e(Ox•/K)-' 0 x! 14 01 02 03.•.•4 x• x [- • M'ode, 3 '• '• 3 23 o 46 o 69 o 920 O(øC) for•--1 HFU is not importantas long as the fault width is small in relation Fig. 5. Surfaceheat flow q (part a) and fault plane temperature0 to the thicknessof the seismiclayer. (part c) for sourcestrengthdistributionshownin Figure 5b. DimenA major problem in applyingmodelsof an infinitesimally sional resultsare for x2 -- 14 km, K -- 6 mcal/cm s øC, and 3, -- 1.6 thin strike slip fault to frictional heating lies in determining m.y. LACHENBRUCHAND SASS:ENERGETICSOF THE SAN ANDREASFAULT ZONE 6189 I0o '7- I10 o I00' ,;50.. WYOMIN t L, HEAT FLOW rnWrn-2 HFU IO4- ••] >2.5 63-104-• 1.5-2.5 $1--65 ] ] O.75--1.5 0 ;•00 I i 600 i KILOMETERS, e, <0 Fig. 6. A contourmapof heatflowin thewesternUnitedStates.Abbreviations: BMH, BattleMountainhigh;EL, Eureka Low; LV, LongValley volcaniccenter;SRP, easternand centralSnakeRiver Plain;Y, Yellowstonethermalarea;and RGR, Rio Grande Rift. in the body of the seismogeniclayer or beneath it. If the former is appreciable, it should be associatedwith a local pare the observedheat flow with a referencemodel obtained from Figure2 by selectingfault slipvelocities(2v) of 2 cm/yr anomalyover the fault, accordingto the modelswe havejust and 4 cm/yr. The linear increaseof Q(x) with depth in this presented;the latter may be associated with a broaderthermal model correspondsto a simplefrictional law in which the efanomaly [seeLachenbruchand Sass,1973]. fectivenormalstressduringfault movementis proportionalto Although both•artsof theproblem will bediscussed, at overburden pressure. We selectQ = ,r/2 HFU sothatthe refpresent,we are interestedin limits to fault frictionimplied by erence model yields a convenientmaximum heat flow anomlocal heat flow observations.For this purpose,we shall com- aly of I HFU (A23b), and we shallassumethat the depthof 6190 LACHENBRUCH AND SASS' ENERGETICS OF THE SAN ANDREAS FAULT ZONE TABLE 1. Summaryof Heat Flow and Heat ProductionValuesNear the San AndreasFault Heat Flow Designation FS-STH FS-AZI FS-AZ2 SB-AZ3 FS-LB 1 FS-ACI MB-YVW FS-SB2 FS-SB3 FS-SB4 TR-PSA TR-PSB MB-PSC FS-LH3 MB-PSD FS-LH2 MB-PSE MB-HVI FS-LHI GF-TEI GF-TE2 GF-TE3 GF-TE4 GF-TE5 GF-TE6 GF-TE7 MB-PSF Locality Reference 2 Latitude Longitude HFU mW/m2. SaltonTrough CO 71 32-48. 115-15. 3.2 ANZA A-I ANZA A-3 ANZA A-2 L.A. Basin LB-I Santa Ana AC-I HW 71 HW 71 HW 71 US 71 US 71 33-30. 33-32. 33-32. 33-53. 33-58. 116-36. 116-36. 116-48. 118-02. 117-38. 1.87 1.76 1.46 1.74 1.60 78.3 73.7 61.1 72.8 67.0 Yucca Valley Water US 79 34-13.5 116-24.2 1.12 46.9 San Bernardino SB-2 San Bernardino SB-10 San Bernardino SB-5 Palmdale Stress A Palmdale StressB Palmdale StressC HW 71 HW 71 HW 71 US 79 US 79 US 79 34-15. 34-15. 34-16. 34-25.6 34-28.1 34-33.2 117-19. 117-20. 117-20. 117-51.8 117-51.2 117-42.9 1.63 1.58 1.08 1.58 1.60 1.51 68.2 66.1 45.2 66.1 67.0 63.2 118-29. 1.68 70.3 1.57 65.7 4.9 2.1 Lake HughesLH-2 HW 71 34-41. 118-26. 1.56 65.3 3.4 Palmdale StressE Hi Vista US 79 US 79 34-43.9 34-43.9 117-4 1.7 117-41.7 1.64 1.60 68.7 67.0 6.2 6.2 1.4 2.6 2.6 Lake HughesLH-I Tehachapi Mt. DH-15A TehachapiMt. DH-70 TehachapiMt. DH-14 Tehachapi Mt. DH-43 Tejon Ran. DH-43 Tehachapi Mt. DH-65, 67 Tejon Ran. DH-61, 65, 67 HW 7I HW 71 HW 71 HW 71 HW 71 US 71 HW 71 US 71 34-44. 34-51. 34-52. 34-52. 34-53. 34-53. 34-56. 34-56. 118-24. 118-44. 118-45. 118-45. 118-46. 118-46. 118-49. 118-49. 1.72 1.48 2.21 2.03 2.02 1.83 1.30 1.36 72.0 62.0 92.5 85.0 84.6 76.6 54.4 56.9 8.7 3.6 1.64 68.7 1.93 80.8 FE-EHI Elk Elk Elk Elk Hills 382-3G Hills 343-4G Hills 344-35S Hills 372-35R Elk Hills 326-28R Elk Hills 385-24Z Elk Hills 366-24Z Temblor La Panza TS-I West of Bakersfield US US US US US US US US US BE 71 71 71 71 35-16. 35-16. 35-17. 35-17. 119-23. 119-24. 119-22. 119-28. 1.26 1.12 1.20 1.30 71 71 71 79 35-17. 35-18. 35-18. 35-21.3 119-31. 119-33. 119-34. 119-49.7 71 47 35-26. 35-28. 120-30. 119-45. 1.26 1.20 1.00 1.45 2.21 1.29 52.7 46.9 50.2 54.4 52.7 50.2 41.9 60.7 92.5 54.0 FE-VCS FW-BHR FE-PRM FE-SEI FE-MPI FE-DMI FE-DEI FE-RCE FE-MSL 1.6 1.0 1.3 1.1 117-50.8 119-26.3 FE-PT0 FE-PT9 FE-PT2 FE-PT3 FE-PT6 FE-PT5 FW-BLM 1.5 34-39. 117-45.7 FW-HT6 FW-HT2 FW-HT7 FE-PTI 3.7 34-39.1 35-04.2 FW-HT4 2.7 HW 71 34-56.1 FW-HTI 6.5 US 79 US 79 FW-MOP 1.5 1.2 0.9 Lake HughesLH-3 US 79 FW-STC 3.6 2.8 2.2 Palmdale StressD Palmdale Stress F FW-PR3 FW-PR4 FW-HT3 FW-HT5 #W/m 3 3.8 2.5 3.0 2.6 Maricopa FW-USL FW-PR2 HGU 134. FE-MLI FE-EH2 FE-EH3 FE-EH4 FE-EH5 FE-EH6 FE-EH7 FE-TLI FW-TSI FE-WB 1 FW-HAR FW-PAT FW-PRI Heat Production 7.7 3.2 2.1 0.9 1.1 0.5 5.7 2.4 5.4 2.3 Harmony US 79 35-29.5 120-58.7 1.80 75.3 2.7 1.1 Patterson et al. PRC-2 USL 1-3 PRC-7 PRC-12 PRC-19 Hollister HO-3 Hollister HO-5 US US US US US US HE HE 79 79 79 79 79 79 68 68 35-55.5 36-01.9 36-02.9 36-03.0 36-03.9 36-05.7 36-32. 36-35. 121-01.7 120-51.6 120-46.6 120-48.7 120-46.0 120-42.5 121-40. 121-27. 2.3 2.0 2.25 2.0 2.4 2.1 1.20 1.90 96. 84. 94. 84. 100. 88. 50.2 79.5 4.1 1.7 2.8 1.2 Monterey Stone Canyon US 79 US 79 36-36.3 36-38.4 121-54.9 121-15.5 1.56 1.79 65.3 74.9 3.8 1.6 3.5 1.5 Hollister HO- 1 Hollister HO-4 Hollister HO-6 Hollister HO-2 & 7 Hollister HO-8 Pacheco Tunnel 208 Pacheco Tunnel 210 Pacheco Tunnel 219 Pacheco Tunnel 212 Pacheco Tunnel 213 Pacheco Tunnel 216 Pacheco Tunnel 215 Ben Lomond HW 71 HE 68 HE 68 HE 68 HE 68 US 79 US 79 US 79 US 79 US 79 US 79 US 79 US 79 36-43. 36-48. 36-50. 36-53. 36-55. 37-01.6 37-01.7 37-02.6 37-02.8 37-03.0 37-03.2 37-03.3 37-06.9 121-24. 121-20. 121-17. 121-35. 120-58. 121-18.7 121-18.4 121-15.1 121-14.2 121-12.8 121-11.3 121-10.8 122-09.1 1.71 2.30 2.30 1.70 1.40 2.03 2.12 2.14 1.90 2.10 1.60 1.63 1.61 71.6 96.3 96.3 71.2 58.6 85.0 88.7 89.6 79.5 87.9 67.0 68.3 67.4 3.4 1.4 3.1 3.1 3.1 3.1 3.1 3.1 3.1 2.2 1.3 1.3 1.3 1.3 1.3 1.3 1.3 0.9 Valley Christ. School US 79 37-14.3 122-04.1 2.15 90. Pescadero BHR-I Permanente US 79 US 71 37-14.7 37-19. 122-23.7 122-07. 2.53 2.20 105.9 92.1 4.6 1.9 3.2 1.3 SunnyvaleC-3 US 71 37-27. 122-02. 2.02 84.6 Melno Park MP-1 US 68 37-27. 122-10. 2.16 90.4 Dumbarton S.F. Bay US 71 37-29. 122-08. 2.25 94.2 DumbartonEarthquakeI US 79 '•7-30.6 122-07.8 2.03 85.0 Redwood City Exp. May School US 79 US 79 37-30.8 37-44.3 122-12.5 121-45.2 1.85 1.52 77. 63.6 LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREASFAULT ZONE 6191 TABLE 1. (continued) Heat Flow Designation • Locality Reference 2 Latitude Longitude HFU mW/m2 FE-TRI TracyDH-2 US 71 37-48. 121-35. 0.96 40.2 FE-DA 1 Danville US 79 37-48.1 121-56.5 1.28 53. FE-MST BerkeleyMSTW US 71 37-52. 122-15. 2.00 83.7 FE-NIC Nicasio US 79 38-05.5 122-44.9 1.84 77.0 FW-PTR PointReyes US 79 38-05.7 122-53.5 1.91 80.0 FE-SEA Sea Ranch US 79 38-42.0 123-25.1 2.2 92. FE-ANP Annapolis US 79 38-43.1 123-20.8 2.16 90. FE-CL 1 Cloverdale US 79 38-46. 122-58. 3.50 146.5 FE-GYS 3 Holesat the Geysers US 79 38-48. 122-50. 10.00 418.6 FE-GUI FE-COL FW-PTA FE-BVL FE-UKI Guinda Colusa Basin Pt. Arena Booneville Ukiah US US US US US 38-50.4 38-53.0 38-55.7 38-59.6 39-03.4 122-12.0 121-53.7 123-42.6 123-20.8 123-09.1 0.90 0.80 1.44 1.27 1.78 37.7 33.5 60.3 53. 74. 79 79 79 79 79 FE-PVY PotterValley US 79 39-20.4 123-06.1 1.8 76. FE-NOY NoyoHill US79 39-24.5 123-44.•8 1.8 75. FE-WIL CR-EG7 CR-EG8 Willitts EC- 1 Cottonwood Glade EG-7 Cold Creek EG-8 US 71 US 71 US 71 39-34. 39-42. 39-42. 123-07. 122-48. 122-53. 1.85 1.20 1.50 77.4 50.2 62.8 FE-LYN Laytonville US 79 39-43.8 123-30.1 1.42 60. FE-CLN FE-LMR FE-SCV FE-GBL Covelo North Lake Mtn. Ranch Shelter Cove Garberville US US US US 39-50.1 39-56.8 40-01.3 40-05.5 123-09.9 123-21.1 124-04.0 123-48.0 1.40 0.66 1.39 1.03 59. 28 58. 43. FE-EBG FE-KET Ettersburg Kettenpom US 79 US 79 40-08.1 40-08.3 124-00.0 123-24.6 1.04 1.2 43. 50. FE-FTS Fort Seward US 79 40-13.0 123-38.4 1.44 60. FE-HON Honeydew US 79 40-14.4 124-07.6 1.8 75. FE-PRA Petrolia US 79 40-19.4 124-16.5 1.82 76. 79 79 79 79 FE-GZB Grizzly Bluff US 79 40-32.8 124-10.6 1.31 55. FE-NOS FE-EUR Nav. Ocean. Sta. Eureka B-69 US 79 US 79 40-33.8 40-43.6 124-21.2 124-12.8 0.82 0.83 34. 34.7 FE-KSM King Salmon US 79 40-44.4 124-12.8 0.92 38. FE-BLK Blue Lake US 79 40-52.6 123-58.8 1.18 49. Heat Production HGU •W/m 3 2.9 5.0 2.1 1.2 0.2 0.1 2.8 2.4 1.2 1.0 ISubregions: FS, FaultSouth;SB,Southern CaliforniaBatholith;MB, MojaveBlock;TR, Transverse Ranges; GF, GarlockFault;FE, Fault East; FW, Fault West; CR, Coast Ranges. 2CO71, Combs [1971];HW 71,HenyeyandWasserburg [1971];US 71,Sassetal. [1971];US 79,present paper;BE 47,Benfield[1947]; HE 68, Henyey[1968];US 68, Sasset al. [1968]. the seismogeniclayer (x2) is 14 km. The reference model thereforerepresentsthe following conditions: q---- 1 HFU--41.8mW/m 2 y--0 Q/2v -- Zavg • 518bars = 1035bars t-- oo (11a) 2v = 4 cm/yr (1lb) 2v = 2 cm/yr (11c) The scaletemperaturefor use in Figure 2c is Qx2/K = 365øC (12a) and the total rate of heat productionper unit length of fault is trated in Figure 6. The westernUnited Statesis characterized generallyby highand variableheatflow,althoughit alsocontainslargeareasof low-to-normalheatflow. High heat flow in regionsof extensionaltectonicslike the Basin and Range provincecan be accountedfor by the upward massflow required for plausiblerates of lithosphereextension[Lachenbruch and Sass, 1978; Lachenbruch,1978]. Some zones of low heat flow like the continentalmargin of the PacificNorthwest may be related to heat sinksassociatedwith subduction;others, like the 'Eureka Low' (EL, Figure 6), may be related to systematicregionalhydrologicphenomena[Lachenbruchand Sass, 1977]. Qx2 -- 2.2 cal/cm s = 0.92 Mw/km (12b) (12c) The part of the California Coastalregion associatedwith the transformboundarybetweenthe North American and Pacific plates has a range of heat flows typical of most other Accordingto theseresults,if the averageslip velocitywere partsof the westernUnited States(Figure 6). This includes 2-4 cm?yr,we shouldexpectto seea local heat flow anomaly measurements offshore in the Southern California Borderland of 1.0-0.2 HFU over the fault trace for each 100 bars of fric[Lee and Henyey, 1975]. tional resistance,unlessthe fault trace were very young or the In this paper, we shall examinea data set (Table 1, Figure frictional heat had been removed or redistributed by hydro7) that includes:the firstheat flow determinationin California logicprocesses. In the next two sections,we shall examinethe (WB 1, Table 1, Benfield[1947]),subsequent measurements by data on heat flow and thermal springsin an attempt to obtain Combs[ 1971],Henyey [ 1968],Henyeyand Wasserburg[ 1971], a limit to frictional resitancefrom this point of view. Sass et al. [1971], Roy et al. [1972], USGS data, which until HEAT FLOW now have receivedonly preliminary mention in the literature The status of published heat flow measurementsin the [Lachenbruchand Sass, 1973, 1977, 1978], and results from westernUnited Statesas of late 1979 [Sasset al., 1980] is i!!us- somerecentdrilling in the Northern California Coast Ranges. 6192 LACHENBRUCH AND SASS: ENERGETICS 35,•'49 OF THE SAN ANDREAS FAULT PTI 38ON 3•#EUR .53 '55 120ow .40 ' 64 •76.75 .60 •.43. .50 x -F _• .59 122•W • .60 63..50 ' Hayward-Calaveras 80-90 •"68 Palo Colorado ,• \ San Gregorio 96 ø59 72 75 65 . ZONE • .74 / No 84-100 36øS+ 96 o 122ow ø54 '-- 50 0 50 1O0 150 \•,•.oR, 42-54 200 KILOMETERS 't•154••., 5. 4S2'54.•ock 81 MA B••62 • 7. 85_••72 _,. .69 72 66..68 7•65 + 35øN 116øW \ Fig. 7. Heatflow(mW/m2) in relationto majorrecentlyor currentlyactivestrikeslipfaultsin California:a, Eureka (EUR) to PointReyes(PTR); b, PointReyesto Maricopa(MA); c, Maricopato SaltonTroughhydrothermal areas(STH) (1 HFU = 41.8mW/m:). Data from the latter two categories(designated'US 79' in Table 1) shouldbe consideredpreliminary and are subjectto revisionafter more detailed study. Severalheat flow determinations within a small area of the Salton Trough [Combs, 1971] have been averagedand are representedas a single point (STH) in Table 1 and the variousillustrations.The entire data setis composedof 106independentdeterminationsat 103 distinctsites(Table 1, Figure 7) over a broad (•100 km wide) area including the main trace of the San Andreas fault. For most of our statistical analyses,we have excluded data from north of Cape Mendocino (region la, Figure 8), the Great Central Valley (EH1 through EH7, MSL, TR1, DA1, HT7, WB1, GUI, and COL, Table 1), the Geysers region (GYS and CL1, Table 1), and the Salton Sea Area (STH, Table 1), leaving a core of 81 heat flow determinationsfrom the Northern and Central Coast Ranges, the Transverse Ranges,and the southernpart of the San Andreas fault zone. Sites more than 100 km from the fault zone in the California borderland [Lee and Henyey, 1975] and the Mojave Block [Lachenbruchet al., 1978] were omitted from our analysis. LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE Heat flow values are shown in relation to the San Andreas 35 N and other recentlyor currently activefault zonesin Figure 7. In Figure 8, the locationsof all sitesare illustratedin relation to the main trace of the San Andreas fault and to a number of 6193 3O subregionsto be discussed below. The core of 81 'fault zone' determinationsform the basisfor Figures 9, 11, and 12 and the associatedstatisticalanalyses. The histogramof Figure 9 indicatesa nearly normal distribution of heat flow, with a mean of 1.75 + 0.04 HFU (72 + 2 25 mW m-'), significantly lowerthantheBasinandRangemean 20 =81 HEAN = SD = 0.36 1 .75 SE = 0.04 HFU of 2.1 HFU [Lachenbruchand Sass, 1977, 1978] but com- parableto the mean of 70 valuesfor the SouthernCalifornia Borderlanddiscussed by Lee and Henyey[1975].An examination of the heat flowreheat productiondata (Figure 10) suggeststhat despitea large amount of scatter,the mean 'reduced'heat flow qr may be comparableto that from the Basin and Range province [see Figure 9-3, Lachenbruchand Sass, 1978].However, there is evidentlyno linear relation between heat flow and heat production,as is found farther east in the Sierra Nevada batholith. The heat flow as a function of distance normal 15 10 to the fault is shownfor the 81 data of Figure 9 in Figure 11. Shown also 2 1 is the pattern of curves representingthe reference anomaly q ,HFU discussed in the last section.As in Figure 2, the top curve representsthe steady state, and the transient curves represent Fig. 9. Histogram of 81 values of heat flow in the San Andreas conditionsat 8 m.y., 2.4 m.y., 0.8 m.y., and 0.3 m.y. after the fault zone (Figures 7 and 8) (excluding hydrothermal areas, STH, start of faulting. It is seenfrom Figure 11 that even the steady GYS, CL1, Table 1, and data from the Great Valley and north of Cape Mendocino(region lb)). state anomaly (representinga frictional resistanceof 1 kbar for 2v = 2 cm/yr or 0.5 kbar for 2v -- 4 cm/yr) might go almostundetectedin suchwidely scattereddata. However, Fig- jected perpendicularlyonto the San Andreasfault and plotted ure 11 obscuresa certain amount of systematicvariation. This as a function of distancefrom Cape Mendocino along the can be seenfrom Figure 12 in which the heat flows are pro- fault trace to a distance of 1000 km, where the surface expressionof the fault disappearsin the vicinity of the Salton Sea.The profile of Figure 12 has beendivided somewhatarbitrarily into subregions (seeFigure 8 and Table 2) for the purpose of discussion.Although Figure 12 has a great deal of scatter,it showsthat north of the MendocinoTriple Junction, wherethe San Andreasfault doesnot exist(region l a), the average heat flow is low, about 1 HFU. Southward from the Triple Junction, the heat flow seemsto increaseover a distance of about 200 km (regions lb and 2) to averagevalues closeto 2 HFU, characteristicof the remaining550 km of the California Coast Ranges(regions3-6). Heat flow is significantly lower (• 1.6 HFU) farther southalongthe 150-kmfault segmentthat boundsthe Mojave Block (region 7). South of region 7, data are inadequateto indicate the relation of heat flow to the fault; however, the few data available are included in Figure 12. The high values from STH near the Salton Trough and CL1 near the Geysersare shownin Figure 12 for 1250 350+ reference. 0 500 KILOMETERS Fig. 8. Map of California showingthe main trace of the San Andreasfault, heat flow control,and the regionsdepictedin Figures1319. Stippled area indicatesthe approximateextent of the Great Central Valley. In Figures 13-19, heat flow in each of the seven regions (Figure 8) is shown(with the referenceanomaly)as a function of distancefrom the fault trace.A casualinspectionof the individual profilesrevealsthat there is no evidencefor a thermal anomalydue to local frictional heatingat any latitude;if such heatingdoesoccur,its magnitudemustbe much smaller(<<1 HFU) than that representedby the referenceanomaly. Attemptingto decidejust how much local heating at the fault tracecouldgo undetectedin an individualprofile probablyis more a matter for judgment than for statisticsbecauseof the uneven areal coverage,the variability from site to site in the quality of temperature and thermal conductivity data, and suchlocally variable conditionsas hole depth, structure,hy- 6194 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE 4 4- Sedimentary andmetamorphicrocks 13 Granitic rock - 131313 a• a a 0 0 I I 2 I I 4 ,I I 6 I I 8 I , 10 Heatgeneration Ao,HGU Fig. 10. Heat flow versusnear-surface heat productionfor the San Andreasfault zone.Referencecurvesare for Basin and Range province(upper) and easternUnited States(lower) [Roy et al., 1968]. the condrology, and microclimate. Making allowance for some of tach suchprecisionto any heat flow measurements, that a local anomalydoesnot exthesedetails,we judge that a local anomalygreaterthan 0.2 sistencyof the data suggests HFU is probablyprecludedin the CoastRanges.Scalingfrom ist even at the 0.1 HFU level. However, this is the region in our steadystatereferenceanomaly,this would imply a maxi- which Thatcher[1979b]found geodeticevidencefor a broad mum frictional resistanceof 50-70 bars assumingthe present- shearzone, and the local slip velocitymay well be only 2 cm/ day rate of fault slip of 3-4 cm/yr; allowinga bit for dis- yr or less.Furthermore, the tectonicsof the region suggests equilibriumandfinitefaultwidth,a figureof-* 100barswould that the main fault trace could have shiftedpositionin the last seem reasonable. few million years.Hence even at this location,the thermal efSome numericalsupportfor a figure of this magnitudecan fectsof frictional resistanceas large as 100barsmight possibly be obtained if we consider the 39 heat flow determinations go undetected. The horizontal lines denoting mean heat flow valuesin reusedto calculatemeansfor regions3-6 (Figure 12), the portion of the CoastRangesin which the fault zone has probably gions 3, 4, and 6 (Figures 15, 16, and 18) terminate between been in existenceon both platesfor severalmillion years.It is the easternmostdrainage divide of the Coast Rangesand the seenfrom thesedata (Figures20a and 20b) that the difference Central Valley. As in the caseof heat flow north of the San betweenthe mean heat flow within 10 km of the fault (Figure Andreas fault zone (i.e., north of Cape Mendocino, Figure 20a) and beyond10 km from the fault (Figure20b) is quite in- 12), the valuesto its eastin the Great Valley fall off abruptly. significant(0.01 HFU). Accordingto a two-sampleT test for (For discussion,seeLachenbruchand Sass[1973].) this data set, the mean heat flow within 10 km of the fault difIn summary, the California Coast Ranges representa refers from the mean heat flow beyond 10 km from the fault by gion of generally high but variable heat flow enclosingthe less than 0.2 HFU at the 95% confidence limit. landward portion of the San Andreas fault zone. There is no The remarkably uniform values in the Mojave segmentof evidencethat the heat flow is greaternear the fault trace than the fault (region7, Figure 19) deservespecialcomment.(Fol- in the surroundingterranesover a 1000-kmdistancealong the lowing Henyey [1968] and Henyeyand Wasserburg[1971], we fault trace southwardfrom Cape Mendocino.With reasonable exclude the single low value, SB4, Table 1, becauseof sus- assumptionsfor fault slip rate and age and assumingconpectedhydrologicdisturbances.)The total range is 1.51-1.74 duction is the only mode of heat transfer, this result suggests HFU, and the means and standard deviations(Table 2, Fig- that the average dissipativeresistanceto motion on the fault ures20c and 20d) are indistinguishablebetweendata setsob- does not exceed --•100 bars. The result supportsconclusions tained within and beyond 10 km from the fault. In this region, from earlier studies[Henyey, 1968; Henyey and Wasserburg, the two-sampleT test yields a differencein mean heat flows 1971; Brune et al., 1969; Lachenbruch and Sass, 1973]. We within and beyond 10 km from the fault of less than 0.08 shall return to the problem of the broad Coast Range anomHFU at the 95% confidencelimit. Although it is risky to at- aly later; in the next section, we shall investigatethe possi- LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE I PACIFIC PLATE i I 6195 i 1 ...... NORTH AMERICAN PLATE - 125 •:2.0 1 [] MEAN 84• [] [] [] - 42 -- I 100 i 80 0 100 Fig. 11. Heatflowasa functionof thedistance fromthemainfaulttracefor 81pointsof Figure9. Patternof curvesis referenceanomalyfrom Figure 2a (see(1 l) and (12)). bility that appreciable amounts of frictional heat might have been generated and subsequently removed by hydrologic the fault zone is removed by water flowing through relatively permeablerocks,and it is dischargedby hot springs. The distribution of thermal springs(those with a discharge temperature >15øF above mean annual air temperature) is HYDROLOGIC TRANSPORT AND THE CONSTRAINT shown in Figure 21 (modified from Figure 8 of Waring ON FRICTIONAL HEATING [1965]). Along the main trace of the fault, spring activity is We have seen that if heat transfer in the fault zone has been very rare [see also Jennings,1975]. Within the Coast, Transpredominantly by conduction,then the absenceof a peaked verse, and Peninsular Ranges,there is a scatteringof thermal heat flow anomaly suggests that the averagestressallocatedto springs,but apart from thoseclearly associated with the Geyfrictional heating on the fault is small (•<100 bars). If, how- sers geothermal zone north of San Francisco, their temperever, substantial amounts of frictional heat were transferred aturesare not high, and their flow rates are small, typically of convectivelyby moving ground water, much larger frictional the order of 50 l/min or less.Many of the springsin southern stresseswould be permissible.There are two possibilitiesfor California occurcloseto currently or recently active strike slip removal of the hypotheticalundetectedfrictional heat: (1) it faults (for example,Elsinore,San Juan, Garlock, San Jacinto) was dischargedat the surfaceby thermal springs,or (2) it was [Jennings,1975],but north of the San Bernardino Mountains redistributeduniformly over a broad zone (--•100 km wide) by (i.e., north of the southernlimit of region 7, Figures 8 and 11), no thermal springsare reported preciselyon the main trace of someunspecifiedhydrologicprocess. processes. the San Andreas Dischargeby ThermalSprings fault. In Table 3 we have compiled data from all of the thermal In attempting toreconcile theapparent lackof a peakedspringsthat we could find describedin the literature that lie heat flow anomaly near the San Andreas fault with laboratory evidencefor high mean shear stress(-• 1 kbar), someworkers [e.g., Hanks, 1977;Scholzet al., 1979] have suggestedthat the San Andreas fault zone is analogous to mid-ocean ridges, where the expectedconductiveheat flow anomaly is literally washed out by convection of seawater within permeable rocks. By this analogy, the heat generatedby friction along within 10 km of the 950-km length of the San Andreas fault trace between Cape Mendocino and the southwestboundary of region 7 [Waring, 1965; Renner et al., 1975; Brook et al., 1979; Samreel, 1979]. The total convectivedischargeover this distanceis 1.25 MW, 80% of which is contributed by a spring at the southern edge of the region (Table 3). The reference anomaly of the last section,which correspondsto a frictional 6196 LACHENBRUCH AND SASS.' ENERGETICS OF THE SAN ANDREAS FAULT ZONE TABLE 2. Mean Heat Flows, Standard Deviations, and Standard The possibility remains that the heat was flushed out in Errorsfor the RegionsIllustratedin Figures8 and 13-19 transientpulsesimmediatelyfollowingunstablelocal heating of groundwater duringindividualearthquakes. However,this shouldleadto conspicuous surfaceactivityon a scalenot generally observed,and the model leavesunexplainedthe absenceof a local heat flow anomalyover creepingportionsof Mean Number Heat Standard Standard Region of Data* Flow O, HFU Deviation, HFU Error, HFU I la lb 2 3 4 5 17 5 12 6 14 15 6 1.23 1.01 1.32 1.66 2.06 1.87 2.17 0.33 0.22 0.33 0.24 0.22 0.31 0.17 0.11 0.10 0.09 0.10 0.06 0.08 0.07 the fault. Lateral Redistribution of Heat by HydrologicCirculation If we take the broadheatflow anomalyin the CoastRange to be 0.8 HFU over a width of 84 km (6 timesthe depthof the 6 4 1.85 0.32 0.16 seismogenic layer), the anomalousheat losswould be equiva7 15 1.58 0.15 0.04 lent to the heat generatedlocally by 2 kbar of frictional stress 14'• 1.62 0.06 0.02 on a fault with a slip velocityof 3 cm/yr. A similarresultapplies in southernCalifornia (region 7), where the regional *Excluding sitesin the Great Valley, stippled,Figure 8. '•Excludingsite SB4 (seetext). anomaly is probablybroaderbut lessintense(•0.4 HFU?) [seealso O'Neil and Hanks, 1980].Thus the thermal budget resistance of 500-1000bars(dependingon slipvelocity),repre- would not be violated by very large frictional stressesif a sentsthe generationof almost I MW of frictional heat per kil- mechanismexistedto redistributethe frictional heat broadly ometer of fault length (13b). Hence if such stresseshad been and uniformly. operative, hydrologic convectionwould have had to remove Thermally driven hydrologicconvectionis not a likely canheat at the rate of almost I MW?km to account for the abdidate.The absenceof thermalspringssuggests that modelsof senceof a local heat flow anomaly at the fault. Stated another thermally driven hydrologic convection should have a noway, heat dischargedconvectivelyby the thermal springs mass-flowboundarycondition.In that case,hydrothermalcircould be generatedon the fault by • 1 bar of frictional resis- culation driven by the buoyancyof water heated on a vertical tance. Allowing an order of magnitude increasefor anoma- fault plain would generallyresult in an increase(not a delously large conductiveflux adjacent to the springs[see,e.g., crease) of near-surfaceconductiveflux near the fault trace. Olmsted et al., 1975], discharge by the spring systemsstill Further evidenceagainstthermally driven convectioncomes would not significantlyaffect our estimatedlimits of frictional from a recentstudyby Murata et al. [1979],who concludethat stressbasedon conductiontheory. Furthermore,sincethe dis- 'the presenceof disorderedcristobalitein shaleadjacentto the tribution of springsdoesnot indicate a concentrationof activ- San Andreas fault suggests that movementsalong the fault ity on the San Andreas fault trace, there is no particular rea- have not induceda sustainedrisein temperatueof even a few sonsto supposethat the fault is the sourceof their heat. degreesin the shale.' 4- CLI STH A CALIFORNIA COAST RANGES-- 3.0 '• Mojave Region I i lb • i 2 I I i 3 I • 4- i 4 • I u_ i 5 i I • 'br ++ A &++ I I •1 • _ + A + 84 • + -- A + A 4- 4- 4- 4- 4- A 4- + ++ A <• 10 km fromfault + • 0 I I I I I 500 125 6 'll. 2.0- - I I >10 km from fault I I I 1000 Distance,km Fig. 12. Heat flow,projectedon to the main traceof the SanAndreasfault, as a functionof distancefrom Cape Mendocino(CM, Figure 8). Regionsare as definedin Figure 8. Pointsin Great Valley (stippled,Figure 8) wereexcluded. LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE I PACIFIC PLATE I I i ........ 6197 T NORTH AMERICAN PLATE +North of CapeMendocino (Regionla) REGION 1 [:]Southof CapeMendocino (Regionlb) 1 3.0 o 125 2.0 84 '• 1.32 [] [][] i MEAN lb [] 55 [] + MEAN la 1.0 [] 42 42 + El , 0 r• , 0 I 20 I • , 40 Distance,km I 60 I I 80 I ! 100 Fig. 13. Heat flow versusdistancefrom the main traceof the SanAndreasfault, regionI (Figure 8), shownwith referenceanomaly.Horizontallinesshowmeansfor regionla (north of CapeMendocino)and regionlb (southof Cape Mendocino). Any mechanismdevisedto accountfor a broad uniform redistribution of frictional heat by water movement must contend with the observationthat on oppositesidesof the fault the heat flows are similar, but the hydrologic regimes probably contrastsharply. In the Coast Rangesthe San Andreas fault is bounded on the eastby metamorphicand sedimentary rocksof the Franciscanformation and Great Valley sequence, which accordingto Berry [1973] are characterizedby superhydrostatic pore pressures,approaching lithostatic in many places. On the west the fault is bounded by the Salinian Block, a fractured granitic terrain that probably has an open (hydrostatic)hydrologicregime [I. Barnes,personalcommunication, 1979; see also Brace, 1980]. It is difficult for us to imaginea hydrologicsystemthat might transportheat to both the east and the west for distances of tens of kilometers in such a way as to produce similar heat flows in such contrasting terrains. O'Neil and Hanks [1980]cite depletionof hydrogenand oxygen isotopesin a narrow band of granitic rocks adjacent to the San Andreas fault as evidencefor profound water circulation in the fault zone. It is known that such depletion can result from interactions with ground water, which must penetrate to depths ---10 km to replace water heated during the coolingof a pluton [e.g., Taylor, 1978].Sincethe origin of the granitic rocks predates the origin of the San Andreas fault (---20m.y.B.P.) by about 50 m.y., however,it is not clear that the observed isotope depletion is associatedwith the fault. O'Neil and Hanks estimate that the isotope exchangein this anomalous band probably took place in the temperature range ---100-200øC, temperatures that obtain today in the depth range ---3-6 km. In the absenceof pervasivesurfacehydrothermal activity, for which there is no evidencetoday, the surface from which their samples were obtained must have undergone3-6 km of erosionsincethe isotopeexchangetook place. It follows that unlesserosion of at least 3-6 km occurred at all of their samplesitesduring the last 20 m.y. or so, the water circulation responsiblefor the anomaly probably predated the fault and is unrelated to it. If such erosion did occur,it still doesnot necessarilyimply that the anomaly is related to the fault; it could be the result of differential uplift and erosionnear the fault, exposinga preexistinganomalous condition that was a function of depth. In any case, the evidencefor isotopedepletion occursonly on the west side of the fault and does not resolve the question of why heat flow is high on both sidesin spite of the contrastinghydrologiccharacteristics. In summary, the similarity of heat flow values and the contrast in hydrologic regimeson either side of the San Andreas fault, taken with the absenceof appreciable convectiveheat dischargeby springsin the vicinity of the fault make it extremely difficult to attribute the absenceof local high heat flow to hydrologic transport in the fault zone. However, substantial undetected discharge at temperatures slightly above ambient is a possibilityin almost any terrain, and more corn- 6198 LACHENBRUCH AND SASS:ENERGETICS OF THE SANANDREASFAULTZONE - I I - 'i---- I I 1 I I iPACIFIC PLATENORTH AMERICAN PLATE I • 2.0 84 • 1.66 mean 69 _ 1.0- - 42 0 0 20 , 40 , I 60 I I O0 Distance,km Fig. 14. Heatflowversus distance fromthemaintraceof the SanAndreasfault,region2 (Figure8), shownwithreference anomaly. prehensivestudiesof the hydrothermalbudgetof the fault rm= /•O,' (14a) o,,'= o,,- P (14b) zone may be worthwhile. IMPLICATIONSOF UNIFORM HEAT FLOW, FRICTIONAL STRENGTH, AND SEISMIC CONSTRAINTS FOR TECTONIC STRESS R =/•,o,,' The foregoingdiscussion suggests thattheaveragefrictional resistance duringstrikeslipmotionin the seismogenic layerof •k the San Andreasfault doesnot exceed• 100bars;it could, of course,bemuchless.To avoida problemin notation(because we shall discussseveraltypesof averages),we expressthis condition generally as average dissipative resistance • 100bars= 10MPa = -- %,,•< %,, (15a) (15b) The symbolR usedpreviouslyis relatedto R and total fault slip u by (13) Sincethisupperlimit is an orderof magnitude lowerthan • _-1•o uRdu u (15c) the dynamicfrictionthat onewouldexpectfrom the mostobviousextensionof laboratoryresultsto faults,(13) suggests The maximumstress(or 'frictionalstrength')that can be susthat specialconditionsmust prevail on the faults to account tained by a fracturedsurfaceon which the normal stressis o, for their unexpectedweakness.In the followingsection,we is denotedby rm,porefluid pressureis P, and effectivenormal consider this'strength paradox'in termsof generalized results stressis o•'. The coefficientof (static) friction is the experifromlaboratoryexperiments onrockfriction.It isfollowedby mentallydeterminedparameter/•. R is the instantaneous disa section in which we consider additional constraints on fault sipativeresistingstress(or dynamicfriction)operatingon the stressimposedby seismicobservations. slidingsurfacewhile motion is in progress.The coefficientof dynamicfrictionis the parameter/•,,(15). Althoughthe experRock Friction, Fluid Pressure,and TectonicStressState imentalbasisfor (15) is quiteuncertainand the two equations We shall consider the relation between the heat flow con- oversimplifya very complexprocess,they provide a useful straintand rock friction in termsof parametersfor the sim- framework for discussion. plest frictional model, describedas follows: The frictionalstrengthrmprovidesan upperlimit to the tec- LACHENBRUCHAND SASS.'ENERGETICSOF THE SAN ANDREASFAULT ZONE 6199 NORTH AMERICAN PLATE PACIFIC PLATE REGION 3 3.0 n E•n I;3n 2.06 2.0 mean [] Ci Ci 86 n n n 1.0 o i 20 o 20 i 40 i, 60 i i 80 i ! lOO Distance,km Fig. 15. Heatflowversus distance fromthemaintraceof theSanAndreasfault,region3 (Figure8), shownwithreferenceanomaly.Horizontal'mean-line'terminates at approximate edgeof CoastRangeprovince. tonic stressthat can be resolved on a fault surface and, con- sequently,on the initial stressprior to an earthquake.The frictionalstrengthis not constrained by heatflow (13), and it may be aslargeasotherconstraints permit.The dynamicfriction R is, of course,constraintedby heat flow, but only for thoseeventsresponsiblefor most of the displacementon the San Andreasfault. We shallreferto sucheventsasthe 'principal events.' During principal events,the value of R may changeif o,' changes (evenif/•k is assumed constant), but its displacement-average R mustnot exceedan averageof-• 100 barsoverthe faultedsurface.In discussing the heatflowconstraintin termsof rockfriction,it is usefulto distinguish betweenexplanations that requiresmallfrictionalstrength and thosethat do not. In the firsttype,the fault is weak all of the time,andin the secondtypeit mustbe weakonlyduring principalevents(the two classes are, of course,intergradational). viewedfor our purposeasa reductionin R (15a) by a dynamic reduction of either the effectivenormal stresso•' or dynamic friction/•k. The mostextensivebody of laboratorydata on sliding,rock surfacescan be interpreted in terms of models of either kind. It has been summarizedby Byerlee[1978],who showsthat for a remarkably wide variety of rock materials (with or without an interveninglayer of disaggregatedmaterial) the maximum shearingstress•'mon the sliding surfaceattained during each experimentis a simple linear function of the effectivenormal stresso•' on that surface.To conform to (14a), we have eliminated the constantin his fit to the experimentaldata and have adjustedthe coefficientof friction/• accordingly(seeappendix B). This resultsin the following value: /• • 0.75 (16) which yields a reasonablygood approximationto Byedee's The simplestexplanationof the first type ('permanent equationfor uppercrustalconditions (On'< 6 kbar). weakness')is that the fault containsintrinsicallyweak materiAfter the frictionalstrengthis reachedin laboratoryexperials, for example,water-rich clay minerals [Wu, 1978; Wang and Mao, 1979,Summersand Byerlee,1977]or quartz grains weakenedby the actionof water [Rutterand Mainprice,1978] for which •-,,and henceR (15b) may neverbe muchin excess of 100 bars. For such models the ambient tectonic stress in the ments,failure may proceedunstablyby 'stick slip,' which is generallyviewedas the analogof an earthquake,or by stable sliding,generallytaken as the analog of a creep event in the earth. In spite of the generality of the laboratory results, $tesky[1978]cautionsthat 'we still don't know why Byedee's law [generalizedin (14) and (16)] holdsnor why stickslip oc- fault zonemustbe very low. Modelsof the secondtype ('transientweakness')generallyinvolve a mechanismfor reducing curs. Until we understand the deformation mechanism we friction while fault motion is in progress.They include the will really have little confidencein extrapolatingour laboramodel of acousticfluidization [Melosh, 1979],which can be tory resultsto the earth.' Nevertheless,the simplicityof By- 6200 LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREAS FAULT ZONE N. AM. PACIFIC PLATE PLATE Monterey-Hollister ( U.S.G.S. ) - 125 ,• PachecoPass(U.S.G.S.) v 3.0 CIT I REGION 4 - 84 I 2.0- [] 1.87 mean I ' 42 I 1.0- .,. - o ' 40 20 0 Distance,Km 0 20 Fig. 16. Heat flowversusdistancefromthe maintraceof the SanAndreasfault,region4 (Figure8), shownwith referenceanomaly.Horizontal 'mean-line'terminatesat approximateedgeof CoastRangeprovince. erlee'sgeneralization((14) and (16)) invitesa comparisonwith the heat flow constrainton the San Andreas fault (13). In experimentson polished granite, Byedee [1970] found that R/q-m• 75%, and he commented that for other media the ratio is usuallylarger. To the degreeof approximationof (14) and (15), this suggests 3 /•---• >•-4 /• Combining(15b) and (18) and differentiating, we obtainexpressions for the increasein dissipativeresistance with depth: dR= 1.2/•--& do3 dP. dx I• dx dx (19a) =0.51•'Idø'v•-•xxl (19b) /• Idøi dx •xx (19c) =0.3 •[•k dP 1 I• (17) dx The approximation((14) and (16)) leads to an expression for the frictional strengthof the most favorably orientedfrac- where the unprimed stressesdenotetotal values.For the case ture in termsof the maximum (oi'), minimum (03%and aver- in which the fluid pressuregradientis normal (• 100 bars/kin; age (o.,,g'),effectiveprincipalstresses (seeappendixB): by 'normal' we shall mean conditionsin an unconfmedcolumn of water with unit specificgravity) and where the rock q'm= 1.2 03' densityp is 2.7 g/cm3, (17) and (19) yield =0.50, vg' = 0.30i' (18)d•- = 0.4(Oi -- 03) dX d• Equations(18) may be usedto estimatefrictionalstrength dx q-mby insertingthe appropriateeffectivestressat the time the appropriateaveragevaluesof effectivestressthat obtain while faulting is in progress.We shall take the secondchoice, since it includesthe first as a specialcase. R (7 km) = 1070bars vertical (20a) - 64 bars/km of failure. It mayalsobeused(with(15),(16),and(17)) dR- tocalculate theaverage dynamic friction • provided weinsert dx 153 bars/km 38 bars/km R (7 km) = 445 bars R (7 km) -- 270 bars o•v• vertical (2Oh) vertical (20c) where the vertical stressis assumedto representthe weight of overburden(pgx);R evaluatedat x = 7 km representsthe av- LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE I I I I 6201 1 l PACIFIC PLATE NORTH AMERICANPLATE REGION 5 125 3.0 ..-., 2.17 mean [] [] 91 [] 84 • 1.0 I 3O 0 10 0 10 0 Distance,km Fig. 17. Heat flow versusdistancefrom the main traceof the San Andreasfault, region5 (Figure 8), shownwith reference anomaly. cragcresistance of a 14-kinscismogcnic layer.The valuesof • in (20) arc perhapsthe most'expectable'onesfor this extreme rangeof tectonicstressstates.For an averageslipvelocity(2v) of 3 cm/yr they correspond (fromtop to bottom,respectively) to local conductive steady state heat flow anomalies of 2.4, 1.0,and 0.6 HFU. The constraint(13) imposedby the absence of such anomaliesimplies, however, that dR •'m(avg) • 100bars x p-p• 133b 100 bars • •< 7• This result saysnothing about the frictional strength •m at the onsetof faulting unlessthe effectivestresses in (18) were the samebefore and during faulting. We shah assumethat this condition is satisfiedif the fluid pressurewas unchangedduring the event. For this case, we obtain the condition of permanent weakness((13), (15b), and (17)): "•15bars/km (23) (21) The fluid pressures (22) would then be a permanent(ambient) conditionof the fault zone. If the uncertainvalue of/4, were Combining this condition with (19) yields the constrainton decreasedby a factor of 2, it would, of course, increase the the averagefluid pressureduring faulting: • limit on frictional strengthproportionally, but the fluid presdP sures (22) required for materials satisfying Byerlee's law 03vertical (22a) would be reducedby only 10% or so. The restriction(23) on dx >• 253bars/km frictional strength stemsfrom the assumptionthat the fluid dP pressuregoverning failure is the same as the one governing •xx>•230bars/km Oavg vertical (22b) dynamic friction. The alternative will be discussedpresently, but first we shah comment on explanations for anomalous dP •xx>•203bars/km O1vertical (22c) ambient fluid pressure. The questionariseswhether the large fluid pressuresimplied by (22) could be generatedby long-termfrictional heatAccordingto (22), fluid pressureduring faulting would have ing of the fault. Water expandingat constantvolume under to exceedthe normal value by a factor of 2 or more, irrespec- upper crustal conditionsundergoesa pressureincrease of tive of the state of tectonic stress. This calculation illustrates about 15 bars/øC [Lachenbruch,1980]. Hence to increasethe the difficulty in trying to reconcilethe heat flow constraint fluid pressureby a factor of 2 from a hydrostaticstate would with Byedee'slaw and the frictional theory of faulting. (A require an averagetemperatureincreaseof 47øC in a 14-kin similar analysishas been madeby Steskyand Brace[1973].) seismogenic layer. However, for a dissipativeresistanceof 100 6202 LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREAS FAULT ZONE NORTH AMERICAN PLATE PACIFIC PLATE REGION 6 125 3.0 I I ,.-., [] I _o 2.0- 84 • [] mean 1.85 ] 77 -- [] [] [] [] [] O•0 I I 60 I I 40 20 0 20 40 D•stance,km Fig. 18. Heatflowversusdistance fromthemaintraceof the SanAndreasfault,region6 (Figure8), shownwith referenceanomaly.Horizontal'meanline' terminatesat the approximate edgeof CoastRangeprovince. bars(13) and slip velocityof 4 cm/yr (Q ~ 0.3 HFU, equa- 1973], which probably account for most of the cumulative tion (10), Figures3c, 4c, 5c, and 6c) the averagetemperature fault displacement.Important contributionsto the transient increaseon the fault plane would be only 10ø or 15oc after 10 pressureincreasecouldalsoariseif therewere relaxationof m.y. of faulting. Hence even if the rock permeabilitywere precusorypore dilatation or dehydrationof clay minerals in detail elsezero,thislong-termsourceof high ambientfluid pressuredoes [Raleigh,1977].Thesealternatives,discussed not seempossible.(Thermal expansionof pore fluid during where [Lachenbruch,1980],allow failure to occurat low amevolution of the broad Coast Range anomaly is, however, a bient fluid pressure(Po), thereby relaxing the constrainton frictional strength'rm (23) and leadingto a 'transientweakpossiblesource;it will be discussed later.) Berry [1973] has presentedevidencefor near-lithostatic ness' model. In this mechanism the shear stress would unfluid pressurein parts of the Franciscanformation, which dergoan initial dropfrom a staticvalue(<• •m)to R(t = O)(dethe pressure P(t) would boundsthe San Andreas fault on the eastin the Coast Ranges. terminedby (15b)),and subsequently, Irwin andBarnes[1975]havenotedthatcreeping portionsof increase astheearthquake proceeded. As P(t) approached On, the fault occur where metamorphic fluids from the Fran- the effectivenormal stressOn'(15) would approachzero, and ciscanformation are confined by an overlying sectionof the sowould R(t). One difficultywith the modelconcernsthe posGreat Valley sequence,and they postulatethat in suchplaces siblelimit imposedon P(t) by 03,the leastprincipalstress.As the confined fluid is forced into the fault zone. However, the P(t) approached03,conditionsfor hydrofracturing wouldob- heat flow data imply negligibledynamic friction in both the tain, and further pressureincreasemight be inhibited. This creepingand seismicportionsand within the Coast Ranges condition can be written as and to the south(Region 7, Figures 8 and 19) where no FranR(t) = I•, (on -- P(t)) ciscanrocksoccur.Hence it seemsunlikely that the heat flow data can be accountedfor by speciallocal geologicconditions. • ]'•k(On- 03) If the small dynamic resistanceis to be reconciledwith By(24) erlee's generalization,high fluid pressuresseem inevitable. (We alsohavealternativessuchasthe intrinsicallyweak fault On or acousticfluidization,previouslydiscussed.)However, such pressures do not haveto be an ambientconditionof the fault >•1-%, (øn-Po)II zone;theymightbe causedby transientfrictionalheatingand Under theseconditions,the minimumresistingstressR(t) expansion of the fluid duringlargeearthquakes [e.g.,Sibson, wouldbe limitedby ((B6a),(15) and (17)) LACHENBRUCHAND $ASS:ENERGETICSOF THE SAN ANDREAS FAULT ZONE I I ! I I 6203 I ß Palmdale(U.S.G.S.) REGION 7 [] Other(U.S.G.S.) San Bernardino (ClT) 3.0 125 Lake Hughes(CIT) PACIFIC PLATE NORTH AMERICAN PLATE 2.0 84 • mean 1.58 , I 60 ß I i0 4 ß 20 0 66 20 I I 40 Distance,km Fig. 19. Heat flow versusdistancefrom the main traceof the San Andreasfault, region7 (Figure 8), shownwith reference anomaly. •'k R(t)•>0.375R(0)-• 0.375--•-,,(25a) • 0.28%,(25b) The averageresistingstressR, of course,would be a larger value obtained by integrating R(t) [Lachenbruch, 1980]. Whether or not this limit (25) appliesdependsupon whether significanthydrofracturingcan developon the time scaleof an earthquake. Resultsfrom the abovecalculationsare comparedin Table 4. The case'dry, P = 0' representsthe maximum stressespermitted by (14)-(18). Case 2, Table 4 may be viewed as the transient alternative if we supposethe ambient pressureis normal and that an increasein fluid pressureor acousticfluidizationduringslipeventscanindeedsatisfy(13), i.e., that the resistance can be reducedfrom an initial value R(t = 0) to yield an averageR •< 100 bars for those eventsthat account for mostof the slip.Case1, Table 4 represents the caseof permanent weaknesscorrespondingto constant ambient fluid pressure. If the permanentweakness resultsfrom anomalously weak gougematerials,the assumption (of (18) and Table 4) that faultingoccurson the mostfavorablyorientedsurfaceis notjustified,but effectsof the assumptionon numericalvalues are probablyunimportantfor theseapproximatecalculations. Hencethe lastthreecolumnsof Table 4, case1, can probably state fluid pressuresor from anomalouslyweak fault materials. Tectonic stress differences in the last column of Table 4 are obtained from (18), and numerical values are based on t•k/t• = 0.75, p = 2.7 g/cm3. The foregoing discussioncan be summarizedas follows: 1. The heat flow constrainton dissipativeresistance(13) has strongimplicationsfor the frictionalstrength,,,, if we interpret faulting in terms of Byerlee'slaw and a frictional theory applied to a randomly fracturedupper crust.The relation betweenR and %, dependsupon two additional variables:the state of tectonicstressand the fluid pressure;it is most sensitive to the latter. 2. The heat flow constraint(13) requiresthat irrespective of tectonicstressstate, fluid pressuremust be near lithostatic while fault displacement is in progressif Byerlee'slaw applies. 3. Such high fluid pressuremay be an ambient (steady state)conditionof the fault zone (case1, Table 4), or it may be a transienteffect developedduring the slip event (for example, case2, Table 4). 4. If steadystate(or if Byerlee'slaw doesnot apply, and the fault is intrinsicallyweak), then frictionalstrength,,,, is of the samegeneral magnitude as dissipativeresistanceR, differing only by the ratio of staticto dynamicfriction,and tectonic stress differences do not exceed a few hundred bars. This ('permanent weakness')alternative leavesthe anomalousfluid pressure(requiredby faults satisfying(14)) unexplained. 5. If fluid pressureduringfaultingis transient(or if acousbe usedwhetherthe low strengthresultsfrom high steady tic fiuidizationoccurs),then ambientfluid pressures may be 6204 LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREAS FAULT ZONE Regions 3 through6 Region7 C a I I Within 10 km of fault Within 10 km of fault - N =16 N=7 • = 1.99 • =1.62 SD = O.O6 SD = O.26 co 0 I o 1 2 i I 1 2 q, HFU q, HFU d b I Beyond10 km Beyond10 km 10 N=23 N=7 =1.98• 1 I 2 1.61 SD= 0.07 3 q, HFU 0 I i 1 2 3 q, HFU Fig.20. A comparison between heatflowwithin10kmof themaintraceof theSanAndreas fault(a andc)andheatflow • at greaterdistances (b andd) for regions3-6 (a and b) and region7 (c a•tdd). normal, friction strength7,, and tectonicstressmay be large, and low dissipativeresistance is explainedasa consequence of the slip events.Difficultieswith this ('transientweakness')alternative lie in our ignoranceof the dynamics of transient eventsduring faulting. It might be argued that the absenceof a local heat flow anomaly over presentlycreepingportionsof the fault implies the condition of permanentweaknessdescribedby condition 4 aboveand that in suchplacesat least,the tectonicstressdifferencesmust be very small. However, it can be shownfrom the resultsin appendixA that even if the resolvedshearstress (and hencethe averagefrictionalresistance) on the fault were -• 1 kbar, creep at plate velocitieswould have to persistfor more than 10,000yr before a measurableheat flow anomaly would result. Consequently,if creep episodesof centuriesor even thousandsof years duration were separatedby much longer episodesof seismicdisplacement,the heat flow constraint could be satisfiedwith large tectonic stress(condition 5, above)if largeearthquakeswereresponsible for mostof the displacementand if they developedhigh transientfluid pressures. In the next section, we shall investigateseismological constraints on this alternative. inc the energeticsof the earthquakeprocessand find that modelsof the secondtypeyieldlargequantitiesof radiatedkineticenergy,in apparentcontradiction with interpretations of a largebody of seismicobservations. We shallintroduceone new unknown,the final elasticstress,and two new partially known quantitiesobtainedfrom seismology,the apparent stressand the stressdrop (defined below). Figure22 is a schematic representation of the localenergy budgetof an earthquake[seealsoAndrews,1978a].Here, 7' and 7' are the initial and final average values of the com- ponentof elasticshearing stress in thedirectionof faultslipu, and Fis the corresponding averagevalueof the dissipativeresistancein the directionoppositeto fault slip. The total slip averagedoverthe faultedsurfaceis u,,. (We shahdenotethe faulted surfaceby •, and its area by Ax.) The straightline joiningthe averageinitial and final elasticstresses represents theunloadingof a linearelasticmedium;the areabeneathit is the total elastic energy (E/Ax) released(per unit area of slippedsurface)at the locusof faulting•. Hence E 1 Axu,,-•(7'+ Y) The area beneath the horizontal Fault Energeticsand SeismicEstimatesof StressDifference In the last section,we found that insofar as laboratory in- (26) line with ordinate F or the curve r(u) (Figure 22b) is energy(E•/Az) lost from the mechanical system.Thus formation on rock friction is concerned, there are two inter- gradationalclassesof simplemodelsthat can accountfor the Er -._ •. (27) low dynamic friction implied by the heat flow constraint:(1) permanent-weakness modelsin which the averagefrictional The area between the elasticstresscurve and the dissipation strengthis also low, and (2) transient-weakness models in curverepresents energy(Ea/A=) allocatedto elasticradiation. which the static friction may be an order of magnitude Consequently, greater, the upper limit being determined by rock strength. Ea I , The first type requireslow tectonic(i.e., initial) stresses, but the secondtype permitslarge ones.In this section,we exam- azu,,, = •-(7+7')-F (28) LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE 124' 122' 120' 118' 6205 116' 114' i O 42' OREGON , IDAHO a a Mt Shastai ' 14,161 ft a fee '• / ß /o ß / / eI t / ß ! Wmnemucca ß •./ t t ß ? t ß t ß ß a . ß ._.* a ß Reno t o•... Lake a tt a % t ß t ß N E tI ß t *% el ß , ß ß Sacramento t ./ ß eo I ß I tt I a oTonopah 38' t Sin Franci t ß a ß 36* ee e•. ß Angeles e• •.• • ß 0 50 I I 0 100 150 200 t ! • MILES <::::) 50 100 150 200 KILOMETERS Fig. 21. Thermal springsin California and Nevada (reproducedwith modificationsfrom figure 8 of Waring [1965]). Heavy line is the main trace of the San Andreas fault. Figure 22a is shownfor three differentvaluesof ?-andFigure 22b for variable r. sional models of the fault, in which case their intuitive meaning is clearer. We define the average elastic stressTe and the apparent This interpretationof areasin Figure 22 is fairly generalas long as we define •-*, •-', and ?-respectively,as weighted aver- stress% by agesof initial stress,final Stress,and R (15c) over the faulted I , surface,the weightingfunction being the local fault slip. For refinedmodelsof faulting in which significantamountsof enI , ergy are dissipatedat the propagatingedgesin complexstress •a--•(• + •')- ?(30) fields,the representationin Figure 22 will not, in general,provide a useful view of the sequentialvalues of the stressesat The energy balance, any typical location. Nevertheless,the three generalized E=E,•+En (31) stresses (•-*, •-',?-)definedin thisway are usefuldescriptiveparameters,whatever their relation may be to the details of the can now be expressed((26)-(31)) as mechanicsof the earthquakeprocess.For the purposeof disWe-- Ta -[- ?(32) cussion,we can considerFigure 22 to representone-dimen- 6206 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE TABLE 3. Data on Thermal SpringsWithin 10 km of the Main Trace of the San AndreasFault BetweenCape Mendocinoand the San BernardinoMountains Name Latitude Flow Rate Longitude I s-• Point Arena Hot Springs Rocky Point Spring Near SergentSanta Clara County FresnoHot Springs Tyler's Bath Springs Arrowhead Hot Springs Waterman Hot Springs 38052.5' 37053' 36056.8' 3608.8' 34ø13.5' 34ø11.2' 34011' 123031' 122037' 121ø33.8' 120033.5' 117029.5' 117ø14.5' 117015.5' Temperature, øC 0.3 0.3 0.5 1.3 0.3 3.2 0.3 44 38 25 31-36 33 94 50 Discharge, MW 0.035 0.030 0.020 0.110 0.015 1.0 0.04 1.25 Thus the deformationenergyreleasedlocally during an earthquake is the work done by the averageoperatingelasticstress %, which can be expressedas the sum of a part allocated to the productionof seismicradiation, %, and a part allocated to the production of heat, E For the transient classof models discussedpreviously, % can be large, and since ? must be small, this would imply large seismicradiation (i.e., large ,a). The permanentlyweak models,on the other hand, would be consistent with small contributions from both ? and %. In what follows, we shall expressthe energy constraint(32) in terms of the friction parametersof the last sectionso that the numerical resultsobtained there can be comparedformally to those obtained from seismology. The relations among the seismicallyrelated stressescan be discussedin terms of the following identity: •(, ,+,')1,-,')- 11 Total placementon the San Andreas fault (the 'principal events'), we can write the heat flow constraint(13) as Y<• 100 bars (38) The preponderanceof valuesof stressdrop A determinedfor crustal earthquakes[Hanks, 1977; Kanamori and Anderson, 1975] satisfy 1 bar < A < 100 bars (39) Although seismicallyestimatedvalues of •-• are lesscertain, Savageand Wood[1971]presentdata from many earthquakes to show that within reasonablelimits of uncertainty, 1 (40) and Kanamoriand Anderson[1975] find that on the average% ~ (1/2)A. Substitutingupper limits given by (38), (39), and (40) in Although seismologyprovidesno reliablemeansof estimating the magnitudesof the three stresses in (33), it doesprovide in- (36a) and using(37), we obtain bracketingconditionsfor the direct estimatesof two stressdifferences[Wyss and Brune, initial stress** precedingearthquakes: 1968;Wyss,1970;Savageand Wood,1971]:the apparentstress 1 %, just discussed(30), and the stressdrop A, which is the dif(41a) ference between the average fault stressimmediately before and after an earthquake: The right-hand inequality is equivalent to the result obtained A = ** -- ,' (34) by Brune et al. [1969] (exceptthat they used % <• 100bars inThe seismicefficiency•/is defined as the fraction of the re- steadof (40)). Their resultfor the casein which displacement ?--**-•(, (33) 0<E%,• A<•**--?+r•+}---A <-200 bars leased energy that is allocated to the generationof elastic TABLE waves• Ea r/= •E Ta -- (35) %+? 4. Ambient Conditions Transient Ambient Pressure Gradient, dP/dx, Combinationof (30), (33), and (34) yields 1 ** - ?--%+ •-A (36a) 1 ,,_ ?m%_ •-A (36b) SinceF,,,•, and A are generallypositiveon physicalgrounds,it follows from (36a) that the average initial stresshas lower bounds given by - v*> r,%,•A >0 (37) If we limit our discussions to eventstypical of thosewhose cumulative effect resulted in hundreds Case of kilometers of dis- bars/km Effects Initial Dynamic Resistance, R(t = 0), bars Tectonic Frictional Strength, Tm, Stress Difference, O1-- 03, bars bars o3 Vertical (Thrust Faulting) Case 1 Case 2 Dry, P = 0 •>253 100 0 <•100 1070 1700 •<133 1426 2265 •<332 3565 5560 <•133 593 990 •<332 1482 2355 OavgVertical Case 1 Case 2 Dry, P = 0 1 in Fault Zone for Three Tectonic StressStates:Case 1. Fluid PressureDoes Not ChangeWith Time, Case2. Ambient Fluid Pressure'Normal' and (13) SatisfiedBy >•230 100 0 <•100 445 710 Vertical (Normal Faulting) Case 1 Case 2 Dry, P= 0 >•203 00 0 <•100 270 430 <•133 360 570 <•332 900 1430 LACHENBRUCH AND SASS: ENERGETICS overshoot null locking 0 U--•- um b r'/r(u) r.a • o f 0 U"" OF THE SAN ANDREAS FAULT ZONE 6207 jumping of one notch in opposingsawteeth).Savage and Wood[ 1971]pointout that if frictionstopsthe faulting,inertia requiresthe overshootmodel,a sufficientconditionfor which is (40) (accordingto (42a) and(36b)).If, however,the resisting stressvariesduringan earthquake,the physicalsignificance of this classificationis lost, sincethe final value, not the average r-•shouldbe usedin (42) to relate r' to friction. This is illustrated in Figure 22b for a one-dimensionallocking model (42c)in whichr increases sharplyat the end of an event(dotted lines,Figure 22b). Brune[1976]suggests that suchan increasein resistancemight result when the particle velocity dropsbelowa criticallimit. The resistance is sensitiveto shear zonewidth and poredilatation,both of whichvary as particle velocitychangesduring an earthquake;unstablepropagation of hydrofracturesmight reduce fluid pressurerapidly, and strengthening due to lossof fluid pressureby dissipationof frictional heat or Darcian flow can probablycontribute[Lachenbruch,1980].For the caseof acousticfiuidization[Melosh, 1979], rapid strengtheningmight occur in the decelerating stagewith the lossof acousticenergy.In any case,however, the classificationin (42) is useful for discussion;for a given stressdrop and resistance, •a and the seismicefficiencywill be greatestfor the lockingmodel.Intuitively,the lockingmodel generatesadditionalelasticradiationby its terminalimpact againsta largehypotheticalfinal resistance. (We make no attempt to reconcilethesesuggestions with fracturemechanical Um Fig. 22. Schematicrepresentation of relationsamongaveraged valuesof initial stressr*, final stress¾, dissipativeresistancer• apparent stressr,,, mean elasticstressre, and stressdrop A for an event with mean total displacementU,n.Figure 22a illustratesthe three casesof (42), and Figure 22b showsa one-dimensionallocking model with variable dynamic friction r. considerationsof fault stoppage.) In one-dimensionalmodels it is natural to equate the aver- ageinitial stress r* to the frictionalstrengthrm.For morerealistic two- and three-dimensional models, it is clear that the initial stressmustequalthe strengthonly at the point at which ruptureinitiates(and thisstrengthmay be lowerthan the laboratoryvaluebecauseof the sizeeffect);elsewhere the stressis is exclusivelyby uniform creep follows from (4 la) by setting probablydeterminedby dynamicfriction [e.g., Berg, 1968; A-ra=0: Burridgeand Halliday, 1971].We shall assumethereforethat the averagevalue of initial stresscan be bracketedby the r* •< 100 bars (4lb) static and dynamic friction: Theresult(4lb) isessentially thesameasthefrictional result (23), sinceboth are controlledby the heat flow constraint. However, the upper limit of 200 bars in (41a), obtained from heat flow and seismicobservationscombined,imposesa much more severeconstraint than rock strength (Table 4), which limited the initial stressin models of transient weakness, dis- cussedin the last section.Becauseof the potential importance of the possibilityof large initial stressto the understandingof earthquakes,we shalleliminatethe leastcertainof the seismic constraints,the one on % (40), substitutethe limit on rock strength(Table 4), and investigatesomeof the consequences. Simpleearthquakemodelscan be discussed in termsof the relativemagnitudesof the final stressand the averageresisting stress[seeBrune, 1976].The three possibilities,illustratedin Figure 22a, are r' - F< 0 -- 0 overshoot (42a) null (42b) •'m -->-•'*>--]O[--k 'Tm (43) We shall consider two cases,•'m small and •'m large, corresponding respectivelyto cases 1 and 2, Table 4. The permanent low strengthin case I can result from high steady statefluid pressureor from intrinsicallyweak fault materials. The high strengthin case2 is a consequence of normal ambient fluid pressure;weakeningduringthe earthquakecan be from a transientincreasein fluid pressureor acousticfiuidization. The limiting valueson the left and right in (43) are representedrespectivelyby columns2 and 3 in Table 4. Thus in case2, (43) becomes (270 bars,445 bars,1070bars) --< r* =<(360 bars, 595bars, 1425 bars) (44) where the valuesfrom left to right in parenthesescorrespond respectively to the stressstates:o• vertical,Oavg vertical,and 03 > 0 locking (42c) vertical. The correspondingresult for case 1 (constantfluid pressureor an intrinsicallyweak fault) is ((23) and (28c)) If it is assumedthat the resistingstressis constant(i.e., r(u) -?-), the null model (the 'Orowan assumption'of Brune et al. 1 [1969]) attributesthe final stressto the dynamic frictional 0< K•'a, •-A< •'•g • -- •-• 133 bars (45) characteristicsof the surface.The locking model [e.g., Brune, we shall use (44) 1976]is unrelatedto averagefriction;it could occurin a one- In evaluatingthe constrainton other stresses, dimensional model with elastic barriers (for example, the or (45), inferredfrom frictionalproperties,the heat flow con- 6208 LACHENBRUCH AND SASS: ENERGETICS straint(38), and the constrainton stressdrop (39). The apparent stress•a, possiblythe least certain of the semimeasurable quantities,will be retainedas a parameter. Using the left-hand inequalityin (44) with (36), (38), (39), (41), and (43), we obtain the followinglower limits for the transient case 2: •-* - ?' •> (170 bars,345 bars,970 bars) (46a) 120bars,295 bars,920 bars) (46b) •-' - F •> (70 bars,245 bars,870 bars)> 0 (0.55, 0.75, 0.90) (46c) (46d) Similarly, for case1, (45) and the foregoingrelationsyield OF THE SAN ANDREAS FAULT ZONE cluding but not necessarilylimited to the principal events ('transientweakness'). 3. For the class I models,the averageresolvedtectonic stresson the fault cannotbe appreciablygreaterthan 100bars. For the class2 models,tectonicstressis limited by the average frictional strengthof upper crustalrocks;under normal fluid pressurethis is -•0.4-1.5 kbar, dependingon tectonic stress state. 4. If the principal eventsare slowcreep,permanent-weakness(type 1) models are required. However, creep eventsat plate velocitieslasting no longer than • 10n yr are not necessarilyprincipal events(i.e., they neednot contributeappreciably to the heat flow even if the ambient stressis large). 5. Physical models for permanentweakness(type 1) include thosein which the gougematerial has anomalouslylow friction coefficients, and those in which the friction coeffi- cientsare normal, but the ambient fluid pressurein the fault zone is close to the overburden pressure.Although there is 0< q'a • 33bars - •A < 33bars (47b) little independent information that either of these conditions is general in the fault zone, thesemodelsare consistentwith the small stress drops and apparent stresses(and, con-50bars < •--•A1 <_•'--y<_33bars--A < 33bars (47c) sequently,small seismicradiation) inferred from earthquake seismology.Models of type I require low seismicefficiency; they are consistentwith all three typesof earthquakemodels: locking, null, and overshoot. 6. Physicalmodelsfor transientweakness(type 2) include Thus the transienthigh-stressalternative(case2) requires those in which the ambient fluid pressure and frictional very large q'a(46b) in violationof (40). This leadsto a high strengthof the fault zone are normal, and resistanceis deseismicefficiency(46d)and the requirementfor a locking-type creasedduring principal events,for example, by a transient earthquakemodel(46c)accordingto the classification of (42). increasein fluid pressure[Lachenbruch,1980] or acousticfiuiIn contrast,the steady state pressurealternative (or the in- dization [Melosh, 1979].For suchmodels,the principal events trinsicallyweak fault) is consistent with the smallvaluesof are probably large earthquakes.These models have the ad(47b) usually reported;it leads to a low seismicefficiency vantagethat the anomalousweakness(necessaryto satisfythe (47d) and couldbe consistent with any of the three alterna- heat flow constraint)is 'explained'asa consequenceof the slip tives in (42) (see (47c)). event. Only models of this type can be consistentwith large Althoughit compoundsthe problemof fault stoppage,the ambient tectonic stressand the heat flow constraint. They transient alternative, with initial stress-• 1 kbar and dissipative have the disadvantagethat for tectonic stresscontrolled by resistance •0.1 kbar, is not inconsistent with the observation normal rock strength, they require apparent stressesin the that the stressdrop A is generallysmall(•0.1 kbar). It is, how- range100bars-lkbar (depending on tectonicstress state).This ever, inconsistentwith the generalinferencethat q'ais small is an order of magnitudegreaterthan the range inferred for and, in particular,with the Savageand Wood constraint(40). crustaleventsfrom earthquakeseismology.For large tectonic However, the physicalargumentfor the validity of (40) is stress,type 2 modelsrequire large seismicefficiencyand lockbasedon the constancyof r and doesnot necessarilyapply in ing modelsof the earthquakeprocessfor principal events. the transient case. The questionthen remains whether the Factors that might control the transientvariations of fluid seismicallydeterminedquantity q'acorrespondsto the one pressureduring an earthquakeare discussed in another paper definedin this section,or whetherthe seismic%a'might repre- [Lachenbruch,1980].Although they might ultimately prove to sent only a small fraction of the energy radiated from the be important to the earthquake process,if the heat flow and locusof faulting during thoseeventsresponsiblefor most of seismicdata are taken at face value (4 la), even suchtransientthe displacementon the San Andreas fault, probably very weaknessmodelsare limited by a fault strengthof only -•200 largeearthquakes. Only in the latter casewouldthe transient bars. In the next section,we shall investigatethe possiblerealternativeand the possibilityof large strengthand tectonic lation betweentheselocal fault stresses and thosethroughout stressbe viable [seealso Brune et al., 1969]. the broader interplate shearzone by consideringthe origin of In summary,the heat flow constraint,taken with measure- the broad CoastRange heat anomaly.It is shownthat if much ments of rock friction and seismic observations, seemsto have of the heat flow anomaly is mechanicallygenerated,it is likely the followingimplicationsregardingstresses on the San An- that the seismogeniclayer is partially decoupledfrom the undreas fault: derlyingmaterialand that the resultingbasalstresses lead nat1. If we acceptthe heat flow constraint,then resistanceto urally to a relatively weak fault, with much larger stresses 1 1-•}•0.25 (47d) fault motionmustbe small(•<100bars- 10 MPa) duringthose events responsiblefor most of the fault displacement(the 'principal events'). 2. This condition can be satisfiedby two simple (inter- gradational)classes of models:(1) Thosein whichthe fault resistanceis small all of the time ('permanentweakness'),or (2) those in which it is small only during short-livedevents,in- elsewhere in the shear zone. THE BROAD HEAT ENERGETICS FLOW ANOMALY OF THE FAULT AND ZONE The heat flow data from the San Andreas fault region have two notable features: (1) the absenceof a local heat flow anomaly over the 1000-km length of fault trace from which LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6209 + CL1 3.0 125 I la lb 2 i _1'0111'3211'66, 3 2.06 4 + ,1.87 + -IF • 2.0 /• 2.7 + 6 1.85 84 + ++½ 1.0 42 •+ 0 200 400 600 Distance,km Fig. 23. Northern portion of Figure 12 showingmodelsfor the transientrise in temperaturecausedby northward migration of the MendocinoTriple Junction.CurvesA, B, C, and D correspondto modelsof equations(38a), (38b), (38c), and (38d), respectively.Mean heat flows (in HFU) are indicatedfor each region. data are available,and (2) the broad anomalyacrossthe Coast Rangesextendingfor more than 500 km along the fault zone (regions3 through6, Figures8 and 12). In the previoussection we discussed someimplicationsof the first featurefor the ley to the eastover a distanceof only tensof kilometers(see, for example,Figures 15 and 18). This suggests that the source of the Coast Range anomaly is in the upper crust and that it might be related to the tectonicsof the San Andreas fault local mechanics of the main fault. In this section we shall con- zone; however, details of the thermal transition are complicated by effectsof rapid sedimentationin the westernportion of the Great Valley. To testthe relation betweenthe fault zone and the heat flow anomaly, we subsequentlyextended measurements northward to and beyond the Mendocino sider possibleimplicationsof the Coast Range anomaly for thermal and mechanicalconditionsthroughouta broadinterplate shearzone surroundingthe main fault. The analysisis limited to the Coast Range anomaly, becauseit is a con- Triple Junction,where the San Andreasfault on the North American plate should,in principle,have zero age. Owing to the nature of the formations and the youthful topography, confidentheat flow measurementsin this northern region (as in many other parts of the Coast Ranges)are difficult to obtain, and this might account for much of the scatter in the data. Nevertheless,the resultsshownin Figure 12 suggestthat the Coast Range anomaly does,indeed, fall off to the north. North of the Triple Junction(region la, Figure 12) the averageheat flow is about 1 HFU, consistent with resultsfrom the Klamath Mountains and northward along a coastal strip to the Canadian border (seeFigure 6), the region in which suband the tectonicsof this complicatedregionhave been dis- duction is believed to be in progresstoday [see also Lachencussedby Lee and Henyey[1975];presentheat flow coverage bruch and $ass, 1977; Blackwell, 1978]. South of the Triple there is inadequateto warrant the constructionof explicit Junction, averageheat flows seemto rise over a distanceof models. about 200 km (regionslb and 2, Figure 12) to valuescharacteristic of the Coast Range anomaly. The CoastRangeAnomalyand Depthto Its Source From plate tectonic reconstructions[Atwater and Molnar, From heat flow measurementsin central California, we 1973],the Triple Junctionis believedto have been migrating noted in an earlierpaper [Lachenbruch and $ass, 1973]that northwestwardat roughly 5 cm/yr for the pastseveralmillion the Coast Range anomalyfalls off from averagevaluesof years.Hencethe ageof the fault zone on the North American about 2 HFU to valuesof perhaps1.2 HFU in the Great Val- plate shouldincreaseabout 1 m.y. for each50 km of distance spicuousfeature with well-defined boundarieson the east and north, and it enclosesa region in which most of the relative plate motion can generallybe attributed to the main trace of the San Andreasfault. However,the possiblewestwardextent of the anomaly onto the continentalshelf is unknown. South of the CoastRanges,conditionsare very complex,the shearing deformationis evidentlydistributedover a broaderzone, and the heat flow anomaly is smaller but distributedover a largeregionincludingthe offshoreborderlandand partsof the Mojave Block and the Salton Trough; an island of low heat flow occursin the southernCalifornia batholith (Figure 6) [seealsoRoy et al., 1972].Possiblerelationsbetweenheat flow 6210 LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREAS FAULT ZONE Fig. 24. Idealizedrelationsamongtectonicplatesnear the MendocinoTriple Junction(intersectionof the Mendocino transform(M. Trans.)and the SanAndreasfault (S.A.F.)). Arrowsindicatemotionrelativeto a fixedPacificplate. Coordinate axesrefer to model in text (seealso Figure 25). southeastwardfrom the Mendocino Triple Junction. (All of the sitesin the arbitrary regionslb and 2 (Figure 12) were, of course,on the North American plate as the Triple Junction migrated northward along an offshore trench.) To obtain a rough idea of the depth of the sourcesthat might be respon- tive if the time constant for Darcian flow were an order of magnitudegreaterthan the time constantfor heat flow, i.e., if the permeability of overlyingrocksdid not exceed---0.1nanoDarcy [seeFigure 8, Lachenbruch,1980].A possiblesourceof water is sedimentscrapedfrom the subductingplate. Hence it sible for the northward transition in heat flow, we have shown is quite possiblethat the thermal evolutionof the broad Coast the effectsof four simple one-dimensionaltransient heat con- Range anomaly could have a substantialeffect on the meductionmodels,curvesA, B, C, and D, in Figure 23. For each chanical evolution of the San Andreas fault zone through the model the time origin is taken at the presentTriple Junction, weakening effect of high fluid pressureon mid- and upperand each anomaly is constrainedto reach 80% of its steady crustal rocks. We can imagine two plausible mechanismsfor the anomastate value of 1 HFU at 3 m.y., which correspondsto a point about 150 km southeastwardfrom the present Triple Junc- lous near-surfacesourceof heat in the Coast Ranges: tion. For each model, the depth of anomaloussources(x -- d) was adjustedto meet theseconditions;the resultsare Temperature (0) Model A: constant planesourceat x ----d; d • 3 1/2 km (48a) Model B: uniformsources in 0 <_-x <_-d; d • 7 km (48b) ModelC: stepchangein flux at x = d; d • 11 km (48c) ModelD: stepchangein temperature at x ----d; d m 20 km (48d) The deeper the source,the greater its initial strengthmust be and the more intense is the associatedtemperature anomaly; for example,model C requiresan initial sourcestrengthtwice as great as model A and a temperatureanomaly greaterby a factor of 3. If the values of d were doubled, the horizontal 0 Os(a ) X\x,o) i ß : (x,o) scaleof curvesin Figure 23 would be extendedby a factor of 4. Thus as in the transitiontoward the Great Valley, a shallow origin (•<20 km) is suggested. Before undertaking further analysis of these results, it is worth noting that the observedincreasein heat flow would correspond to an increase in thermal gradient above the sourceof >•15øC/km. The expansionof pore water at constant volume under upper crustal conditionsleads to a fluid pressure increase--- 15 bars/øC [see,e.g., Figure 3, Lachenbruch, OH(a) 1980]. Consequently,an increasein thermal gradient of only ,--13øC/km could increase the fluid pressure gradient by Fig. 25. Dashedcurvesshowtemperatureat z -- 0+, Figure 24, ---200bars/km, enoughto raisethe fluid pressurefrom hydro- the initial condition for model in text.' Solid curves show initial temstatic to lithostatic. In the Coast Ranges, this could be effec- peraturewith contributionfrom radioactivityremoved. LACHENBRUCH AND SASS; ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6211 I-<-Mojave Region-• CoastRanges 125 3.0 84 30KM 42 1.0 I 0.5 0 I I 250 I 500 I 750 lOOO I iPresentdistancefrom Mendocino Triple Junction(KM) 0 5 10 15 20 25 (30) I (35) Timesincepassageof Mendocino TripleJunction(m.y.) Fig. 26. Heat flow as a functionof distancefrom the MendocinoTriple Junctionfor selectedinitial thicknesses a of the North Americanplate (seeFigures24, 25, and (22)). Horizontalsolidlinesare meanheat flowsin the fault zonefor regions delineatedin Figure 8. Horizontal dashedlinesare mean (corrected)heat flowsinterior to Great Valley. Seetext for further explanation. 1. As the North American plate movessoutheastward(relative to the Pacificplate) pastthe projectionof the Mendocino transform(z - 0, Figure 24), its baseat depth x = a (Figure 24) slidesfrom the Gorda plate,cooledby subduction,to a re- gion of warmer mantle material, the northernedge of the 'window' of Dickinson and Snyder [1979]. Conduction from this material might supplythe anomalousheat provideda •< 20 km (as required by (48)). 2. A secondplausible shallow source of the anomaly is shear-strainheating associatedwith distortion and possible basaldecouplingof the seismogenic layer of the San Andreas fault zone, which originatesin the North American plate at this location (z = 0, Figure 24). In an earlier stage of this study [Lachenbruchand Sass, 1980] we discountedthe first possibility,sincewe thought it unlikely that the top of the subductingplate would be at midcrustal depths(a •< 20 km) for distances~100 km from a trench at the continentalmargin, and even if it were, low heat flow in the Great Valley would be left unexplained.However, preliminary resultsfrom a recent gravity study near Cape Mendocino by Robert Jachensand Andrew Griscom (oral communication,1980) suggestthat the North American plate might presentlybe thin enoughin the CoastRangesto satisfy this condition. Since an understandingof the nature of the San Andreas fault zone might depend upon a distinction between the relative contributionsof these two processes,we shall consider models for each. Models for Conductive Heating oftheCoast Ranges bya War.• Mantle Accordingto Dickinsonand Snyder[1979], shortly after the Mendocino Triple Junction contactedNorth America, the Farallon plate began to disappearbeneath the continental margin, leaving a growing triangular region, the 'window,' in which upwelled asthenospherewas left in contact with the originalbaseof the North Americanplate. The northern edge of the window is the inland projectionof the Mendocino transformfault; it moveswith the Pacificplate and hasnot been occupiedby a subductingslabfor more than 20 m.y. The temperature in the window (region x > a, z > 0, Figure 24) dependsupon the motion within it during that period, and this, in turn, dependsupon how the window material has been 6212 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE where we have used a thermal conductivityK of 6 mcal/cm s øC. In westernCalifornia, OH(a)is probablysmall, -•30øC + We shall consideronly the most extreme case of heating 50%. Allowing 15øC for the mean annual surfacetemperature whichoccurs whenthemechanical coup!ing at x -- a (z > 0) is and assumingthe mantle solidustemperatureat atmospheric complete(i.e., no slip); after the plate of thicknessa slidesoff pressureto be 1045øCyields the Gorda plate, it dragsthe underlyingmantle material with Os(O)- OM•a)-• 1000øC (54) it. This would generatea one-sidedburied spreadingcenterat the inland projection of the Mendocino transform as illus- We assumefurther that the solidusgradient ¾is 3øC/km and trated in Figure 24. Upwelling asthenosphere at z = 0 would the thermal diffusivitiya is 0.01 cm2/s. Equation (52) givesthe reducedheat flow at any point in be cooledand accreteto the thickeningNorth American plate as it moved southeastward. the North American plate at time t after passageof the MenThe 'initial' condition (at z -- 0q-, Figure 24) for quasi-one- docino transform (provided the Farallon plate disappeared dimensionalconductivecoolingis shownby the dashedcurves there:•rst and 'openedthe window'). In Figure 26, resultsare in Figure 25. The temperatureOs(x) in the North American shown for selectedinitial plate thicknessesa, with the time plate (of thicknessa) is the sumof a steadystatecontribution convertedto distancez by assumingthe relative velocity of OH(x)from radioactivityin the plate and a linear part I'x asso- the Pacific and North American platesto have been 5 cm/yr ciated with reduced heat flow. The temperaturein the up- for the past 10 m.y. and 2 cm/yr previously[Atwaterand Molwelling mantle below a is a linear function,0,,,-- Os(a)q- G(x nar, 1973]. Theoretical reduced heat flows have been increased0.2 HFU for radioactivity. Horizontal solid lines de- a), where Os(a)is the mantle solidus(at depth a) given by note means(Table 2) for the variousregionsof the fault zone Os(a)---Os(O) q- ¾a (49) delineatedin Figure 8. The two dashedlinesdenotemeansfor and G is a gradient, artificially introduced to accommodate data interior to the Great Valley in regions3 and 6; thesevalheat flow from below a. The transientpart of the problem in- ueswere adjustedby 0.2 HFU for effectsof rapid sedimentavolvesthe solid lines in Figure 25, which are denoted collec- tion. The mean for the Mojave region is shownonly for curitively by Or(x,0); it is the temperaturewhose(transient)gradi- osity; it could not be accommodatedby the model for the ent yields the reduced heat flow qr(t). The radioactive chronologywe have assumed,sinceits time origin (-•30 m.y. B.P.) predatesthe arrival of the Triple Junction at the conticontribution is additive. Thus the initial condition is: nental margin. Or(X, 0) • 0(X, 0) -- OH(X ) -• FX X< a (50a) It is seen that a simple model of this type fits the Coast Range data (solid lines, Figure 26) fairly well, provided the = 0(0) - (GNorth American•platethicknessa wasonly about 20 km there + Gx - OH(a) X> a (50b) while subductionwasin progress.The Great Valley data fit an initial plate thicknessa -• 40 km (dashedlines, Figure 26). The correspondingsolution to the differential equation of Hence accordingto the model, the rapid drop in heat flow beheat conductionwith zero surfacetemperatureis tween the Coast Rangesand Great Valley representsa rapid 1 steepeningof subductionthere. (This model could be used Or(X , t)--'I'X+ • [Os(O) -- (G- ¾)a -- OH(a)] also to estimatethe heat flow at a fixed point on the North American plate following disappearanceof the Farallon plate beneath southernand central California; for this purpose a larger initial heat flow might be appropriate;see Pilger and Henyey [1979].) The foregoingmodel illustratesthat extremeconditions(for + •(I'-G)a -a erf•-erf•-2 (4at)!/2 (4at)!/2 example, very shallow subductionand a full-fledged spreading center in the mantle at a depth •20 km) are required to account for the heat flow transition in the Coast Ranges by (*r)!/2 h heat conductionfrom a warm mantle. The consequences of the model are also extreme (a 'hot line' sweepingnorthward and the reducedheat flow at time t is given by under the edgeof North America), and perhapsthey could be confirmed by studiesin the Cape Mendocino area. If such • Os(O) - OH(a) G- 3* coupled to the original North American plate at their interface x ---- a. ßIerf(4at)•/2 +erf(4at)•/2 x+a x--a 1 1 {xlx+a (x-a)l 1 (4at)!/2 [e -l(•-"):/(4"t)l --e-l(•+"):/(4"t)l] } (51) qr(t) =1+ I'h qr(O) I' conditions do not exist, less extreme warm mantle models ß2 a e-[•/'4"ø'1-•-) (•!/2 (4at)!/2 ßerfc a q- 2 a (4at)!/•_ (•)!/2 (4at)!/: e-[a•/4at)] (52) Taking the initial reduced heat flow qr(O) to be the value today at Cape Mendocino (-•0.8 HFU) and assumingthe steady heat flow from below a to be 0.5 HFU, we obtain I' = 13.3øC/km (53a) G = 8.3øC/km (53b) would apply, and it is likely that a contribution from shearstrain heating (discussedin the next section)would have to be invoked, at least in the northern Coast Ranges.Lessextreme modelsinclude thosewith steepersubduction(for example, a >• 30 km, Figure 26) or those in which the North American plate becomesonly weakly coupledto the mantle (at x = a) when it passesthe Mendocino transform(z -- 0, Figure 26), thereby requiring lessintensedivergenceand upwelling. (If a shouldturn out to be even lessthan 20 km, weaker-coupling modelscould, of course,accountfor the anomaly without invoking shearheating.) Taken literally, the model of Figures 24 and 25 leads to a deep San Andreas fault that penetrates the entire North LACHENBRUCH AND SASS.'ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6213 and velocity is continuousacrossthe projectedplane of the main fault trace y -- 0. For physicalreasons,there must be a transitionallayer betweenh andf, we generallyexpecth to lie at or beneath the baseof the seismogeniclayer and j to lie at or above the top of the fluid asthenosphere. Anticipating the problemof accountingfor shallowheat sources,in Figure 27b a we illustrate the extreme case, with h at the base of the seismo- geniclayer and j only a few kilometersbeneathit at the base of a 'partial decouplinglayer' with large verticalvelocitygradients. Displacementsoccuronly in the positiveand negativez directions,and they are anti-symmetricin y. Hence our referenceframe is fixed in spaceas the primed and unprimed regions move past it in a right lateral sense.Since we are interestedin effectsof interactionbetweenthe primed and unprimed lithosphereplates, we assumethe material at great depth to be at rest relative to the referenceframe. In the seismogeniclayer AA', in which uniform fault slip is permitted, we assumethat v is independentof depth, i.e., plasticdeformation or secondaryslip can occuronly acrossplanesparallel to y -- 0. We assumefurther that no distortion of the seismogeniclayer occursbeyondy •- :l:y*. The regionsC and b C' are rigid and movewith velocities_+vp characteristic of the respectiveplate interiors.In summary, Ov Ov' ox •ox -- 0 vO')-----v'(--y) x< h (55a) all x (55b) vO'*)= -v'(-y*) = v•, Fig. 27. Steadystatemodelsfor plate interactionat a transform boundary;C and C' movewith velocityof plate interiors(+v•,):(a) v(O+) = -v'(O-)-- Vo minordecoupling at baseof seismogenic layerAA', (b) majorde- x< h (55c) x< h (55d) couplingat baseof seismogenic layer. > 0 x<j (55e) American lithosphere,a reasonablecondition during early stagesof the fault boundary betweenthin plates. However, subsequenteastwardjumps of the locus of surfacefaulting [Atwater and Molnar, 1973] may be more compatible with -- 0 x >-j (55/) continuous deformation across the North America-Pacific plate boundary at depth in the thickeninglithospheres,with shallow faulting confined to a superficialbrittle layer, partially decoupledat its base. (Shallow decouplinghas often beensuggested in the past;see,for example,Anderson[1971].) Whether or not this is so, it is shown in the next sections that h<j=• H v, v' --> 0 x --> oo (55g) (55h) In discussing this model, we shall refer only to the unprimed (y positive) portion, with the understandingthat the correspondingprimed velocitiesand stresses are equal in magnitude and reversedin signunlessotherwisenoted. We shahuse the outer normal; tractionsthat drive the fight lateral motion of A are positive,and thosethat drive fight lateral motion of sucha superficialfault configurationis probablynecessaryif shearheatingis to contributeappreciablyto the CoastRange anomaly without violating the heat flow constraint(13); the model has testableimplicationsfor fault zone stresses. A' are negative. The displacements illustratedin Figure 27 are valuesaveragedover timesvery muchlongerthan the cycleof recurrence yet know how large a contribution is required from a frictional source,it is worth investigatingthe conditionsunder which it could be appreciableand what they might imply during earthquakesare as small as seismicestimatesindicate, this shouldbe a usefulapproximation. The complexdeformationin the regionof plate interaction, AA', BB', DD', is a consequenceof the vertical variation in failure and deformationcharacteristics of the outer layersof the earth. The stresses ?cand ?½b exertedby C on A and B, respectively, are reactions to resistance in the shear zone of earthquakes. The associated stressstateis alsoa long-term averagecondition,and it is unchangedby the deformationasModelsfor a Mechanical Contributionof AnomalousHeat in sociatedwith thesevelocities.The work doneon sucha systhe CoastRanges temmustbe dissipated, andwe shahassume that the principal If shearheating contributesto the Coast Range anomaly sourcesof heat and seismicradiation dissipatedin the fault and friction on the fault is negligible,resistingstresses mustbe zone can be accountedfor by a steadymodel of this kind. If appreciablesomewherein the fault zone.Although we do not mostof the transienteffectsare elasticand if stresschanges about the mechanics of the fault zone and the resistance to plate motion at a transformplate boundary. For this purpose, we considerthe quasi steady state mechanical models of Figure 27 in which H representslithospherethicknessand h representsthe depth to which slip on the main fault trace is uniform and discontinuous. At some depthj (Figure27a) beneathh, the mediumdeformsductRely, AA'BB'DD'. We shah assume that the boundaries of C and C' have been selected so that %, •cb> 0 (56) 6214 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE (57a) va-- vb-- v•, •A A'I • b JD (57b) where y* is the half-width of the fault zone, and the correspondinganti-symmetricrelations apply for negativey. For small y, (57a) is required if vois continuousat y -- 0 and v• is not (Figure 27b), but the inequality could be reversedfor larger y (< y*) if most of the fight lateral shearingof B were concentratedin a thin shearzone at y -- 0, as is often assumed [see, e.g., Savage and Burford, 1973; Hanks, 1977]. In that case,the seismogeniclayer would be driven by the basal stress (Figure 28b), a condition that we shall show to be incompatiblewith the heat flow constraintif there is appreciable heat generation in the fault zone [see also Lachenbruchand Sass, 1973]. The fault slip velocity in the seismogeniclayer is given by • tA y = y* Vo= v•,- v,,(0+) (58a) Vo'= v; + (58b) • • The differencebetweenv•,and Vorepresents inelasticdeformation in A (vp,vp',Vo,andVo'aretreatedaspositivequantities.) The long-term resistanceto plate motion offered by the fault at y -- 0, x < h is the averageelasticfault stress•e, which in the notation of an earlier sectionis given by ! Fig. 28. Major alternativesfor flow of mechanicalenergyin zone of plate interaction:(a) subcrustBB' movesslower(in right lateral sense)than seismogeniclayer AA', which it retards,(b) BB' moves faster than AA', which it drives. Eobsrepresentslossof heat and seismic radiation energy.Arrows denotethe directionof steadyflow of energyfrom one regionto another. and that the neglectedbasal traction betweenC and D is unrelated to plate interaction. We are left with two principal optionsfor the steadyflow of mechanicalenergyinto and throughthe zone of plate interaction(Figure 28). They are distinguished by whetherB and 1 =+ from A and B to C. We denotethe shearingtraction at the baseof A by ,b and the velocityof A by va(y).The velocitya shortdistance(relative to h) belowA will be denotedby vb0')(Figure 27). Unless vb(y)-- va(y),A will be partiallyrecoupledfrom the restof B by thisthin zone,and energywill be dissipated within it at the rate ,b(vb- v•), a positivequantity, because•b will have the sign of the velocitydifference.Accordingto (48), heat generated in this partial decouplingzone beneaththe seismogenic layer might contributeappreciablyto the CoastRange anomaly, but heat generatedat greaterdepth (say >•20 km) probably couldnot. The rate of deformationin this zonemay be substantial,with vb(y)continuousat y = 0 (as in Figure 27b), or it may be slight,as in Figure 27a. Furthermorev,,may be greateror lessthan vb,and the signof their differencecould changethroughoutthe half-width y* of the fault zone. We shall be interestedprimarily in the followingcase(illustrated in Figures 27): (59a) -" ? -I'-'r,• (59b) •< 100 b + ,a (59c) We continueto usethe apparentstress*a and dissipativeresistance ? as positive parameters.Equation (59b) expressesthe fact that insofar as plate motion is concerned,seismicradiation and frictional heat both representenergy dissipatedby the mechanicalsystem. The average(tensor)componentof sheafingstress,y• in A is constrainedby the conditionof equilibrium: ,y:(y) = -%- •- B' drive(Figure28b)or resist(Figures28a and 27) the right lateralmotionof A andA', or equivalently,whetherA andA' are movingslower(Figure 28b) or faster(Figure 28a) than B and B' in a right lateral sense.(The 'brittle' modelof Lachenbruch and Sass[ 1973]is an exampleof conditionsin Figure 28a, and the 'ductile' model is an example of Figure 28b.) Condition (56) excludesthe possibilitythat arrowsin Figure 28 point = 0'*+ ,• dy (60) For purposesof illustration, we shall assumethat • is con- stant,determinedby a plasticyield stress; for conditions(57), ß• is negative,and ,b' is positive,asin Figure 28a. (The caseof ß• proportionalto (vo- v,,)hasbeendiscussed by Lachenbruch and Sass [1973]; T. C. Lee (personalcommunication,1979) has pointed out that a factor of 1/2 is missing before the bracesof equation 7 in that paper.) With this simplification, (60) yields Y *y,(y) =-%- •-*• (61a) and at y = y*, y* %----Ye- -•-Y• (6lb) Thus if •b is positive(Figure 28b), the greatestshearingstress in the seismogeniclayer occurson the fault and the least on the periphery;if ,• is negative(Figure 28a and Figures27), the minimum occursat the fault and the maximumat the periphery. Energeticsof Plate Interaction Let Er be the total rate of energydissipationper unit length at a symmetricaltransformplate boundary,and let Rr be the LACHENBRUCHAND SASS:ENERGETICSOF THE SAN ANDREASFAULT ZONE corresponding resistance (per verticalunit area)to platemo- Combining (59c), (6lb), and (67a), we can now expressthe heat flow constraint(13) as follows: tion. Then for an antisymmetricshearzone, Er-- 2hvp% + 2fcB%bv v ds -- 2Hv•,Rr 6215 (62a) 0<e= (62b) wherethe integralis takenoverthe boundarybetweenC and B, and %•,is the tractionthere. Applicationof (56) yieldsa + -•- 1- lower limit to plate resistance: •'•,•<1• bars (68) For the Coast Rangeswe selectthe follow•g values: ' h Rr> •% (63) 2v• =5cm/yr v•=0.6 2Vo= 3 cm/yr Let Eobs bethesumof the 'observable' losses of mechanically generatedheatand seismicradiationenergy.Assuming all of (69) the seismicenergyoriginateson the main fault trace,we have Eobs---2y*Aq+ 2hvo•a h = 14km y, 2y*=84• (64a) -- 2hv•obs v• T =3 (64b) whereAq is the averagevalueof the mechanically generated part of the broadheat flow anomaly,and Robs is the resistance to plate motion that is, in principle,observablefrom studiesof heat flow and earthquakes: which leadsto the follow•g relation(see(10)): 0<F[bars] 910Aq [HFU] 0.4%+3(1 -- - - •'•,_<-100 where stresses are expressedin bars. If the basalstress•-•,vanished, a large heat flow contribution would require a very largeapparentstress(--•2kbar/HFU); if •-•,werepositive(FigSincethe portion of Er generateddeep within the lithosphere ure 28b), % would have to be larger still. Hence a substantial cannot contributeto Eobs,(62b) and (64) yield a secondlower heat flow contribution from the shear zone would seem to imlimit to plate resistance: ply basal drag on the seismogeniclayer (i.e., •-•,negative,as in Figures27 and 28a). This requiresthat the minimum shearing h stressin the seismogeniclayer be the value at the main fault y* Aq Vo =- • • + •'ra (64c) Robs h vv vv Rr> • Robs (65) Accordingto (48), Eobsshould be approximatelyequal to the net rate at which work is done on the seismogeniclayer plus the underlying decouplinglayer after a few million years of transformfaulting. If we assumev•,is continuousat y -- 0 (as in Figure 27b), then Eobs --•2hvvr• + 2f oY'V•dy (66a) • 2hv•%+ 2y* •,r•, (66b) and the maximumbe the value at the periphery% (61). Using the equilibrium relation (61) and the energybalance (67), we can expressthe basal and peripheral stresseson the seismiclayer in terms of heat flow and the fault stresses and r-)as follows: -%,-1--• { vvif. 1---%+} vv -- 1- • (70a) {303Aq [HFU]-3[0.6% + r-'I} (70b) where •, is the averagevalue of v•,,and we have assumed•'•,to be constant. Combining (64) and (66) yields the energy balance for the seismogenicand decouplinglayers: hvv%= y*Aq + hvo%- y*•b y* Aq Vo %= --h •vv+ --•',, vv ..... y* V• --Robs h vv•'•, y* • - -- --•'•, h vv (67a) %...Ii_•_v'l-'{y' Aq(V•, Vo I v-• } vv ..... 1 • - 910Aq [HFU] •' v•, •' v•, %- --? (71a) (7lb) (67b) where stressis in bars in the dimensional forms, which incor- (67c) porate (69). It is seenthat thesestresses are very sensitiveto the ratio (vo/vv)of the averagespeedof the baseof the decoupling layer to the speedof the rigid plates. For convenience,we representv•,by The term on the left in (67a) is work doneby the plate in driving the seismogenic layer; the first two termson the fight are n> 0 (72a) the heat and seismicradiation that leave the system,and the last term is work done by the decouplingzone on the underfrom which it follows that lying crustif •'•,is negative(Figure 28a) or on the decoupling zone by the underlyingcrustif •'• is positive(Figure 28b). It is (72b) seen from (67c) that the sign of % determineswhether •-• is vv n+l greater or lessthan Robs,and consequently,whether (63) or (65) imposesa strongercondition on resistanceto plate mo- Small n implies large right lateral velocitiesin the subcrust, tion. and large n impliessmall ones,a conditiongenerallyassoci- 6216 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE ated with substantialdecouplingof the seismogenic layer (the amountdependingon the form of Va(y)).This relationis illustrated in Figure 29 where vois shownfor n = 1/2, 1, and 2. The solid curve labeled vavJreprese. nts motion of the seismo- flow anomaly for stressconditionson minor peripheral faults will depend upon more detailed considerationsof apparent stressand dynamic friction on suchfaults and their possible observableseismicand geothermalconsequences. If the pringeniclayer for a centralfault. As we have assumed•'bto be cipal interplate slip shouldoccurat the periphery of the shear constant,the form of va(y)entersthe analysisonly throughits zone, the high stressesrequiredby the central fault would not effecton the signof •'b(whichis the signof vb(y)- Va(y))and be eliminated; they would only be moved elsewhere [see on the fault slip Vo. Lachenbruchand Sass, 1980].For the symmetricalperipheral The leastvalue of •'coccursin the physicallyunrealisticcase faults illustratedby the dashedcurve VaVa', Figure 29, the central stress(at y -- 0) is the sameas the peripheralstress•'c(at y of completedecouplingTo-- 0, which leadsto -- +_y*) for the correspondingcentral fault provided n is re•'c> 910Aq [HFU] + 0.6•'a----Robs (73) placed by its reciprocal;under these conditions,the magniwhere stressis in bars. The greatestvalue of •'c occurs with tudesof the basal stressesare also the same.For a single pecompletecouplingat the baseof the seismogenic layer,i.e., v-o ripheral fault (at y -- -y*), Figure 30, which approximates • v%.Judgingfrom the measurements of Thatcher[1959] in conditionsat the southernend of the Coast Ranges,the stress at the oppositeside of the shearzone (y -- +y*, Figure 30) is the CoastRanges,•a/Vp"• 0.8--0.9,whichleadsto (7lb) the sameas that for the correspondingcentral fault (i.e., with (74) the samen), but the basalstressis reducedby a factor of «. In [bars]• (5-10 x 910Aq [HFU] Hence in the absenceof decoupling,appreciableheat genera- each of thesecases,the basal stressopposesfight lateral motion would require peripheral stresses of.several kilobars. If tion on the fault(s), for otherwise,the heat flow constraint the averagespeedat the baseof the decouplinglayer were 1/2 would be violated. Henceno matterwherein thesl•ear zonetheprincipal platevelocity(n = 1,Figure29), a conditionthatwouldrepre- faults are located, an appreciable mechanical contribution to the broad heat flow anomaly is possibleonly if the average shearstressin the seismogeniclayer increasesrapidly with distance from the fault(s), a testableconsequence. Decoupling at the baseof the faulted blockswould generallybe in the sense that resiststheir right lateral motion. Under theseconditions, the unobservedmaximum stressaway from the fault(s) 0'c for quired basalstress(70) is generallylessthan the peripheral the central fault) would determine the resistanceoffered by stress by a factor •1/3, becoming somewhat smaller for the seismogeniclayer to plate motion. If there is an appreciable heat flow contribution,this resistancewill generallybe larger %. Rearranging(71a), we can expressthe heat flow anomaly substantiallylarger than Robs(68) estimatedfrom measurements of seismic radiation and heat flow. The unobserved disby sipated energy representswork done by the decoupling layer sent substantialdecouplingin the Coast Ranges,peripheral stresses are still large for modestheat flow contributions.It is clear from (71) that the permissible amountof mechanically generatedheat will be limited by the conditionthat the peripheral stress•'c not exceedthe strengthof upper crustal rocks;the stressis insensitiveto '}'aand hence to whether the fault is weak always or only during earthquakes.The re- Aa=f;hvp{•c1-¾ø '•-q'a -- (75) on the material beneath it. The Form of the Heat Flow Anomaly;Independence of Distance Taking an upper limit for rock strengthunder normal fluid pressureof 1.5kbar (Table 4), the maximumpermissible contributionof mechanicallyderivedheat in the CoastRangesis Aq[HFU]--0.65 1.51-• +•'a•-- +?'•-•'•(76a) -•1-•}[HFU] (76b) where stressis in kilobars.This yields0.5 HFU for conditions illustratedby n = 1 (equation(72), Figure29), 0.33HFU for n = 1/2, and 0.1-0.2 HFU for completecoupling(n -• 1/4 - 1/ 5). Thus, unlessmost of the plate motion in the interplate shearzone is confinedto the seismiclayer (i.e., roughly, unless n > 1), it is unlikely that shearheating can accountfor more than about half of the steady state heat flow anomaly in the Coast Ranges. Even this amount of heating requires substantial decouplingat the base of the layer and large stresses at the periphery of the shearzone. These resultsraise questionsabout stressconditionson peripheral faults. In deriving them, we assumedthat all of the seismicenergy was generatedon the principal central fault, equivalentto assumingthat the heat-generatingstresson minor peripheral faults is the average ambient tectonic stress there. Implications of a broad mechanically generated heat From the Fault The foregoingdiscussionof energeticsdoesnot addressthe details of the heat production processnor, indeed, whether models of the type consideredcould reasonablycontribute a broad uniform heat flow anomaly throughoutthe fault zone, suchas the data from the California Coast Range suggest. The heat flow anomaly Aq(y), discussedonly as an average Aq in the last section,can be viewed as the sum of three contributions [seeLachenbruchand Sass, 1973]: Aq(y) • qyz(Y)+ qb(Y)+ q(Y) (77) For convenience,we consideragain the centrally faulted antisymmetricshear zone, for which it is only necessaryto describethe y positiveside.The termsin (77) can be expressed as -•--•-y = _h{•.e+y • %' dy qyz(y)._h•.yzdVa }dva (78a) qo(y)=--'}'b(Va- lib) (78b) .fo?q dy= ?Vo (78c) Here, qyzis the rate of heat generationper horizontalunit area resultingfrom inelasticdistortionof the seismogenic layer (for LACHENBRUCH AND SASS.'ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6217 Aq-- q----qy•+ qb vp • ,- h { Y} (l-y} (80a) • =- f;(v•,-Vo) %+•-•'•, -•Vo y, .... • 0 •' • • •• • • • • I =-•(v• v•)%-Vo•y which results • thedesired uniform anomaly (•dependent of • v0'• y)iftheslipvelocity Vo ishalftheplate velocity v•(80b). There isatriangular m•um at y=0ifthe slip velocity isfaster than half the plate velocityand a maxxurn if it is slower.The '•* Fig. 29. Velocity of seismogenic layer shownfor a central fault (solidcurveVa,Va')and symmetrical peripheralfaults(dashedcurve Va,Va')with three different velocitydistributionsVb,Vb'at baseof decoupling layer. possibility ofa m•um • qy•+ qoraises thequestion of whetherthis might 'hide' a more substantialcontributionfrom fault frictionq than we haveac•owledged• defiv•g the heatflowconstra•t ((13) and(38)) [seeLachenb•chandSass, 1973].In any case,the local anomalycouldnot exceedthe total broad anomaly of about 0.8 HFU, a conditionthat would relax the constra•t example,by plasticflow or subsidiaryfaulting), qois the rate of heat generation by flow in the hypothetical decoupling layer beneaththe seismogenic layer, and q is heat generated by friction on the main fault trace, discussedin detail in an earlier section.In (78a) and (78b) we have neglectedconductivesmoothingbetweenthe sourceand the surface,justified by the approximatenature of the calculationand the fact that the depth of the sources(x •< h) is small relative to the width of the shearzone (2y* ,-• 6h). From (78a) it is seenthat computationof the form of the heat flow contribution requiresinformation on the distortion of the seismogenic layer not requiredin the calculationof energy balance.The heat flow anomalyis more sensitivealsoto the y dependenceof vo,formerly enteringthe analysisas an average, and of •, which we have assumedto be constant. Reasonable assumptions regarding the formsof Va(y ) and vby), on fault friction to 400 bars (81) However,sucha minimumrequiresratherspecialconditions, and the absenceof an observable local anomalyin the contrastingtectonicand thermal regimesof northernand southern Californiamakesthisalternativeextremelyunlikely. SUMMARY We have summarized heat flow data from more than 100 sites within about 100 km of the San Andreas fault in Califor- nia. From thesemeasurements we havemadethreegeneralizations, the mechanicalimplicationsof which we have attempted to interpret: 1. Thereis no measurable localheatflow anomalyalong the 1000-kinlength of fault trace (Cape Mendocinoto San Bernardino) from which measurementsare available. 2. The heatflowis anomalously high(~2 HFU, ~80 mW and of •b, which will depend upon them both, must come m-2) throughout thesouthern 550km of theCoastRanges, a from an analysisof the controllingconstitutiverelations,an region that enclosesthe San Andreas fault in central Califoreffort beyond the scopeof this study. (Substantialdecou- nia. pling nearthe baseof the seismogenic layer will probablyre- 3. This broadCoastRangeanomalyfalls off rapidly toquire effectiveviscosities <•1022P at temperatures --•750øK, ward the Great Valley to the east,and northwestwardover a conditionsthat mightbe compatiblewith 'wet quartzite'and 200-km distanceto Cape Mendocino. 'limestone'but not the other mediadiscussed by Yuenet al. The first generalization(first discussedby Brune et al. [1978].)Without pursuingunwarrantedrefinements, we pre[ 1969])imposesa severeconstrainton the localenergetics of sent only one simple example to illustrate some general points. We haveshown(61)thatfor • positive, •y•increases asthe the main fault, and the secondand third might containinformation on the energeticsof plate interactionand the mechanical behavior of the broad shear zone. centralfault is approached, andfrom(78a) and(78b)it is seen The importance of generalization I to the mechanicsof that both qy•and qowould generallytend to contributeto a heat flow maximum(not observed)at y = 0. For • negative (Figure 28a), the two contributions tend to vary in opposite Vp senses,and •y• is a minimum at the fault. Hence a uniform heat flow anomalyis consistent with a relativelyweak fault. This is illustratedfor a simplemodelwith linear Vaand voin whichVa> Vband hence• is negative(Va,Va'is the solidcurve in Figure 29, and vo,vo'is the curven -- 1): Va=V0'•(VpVoy)• , v•--vpyY--, From (66), (79a) -Vp y* (79b)Fig. 30. 0 -y* Velocity of seismogenic layer Va,Va'for a one-sidedperipheral fault and three differentvelocitydistributionsvb,vo'at base of decoupling layer. 6218 LACHENBRUCH AND SASS: ENERGETICS faulting dependsupon the applicabilityof two conditions:(1) heat transfernear the fault is primarily by conduction,and (2) work done againstresistingforceson the fault surfaceis dissipatedprimarily as heat. If both conditionsapply, then generalization 1 implies that during events responsiblefor the preponderanceof fault motion, the average(in,stantaneous ) resistingstresson the fault surfacesdid not exceed,--100bars; we refer to this limit as the 'heat flow constraint.' If heat were removed convectivelyby water circulating in the fault zone (in violation of condition (1)), much larger resistingstresses would be permissible.However, we considerthis unlikely, becausethe heat dischargeof thermal springsnear the fault is negligible and heat flows are similar on each side of the fault, but hydrologicconditionsprobably contrastsharply. Undetected heat dischargeinto surface drainage at temperatures slightly above ambient is a possibility that may deservefurther study. Larger resisting stressesare permissiblealso if much of the energy dissipatedin overcomingthem is con- sumedby creatingnew surfacesor by other energysinks(in violation of condition2). Our estimate(basedon specificsurface energiesfrom Brace and Walsh [1962] and an average three-dimensionalfracture spacingof 1 micron in the gouge zone) suggests that conversionof work to surfaceenergyis too small by 2 orders of magnitude to affect the heat flow constraint. There is, however, a large uncertaintyin the rate of creation of surfacearea in gouge by faulting. Other possible energy sinks such as chemical and metamorphic reactions should be investigated,although we considerit unlikely that they will invalidate the heat flow constraint. Our interpretations are based on the assumptionthat the constraint is valid. The heat flow constraintprovidesan upperlimit (,--100bars) to the averagedynamic frictional resistancethat obtainedduring thoseeventsresponsible for mostof the slip on the San Andreasfault; this probably excludessmall earthquakesand isolated creep episodeslastingno more than 10,000yr. The importanceof the constraintlies in the informationit provides on the ambient tectonic stressstate. If almost all of the slip took place by steady uniform creep, the implications are straightforward;the resolvedtectonicshearstresson the fault could not, on the average,have exceeded,--100bars and consequently,tectonic stressdifferencescould not have exceeded a few hundred bars. If, however, most of the displacement occurred during earthquakes,it is the initial stressprior to the event, not the frictional stressduring the event, that indicates the ambient state of tectonic stress. The relation between the two is most easily discussedin terms of the frictional theory of faulting wherein the initial stressis bounded above by the static frictional strengthand boundedfrom below by the dynamic friction. The frictional strength is given approximately by the productof the coefficientof staticfrictionand the ambienteffective normal stresson the fault; the averagedynamic friction (constrainedby the heat flow) is the productof the coefficient of dynamicfriction and the (displacementaveraged)effective normal stresswhile faulting is in progress.The implicationsof the heat flow constraintare different dependingupon whether or not the ambient effectivenormal stresspersistsunchanged during the faulting event, i.e., upon whether the fault is weak all the time or only during an earthquake. If the effectivenormal stressis the samebefore and during an earthquake, then the initial stresswill not exceedthe dynamic frictional stressby a factor greater than the ratio of the static and dynamic coefficientsof friction, a quantity that is OF THE SAN ANDREAS FAULT ZONE probably of order unity (we have used 4/3). In this case,the frictional strengthand the resolvedtectonic(i.e., initial) shear stresscan be only slightlygreaterthan the • 100 barspermitted by the heat flow constraint, and the tectonic stressescannot be significantly greaterthanthosepermittedfor steadystate creep. Such low frictional stresses, are, however, lessby an order of magnitudethan thoseexpectedfrom laboratory results (Byerlee'slaw) in upper crustal rocks with normal (hydrostatic) fluid pressures.This strengthparadox can be resolved by postulating that the fault contains materials with anomalously low staticand dynamic friction (for example,hydrated clay minerals) or that near the fault the ambient fluid pressuresare near lithostaticand, consequently,the effectivenormal stressis very small. At present,however, there is little independent evidence that either of these conditions occur generally on the San Andreas fault. The paradox cannot be resolvedby explanationsof low strengththat do not also explain low dynamicfriction. For example,the strengthof small laboratory samplesmight be much greater than the strength of a large heterogeneous fault becauseof the greaterprobability of the juxtaposition of randomly distributed flaws and local stressconcentrationson the latter. This does not explain the low averagedynamic frictional resistanceimplied by the heat flow constraint. The paradox could, of course,be resolvedwithout contradictinglaboratory results,or by assuming a changingeffectivenormal stress,if the ratio of the coefficientsof staticto dynamic friction were an order of magnitude greaterthan we have supposed.However, this alternativeruns into the sameobservationaldifficulty as the oneswe shall next discuss. If the averageeffectivenormal stresswere an order of magnitude lessduring an earthquakethan it was beforethe earthquake, then the heat flow constraintwould be consistentwith strengthson the order of ,--1 kbar (as suggestedby laboratory resultsfor upper crustalrockswith normal fluid pressure)and with ambient tectonic stress differences of a few kilobars. There are severalplausible physical mechanismsthat might cause such large transient reductions in effective normal stress.The fluid pressuremight increaseduring an earthquake as a resultof frictional heatingand thermal expansionof pore fluid or by reduction of pore volume or dehydrationof clay minerals [Lachenbruch,1980;Sibson,1973;Raleigh, 1977],or perhapsthe effectivenormal stresscould be reduceddynamically by 'acousticfluidization,' as recently suggestedby Melosh [1979]. However, modelsof this kind are constrainedby implications from seismologyfor the energeticsof the earthquake process.The work done by local elastic forcesduring an earthquake must be accountedfor by heat and seismicradiation energy.Stated another way, the averageelasticstress operatingduring an earthquakeis the sum of a part allocated to the productionof heat (the dynamicfrictional resistance) and a part allocatedto the productionof seismicenergy(the 'apparent stress').It is known from seismicstudiesof many crustal earthquakesthat the initial stresscannot exceed the averageelasticstressby more than a few tensof bars(half the 'stressdrop'). Hence to a few tens of bars the initial (resolved tectonic) shear stresscannot exceed the frictional resistance (constrainedby heat flow to be • 100 bars) by more than the magnitudeof the apparentstress.Hence modelsleadingto an initial stressof ,--1 kbar require that the apparent stressbe of similar magnitude; they are inconsistentwith seismicestimates of apparent stress,which are generally •50 bars. Consequently,large ambient tectonic shear stresseson the fault are not permitted by the heat flow constraintunlessthe seis- LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE 6219 mic estimatesof apparent stress(and hence of total radiated energy) are very much in error for thoseearthquakesresponsible for most of the cumulative displacement,probably the very large ones. If we take the thermal and seismicobservational constraintsat face value, it is difficult to justify average tectonic shearingstresseson the fault much in excessof-,200 the foregoinginterpretationof generalizations2 and 3, mechanicalwork in this regionmustaccountfor the contribution to anomalousheat lossin the Coast Rangesand for seismic energyradiatedfrom the fault(s).Insofarasthe tectonicplates are concerned,both of theseforms of energyare dissipated and are associatedwith resistanceto plate motion.We identi- bars. fiedtheselosses with the work doneby the long-termaverage stresses operatingthroughthe long-termaveragevelocitiesat the boundariesof this region(the seismogenic layer plusthe hypotheticalunderlyingdecouplinglayer). We found that in In addition to theseimplicationsfor the stressregime in the immediate vicinity of the fault, the heat flow constraint has implications for stressesin the broad shear zone surrounding the fault when it is applied to interpretation of the broad Coast Range heat flow anomaly. The growth of this broad anomaly southwardfrom Cape Mendocino (generalizations2 and 3) may be interpreted as the transienteffectof the change in thermal conditionsthat took place at the migrating Triple Junction when subductiongave way to transform faulting. Since the transition to the high heat flow characteristicof the Coast Range is complete •200 km southeastof Cape Mendocino and sincethe Triple Junctionhasbeen migratingnorthwestward •50 km/m.y., this suggeststhat a steady state is reached in •4 m.y. and, consequently,that the sourcesof the anomaly are in the mid- and upper crust(•<20 km). This conclusion is consistentwith the rapid fall-off of heat flow toward the Great Valley to the east. There are two simple processes that might accountfor the anomalousobservedheat flow: (1) conductiveheating of the North American plate from below when it passessouthwardfrom a regionin which its baseis in contact with a cool subductingslab to a region in which it contactswarmer mantle material; (2) shear-strainheating associatedwith distortion and possiblebasal decouplingof the seismogeniclayer of the San Andreas fault zone which originates near the Mendocino Triple Junction. An understanding of the nature of the San Andreas fault might depend upon a distinction between the relative contributions of these two order to generate an appreciable heat flow contribution with- out violatingthe heat flow constrainton the main fault, it is necessarythat the (steadystate)basaltraction on the seismogeniclayerbe in the sensethat resists(not drives)right lateral fault motion.Under thesecircumstances, the shearingstress on planesparallel to the fault in the seismogenic layer is a minimumon the mainfault and a maximumat the periphery of the fault zone.The maximumpermissible mechanically generatedheatflowanomalyis limitedby howlargetheseperipheralstresses may be and by the amountof decoupling near the base of the seismogeniclayer. If the peripheral stresses are limited by the strengthof uppercrustalrockswith normal fluid pressure(• 1.5kbar), then the mechanicalcontri- butionto the broadheatflowanomalyprobablycouldnot exceed0.1-0.2 HFU unlessthe seismogenic layerwerepartially decoupled fromthe underlyingcrustat itsbase.If the average fight lateralvelocityof the baseof the decouplinglayer were about« plate velocity,shearheatingcouldprobablyaccount for about« of the steadystateheatflow anomalyin the Coast Ranges(•0.5 HFU). Theseresultsare rather insensitiveto the stress allocatedto seismic radiation0'a),andhencetheyapply whetherthe fault is weak all the time or only during earthquakes.Similarresultsapplyto a fault zonecontainingmore than one fault. Hence accordingto this simplifiedmodel,an processes. appreciable mechanical contribution to the broad heat flow To explain the heat flow transition by the first processwithout a contribution from shearheating, extreme conditionsare necessary,for example,shallowsubductionat a depth of only 20 km beneath the Coast Ranges and complete coupling of the North American plate to the underlying mantle when subduction stops.This would resultin a buried one-sidedspreading center migrating northward (with the Mendocino transform) as a 'hot line' beneath western North America. The lithosphere beneath the Coast Ranges would thicken southward (from 20 km near Cape Mendocino today) by thermal accretion.Taken literally, the model leadsto a deep San Andreas fault penetratingthe entire lithosphere;it is too simple to account for the inland migration of the locus of surface faulting. Nevertheless,it accountsreasonablywell for the distribution of observedheat flow; the rapid transition toward the Great Valley could be accommodatedby a rapid steepening of subductionto 40 km (from 20 km beneath the Coast Ranges). If field evidence for extreme conditions of the sort implied by this model cannot be found, it is likely that an appreciable contribution from shear-strainheating will have to be invoked. Although this requirementis speculativeat present, it is of interest to considerits implications, when taken with the heat flow constraint, for the mechanics of the fault zone and plate interaction. We consideredmodelsin which the anomalousheat is generated in a broad shear zone (correspondingto the Coast Ranges)by distortionin the seismogenic layer (• 14 km thick) combined with a possiblecontribution from shear flow in a horizontal partial decouplinglayer at its base. According to anomalyshouldbe associated with a rapid increasein mean shearingstress with distance fromthe principalfault(s),a testable consequence. If heat generationis significant,the seismogenic layer couldoffer appreciableresistance to plate motioneventhoughstresson the main fault were negligible; the principal resistingsurfacewould be the horizontal baseof the seismogenic layer. APPENDIX A: GROSS CONDUCTIVE OF FRICTIONAL VERTICAL HEAT FROM EFFECTS A STRIKE SLIP FAULT We consider effectsin the half-space x > 0 of heat conductedfrom a verticaldistributionof sourcesof strengthQ(x) commencingat t = 0 on the fault plane y - 0. The medium is initially at zero temperature, and the surface x -- 0 is maintained at zero temperature (Figure 1). A general expressionfor temperature 0 is obtained by integration of the transient result for continuoushorizontal line sourceslying in the fault plane [Carslawand Jaeger, 1959, p. 261, equation 5]. O(x, y,t)=4--• I --El-- nat +Ei{-ya+ (x' +x):.}-Q(x')dx' (AI) 4at wherethesecond termin theintegral represents a negative imagesourceas requiredto meetthe zero-temperature condition on x = 0. 6220 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE The Caseof ConstantSourceStrengthin the Depth Range (x,,x9 S(u,m) -- -•r1 For this case,Q(x) = Qc,a constant,whereXl -< x _<x2, and Q(x) = O,where0 < x < Xl, x2 < x. A changeof variablesin (A1) leadsto (A1la) v2+ 1 [tan-' u-um2] u:m 2<<1 (A1 lb) 4•rk o(x,y,t ) 4•t[ f{x•+x,/{4•t,l,• Qc • (Xl+X)/(4at)l/2• (Xl--X)/(•t)l/2 1 -• •-(1- erfm) u-• oo .[_(m: +f12)] dr} where e•m2(v2+l ) d¾ •o u (A2) (A1lc) Tables of the exponentialintegral and the error function are, of course, available in standard handbooks. The function S m2= •/(4at) (A3) hasbeendiscussed and tabulatedby Smith[1953]and Lachenbruch [ 1957]. The •tegral • (A2) may be •tegrated as foHows: Applying (A9), (A10), and (A11) to (A8) yieldsthe steady •E,(_m •_•2) d•=[bE,(-m •- b•)- aE,(-m 2- a2)] state result: O(x,y, oo)= Qc {(x2q-x)il'l[y2q-(x2q-x)2]_ (xI q-x) (A4) The integral on the right • (A4) may be written ßIn [y2q-(x I q-x)2]_ (x2_ x) In [y2q-O•2-- X)2] -2 + (x, - x) tn Iv2 + (x, - x):1 m:+ m2 ---2fab e-m2-t•2d• +2fab e-m2-t•2 m2 q'fl•dfl (A5) +2y Itan-I x2 +x tan2x•+x tan_ Ix2 - x XI--X]} Y but --2 be-m2-t•2d• -' -e -m2• [erfb - erfa] -- (A6) and • Y + tan-• Y , (A12) Y The temperaturein the planeof the fault is obtainedby setting y = 0 in (AS) 2f•t, e-m:-t•: m2 +2 fl•dfl =2•m ,m m IS{mb-}- S{•,m)1(A7) where the function S is descried below. ApplyMg (A4), (A5), and (A6) to (A2) yieldsthe desiredgeneralresultfor the transient temperaturerise M the half-space: -•- (4ott)l/2 4ott 4at2 (x:x)Ei-{ (x:x) x•+x{ (x•+x) (A8a) + •I'4'• --X)2}I 1I- erf(4at),/2 X2q-X Xl--X { (X (4at)l/2 Ei- where F(a; y, t) ---aEi y2+ a2 4at -- • e-?/(4at) erf • (4at)•/2 + 2•ryS a y y' (4at)'/• XIqX+erf(4at)•/-----X2--•X-erf(4at),/2 Xi--X•j } (A13) +erf(4at)l/2 (ASb) The functionsin the three termson the right in (A8b) are respectivelythe exponentialintegral, the error function, and the 'Smith' function; they can be representedas follows: •o dv E•(-u) = - 0.5772- In u - u erf u-=• = • (A9a) e-• • 2 e-v• dv u u2<< 1 u2 << I 4at (4at),/2 Ei- O(x, y,t)= 4• Qc[F(xe + x;y,t) -- F(x•+ x;y,t) -- F(x 2 -- x; y, t) + F(x, - x; y, t)] 2(•r)1/2 (4at)l/2 Ei- The factor in front of the bracesis the correspondingresultfor an infinite plane source,and the expressionin bracesrepresentsedgeeffects,which,of course,ultimatelydominateas the steadystate is approached. The Caseof a Linear Increasein SourceStrengthWith Depth For convenience, we take the linear increase to be of the following form: (A9b) (A10a) x Q(x)= Q*•-; Q(x) = o XI < X< X2 o < x < x, x2 < x The completetransientsolutionfor this caseis unwieldy,and (A lob) we solveonly for the temperatureon the fault plane y -- 0. Substituting the aboveexpression in (A l) yields LACHENBRUCH AND SASS'.ENERGETICS OF THE SAN ANDREAS FAULT ZONE In the steadystate,this reducesto • fxx2 O(x, O,t)--4,rkx* (x+x')2 I _E,I_ (x'- E,{ • 4at + - 4at x'dx' The transient build-up of heat flux q at the earth's surface (x -- 0) is obtainedby differentiationof (A1): Q..O(x, O,t)--4at x (Xl+X)/(4at)1/2 q(y, t)---1•x:e-{x':+?•/{4"'• x,2x' +y•Q(x') dx' (4at)!/2 • (xl--X)/(•t) / (A21) For the caseof a constantsourcestrengthQo this yields •f(x•+x)/(4•t) '/• f(x•-x)/(•t) '/• • (A20) Heat Flow at the Earth's Surface(x -- O) 4•rkx* [ • (x 1+x)/(4at)1/l 8,rk (x• -+x) +8x•x O(x, O, oo) _-Q*x•I I(x,_'_x') ln'•Y (X'x)'1 (AI4) and a changeof variablesleadsto ,r/, ,-.,, 6221 V(Xl--X)/(•t) 1/2 From (A4) and (A6) the•tegrals with• the bracketsyieldresults of the fore: • E E,•t ]-E'4at q(y,t)---Qc (x•2+y21 ( x!2+y2}l (A22a • 2,r Q__sc Inx22 +-••-• t--}oo x,'-+y: x E,(-•2)d•-- (4at)!/2 x (4at)!/2 - aE,(-a 2) - • [erf b - erf a]} (A16) In the firsttwo integralsof (A15), integrationby partsyields (A22b) For the case(correspondingto (A19)) of a linearly increasing sourceextendingfrom the surfaceto depth x,, manipulations similar to those used above lead to resultsof the following form: Q*{4•• e-?/•4'"> X2 q(y, t)=2•--•x• erf(4at), a E,(-•2)• d•--• [b2E,(-b •)- a2E,(-a 2)+e-b:- e-a:] (A17) (A23a) - 2½ryS and the final result is • O(x,O,t) = I - • tan-n • Q*1{(x•: _x• ) t '-• oo (A23b) x2 8,rk x* where S is the 'Smith' function describedearlier (A11). ßIEi(-(x•+x)•-Ei I-(x2-x) • APPENDIX B: AN APPROXIMATION LAW FOR FRICTIONAL FROM BYERLEE'S STRENGTH IN THE UPPER CRUST Byedee's[1978] summary of extensivelaboratory measurementsof stresson slidingrock surfacesshowsthat for a wide variety of rock materials,the maximumshearingstresson the slidingsurface•m attainedduringeachexperimentcan be expressedin termsof the normalstresso. on that surfaceas fol- + 4at[e-(x:+x):/(4at) -- e • e-(X•+X):/(4at) .{.e-(X•-X):/(4at)] X2 t lows: X a +2•4•/--•-•,at x•err(4at), a+erf(4at)l X2+X x,+x - erf (4at),/-------• - erf (4at),al •'m= C +/•(o,,- (A18) where C= 0 /• -• 0.85 if The most useful case,in which the sourcestrengthincreases from zero at the surfaceto Q* at depthx2, resultsin a slight C = 0.5 kbar simplification (by settingxl -- 0, x2 -- x*): Q*x-• 1{(x22_x2 ) O(x, o,t)- 8•rk .IE,{_ (x: +x): -E,((x:x):'}l 4at (B la) 0.1 kbar < (o, - P) < 2 kbar (Blb) /• = 0.6 if 2 kbar <(o, - P) < 15 kbar (Blc) Pore fluid pressureis denotedby P, and the expressionin parenthesisis effectivenormal stressdenotedhereafter by For analytical convenience,we use the following approximation for (B 1) [seealso Zoback and Zoback, 1980]: 4at + 4at[e-•: +•?/•'• - e-•: -•:/{•'•] +2• P) xIerf(4at), a+erf(4at),/2 •t (A19) x,-x •m • •o,,' o,,' < 6 kbar (B2a) 0.75 (B2b) This relation doesnot depart from (B 1) by more than about 11% for a range of effectivenormal stressfrom 0 to 6 kbar, 6222 LACHENBRUCH AND SASS: ENERGETICS OF THE SAN ANDREAS FAULT ZONE sufficientto accommodateconditionsin the upper 15 km or so of the crust.If we let 4, be the angle formed by the trace of the fault and the directionof greatestprincipal stressOl, the effective normal and shearstresses on the fault plane are given by [e.g., Jaeger, 1962] on the TectonicProblemsof the San AndreasFault System,pp. 136148, StanfordUniversityPress,Palo Alto, Calif., 1973. Benfield, A. E., A heat flow value for a well in California, Am. J. Sci., 0.'= « (0•' + 03') - « (Ol' - 03')cos20 (B3a) Berry,F. A. F., High fluid potentialsin California CoastRangesand their tectonic significance,Am. Assoc.Pet. Geol. Bull., 57, 1219- rm = « (Ol' - 03) sin 20 (B3b) 245, 1-18, 1947. Berg, C. A., Relation betweenstressdrop, fault friction, and crustal strengthin shallowfocusstrike slip faulting, J. Geophys.Res., 73, 2217-2223, 1968. Combination of (B2a) and (B3) yields ' O)• ø-2L = 1- p, 2(sin o. sin 20 (B4) Assumingfault failure occursin the most favorably oriented fracture direction, 1249, 1973. Blackwell, D. D., Heat flow and energy lossin the Western United States, Geol. Soc. Am. Mem., 152, 175-208, 1978. Brace, W. F., Permeability of crystalline and argillaceousrocks: Statusand problems,Int. J. Rock Mech. Min. Sci., in press,1980. Brace,W. F., and D. L. Kohlstedt, Limits on crustalstressimposedby laboratoryexperiments,J. Geophys.Res.,85, this issue,1980. Brace, W. F., and J. B. Walsh, Some direct measurementsof the surface energyof quartz and orthoclase,Am. Mineral., 47, 1111-1122, 1962. 0= •- •-- tan -1 (B5a) Brook, C. A., R. H. Mariner, D. R. Mabey, J. R. Swanson,M. Guffanti, and L. J.P. Muffler, Hydrothermal convectionsystemswith reservoirtemperatures_>90øC,U.S. Geol. Surv. Circ., 790, 18-85, 1979. -- 26.6ø • = 0.75 (B5b) which leadsto the following generalrelations[seealso Brace and Kohlstedt,1980]: Brune, J. N., The physicsof earthquakestrongmotion, in Seismic Risk and EngineeringDecisions,Develop.in Geotech.Eng., 15, edited by E. Rosenbluthe and C. Lomnitz,pp. 141-177,Elsevier,New York, 1976. Brune, J. N., T. L. Henyey, and R. F. Roy, Heat flow, stress,and rate of slip alongthe SanAndreasfault, California,J. Geophys. Res., 74, 3821-3827, 1969. 1.6 Burridge,R., and G. S. Halliday, Dynamic shearcrackswith friction as models for shallow focusearthquakes,Geophys.J. Roy. Astron. (B6a) Soc., 25, 261-283, 1971. 01t 0'3t - 4.0 (B6b) Byedee,J. D., Staticand kinetic friction of granite at high normal stress,Int. J. Rock Mech. Min. Sci., 7, 577-582, 1970. Byedee,J. D., Friction of rocks,Pure Appl. Geophys.,116, 615-626, 1978. Oavg ! 0.64 (B6c) Cardwell, R. K., D. S. Chinn, G. F. Moore, and D. L. Turcotte, Fric- tional heatingon a fault zonewith finite thickness,Geophys. J. Roy. Astron. Soc., 52, 525-530, 1978. where Oavg t •---' «(01 ! "['03t) (B7) Thus the frictional strength(B 1) can be expressedin terms of the effectivenormal stress,the greatest,least, and averageeffectiveprincipal stresses and the differencein horizontalprincipal stressesas follows: rm'--'0.75 o,' (B8a) = 1.2 03t (a8b) = 0.30 Ol' (B8c) = 0.48Oavg' (B8d) ---0.40 (o, - 03) (B8e) Carslaw, H. S., and J. C. Jaeger,Conductionof Heat in Solids,386 pp., Oxford University Press,New York, 1959. Combs,J., Heat flow and geothermalresourceestimatesfor the Imperial Valley, in CooperativeGeological-Geophysical-Geochemical Investigations of GeothermalResourcesin the Imperial ValleyArea of California, pp. 5-27, Education ResearchService, University of California, Riverside, Calif., 1971. Dickinson,W. R., and W. S. Snyder,Geometryof subductedslabsrelated to San Andreas transform, J. Geol., 87, 609-627, 1979. Eaton, J. P:, W. H. K. Lee, and L. C. Pakiser, Use of micro-earth- quakes in the study of the mechanicsof earthquake generation alongthe San Andreasfault in centralCalifornia, Tectonophysics, 9, 259-282, 1970. Graham, S. A., and W. R. Dickinson, Evidence for 115 kilometers of right-slip on the San Gregorio-.Hosgrifault trend, Science,199, 179-181, 1978. Hanks, T. C., Earthquakestressdrops,ambient tectonicstresses and stresses that drive plate motions,PureAppl. Geophys., 115,441-458, 1977. Acknowledgment. We are very grateful to Robert Jachensand Henyey, T. L., Heat flow near major strike-slipfaults in California, Andrew Griscom for advisingus of the unpublishedresultsof their Ph.D. thesis, Calif. Inst. of Technol., Pasadena, Calif., 1968. recent gravity study near Cape Mendocino. In addition, we thank Henyey,T. L., and G. J. Wasserburg, Heat flow nearmajor strike-slip ThomasHanks, Barry Raleigh, Wayne Thatcher,Malcolm Johnston, faults in California, J. Geophys.Res., 76, 7924-7946, 1971. JamesSavage,Joe Walsh, and Tanya Atwater for helpful discussion Irwin, W. P., and I. Barnes,Effect of geologicstructureand metamorand suggestions. phic fluidson seismicbehaviorof the San Andreasfault systemin central and northernCalifornia, Geology,3, 713-716, 1975. REFERENCES Jaeger,J. C., Elasticity,Fractureand Flow, Methuen, London, 1962. Anderson, D. L., The San Andreas fault, Sci. Am., 225, 52-68, 1971. Jennings, C. W., Fault map of California,scale1:750,000,Calif. Geol. Andrews, D. J., Application of rupture propagationtheoriesto earthMap Set., map 1, 1975. Johnson,A. G., R. L. Kovach, and A. Nur, Fluid-pressurevariations quakes, U.S. Geol. Surv. Open File Rep., 78-380, 33-70, 1978a. and fault creep in central California, Tectonophysics, 23, 257-266, Andrews, D. J., Coupling of energy betweentectonicprocesses and 1974. earthquakes,J. Geophys.Res., 83, 2259-2264, 1978b. Kanamori, H., and D. L. Anderson,Theoretical basisof someempiriAtwater, T., and P. Molnar, Relative motion of the Pacific and North cal relationsin seismology,Bull. Seismol.Soc.Am., 65, 1073-1095, American plates deducedfrom sea-floorspreadingin the Atlantic, 1975. Indian, and south Pacific Oceans,in Proceedingsof the Conference LACHENBRUCH AND SASS:ENERGETICS OF THE SAN ANDREASFAULT ZONE Lachenbruch,A. H., Thermal effectsof the ocean on permafrost, Geol. Soc. Am. Bull., 68, 1515-1530, 1957. Lachenbruch,A. H., Heat flow in the Basin and Range Province and thermal effectsof tectonicextension,Pure Appl. Geophys.,117, 3450, 1978. Lachenbruch,A. H., Frictional heating,fluid pressure,and the resistance to fault motion, J. Geophys.Res., 85, this issue, 1980. Lachenbruch,A. H., and J. H. Sass,Thermo-mechanicalaspectsof the San Andreasfault system,in Proceedings of the Conferenceon the TectonicProblemsof the San Andreas Fault System,pp. 192205, Stanford University Press,Palo Alto, Calif., 1973. Lachenbruch, A. H., and J. H. Sass, Heat flow in the United States 6223 Sass,J. H., D. D. Blackwell,D. S. Chapman,J. K. Costain,E. R. Decker, L. A. Lawvet, and C. A. Swanberg,Heat flow from the crustof the United States,in PhysicalProperties of Rocksand Minerals,edited by Y. $. Touloukian,W. R. Judd, and R. F. Roy, McGraw-Hill, New York, in press,1980. Savage,J. C., and R. O. Burford, Geodetic determination of relative plate motion in central California, J. Geophys.Res., 78, 832-845, 1973. Savage,J. C., and M.D. Wood, The relationbetweenapparentstress and stressdrop, Bull. Seismol.Soc.Am., 61, 1381-1388, 1971. Savage,J. C., W. H. Prescott, M. Lisowski,andN. King, Deformation acrossthe SaltonTrough,California,1973-1977,J. Geophys. Res., and the thermal regime of the crust,in The Earth's Crust, Geophys. 84, 3069-3079, 1979. Monogr. Ser., vol. 20, edited by J. G Heacock,pp. 626-675, AGU, Scholz,C. H., J. Beavan,andT. C. Hanks,Metamorphism, argondeWashington,D.C., 1977. pletion,heat flow, and stresson the Alpine fault, J. Geophys. Res., Lachenbruch,A. H., and J. H. $ass,Models of an extendinglitho84, 6770-6782, 1979. sphereand heat flow in the Basin and Range province, Geol. Soc. Sibson,R. H., Interactions betweentemperature and pore-fluidpresAm. Mern., 152, 209-250, 1978. Lachenbruch,A. H., and J. H. Sass,Heat flow and energeticsof the San Andreas fault zone, U.S. Geol. Surv. Open File Rep., 80-625, 47-146, 1980. Lachenbruch, A. H., J. H. $ass, and $. P. Galanis, Jr., New heat flow resultsfrom southernCalifornia (abstract),Eos Trans. AGU, 59, 1051, 1978. sureduringan earthquakefaultingand a mechanism for partial or total stressrelief, Nature, 243, 66-68, 1973. $ieh, K., Study of Holocenedisplacement historyalong the southcentralreachof the San Andreasfault, Ph.D. thesis,219 pp., Stanford Univ., Stanford, Calif., 1977. Smith, R. C. T., Conduction of heat in the semi-infinite solid with a shorttableof an importantintegral,Aust.J. Phys.,6, 127-130,1953. Lee, T. C., and T. L. Henyey, Heat flow through the southernCaliforStesky,R. M., Rock friction-effectof confiningpressure,temperature, nia borderland, J. Geophys.Res., 80, 3733-3743, 1975. and pore pressure,PureAppl. Geophys., 116, 690-704, 1978. McGarr, A., S. M. Spottiswoode,and N. C. Gay, Observationsrele- Stesky,R. M., and W. F. Brace, Estimation of frictional stresson the vant to seismicdrivingstress,stressdrop,and efficiency,J. Geophys. SanAndreasfault fromlaboratorymeasurements, in Proceedings of Res., 84, 2251-2261, 1979. Melosh, H. J., Acoustic fiuidization: A new geologic process?,J. Geophys.Res.,84, 7513-7520, 1979. Murata, K. J., T. W. Dibblee, Jr., and J. L. Drinkwater, Thermal ef- fectsof largebodiesof intrusiveserpentiniteon overlyingMonterey shale; Southern Diablo Range, Cholame area, California, U.S. Geol. Surv. Prof Pap., 1082, 1979. Nut, A., and J. D. Byedee,An exacteffectivestresslaw for elasticdeformation of rock with fluids,J. Geophys.Res., 76, 6414-6419, 1971. Olmsted, F. H., P. A. Glancy, J. R. Hatrill, F. E. Rush, and A. $. VanDenburgh,Preliminaryhydrogeologicappraisalof selectedhydrothermal systemsin northern and central Nevada, U.S. Geol. Surv. Open File Rep., 75-56, 1975. O'Neil, J. R., and T. C. Hanks, Geochemical evidence for interaction along the San Andreas and Gatlock faults of California, J. Geophys.Res., 85, this issue,1980. Pilger, R. H., Jr., and T. L. Henyey, Pacific-North American plate interaction and neogene volcanism in coastal California, Tectonophysics,57, 189-209, 1979. Rennet, J. L., D. E. White, and D. L. Williams, Hydrothermal convection systems,U.S. Geol. Surv. Circ., 726, 5-57, 1975. Richards,P. G., Dynamic motionsnear an earthquakefault: A threedimensional solution, Bull. Seisrnol.Soc. Am., 66, 1-32, 1976. Roy, R. F., D. D. Blackwell,and E. R. Decker, Continental heat flow, in The Nature of the Solid Earth, edited by E. C. Robertson, pp. the Conference on TectonicProblemsof the SanAndreasFault System,pp. 206-214, StanfordUniversityPress,Palo Alto, Calif., 1973. Summers,R., and J. Byedee,A noteon the effectof fault gougecomposition on the stability of frictional sliding, Int. J. Rock Mech. Min. Sci. Geornech.Abstr., 14, 155-160, 1977. Taylor, H. P., Jr., Oxygenand hydrogenisotopestudiesof plutonic granitic rocks,Earth Planet. Sci. Lett., 38, 177-210, 1978. Thatcher,W., Systematicinversionof geodeticdata in centralCalifornia, J. Geophys.Res., 84, 2283-2295, 1979a. Thatcher,W., Horizontalcrustaldeformationfrom historicgeodetic measurements in southernCalifornia, J. Geophys.Res., 84, 23512370, 1979b. Wang, C.-Y., and N.-H. Mao, Sheafingof saturatedclays in rock joints at high confiningpressures, Geophys. Res.Lett., 6, 825-828, 1979. Waring, G. A., Thermal springsof the United Statesand other countriesof the world, U.S. Geol.Surv.Prof. Pap., 492, 1965. Wu, F. T., Mineralogyand physicalnatureof claygouge,PureAppl. Geophys.,116, 655-689, 1978. Wyss,M., Stressestimatesfor South Americanshallowand deep earthquakes,J. Geophys.Res., 75, 1529-1544, 1970. Wyss, M., and J. N. Brune, Seismicmoment, stress,and sourcedi- mensionsfor earthquakesin the California-Nevadaregion, J. Geophys.Res., 73, 4681-4694, 1968. Yuen, D. A., L. Fleitout, G. Schubert,and C. Froidevaux, Shear de- 506-543, McGraw-Hill, New York, 1972. formationzonesalongmajortransformfaultsandsubducting slabs, Rutter, E. H., and D. H. Mainprice, The effect of water on stressreGeophys.J. Roy. Astron. Soc., 54, 93-119, 1978. laxation of faulted and unfaulted sandstone,Pure Appl. Geophys., Zoback,M. L., and M.D. Zoback,Faultingpatternsin north-central 116, 634-654, 1978. Samreel,E. A., Occurrenceof low temperaturegeothermalwatersin the United States, U.S. Geol. Surv. Circ., 790, 86-131, 1979. Sass,J. H., A. H. Lachenbruch, R. J. Munroe, G. W. Greene, and T. H. Moses,Jr., Heat flow in the westernUnited States,J. Geophys. Res., 76, 6376-6413, 1971. Nevada and strengthof the crust,J. Geophys. Res., 85, 275-284, 1980. (ReceivedFebruary 27, 1980; revisedMay 2, 1980; acceptedJune 25, 1980.)