JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 103, NO. B12, PAGES 30,101-30,115, DECEMBER 10, 1998 Location of the southern edge of the Gorda slab and evidencefor an adjacent asthenosphericwindow: Results from seismic profiling and gravity Bruce C. Beaudoin,John A. Hole,• and Simon L. Klemperer Department of Geophysics,StanfordUniversity,Stanford,California Anne M. Trdhu Collegeof Ocean and AtmosphericSciences,Oregon State University,Corvallis Abstract. As the MendocinoTriple Junctionmigratesnorthwardalong the California margin it is widely presumedto leave a "slab-free"or "asthenospheric" window in its wake. A 250-km-longsouth-northseismicrefraction-reflectionprofile crossingthe transitionfrom transformto subductionregimesallowsus to compareand contrastcrust and upper mantle of the North American margin before and after it is modifiedby passageof the MendocinoTriple Junction.From the seismicdata we have determined that (1) the crustis laterallyhomogeneous in velocityto a depthof 20 km (interpretedby us as Franciscancomplex),(2) below20 km depththe crustis characterizedby velocities of >-7.0 km/s for the southernhalf of the profile and by velocitiesof -<7.0 km/s for the northernhalf, (3) regionsof high reflectivityin the crustoccurbelow -20 km depth throughoutthe profile,and (4) the North Americancrustis thickest(-35 km) in the center of the profile and thins to -25 km at either end. From the gravitydata we have determined that(1) asthenospheric densities (3.2g/cm 3) occursubjacent to theNorth Americancrustin the centerof the profile, and (2) a wedgeof lithosphericmantle density material(>-3.2g/cm 3) isrequired onthesouthern endof theprofile.We interpret these combinedresultsto indicatethat our profile crossesthe southernedge of the Gorda plate and that directlyadjacentto this edge is an asthenospheric windowwith overlyingmafic rocksin the crust.These mafic rocksand a reforminglithosphericmantle increasein thickness southward. lithosphericscale,however,is only now being examinedwith seismicrefractionand reflectiontechniques. Northern California tectonicsare governedby the interacIn the currentlymost popular model for evolutionof the tions of the Pacific,Gorda, and North American plates.(We MTJ, first formulatedin a rigid-platemodelby Dickinsonand use the term Gorda plate while recognizingthat recentwork Snyder[1979],an importantconsequence of MTJ migrationis suggests that Gorda deformationzone may be more appropri- that the North American plate slidesoff the Gorda plate, ate [Wilson, 1989].) The three plates come together at the leavingin its wake a void that is filled by upwellingasthenoMendocinoTriple Junction(MTJ) (Figure 1), a transform- sphere,referred to as the slablesswindow or slabgap [Dicktransform-trenchtriple junctionthat wasinitiated at 25-29 Ma insonand Snyder,1979;Severinghaus andAtwater,1990].Supby the collisionof a Pacific spreadingcenter with the North porting evidencefor a slablesswindow existsfrom heat flow American continent [Atwater,1970]. As the Juan de Fuca- data [Lachenbruch and Sass,1980],gravityand magneticdata Pacificplate boundarymigratesnorthward relative to North [Griscomand Jachens,1989; Jachensand Griscom, 1983], America, subductionalong the Cascadiasubductionzone is teleseismicP wave delay studies[Benzet al., 1992; Verdonck replacedby the transformboundaryof the San Andreasfault andZandt, 1994;Zandt, 1981;Zandt andFurlong,1982],shear system(Figure 1). Today, the MTJ is definedby a broad on- wave velocities[Levanderand Kovach, 1990], and changesin shoreregion near Cape Mendocino[Clarke, 1992] and pro- volcanism[Dickinsonand Snyder,1979;Fox et al., 1985;Furvidesa uniquelaboratoryto studyplate interactionsat a triple long,1984;Johnson andO'Neil,1984;ZandtandFurlong,1982]. junction.The transitionfrom a subductionregime to a trans- Thesestudiesfurthersuggest that (1) the observeduppermanform regimehaslongbeen recognizedfrom geologicalstudies, tle low-velocityzone is smaller and displacedto the east comand its subsurfacemanifestationshave been recognizedfrom pared to that predictedby rigid-platemodels;(2) the Gorda upper mantlevelocitytomographyand potentialfield studies. slabis fragmentingalongits southernedge;(3) arcvolcanism How this transition is accommodated on a detailed crustal and is being replacedby bimodal volcanismthat exhibitsa northward decreasein age; and (4) the easternboundaryof the •Nowat Departmentof Geological Sciences, VirginiaPolytechnic Pacificplate hasmigratedeastwardthrougha seriesof jumps Institute and State University,Blacksburg. of the MendocinoTriple Junction.Someof thesecomplexities Copyright1998 by the American GeophysicalUnion. maybe explainedby introducingmore complexrheologiesand thermaldistributionsinto the simplerigid-platemodelof DickPaper number 98JB02231. 0148-0227/98/98 JB-02231 $09.00 inson and Snyder.For example,thermomechanicalmodeling 1. Introduction 30,101 30,102 BEAUDOIN 125 o 42 ø ET AL.: SOUTHERN 124 ø EDGE OF THE GORDA SLAB 123 ø 128 ø 125 ø 1.22ow Juan de Fuca Miocene &sediments younger Eel River __[•_ Klamath [---] Mesozoic Great [• Tertiary volcanics Valley • :.i_:-• Central &Eastern_['7-7] Sierra Nevada ..............Belt Franciscan [-•ß KingRange & Coastal BeltFranciscan 41 ø Line 6 Redding •\#• //\\ . . Line 9 601 40 ø Line 1 Willows 39øN -. ß ßß Lk Clear Figure 1. Generalizedgeologicmap of northern California. Squares,instrumentlocations,whichcommonly overlapto form solidlines;stars,shotpoint locations;BSF, Bartlett Springsfault; CSZ, Cascadiasubduction zone; MAF, Maacama fault; MFZ, Mendocino fracture zone; SAF, San Andreas fault; SEDGE, southern edge of the Gorda slab as determinedfrom this study.Inset showstectonicsettingof the Mendocino Triple JunctionSeismicExperiment.Wigglylines,seismicprofiles;large arrows,plate motion relativeto fixedNorth America; MTJ, MendocinoTriple Junction;GV, Great Valley. suggeststhat the boundaryat depth between the Pacificand North American plates shouldstep inboardwith time as the slablesswindowcoolsand new material is accretedto the edge of the Pacificplate [Furlonget al., 1989];recentresultssuggest that sucha processmay be occurringin the San FranciscoBay region [Brocheret al., 1994].Alternatively,the slablesswindow model maybe fundamentallyincorrectfor northernCalifornia. Bohannonand Parsons[1995] recently questionedthe longstandingparadigmof a slabless windowfor the MTJ region.On the basisof analysisof plate motionsand geometries,Bohannon and Parsonssuggestthat coastalCaliforniashouldbe undedain by a stalledslab rather than a slablessgap; in their model a slablesswindowmay existbut would be much farther inland. Our study focuseson the transition from subductionto transformtectonicsby analyzinga SE-NW trendingwide-angle reflectionprofilehereinreferredto asline 9 (Figure 1). These data were part of the first phase of the Mendocino Triple Junction Seismic Experiment (MTJSE) collected in 1993 [Beaudoinet al., 1996;Godfreyet al., 1995;Trdhuet al., 1995]. During this 1993 field seasonwe collected650 km of onshore seismicrefraction data: line 1 samplingthe transformregime; line 6 samplingthe subductionregime;and line 9 samplingthe transitionfrom transformto subductionregimes(Figure 1). The MTJSE is a multiyear, multiinstitutionaleffort to study the lithosphericresponseto MTJ passageby comparingand contrastingthe subductionregime to the transformregime. 2. 2.1. Tectonic and Geologic Setting Plate Geometry and Cascadia Seismicity As the Gorda-Pacific plate boundary migrates northward relativeto North America, subductionalongthe Cascadiasub- BEAUDOIN ET AL.: SOUTHERN ductionzone is replacedby the transformboundaryof the San Andreasfault system(Figure 1). North of the MTJ, the tectonics are governedby the subductionof young (---5 Ma) Gorda plate obliquelybeneathNorth Americanplate alongthe Cascadia subduction zone. South of the MTJ the tectonics are governedby the transformboundarybetweenthe Pacificand North American plates, the San Andreas fault system.Offshore, the boundary between the Pacific and Gorda plates defines the Mendocino fracture zone which juxtaposesthe young Gorda plate with the older (---26 Ma) Pacific plate. Kinematic modeling of northward convergenceof the Pacific plate with the Gorda plate indicatesthat the southeastern portionof the Gorda plate, nearestthe MTJ, is undergoingthe greatestamount of compressionalstrain in a north-southdirection[Wilson,1986],further complicatingour understanding of the MTJ region. The MTJ regionhasexperiencedover60 earthquakesof Ms -> 5.5 since1853 [Oppenheimer et al., 1993].Recent analysisof the physicalcharacteristicsof the Cascadiasubductionzone and seismicityof the Juan de Fuca-Gorda plate and surface deformationindicatethe Cascadiasubductionzone may be a well-coupledor fully locked systemwith potential for large (Ms -> 8) subductionzone earthquakes[Carverand Burke, 1989; Clarkeand Carver,1992;Heaton and Hartzell, 1987]. In the subductionregime,earthquakesare commonto depthsof 40 km and largely result from internal deformation of the subductingGorda slab and slip alongthe Mendocinofracture zone [Castilloand Ellsworth,1993]. In contrast,in the transform regime,earthquakesare normallyshallowerthan ---15km and reflectthe right-lateralmotionbetweenPacificand North American plates [Castilloand Ellsworth,1993]. 2.2. Geologic Overview EDGE OF THE GORDA SLAB 30,103 wedgingbeneaththe older Great Valley forearc basin and its ophioliticbasement[Godfreyet al., 1997]. 3. Modeling: Technique, Results, and Errors The modelpresentedhereinis the resultof inversemodeling of travel times of both refractedand reflectedarrivals[Hole, 1992;Hole and Zelt, 1995](Plate 1). The modelpresentedhas an overall absolute value rms fit of 82 ms to first arrivals. In addition to travel time modeling,we have alsoincorporateda depth-migratedsingle-foldimageof the precriticalreflections from offsets of _<30km, and gravity modeling, to constrain possibletectonicmodels. 3.1. Velocity and Interface Modeling The velocity modelingwas carried out in two stages.First, travel times for P•7, defined as the divingwave in the crust, were picked, and thesetravel timeswere inverted for velocity. The final inversion model has a 1 km square grid velocity parameterization and a 20-km-horizontal x 4-km-vertical smoothingoperator applied.The Franciscancomplexhas few coherent seismic reflectors. Hence we are unconcerned with midcrustaleventsin this data set sincethey are isolated features (observableon fewer than three shotgathers;for a detailed discussionof the effect the Franciscancomplexhas on the wavefield,seeLendlet al. [1997]).Oncethe crustalvelocity structure is established,we can infer first-order discontinuities from the presenceof steep gradients,althoughthe method is not designedto have explicitvelocitydiscontinuities. Our inversionmethod doesallow the placementof a reflectorwithin the model by "floating"the interfaceover the velocityfield to minimizereflectiontravel time residualswithout updatingthe model velocities. Hence we introduced interfaces for the basal crustaland Moho reflectionsbasedon the wide-anglereflecNorthern coastalCalifornia recordsa historyof subduction tionsalongthe profile. The modeledlocationsof thesereflecand accretion that has been active since at least the Early tors coincidewith many of the high-amplitudereflectionson Cretaceous[Dickinson,1981].The overridingNorth American the depth-migratedsingle-foldimage(seebelow). plate in the MTJ regionconsistsof accretionarywedgesof the 3.1.1. The subduction regime (model coordinates --•110Mesozoic-CenozoicFranciscan complex [e.g., Blake et al., 250 km). The Franciscancomplexbetween shot points 906 1985]overlainby the Eel River Basin,a Cenozoicforearcbasin and 911 exhibitsthe largestamount of lateral variation along [Clarke,1992](Figure 1). The Franciscancomplexincreases in the profile (Plate 1). Lateral variationsin velocityas large as age and metamorphicgrade inboard and is largelycomposed 0.5 km/s in 25 km are modeled from refracted arrivals in the of metasedimentaryrocks.It hasbeen dividedinto three belts: upper 10 km beneathshotpoints907, 908, and 909. There is no the Coastalbelt, consistingpredominantlyof a Tertiary accre- evidenceof higher-velocityKlamath rocksbeneaththeseshot tionarycomplex;the Central belt, a tectonicm61angeof Early points[Beaudoinet•al.,1996;Trdhuet al., 1995].Beneath10 km Jurassicto Tertiary(?) age; and the Easternbelt, Jurassicto depththe velocityfield is laterallyuniformto 20 km depth(the Cretaceousblueschist-facies rocks (Figure 1) [Blake et al., 6.5 km/s contour).However,it is likely that variationsof the 1985]. Onshore,the Eel River Basinconsistsof up to 4 km of scaleobservedin the upper 10 km are presentbut not resolved Miocene and youngersedimentaryrocksthat lie unconform- [Lendl et al., 1997]. Refracted arrivalssamplingdeeper in the ably on Coastaland Central belt Franciscancomplex[Clarke, model definea gentlySE dippingvelocityboundaryto --•35km 1992]. Line 9 is roughlyparallel to the geologicstrike of the depth. Mantle diving rays, arrivals with apparent velocities region and is therefore approximatelyin the sameFranciscan >-7.6 km/s, are observedfrom shot points 903, 905, 906, 909, subterranethroughoutits length(Figure 1). and 911 (Figure 2). Lower crustaland mantle ray coverage North of the present-dayMTJ, the Franciscanbelts along from thesearrivalsis limited to a regionbetweencoordinates the Cascadia subductionzone are overlain by the Klamath --•95 km and •215 km. Reflections from the lower crust and Moho are observed in terrane, composedof arc-relatedrocks of Paleozoicto Late Jurassicage [Harper,1980].Accretionof the Klamath terrane both the wide-angle(Figure 2) and depth-migratedsingle-fold to the North American margin occurredduring the Late Ju- data (Figure 3). The depth-migratedsingle-foldimagedefines rassic[Blakeand Jayko,1986].The contactbetweenthe Klam- a segmentof an intracrustalinterfacewith an apparentdip of ath terrane and the Easternbelt Franciscanis thoughtto be an --•10ø SE. Wide-anglereflectionsfrom shotpoints906 and 907 eastdippingthrustfault [cf.Beaudoinet al., 1996;Tr•hu et al., (Figure2) are modeledas an interfacethat is coincidentwith 1995].Southof the presentMTJ and Klamath belt, continent- the depth-migratedsingle-foldimage at a depth of 25-29 km ward accretionof the Franciscanwas limited by crustal-scale between model coordinates 125 and 150. The inverse method 30,104 BEAUDOINET AL.:SOUTHERNEDGEOF THE GORDASLAB L_ o o o o o o o o i i'•..... ! , •', • o o o (tull) 41doCl o o 0 BEAUDOIN ET AL.' SOUTHERN EDGE OF THE GORDA SLAB 30,105 SE NW 901 50 100 150 200 PmP 6 250 km ' 902 . 2 0 o 50 lOO 15o 200 km 6 2 0 -50 0 -50 50 0 100 50 150 km 100 150km Figure 2. Record sectionsfrom line 9. Heavy black lines representcalculatedtravel times for all arrivals usedin the velocityand interfaceinversion.Thin black lines representcalculatedtravel times for the gravity Moho (seetext). Data are filteredfrom 2 to 16 Hz andtracenormalized.Time is reducedby 6 km/s.Pg, crustal divingwave;PcP, reflectionsfrom the lower crustallayer;PmP, reflectionsfrom the Moho; Pn, mantle diving wave. models this interface with an rms residual of +_0.31 km in depth (Figure 4a). Wide-angle reflectionsfrom the Gorda plate Moho are observedon shotpoints904-911 (Figure 2). Both forward and inversemodelingpositiona dipping inter- face that is 35-36 km deep at model coordinate100 km and shallows northwestward km. The inverse method to 26-27 models km at model coordinate this interface residualof +0.96 km in depth (Figure 4b). with 230 an rms 30,106 BEAUDOIN ET AL.: SOUTHERN EDGE OF THE GORDA SLAB -50 0 50 100 150 km 2 0 -100 -50 0 50 100 km 1907 6 -150 -100 -50 50 km 0 8 6 4 _ 2 0 -150 -100 -50 0 50 km Figure 2. (continued) these data are Pn arrivals that sample the mantle south of model coordinate--•100km (shotpoints901 and 902 did not propagateto distances->140 km). Basalcrustalreflectionsfrom the depth-migratedsingle-fold depth.Thisdepthcoincides with the 6.5 km/scontourasin our model for the subductionregime. Refracted arrivals from data are groupedin two sets,a shallowsetthat roughlyfollows deeperin the model are confinedto an arrival observedon the 6.5 km/scontouranda gentlyNW dippingsetthat is --•5km shotpoints901 and907 that definea highervelocity,--•7.0-7.5 deeper (Plate 1 and Figure 3). Wide-anglereflectionsfrom km/s, between model coordinates--•30-90 km. Absent from shotpoints901, 904, and 905 definea shortreflectorbetween 3.1.2. The transform regime (model coordinates--•0-110 km). Crustal refracted arrivals from the transform regime define a laterally uniform Franciscancomplexdown to 20 km BEAUDOIN ET AL.' SOUTHERN EDGE OF THE GORDA SLAB 30,107 o -1•0 -2oo km -100 -50 0 6 4 2 0 -200km -150 -100 -50 0 6 40 911 4•5.0' - I!!! t 6.0 0 -2 I I I I I I I I i -200 km I I I I -150 Source-Receiver I I I I I I I -100 Offset I I I I -50 I 0 Figure 2. (continued) 901 902 SL 903 iii i 904 iiiiii 905 i i ii 906 -- i 907 llllll III I 908 III i i1111 909 ii i 910 II 911 - I I I I .......... 10 20 30 50 SE "'" 0 I' 20 ' .......I .........' 40 I' ' 60 ' I 80 ' I 1• " i ....... ' 120 I 140 , ' t 160 , ! 180 ' I 2• , I 220 ' I 2• Figure 3, Depth-migratedsingle-foldimage from precriticalreflectionsat offset of •30 km; a arrowsare reflectionsinterpretedas top or near-topof Gorda crust,and b and c arrowsare basalcrustalreflections. ' • 30,108 BEAUDOIN ET AL.' SOUTHERN lower-crustal o _midpoints * •Sz 5 VE5:I 10 refine the crustal structure used in these earlier models and • 25 30 -2 35 40 SLAB tomography[Verdonckand Zandt, 1994], and mantle tomography [Benzet al., 1992].This restrictedmodelsof the gravity field to makingbroad assumptions basedon thesesparsedata and observations of Benloft zone seismicity[e.g.,Jachensand Griscom, 1983]. With our new seismicdata we are able to reflector wmodel interface 0' EDGE OF THE GORDA rms: _+0.93 km ;• o hence to utilize the gravity data to further constrainvarious modelsof lower crustal/uppermantle geometry. 3.2.1. Jachensand Griscom model (Figure 7a). Jachens and Griscom[1983]modeledthe isostaticresidualgravityfield along six profiles oblique to line 9 (Figure 5). Five of their profilesand our line 9 all samplea pronouncedgravitylow of ->50 mGal below regional average.Lacking any control on crustal rms: _+0.31 km -3 ...... 1o'' "'264" Model Coordinate (km) thickness Jachens and Griscom assumed a uniform North American(Franciscan)crustalthickness of 20 km (Plate 2a) aboveGorda plate lithosphereto the north, and asthenosphere (slabless-window model) to the south. The essential componentsof their model are (1) a low-densitytabularbody (Gorda crustand mantlejuxtaposedagainstasthenosphere to the south)creatinga gravitydecreaseto the north; and (2) a wedgeof higherdensity(relativeto Franciscan)Klamathrocks on the northern end of their profiles to create a gravity in- .... m ødei'interface basal-crustal reflector 0 o •midpoints 3 * bz VE5:1 •i •E I • • a crease at the northern 15 I + 2 1 0 35 ms: •• 0 • _. 50 1• • -2 rms: •.96km 150 Model Coordinate (km) 200 250 end of the model. Their introduction of the Klamath terraneinto the modelwaslargelyunconstrained at the time of their writing.Althoughthe Jachensand Griscom assumptionof constantcrustalthicknesshas the merit of simplicity, here [Beaudoinet al., 1996; Godfreyet al., 1997; T. J. Henstockand A. Levander,Lithosphericevolutionin the wake of the Mendocino Triple Junction:Structureof the San Andreasfault systemat 2 Ma, submittedto Journalof Geophysical Research, 1998, hereinafter referred to as Henstock and Levander,submittedmanuscript,1998] as elsewherein California [e.g.,Fuis and Mooney,1990]crustalthicknessincreases by 5-10 km from the coastto 100 km inland, rendering their conclusions invalid. As seen in Plate 2a, Bouguer gravity along line 9 can be Figure 4. Splinedinterfacesfor (top) lowercrustaland (bot- modeled incorrectlyusing a model similar to that of Jachens tom) basal crustalreflectors(shown as hea• black lines in and Griscom(rms residualof 4.8 mGal). "Incorrect"because, Plate 1), midpoints,and residualsfor the interfacesmodeled unlike the profilesof Jachensand Griscom,line 9 does not usingthe inversiontechnique. crossany exposedKlamath rocks.Furthermore,the velocity field presentsno indicationof significantvelocitydifferences, and henceinferred densitydifferences,in the upper --•20km of model coordinates --•40-55 km that coincides with the shalcrust(Plate 1) [cf.Beaudoinet al., 1996].Without the Klamath lower set of reflectionsfrom the depth-migratedsingle-fold body,this "Jachensand Griscom"-typemodel doesnot fit the section. This interface has an rms residual of _+0.93 km in observedBouguergravity (Plate 2a). We are therefore comdepth(Figure4a). Two deeperreflectorsare modeledfrom the pelled to evaluate a suite of models,based on our velocity wide-anglereflectionsin this region of the transformregime. model, that vary only in lower crustalor upper mantle geomThe first,sampledby shotpoints901 and 903, is at a depthand etry in order to determinepermissibleplate geometrybeneath a dip that agree with the single-folddata. This segmentis line 9. modeledwith an rms residualof _+0.64km in depth (Figure 3.2.2. Gravity models derived from the velocity model. 4b). The secondsegment,definedby shotpoints903 and 906, Belowwe discussadditionalconstraintsthat gravitymodeling, dipsgentlyto the NW, coincideswith the single-folddata, and basedon our velocitymodel, offers to understandinglithosextends from model coordinate --•55-90 km. The inverse phericstructurein the vicinityof the MendocinoTriple Juncmethodmodelsthis segmentwith an rms residualof _+0.46km tion and in particular along the southernedge of the Gorda in depth (Figure 4b). Ray coveragefor this segmentdoesnot slab.A smoothedversionof our velocitymodel was converted overlapwith the deepestreflectionnorth of modelcoordinate to MacRay format for the gravity modeling [Luetgert,1992]. 100 km. Featuresin thesedensitymodelsthat are held constantare the densitystructureof the Franciscanrocksabove20 km depth 3.2. Constrained Gravity Modeling and the apparentdip of the Gorda plate. In all of these2.5-D Prior to collectingthe MTJSE data, little was known about models, the thicknessof Gorda lithosphereis based on a crustalstructurein northernCalifornia.Existingdata included trenchageof 5.3 Ma and a half spreadingrate of 38 mm/yr for unreversedseismicrefractiondata [Warren,1981], seismicre- the last5 Myr [Riddihough,1984]givingan approximateage of fractiondata north of the MTJ [Beaudoinet al., 1994],crustal 8 Ma and thicknessof --•30-35 km for the lithospherebeneath BEAUDOIN 125 ø ET AL.: SOUTHERN EDGE OF THE GORDA SLAB 124ø 123 ø 30,109 122 ø 41 ø 909 40 ø 39øN Figure 5. Bougucrgravityin the regionof line 9 [after Oliveret al., 1980]. Dashedlinesrepresentprofiles modeledbyJachens and Griscom[1983].Shadedare regionsof -< -100 mGal anomaly.Stippleis regionof -> -50 mGal anomaly. line 9. A densitycontrastof 0.04 g/cm 3 is usedacrossthe 3.2.2.3. Preferred model (Plate 2d): Finally,we introduce a lateral changein upper mantle densityas predictedby the 3.2.2.1. Direct conversion (Plate 2b): Our first model is a slabwindowmodel. The lack of Pn arrivalsfrom shotpoints direct conversionof the velocitymodel to densityusinga sim- 901 and 902 (Figure2) and the extentof modeledray coverage ple velocity-densityconversion appropriate for Franciscan (see Plate 1) suggesta zone of increasedscattering,intrinsic rocks[seeGodfreyet al., 1997,Figure 7) (Plate 2b). From this attenuation,or both that affectsenergy propagationthrough model it is evident that the apparent dip of the Gorda slab the centerof our model[cf.Beaudoinet al., 1996].We attribute (corresponding to yellow and orangecolorsin Plate 2b) will thiszone of attenuationto the southernedgeof the Gorda slab provide much of the needed northward increasingBougucr (SEDGE). With this in mind, our preferredmodel introduces anomalyon the northernend of line 9. The large mismatchon a slabedge(SEDGE) that coincideswith the envelopeof ray the southernend indicatesthat further adjustmentto the lower coverage(Plate 1; dashedwhite line, Plate 2). By introducing 3 tabularbodyonthenorthernend,we are crustor uppermantleis requiredin thisregion.This modelhas this3.20-3.23g/cm lithosphcr½-asthcnosphcr½ boundary. an rms residual of 26.4 regal. 3.2.2.2. Adjustedgravity Moho (Plate 2c): The first logical modificationto our direct conversionmodel is to adjustthe gravity Moho to coincidewith the modeled reflector depths from the wide-anglereflections.This is justified becausethe tomographicinversion smoothesthis contrast, as discussed above. A density discontinuitywas therefore introduced to coincidewith the basalcrustalreflectors.With this change,the observedBougucranomalyis remarkablywell matched(rms residualof 9.7 mGal), implyingthat mostof the gravityanomaly can be accountedfor by structurein the lower crust.The predictedanomalyis only weakly sensitiveto the variationsin mantle geometryif the mantle densityremainsconstantalong the profile. requiredalsoto includea sub-Moho3.20-3.23g/cm 3 highdensitywedge on the southernend of the profile to fit the observedgravity.The boundarybetween Gorda crust and the southern basal crustal layer was also adjusted from model coordinate110 km to 95 km to better fit the gravitylow in the centerof the profile.The final modelhasan rmsresidualof 5.8 regal with the largestmisfitsbetweenmodel coordinates90130 kin. Removing either the high-densitywedge at Moho depthson the southernend or the Gorda mantle slab on the northernend degradesthe fit (rms residualsof 11.7 regal rms and 20.0 regal rms, respectively,Plate 2d). Our preferredmodelhasfour main features:(1) The North American crust is thickestin the center of the profile. This thickeningof low-densityrocksin the middle of the profile is 30,110 BEAUDOIN ET AL.: SOUTHERN EDGE OF THE GORDA SLAB -20- _40'• • -60: Gorda Plate w/Klamath ß ß C9 -802 •"•-•oo2 -120- rms: 4.8mGal •4o5.A--] 20 o 40 lO 60 80 2.67 100 i":•0 l•t'0' 'l(•0 1'•'0 "•"60" 'Kl"im,4th'" North American crust 2O 3. e5 E 30 .c 40 • 50 2.67(Gorda upper crust) ewer crust or, a 60 70 ' 0 20 40 60 80 100 120 140 160 180 200 220 240 km -20- •E -8o •,o - 100- -] 2o-' .140.,., •..,,B•. Direct Conversion (rms: 26.4 mGal) 0 20 40 60 0 80 2.60 100 120 140 160 180 200 220 240'km - North America•'•rust 20_. 30 ß ,. 3.•) 40 ........ 50 ß........... Gordamantle 60 70 SE 0 NW 30' 20 40 60 80 100 120 140 160 180 200 220 240 km Plate 2. Gravitymodelsalongline 9. Dottedwhite line, the ray coverageof our velocitymodel(see Figure2). Solidwhitelines,densitycontours.Red circles,the observedBouguergravityalongline 9. (a) Jachensand Griscom[1983]model (profileE-E' in Figure5) rotatedand modifiedto fit the Bouguer gravityalongline 9. (b) A directconversion of the velocitymodelto density.(c) Sameas Plate2b with the Moho adjustedto match basalcrustalreflectors.(d) The best fit model with changesmade to densitiesonly beneath20 km depth. Also shownare the effectsof removingthe Gorda lithospheric mantleon the north and postulatedlithosphericmantlewedgeon the south.(e) Variationsin location of southernedge of Gorda slab. -20 -40 • -60 o -80q Og - e•. ß© - 140 Mo, hoCorrected !rms: 9..7 mOa!) ..... 0 20 40 60 80 loo 120 140 160 l•o 260 "2'•.0240'km ! " North American ". •- 20 • 30 •, 40 • crust 2 80 ø'ø - ' • .'7 .,, 50 60 '-,i 70 %% 0 20 40 60 80 -20L•_ 100 120 140 160 180 200 220 240 km pretErred (black) -40_[..__• nomantle wedge SE(red; rms: 11.7 mGal) •o -60q •x.,.• *•••••_• -80-1 '•Y•-'•-•• no Gorda mantle NW (green; rms: 20.0 regal)) ©ß ß© ß •,'• -! 00 i14020.... Preferred odcl ( 0 20 40 60 ' ............. , 80 ........, ......... 100 120 140 160 180 2(50 2•.0 240km 0 ioJ North American crust ,.. •. • . •,.r,, ,,o••---•-•, 30 •" .. 2.80 "''•"•' ':'*/'•//'"-• Gd(({a crust"" ßß E• 40 ____ ß .. ........ . 320 Gorda mantle . c2 50 60 \ \\ 70 0 20 40 60 80 100 120 140 160 180 200 220 240 km -20- preferred(black;rms:5.8 mGal)) slabedge+25 km NW (green;rms:6.3 mGal) slabedge-25 km SE (red;rms:6.4 Gal) -40-• . • -6o co -80 regal) . .. •o -100- -I 2o-' . 140 ,E---• Extent ofSouthern edge 0 20 40 60 80 160 1•0 l•i0 l•J0 1•0 2130 2"•. 0 240'km i North American crust 1o •' 20 -.-- 30 • "• -.- -- z/.• crust "•orca '....... .....,"' ,/•. / • 7 E. 40 t2 ....... Gorda mantle 50 60 70 20 40 60 80 100 120 140 160 180 200 220 240 VE 2.50 2.75 3.00 Density(g/cm3) Plate 2. continued 3.25 km I' 1.25 30,112 BEAUDOIN ET AL.: SOUTHERN EDGE OF THE GORDA SLAB A) Line9 Interpretation 901 902 903 V 904 905 906 V ! _ v 907 908 909 V . 910 v 911 v . .V- 1o t,,:..Oo ø"-'7:.'? .:•:..•....-.,.?..øo•o ø:q}"•.,",a•. •'•:0, • • o• •/,•,.•,,.,,•,.,,• o.Y,.,,,J, -oO: ø•' ø North Ooo American crust ' oo'..'.,•- .:c.--;'-'-'.",,..'.:,,:,: •, 20J. •,:' :,.,,•:. - - - ',,', m ••..,•, -7" ...-'• ........ 40 ß '" tectonicall .-.•,,. . "o .,' g, ,.,,.,•.•, ,,,,•o: •, ,;,3,3,z .: •, :_-?.=-, .,.•,,. o o .'zz.• o _ oo •7 .. .... ,.,.•.•.:."..•.....,.z•.o - o o•5o,•x"'•__9oes 8'••oo z.,'zZ/// rdaOcrust •,• :'//' •da½cfust - ". . o o .. -o ' o _ .... '3oceanm .crust or ""•.•..? 2'.'"".;:.•. i............................................ Oorda mantle - 50_] rnagmatm newly accreted •' ',' 6o-]underplate &thickened nartial melt/ m SE -•a'} '""1 " ' ' I 0 ' 20 ! ' ' '-I 40 ' I" 60 •' erosion 80 ' I ' 100 oo I 120 ' ! 140 •oø• asthenosphere 2øø'/ NW Iso' ]- ' i ' 160 I 180 '' I " 200 I 220 Magnitudes'o2+ 03+04+ 05+ C)6+ ' I 240 km VE1.1 B)decompression melting & quenching: requires extreme rates D)slow decompression melting + tectonic scraping & assembly NorthAmerican Crust(fixed) NorthAmerican Crust(fixed) ...... // .•ew.Pacific' %• t • • •'///•//GordaPlate•/ C)preexisting feature: requires extreme coincidence //Gorda Plate/•/ preferred NorthAmerican Crust(fixed) _•pre•North America' • Gorda Plate •/ Plate 3. (a) Interpretation of our velocity model. Overlain are interfacesdetermined from wide-angle reflections(heavyblack lines), regionsof high reflectivityfrom the depth-migratedsingle-foldimage (hachured),velocitymodelray coverage(dottedwhite line), and hypocenters for 1984-1994from the Northern CaliforniaSeismicNetwork catalogwithin 7.5 km of the profile (octagons,hypocentershaveerrorsin depth of <-2.5 km and lateral positionof <-1.5 km). Below are three possibleexplanationsfor the body of mafic materialat the baseof North Americancrustand immediatelyadjacentto the Gorda slab.(b) Decompression meltingand rapid quenchingaddingmaterial to the North Americanplate. (c) A preexistingfeature of the North Americanplate.(d) Slowdecompression meltingcoupledwithtectonicoff-scraping andassembly of the basal crustalmaterial as part of the Pacificplate. to the location of the southern edge of the Gorda plate (SEDGE). Usingour preferredmodel (Plate 2d) as a starting velocities at 23 km deptharemodeled with0.1g/cm 3 density point, we have calculatedgravity along our profile for three increase relative to the Gorda crust densities.This layer's models and compareit to both the observedBouguergravity thicknessis basedon reflectionsfrom both the wide-angleand along line 9 and the calculatedgravity from our preferred depth-migratedsingle-foldimage.(3) Gorda lithosphereis ex- model. In thesemodelsthe southernedgeof the Gorda slabis tendedonly as far southas model coordinate---110km where movedfrom modelcoordinate105 km by -50 km (rms residmantle lithospheredensitiesgradeinto asthenospheric density ual of 7.9 mGal), -25 km (rmsresidualof 6.4 mGal), and + 25 over 10 km (e.g., centeredabout model coordinate105 km). km (rms residualof 6.3 mGal). We concludethat our preferred (4) A high-densitywedgeis presentat Moho depthsbeneath location of SEDGE could be in error by _+25km but that a the southernend of our profile. However,the thicknessof this shift of 50 km is precludedby our modelling. wedge is poorly determined. In summary,first-order features from the velocity model 3.2.2.4. Southern edge of the Gorda slab (Plate 2e). In that we demonstrateare needed to fit gravity data are the this final modelwe test the sensitivityof our preferredmodel thickeningof the overlyingFranciscanterrane near the center basedon the well-sampled,low velocitiesobservedin the inversemodel. (2) On the southernend of the profile the high BEAUDOIN ET AL.: SOUTHERN EDGE OF THE GORDA SLAB 30,113 of our profile, the existenceof a higher-density(relative to overlyingFranciscan)lower crustallayer on the southernend of our profile, and the truncationof the Gorda plate near the center of our profile. Furthermore, given these featuresconstrainedby our seismicstudy,the gravity model requiresthe presenceof high-densitymaterial below the Moho on the migrated single-fold section, we can trace the Gorda crust reflectionsfrom shotpoint 911 to just southof shotpoint 907, coincidentwith the terminationof Gorda slabseismicity(Plate 3a). At this point we are unable to determinewhether the reflectivitybeneathshotpoint 906 is geneticallyassociatedwith southernend of our profile. locally strong, reflectivity to the south. However, based on first-arrivalray coverageand the wide-anglereflections,Gorda Moho extendsbeyond these Gorda crust reflections to between shot points 905 and 906 (Plate 3a). Finite difference modeling of amplitudesof Pn and PmP on shot point 911 supportsthis interpretation[Lendl, 1996]. We interpret the deep structurebeneath line 9 determined from seismicand gravity data to indicate that the southern edge of the Gorda slab is between shot points 905 and 906 (Plate 3a). Changesin Gordacrustreflectivityandcessationof slab-associated seismicity--•30 km north of this edge may be due to either changesin slab continuityor a changein lithologies and therefore impedancecontrastand coupling,across the North American-Gorda plate boundary. On the basis of our velocity and gravitymodelingour model is generallycompatible with that of Jachensand Griscom[1983], in which the Gorda slab extendsas far south as approximatelyshot point 904, althoughour preferred model placesthe southernedge Discussion Travel time tomographyusingfirst arrivalsalongline 9 producesa modelof relativelyuniformvelocity(<6.5 km/s) crust to depthsof--•20 km (Plate 1). We interpretthesevelocitiesto indicate that Franciscan rocks extend to at least 20 km beneath our profile [Thompsonand Talwani, 1964]. At depthsgreater than 20 km, south-to-northvariationsin both the velocityfield and model ray penetration are observed.In our "transform regime"betweenshotpoints903 and905, highvelocities(7.07.5 km/s) occur at 23 km depth. A similar onset of higher velocities(an increasefrom ---6.5to 7.0 km/s)is modeledalong crossingprofile line 1 [Godfreyet al., 1997; Henstock and Levander, submittedmanuscript,1998]. Dip on this "layer" is poorly constrained.In our "subductionregime" from shot points 906 to 910, velocities>7.0 km/s define a southward dippingzone from 22 km depth beneathshotpoint 910 to 28 km depth beneath shot point 906. Within the North American crust, line 9 indicatesa deepening(-10 km) of the 7.0 km/svelocitycontourin the center of our profile.A similarresult,interpretedasthickenedNorth American crust along the eastwardprojection of the Mendocinofracturezone, hasrecentlybeen reportedby Verdonckand Zandt [1994]. Verdonck and Zandt interpret their model to indicate that the Gorda slabis flexing downward6-12 ø to the south and that the local thickeningof the North American crust is either due to compressionalforces associatedwith Pacificplate convergenceor underthrustingof the accretionary complex along the CascadiaSubductionZone. Our line 9 model is only 10ø oblique to the local trend of the Cascadia SubductionZone, and therefore we expectno more than half of the crustalthickeningto be attributed to moving downdip, southwardalong our profile (-4-5 km depth variation between shotpoints911 and 906 for a 10øobliquityand a 12ødip on the Gorda slab). If, however,the Gorda slab is flexing southward as suggestedby Verdonck and Zandt, our data suggesta maximum of 5ø southwardflexing. We prefer to interpret the overthickenedNorth American crustas tectonic thickeningin responseto the northward convergencebetween Pacificand Gorda plates rather than underthrustaccretionary complex[Beaudoinet al., 1996;Tr•hu et al., 1995]becausewe are forced (Plate 3) to adopt a similar tectonic thickening model to explainour uppermostmantle observations. the reflective band to the north about 25 km farther or with the more diffuse but north. The lower crust/uppermantle reflectivityand velocitystructure observed near the southern end of line 9 is consistent with that modeled for line 1, which intersectsour profile north of shotpoint 902 (Plate 1) [Godfreyet al., 1997;Levanderet al., 1998; Henstock and Levander, submittedmanuscript,1998]. The -5 km of maficmaterial(6.5 km/s<_Vp <- 7.5 km/s) overlyingmantle can be interpretedas either tectonicallyunderplatedoceaniccrust [cf. Benzet al., 1992] or magmatically accretedmafic material [cf. Furlonget al., 1989]. We cannot distinguishbetweenthesetwo modelson the basisof velocity alone. The modeledshallow(---23km) highvelocitiesand densities in the center of our profile suggestup to 10 km of mafic material exists at the base of the North American crust imme- diately adjacentto the southernedge of the Gorda slab.Here we explorethree possibleexplanationsfor this body of mafic material. The first explanation is a localized region of basal underplating(Plate 3b). Thermomechanicalmodeling indicatesthat in -5 m.y. 5-10 km of basalticunderplatecould be generatedafter passageof the MTJ by decompression melting of the rising asthenosphere[Furlongand Fountain, 1986;Liu and Furlong,1992].However,for decompression meltingto be the mechanismcreating this layer requires that we invoke maximumdecompressionmelting, completemelt segregation, and nearly instantaneousquenchingto producethe structure observed,i.e., conditionsthat seemunlikely.A secondexplaReflections from the lower crust and Moho are observed in nation is that this body is a preexistingstructureof unknown both the wide-angleand depth-migratedsingle-folddata (Plate origin (Plate 3c), which also seemsunlikely sincewe see no 1 and Figure3). North of shotpoint 905, a southdippingzone evidencefor similarlower crustalstructurenorth of thisregion. of high-velocitymaterial is coincidentwith south dipping re- It seemsan unlikelycoincidencethat sucha preexistingfeature flectorsinterpretedas Gorda plate Moho. The high-amplitude would coincidewith the southernedgeof the Gorda slabtoday, reflectionsaboveGorda plate Moho (arrowson Figure 3) are rather than arbitrarily in the past or future. Our preferred interpretedto originateat or near the North American-Gorda explanationis a modelwherebybasalunderplatingis occurring plate boundary.This interpretationis consistentwith models at expectedratesbut at a lower rate than requiredto createthe for MTJSE line 6 [Beaudoinet al., 1996;A.M. Tr•hu et al., entirethicknessof maficmaterialmodeledhere (Plate 3d). As manuscriptin preparation,1998],earthquakehypocenters, and this new material cools,it is accretedto the easternedgeof the projectedGorda slabdepth[cf.Beaudoinet al., 1994].Utilizing Pacificplate [cf.Furlonget al., 1989]and movesnorthwardwith both the wide-angle reflection modeling and the depth- respectto North America. With this northwardmovementthe BEAUDOIN 30,114 ET AL.: SOUTHERN EDGE OF THE GORDA SLAB Important in our observationsand interpretation are the variationsof lowercrustalreflectorsalongline 9 and the thickner, the mafic body is constructedby both magmaticunder- ening of mantle lithosphereto the south. The variationsin plating and tectonic thickening.Although this model can lowercrustalreflectorsand the apparentabsorptionof seismic explain observationsalong our profile, inconsistencies exist energyfrom the southernshotssupporta zone of increased whencomparingto observations from otherwide-angleseismic scattering,intrinsic attenuation,or both, resultingfrom meprofiles [Hole et al., 1998]. Most notable is evidencefrom chanicalmixingof lithologiesand/orpartial melt in the central onshore/offshore line 1 that suggests the southernbasalcrustal region of line 9 beneaththe onshoreprojectionof the Menlayer, like the overlyingNorth American crust,is faulted and docino fracture zone. This zone of attenuation combined with thereforemostlikely part of the North Americancrust[Hen- the southwardthickeningmantlelithosphereindicatesthat our stocket al., 1997].Thisrequiresthat the basallayerbe coupled profile crossesthe southernedgeof the Gorda plate and that to the North Americanplate and henceunableto movenorth- directlyadjacentto this edgeis an asthenospheric window. ward to act as a bulldozeroff-scrapingmaterial alongits leading edge. Acknowledgments. Supported by National Science Foundation In our model (Plate 3a) the lithosphericmantle is thinnest grants EAR-9218209, EAR-9219870, EAR-9218968, and EARjust south of the southernedge of the Gorda plate and in- 9219598and by the U.S. GeologicalSurveyDeep ContinentalStudies creasesin thicknessto produce a wedge to the south. This program.Volunteersfrom Stanford,Rice, OregonState,Lehigh,and geometryis consistentwith a model in which the astheno- Humboldt State Universitiesand the U.S. GeologicalSurveywere to the success of thisproject.We particularlythank Ed Criley sphererises to the base of the crust in a slablesswindow crucial and the Crustal StudiesGroup at the USGS. Seismographs from the createdby the northwardmovementof the Gorda plate and Program for Array SeismicStudiesof the Continental Lithosphere then cools,creatingmagmaticunderplatingand new mantle (PASSCAL), StanfordUniversity,and the GeologicalSurveyof Canlithosphere,as a functionof time sincepassageof the triple ada were used in this research.Land use, graciouslypermitted by junction.Our proposedmodel for the thickeningof the lower individuals,corporations,and government,was essentialto our success.Specifically,we would like to thank Bob Bremner,Craig Brown, crustalmafic layer overlyingthe region of thin mantle litho- Lee Bucknell, David Dodd, Bill Feeney, Gilbert Martin, Michael sphereimpliesthat this mantle lithosphereand the overlying Ward, AndersonSolid Waste (Rick Morton), Bureauof Land Manmaficcrustareweldedto the Pacificplate andmovenorthwith agement(Michael Truden), CircleX Ranch(Marilyn Brooks),Georit relative to North America. This result supportsthe offset- giaPacificCo. (Tom Ray), Hoff Trust(Marie Hoff), LouisianaPacific (Dave Shantzand Roger Krueger),MendocinoNational Forest plate-boundary model,as developedby Furlonget al. [1989], Co. (DavidRoak), PacificLumberCo. (Tom Herman),RedwoodNational rather than the modelof Bohannonand Parsons[1995]which ParkService(EdwardHaberlin),SierraPacificIndustries(JackFrost), predictssignificantly older, and thereforethicker,lithosphere SimpsonTimber Co. (JuergenMomber), Six RiversNational Forest (Meredith Smith), Soper-WheelerCo. (James Soper), and Stover adjacentto the subductedGorda slab. South of our interpretedSEDGE, the region of thickened Ranch (Libby George). For use of their facilitiesduring our field operationwe extendthanksto HumboldtStateUniversityand Blosser Franciscancomplexand basalunderplatingoverliesa zone of Lane ElementarySchool(CharlesDavison). increasedscattering,intrinsicattenuation,or both, resulting from mechanicalmixingof lithologiesand/or partial melt left in the wake of the Gorda plate. Our modelsuggests that there References is newlyaccretedmaterial at the baseof the Franciscanrocks, Atwater, T., Implicationsof plate tectonicsfor the Cenozoictectonic perhapscoupledwith alterationof the lower crustor assimievolution of western North America, Geol. Soc.Am. Bull., 81, 3513lation into melt triggeredby upwellingasthenosphere[e.g., 3536, 1970. Johnsonand O'Neil, 1984;Liu and Furlong,1992].This newly Beaudoin,B.C., M. Magee, and H. Benz, Crustal velocitystructure northof the MendocinoTriple Junction,Geophys. Res.Lett.,21(21), accretedmaterial is at leastlocalizedalongthe southernedge 2319-2322, 1994. of the Gorda slaband mayextendsouthwardto the end of our Beaudoin, B.C., et al., The transition from slab to slabless:Results profile. Melts penetratingup into the crust,or lower crustal from the 1993 Mendocino Triple Junction Seismic Experiment, fluids releasedduring progrademetamorphismof overlying Geology,24, 195-199, 1996. North America, give rise to seismicbright spotsbelow shot Benz, H. M., G. Zandt, and D. H. Oppenheimer,Lithosphericstructure of northern California from teleseismicimagesof the upper point 903 [cf.Levanderet al., 1998]. leading edge of this "new Pacific"beginsto scrapeoff additional material 4. from the base of North America. In this man- Conclusions The MTJSE providesnew seismicdata that imagechanges in crustaland uppermostmantle structureassociated with the passageof the MendocinoTriple Junction.The crustalvelocity modelshows,on average,a laterallyhomogeneous velocityfor the Franciscanterraneand the greatestthicknessof -<7.0km/s lithologiesin the centerof our profile.The high densitiesand velocitiesat 23 km depth in the center of our profile are interpretedasnewlyaccretedmaterialat the baseof the Franciscanrocks,perhapscoupledwith alterationof the lowercrust or assimilationinto melt triggeredby upwellingasthenosphere. This newlyaccretedmaterialhasbeentectonicallythickenedin responseto eitherunderthrustoceaniccrustor newlyaccreted material that is coupledto the Pacificplate and is therefore movingnorth relativeto the overlyingNorth Americanplate. mantle,J. Geophys.Res.,97, 4791-4807, 1992. Blake, M. C., Jr., and A. S. Jayko,Tectonic evolution of northwest California and southwestOregon,Bull. Soc. Geol. Ft., 6, 921-930, 1986. Blake, M. C., Jr., A. S. Jayko, and R. J. McLaughlin, Tectonostratigraphicterranesof the northernCoastRanges,California,in Tectonostratigraphic Terranesof the Circum-PacificRegion,edited by D. G. Howell, pp. 159-186, Circum-Pac.Counc. for Energy and Miner. Resour., Houston, Tex., 1985. Bohannon,R. G., andT. Parsons,Tectonicimplicationsof post-30Ma Pacificand North Americanrelativeplate motions,Geol. Soc.Am. Bull., 107, 937-959, 1995. Brocher, T. M., J. McCarthy, P. E. Hart, W. S. Holbrook, K. P. Furlong,T. V. McEvilly, J. A. Hole, and S. L. Klemperer,Seismic evidencefor a lower-crustaldetachmentbeneathSanFranciscoBay, California, Science,265, 1436-1439, 1994. Carver,G., and R. B. Burke,Active convergenttectonicsin northwestern California, in GeologicEvolution of the NorthernmostCoast Ranges and WesternKlamath Mountains, California, Field Trip Guide., vol. T308, edited by K. R. Aalto and G. D. Harper, pp. 64-82, AGU, Washington,D.C., 1989. BEAUDOIN ET AL.: SOUTHERN Castillo, D. A., and W. L. Ellsworth, Seismotectonicsof the San An- dreasfault systembetween Point Arena and Cape Mendocino in northernCalifornia:Implicationsfor the developmentand evolution of a youngtransform,J. Geophys.Res.,98, 6543-6560, 1993. Clarke, S. H., Jr., Geologyof the Eel River basinand adjacentregion: Implicationsfor late Cenozoic tectonicsof the southernCascadia subductionzone and MendocinoTriple Junction,AAPG Bull., 76, 199-224, 1992. Clarke, S. H., Jr., and G. A. Carver, Late Holocene tectonics and paleoseismicity,southern Cascadiasubductionzone, Science,255, 188-192, 1992. Dickinson,W. R., Plate tectonicsand the continentalmargin of California,in TheGeotectonic Development of California,editedby W. G. Ernst, pp. 1-28, Prentice-Hall, EnglewoodCliffs, N.J., 1981. Dickinson,W. R., and W. S. Snyder,Geometry of triple junctions related to San Andreas transform,J. Geophys.Res., 84, 561-572, EDGE OF THE GORDA SLAB 30,115 Lendl, C., Finite differencewavefieldmodelingof large-aperturedata from the 1993 Mendocino Triple JunctionSeismicExperiment, Master thesis,Oreg. State Univ., Corvallis,1996. Lendl, C., A. Tr•hu, J. Goff, A. Levander,and B. Beaudoin,Synthetic seismograms throughsyntheticFranciscan:Insightsinto factorsaffectinglarge aperturedata, Geophys.Res.Lett., 24, 3317-3320, 1997. Levander, A. R., and R. L. Kovach, Shear velocity structure of the northernCalifornialithosphere,J. Geophys. Res.,95, 19,773-19,784, 1990. Levander, A., T. J. Henstock, A. S. Meltzer, A.M. Tr•hu, B.C. Beaudoin,and S. L. Klemperer,A windowthroughthe lithosphere: Seismicimagesof active magmaticcrustal accretion,Geology,26, 171-174, 1998. Liu, M., andK. P. Furlong,Cenozoicvolcanismin the CaliforniaCoast Ranges:Numerical solutions,J. Geophys.Res.,97, 4921-4951, 1992. Luetgert,J. H., Interactiveseismicraytracingfor the MacintoshT•, U.S. 1979. Geol. Surv.OpenFile Rep., 92-356, 1992. Fox, K. F., Jr., R. J. Fleck, G. H. Curtis,and C. E. Meyer, Implications Oliver, H. W., R. H. Chapman,S. Biehler,S. L. Robbins,W. F. Hanna, of the northwestwardly youngerageof volcanicrocksin west-central A. Griscom,L. A. Beyer,and E. A. Silver,Gravitymap of California California, Geol. Soc.Am. Bull., 96, 647-654, 1985. and its continentalmargin, Calif. Div. of Mines and Geol., Sacramento, 1980. Fuis, G. S., and W. D. Mooney, Lithosphericstructureand tectonics from seismic-refraction and other data, in The San Andreas Fault Oppenheimer,D., et al., The Cape Mendocino, California, earthSystem,California,U.S. Geol. Surv.Prof Pap., 1515, 207-236, 1990. quakesof April 1992:Subductionat the triple junction,Science,261, 433-438, 1993. Furlong, K. P., Lithosphericbehaviorwith triple junction migration: An examplebasedon the MendocinoTriple Junction,Phys.Earth Riddihough, R. R., Recent movementsof the Juan de Fuca plate system,J. Geophys.Res., 89, 6980-6994, 1984. Planet.Inter.,36(3-4), 213-223, 1984. Furlong,K. P., and D. M. Fountain,Continentalcrustalunderplating: Severinghaus, J., andT. Atwater, Cenozoicgeometryand thermalstate Thermal considerationsand seismic-petrologic consequences, J. of the subductingslabsbeneath westernNorth America, in Basin Geophys.Res., 91, 8285-8294, 1986. and Range ExtensionalTectonicsNear the Latitude of Las Vegas, Furlong,K. P., W. D. Hugo, and G. Zandt, Geometryand evolutionof Nevada, edited by B. P. Wernicke, pp. 1-22, Geol. Soc. of Am., Boulder, Colo., 1990. the SanAndreasfault zone in northernCalifornia,J. Geophys.Res., 94, 3100-3110, 1989. Thompson, G. A., and M. Talwani, Crustal structure from Pacific Godfrey, N.J., B.C. Beaudoin, C. Lendl, J. H. Luetgert, and A. S. Basin to central Nevada,J. Geophys.Res.,69, 4813-4837, 1964. Meltzer, Data report for the 1993 MendocinoTriple JunctionSeis- Tr•hu, A.M., and the MendocinoWorkingGroup,Pullingthe rug out mic Experiment,U.S. Geol. Surv.OpenFile Rep., 95-0275, 83 pp., from under California: Seismicimagesof the Mendocino Triple 1995. Junctionregion,Eos Trans.AGU, 76(38), 369, 380-381, 1995. Godfrey,N.J., B.C. Beaudoin,and S. L. Klemperer,and the Mendo- Verdonck, D., and G. Zandt, Three-dimensional crustal structure of cino Working Group, Ophiolitic basementto the Great Valley the MendocinoTriple Junctionregionfrom local earthquaketravel times,J. Geophys.Res.,99, 23,843-23,858, 1994. forearcbasin,California,from seismicandgravitydata:Implications for crustalgrowthat the North American continentalmargin,Geol. Warren, D. H., Seismic-refraction measurementsof crustal structure Soc.Am. Bull., 109, 1536-1562, 1997. near Santa Rosa and Ukiah, California, in Researchin the Geyser. Griscom,A., and R. C. Jachens,Tectonichistoryof the north portion Clear Lake GeothermalArea, NorthernCalifornia,edited by R. J. of the San Andreasfault system,California,inferred from gravity McLaughlin and J. M. Donnelly-Nolan,U.S. Geol. Surv.Prof Pap. 1141, 167-182, 1981. and magneticanomalies,J. Geophys.Res.,94, 3089-3099, 1989. Harper, G. D., The JosephineOphiolite: Remainsof a Late Jurassic Wilson, D. S., A kinematic model for the Gorda deformation zone as marginalbasinin northwesternCalifornia,Geology,8(7), 333-337, a diffusesouthernboundaryof the Juande Fuca plate,J. Geophys. 1980. Res., 91, 10,259-10,269, 1986. Heaton,T. H., and S. H. Hartzell, Earthquakehazardson the Cascadia Wilson,D. S., Deformationof the so-calledGordaplate,J. Geophys. subductionzone, Science,236, 162-168, 1987. Henstock, T. J., A. L. Levander, and J. A. Hole, Deformation in the lower crustof the San Andreasfault systemin northern California, Res., 94, 3065-3075, 1989. Zandt, G., Seismicimagesof the deep structureof the San Andreas fault system,central Coast Ranges,California,J. Geophys.Res.,86, Science,278, 650-653, 1997. 5039-5052, 1981. Hole, J. A., Nonlinear high-resolutionthree-dimensionalseismictravel Zandt, G., and K. P. Furlong, Evolution and thicknessof the lithotime tomography,J. Geophys.Res.,97, 6553-6562, 1992. spherebeneathcoastalCalifornia, Geology,10, 376-381, 1982. Hole, J. A., and B.C. Zelt, 3-D finite-difference reflection traveltimes, Geophys.J. Int., 121,427-434, 1995. B.C. Beaudoinand S. L. Klemperer, Department of Geophysics, Hole, J. A., B.C. Beaudoin,and T. J. Henstock,Wide-angleseismic Stanford University, Mitchell Building, Stanford, CA 94305-2215. constraintson the evolutionof the deepSanAndreasplateboundary (e-mail: bruce@geo.stanford.edu) by Mendocino Triple Junctionmigration, Tectonics,17, 802-818, J. A. Hole, Department of Geological Sciences,Virginia Tech, 1998. Blacksburg, VA 24061-0402.(e-mail: hole@vt.edu) Jachens,R. C., and A. Griscom,Three-dimensionalgeometryof the A.M. Trehu, Collegeof Ocean and AtmosphericSciences,Oregon Gorda plate beneathnorthernCalifornia,J. Geophys.Res.,88, 9375State University,Corvallis,OR 97331. 9392, 1983. Johnson,C. M., and J. R. O'Neil, Triple junction magmatism:A geochemicalstudyof Neogenevolcanicrocksin westernCalifornia, EarthPlanetSci.Lett., 71(2), 241-263, 1984. Lachenbruch,A. H., and J. H. Sass,Heat flow and energeticsof the San Andreas fault zone, J. Geophys.Res.,85, 6185-6222, 1980. (ReceivedAugust11, 1997;revisedJune 18, 1998; acceptedJune 19, 1998.)