) GEOPHYSICALRESEARCH LETTERS, VOL. 21, NO. 4, PAGES293-296, FEBRUARY 15, 1994

Hydro-isostaticdeflection and tectonic tilting in thecentral Andes:Initial resultsof a GPS surveyof Lake Minchin shorelines

BruceG. Bills,•'2 Shanaka L. deSi!va, 3Donald R. Currey,4Robert S. Emenger,4 KarlD. Lil!quist,4's Andrea Donnellan, 6 and Bruce Worden 7

Abstract.Sufficiently large lake loads provide a meansof probing to 3790m. Thespatial pattern of netvertical motions (accumulated since rheo!ogicalstratification of the crustand upper mantle. Lake Minchin the shorelinewas formed roughly 17,000 years ago), contains important wasthe largest of the latePleistocene pluvial lakes in thecentral Andes. new informationabout the responseof the crustand uppermantle to Prominentshorelines, which formed during temporary still-stands in the normal loadsfrom two sources(variations in the lake load itself, and climaticallydriven lake level history, preserverecords of lateral fluctuationsin sea level during the last deglaciation)and places variationsin subsequentnet vertical motions. At its maximumextent the constraintson the venica! tectonismassociated with subductingthe lakewas 140 m deepand spanned400 km N-S and 200 km E-W. The Nazcaplate under the SouthAmerica plate. load of surficial water contained in Lake Minchin was sufficient to Depositionalshorelines of largeendorheic lakes have three essential depressthe crustand underlying mantle by 20-40m, dependingon the propertiesthat make them usefulfor geodynamicstudies: they form subjacentrheology. Any otherdifferential vertical motions will alsobe quickly,they disappear slowly, and they are initially horizontal.As a recordedas departures from horizontality of theshorelines. We recently result,the climaticallyforced oscillations in depthand areal extentof conducteda surveyof shorelineelevations of Lake Minchin with the majorlakes in hydrologicallyclosed basins throughout the world provide expressintent of monitoringthe hydro-isostaticdeflection and tectonic a uniqueopportunity to investigatethe Earth'sresponse to normalloads tilting.Using real-timedifferential GPS, we measuredtopographic on lengthscales of 1O's to 100'sof kilometersand time scalesof decades profilesacross suites of shorelinesat 15 widelyseparated locations to millennia.Gilbert's (1890) pioneeringwork on (in throughoutthe basin. Horizontaland vertical accuraciesattained are western Utah and parts of Nevada and Idaho) establishedthe basic roughly30 and 70 cm, respectively.Geomorphic evidence suggests that frameworkfor limnologicneotectonics and has exerted a majorinfluence thehighest shoreline was occupiedonly briefly (probably less than 200 on subsequentstudies (Crittenden, 1963; Currey, 1982, 1990; years)and radiocarbondates on gastropodshells tbund in association Nakibogluand Lambeck, 1982; Bills and May, 1987,Bills and Currey, withthe shoredeposits constrain the age to roughly17 kyr. The basin- 1993). widepattern of elevationsof the highestshoreline is composedof two Thoughmuch of the literatureon rheologyof the crustand upper distinctsignals: (27 + 1) m of hydro-isostaticdeflection due to the lake mantleis concernedwith responseto majorglacial loads (Peltier, 1974, load,and a planartilt witheast and north components of (6.8 _+ 0.4) 10-5 1976; Wu and Peltier, 1983; Mitrovica and Peltier, 1991, 1992; and(-5.3 + 0.3) 10-5. Thisrate of tiltingis toohigh to be plausibly Lambeck,1990), lake loadshave several advantages over glacial loads. attributedto steadytectonism, and presumably reflects some unresolved Primaryamong them is thatthe complexspatio-temporal pattern of the combinationof tectonismplus the effectsof oceanicand lacustrine loads load can be easilyand accuratelyreconstructed from two ingredients: on a laterally heterogeneoussubstrate. The history of lake level present-daytopography and a history of lake level fluctuations.In fluctuationsis still inadequatelyknown to allow detailedinferences of contrast,reconstruction of an ice sheet load requiresat least three crestand mantlerheology. However, it is alreadyclear that the effective ingredients:geologic constraints on the temporalevolution of the ice elasticplate thicknessis closerto 40 km than the 60-70 km crustal margin, observations(or guesses)to constrainthe variableboundary thicknessin the centralAndes and the effectiveviscosity is lessthan 5 conditionsat the ice-till interface,and a complexsimulation of the non- 1020Pa s. linear flow of ice (Payneeta!., 1989; MacAyeal, 1989; Boultonand Introduction Clark, 1990). Lakesare alsomuch better recorders of climatichistory thanare glaciersand ice-sheets (Currey, 1990). The altip!anoof Bolivia is presentlya cold high desert,but at The early work of Agassiz(1876), Steinmannet al. (1904) and elevationsbelow 3800 m the landscapeis dominatedby coastal Bowman(1909) established that a seriesof largelakes have, at various geomorphicfeatures. During colderand/or wetter episodes in the past, timesthroughout the , occupied the presently semi-arid basins the hydrologicallyclosed basins of the altiplanohave contained of the Bolivianaltip!ano. The northernbasin, with a total area of 57.1 substantiallakes. The largest of thesewas lake Minchin, which extended 103km 2, presentlycontains (8.7 !03km 2 in area,3810 m 200km E-W and 400 km N-S, and was 140 m deep.Despite having surfaceelevation) and previouslyheld a somewhatlarger lake, (13-14 formedas a level surface,the highestshoreline of lake Minchin is 103km 2 area,,-,3900 m surfaceelevation) which Bowman (!909) called presentlyfound at orthometric(geoidal) elevations that range from 3750 lake Bal!ivian.The southernbasin, with a totalarea of 134 103km 2, presentlycontains only the shallow lake Poopo (34 103km 2 area,3660 m surfaceelevation) and the playasof Empexa,Coipasa and Uyuni, but t GeodynamicsBranch, NASA Goddard Space Flight Center previouslyheld a singlelarge lake (48-50 103 km 2 area,3760 m surface Dept.of Earthand Planetary Science, Johns Hopkins University elevation)which Steinmannet al. (1904) calledlake Minchin. It is this Dept.of Geographyand Geology, Indiana State University southernbasin lake which is thefocus of ourpresent study. Dept.of Geography,University of Utah The chronologyof fluctuationsin the level of lake Minchin is still Dept.of Geographyand Geology, Drake University only poorly known. Early workerssuggested that the large JetPropulsion Laboratory, Institute ofTechnology lakesmay haveresulted from meltingof glaciersin the surrounding SeismologicalLaboratory, California Institute ofTechnology mountains.However, Hastenrath and Kutzbach (1985) showedthat there wasinsufficient water stored in the glaciersto be a majorsource for the lake. The work of Servant and Fontes (1978) providedthe first radiocarbondates on shorelinesof lakeMinchin. Theyestimated that the higheststand of the lake (at closeto 3800 m) occurredprior to 28 kyr Copyright1994 by the American Geophysical Union. ago.More recent work, summarized by Lavenuet al. (1984),suggests an agerange of 22-27kyr. Papernumber 93GL03544 Servantand Fontes(!978) also pointedout that there was a late 0094-8534/94/93GL-03544503.O0 episodeof pluvial activity,which they calledthe Taucastage, that

293 294 Billset al.: Rebound and Tilting in CentralAndes culminatednear 3720 m elevationat about12 kyr BP. This roughly topographicantennawere traverse,obtained real-timeusing the estimatesL 1 signalsoffrom the position 5-8 GPSofsatellites the"rover" and coincidesin agewith the YoungerDryas climatic event, which was a briefreturn to full glacialconditions, and is seenin ice corerecords at an RTCM signaltransmitted from the "base"receiver. This allows bothpoles (Jouzel et al., 1987;Dansgaard et al., 1993)and in globalsea differentialcorrection of the pseudo-rangesignals from the satellites, level variations(Fairbanks, 1989). The Gilbert shorelineof Lake and enablesthe positionof the "rover"to be estimated,relative to the Bonnevillealso reflects this same global climatic event (Currey, 1990; "base",with horizontaland vertical accuraciesof 30 and 70 cm, Oviattet al., 1992). respectively.Distance between "base" and "rover" never exceeded 2 km Duringmuch of thelast deep-lake cycle on thealtiplano, global sea- duringthese surveys. levelstood -•120 m lowerthan at present(Fairbanks, 1989). The large- Afterreturning to the fiducial site, we processed the data collected by scaleand long-term net effect of a---120m deglacialrise in sea-levelis the "fiduciar'and "base"receivers to obtainan improvedestimate of the to depressthe oceanbasins relative to the continentsby --,40 m. positionof the "base"antenna during the surveysconducted that day. Interpretationof relative sea-levelhistories requires accurate Differencesbetween the improvedestimate and the initial estimatewere computationof deflectionat the coastline(Peltier, 1976; Lainbeck, then added to the real-time estimatesof "rover" locations. Distances 1990),but thereare relativelyfew directobservational constraints on between"fiducial" and "base"systems ranged up to 160 kin. As the howthe sea-level induced deflection decays with distance from the coast. shorelineswere initially formedon level (equipotential)surfaces, we The spatialand temporal scales over which this deformation occurs are correctedthe ellipsoidheights initially obtainedfrom the GPS datato determinedby the strengthof the crustand uppermantle. For any orthometric(geoidal) heights. significantocean loading signal to be presentin the LakeMinchin At each of the traverse sites, we measuredelevations of numerous shorelines,which lie at distancesof 100-300 km from the coastline, (10-30)shoreline features. However, the task of correlatingmost of these wouldrequire an effectivelithospheric thickness of order100 kin. featuresthroughout the basin will likely prove difficult. The one Thoughmany modelsof the tectonicevolution of the centralAndes conspicuousexception is the highestshoreline, which in everycase was exist (Lyon-Caenet al, 1985; !sacks,1988; Kono et al., 1989; easilydiscernible by the completelack of coastalgeomorphic features at Wdowinskiand O'Connell,1991; Baby et al., 1990),understanding of higherelevations on the surroundingterrain. As a result,we will limit the spario-temporalpattern of vertical motionsassociated with the ourdiscussion to thepattern of elevationsseen on this highestshoreline. subductionof theNazca plate and the accompanying deformation of the Results SouthAmerica plate is really quite limited.Existing observational constraintsare few and mostlyindirect. One of the mostcompelling Table 1 displaysour measurementsof the heightsof the high observationsis that the dip of the subductingslab (as indicatedby shorelineof lake Minchin at 15 locationswhich are reasonablywell patternsof earthquakehypocenter depths) varies significantly along the distributedaround the basin.Figure ! locatesthe surveysites and also strikeof thetrench (james, 1971; Barazangi and Isacks, 1979; Hasegawa showsthe configuration of the3800 m topographiccontour, which gives andSacks, 1981), and the pattern of reliefis significantlydifferent over a roughidea of the shapeof the lake. The contourinformation was the steeplyand shallowlydipping segments of the slab(Jordan et al., extractedfrom 1:250,000scale maps published in the early 1970'sjointly 1983; Dewey and Lamb, 1992). Marine terracesalong the Pacific by the U.S. DefenseMapping Agencyand the Bolivian Instituto coastlinesof Chileand Peru provide some constraints on upliftnear the GeograficoMilitar. trench,and indicatethat ratesare spatiallyand temporallyvariable We have obtainedradiocarbon dates on samplescollected from (DeVries, 1988; Hsu et al., 1989;Machare and Ortlieb, 1992). Fission severalshoreline contexts. Two of our samplespertain to the high track recordsof exhumationrates of plutonicrocks farther from the shoreline.Both were obtainedon gastropodshells collected from fine trenchindicate temporal variability (Crough, 1983; Benjaminet al., sandunits lying 1-2 m belowthe crestof the high shoreline.One of the 1987),but are oftenquoted as thoughthey were usefulmeasures of the samplelocalities is nearour surveysite Yonza, the othersample was uplift rate of the centralAndes, rather than singlepoint samplesof a takennear the village of SanPedro de Quemez(20 ø 44' S, 68ø 04' W). complexspatial pattern. Age estimatesfor bothsites were identical: (13,790 _+ 70) radiocarbon Risacherand Fritts(1991) recentlypresented evidence for regular years.Using the recent radiocarbon calibration of Bardet al. (1990),this variationsin thicknessof the uppersalt layer in the , correspondsto 16.8 kyr. This suggeststhat the high shorelinewas whichpresently contains most of the saltdelivered to lakeMinchin by its occupiedrather more recently than previous published estimates would tributaryrivers during the lastdeep-lake cycle. The top of thesalt unit is indicate. maintainedas a levelsurface by annualepisodes of rewettingand hydro- One of the primaryincentives for our studyis a desireto usethe aeolianplanation (Ericksen et al., 1978;Currey, 1990). The bottomof shorelineelevation patterns to refinethe rheologicalmodel of the crust the saltunit occursat depthsthat increasenearly linearly from <1 m on andupper mantle in thecentral Andes. However, attainment of thatgoal thewest margin to ~10 m nearthe east edge. The variationsin thickness mustawait further progress in two areas:additional clarification of the of thissalt unit couldbe taken as an indicationof 10 m of tiltingdown to spario-temporalpattern of deformation,and improveddefinition of the theeast over the 100km widthof thesalar since the endof thedeepqake lake levelfluctuation history. At present,the bestwe can achieveis a cycleroughly 10-12 ky.r ago. This tentativeinterpretation first suggested to us that we mightsee amounts of tectonictilting comparableto the lacustrinehydro-isostatic rebound signal. Table 1. Lake Minchin Hi•;h Shoreline feature observat:ions Method latitude(S) longitude(W) elevation,, sitenumber & name dd mm ss.ss dd mm ss.ss (m) Our topographicsurveys of the shorelinesof Lake Minchin were 1. Cerro Tres Cerrillos 20 33 00.79 66 48 07.85 3781.2 conductedusing three Trimble 4000 SSE P-code GPS receivers, 2. LomaCapilla 20 28 58.42 66 30 55.96 3787.6 correspondingantennas and ancillaryequipment. Each of the receivers 3. Cerro Paco Kkollu 20 05 07.28 68 13 58.73 3774.8 playeda differentrole. The "fiducial"receiver and its geodeticantenna 4. CerroCampanani 20 10 32.90 67 56 57.84 3784.0 occupiedour two fiducialsites, one for 10 days,the otherfor 4 days. 5. CerroAgua Castilia 20 52 36.96 67 03 47.09 3779.2 Data were collectedat the fiducial sitesfor 10 hours each day. The 6. Cerro Colorado 21 05 38.68 68 04 28.97 3777.6 "base"receiver and its geodeticantenna were transported to eachsurvey 7. Cerro Caldami 19 37 10.07 67 45 35.91 3786.0 locality,and remained fixed for the 1-2 hourduration of thetopographic 8. CerroKhasoj Saya 19 33 45.29 67 52 08.47 3782.8 traverseof the suiteof shorelines.The "rover"receiver and its compact 9. CerroMaraga 19 26 45.72 67 35 49.48 3789.2 domeantenna were mounted in a back-packand weretransported across 10. Cerro Salli Kkollu 19 23 36.83 67 30 46.43 3783.3 the terrain to be measured. l l. CerroPiquina 20 13 43.06 66 54 45.27 3787.4 Beforestarting a topographictraverse, the positionof the "base" 12. Cerro Ochkha 18 48 59.95 68 26 01.24 3753.0 receiverwas roughly estimated (horizontal position was taken from a 10- 13. Cerro Chacacheta 18 52 18.77 66 45 26.60 3768.1 20 minute GPS solution,vertical was estimatedfrom a 1:50,000 scale 14. Cerro San Pedro 18 06 51.90 67 00 11.98 3772.3 topographicmap with 20 m contours).During the courseof the 15. Yonza 20 37 16.33 68 0! 51.57 3783.7 Billset al.: Reboundand Tilting in CentralAndes 295

•oo For theinviscid model, we examined plate thicknesses from 10 to 75 kin. The bestfitting value was 38 km, wherethe residualvariance was

200 32.8% of the datavariance. All platethicknesses from 26 to 63 km producedre/sidua! variances below 40% of data variance.While it is clearthat the inviscid substrate model is overlysimplistic, this modeling exercisedoes clearly demonstrate that the mechanicalresponse of the loo crustand uppermanfie in the centralAndes to appliedsurface loads reflectsa "strong"layer with thickness less than that of thecrust alone. Valuesfor theother model parameters, obtained via a simpleleast- o squaressolution, are: A = (9.19+ 0.71)m, Sx TM(6.8 _+ 0.4) 10 4, and Sy= (-5.3+_ 0.3) 10 '5. Wealso considered a model which included quadraticterms [Sxx*(Ax)2 + Sxy*(Ax)(Ay) + Syy*(Ay)2], butit did not -!00 providea significantlyimproved fit tothe data. In the viscoussubstrate model, with observationalconstraints from only one shoreline,there is an unresolvedtrade-off between plate -200 /l thicknessand substrateviscosity. We can clearly reject substrate viscositieshigher than 5 1020Pa s. For lowerviscosities, our -'2 :o observationsdefine a curve in the two dimensionalparameter space east(kin) alongwhich the residualvariance is essentiallyconstant. Representative pointson that curve are: {(38,0), (36,10•8), (35, 10m), (30, 10•ø)}, Fig1. LakeMinchin Sun'ey Sites and Deflection Contours. wherelocations are given in termsof platethickness in km andsubstrate Solidcircles are surveysites. Irregular solid line is 3800 m elevation viscosityin Pa s. contour,x• hich approximatesthe maximun•extent of LakeMinchin. Taken together,these observations imply that: (a) the lake surface Smoothlines are countoursof net verticalmotion. including lacustrine elevation(corrected for tilt andrebound) at the high shorelinewas 3750 reboundand planartilting. Distancesare measuredfrom an arbitrar).' + A = (3759.2 + 0.7) m, CO)the computeddeflection profile was referencepoint with UTM coordinates 783øøøø N and 63 øøøø E. This essentiallycorrect, (c) the net motions,other than lacustrinerebound, correspondsto i9 ø 37' 19"S, 67ø 45' 37" W. havea remarkablyplanar aspect, and (d) net tilt in thecentral altiplano overthe last 16-17kyr hasbeen up to the eastby 6.8 cm/kmand down to simpleexploration of the sensitivityof our measurementsto the the north by 5.3 cm/km. The magnitudeof the observedtilt is theologicalstratification and a testof the hypothesisthat the observed comparableto oura priori expectations(from Salar de Uyuni salt wedge patternof elevationscan be reproducedby a linearsuperposition of two arguments)but is orientedroughly to the northwestrather than to the effects:a simplerebound model and a planartilt. Thoughthe viscosity of crustaland upper mantle rocks is a complex Interpretation functionof (at least) compositionand temperature,we confineour presentanalysis to two simplecases, involving either inviscidor The observedrebound signal is verysimilar to that computedfrom uniformlyviscous substrates, each overlain by an elasticlithospheric anyof a suiteof Earth models with elasticplate thicknesses of (40 + 10) plate.In bothcases the hydro-isostatic rebound was computed using the km andhalf-space viscosities less than 5 1020Pa s. The history of lake algorithmof Bills andMay (1987),applied to an Earthmodel consisting levelfluctuations is still inadequately known to allowdetailed inferences of an elasticplate at the surfaceand a Maxwell visco~elastichalf-space of crust and mantle theology.However, it is alreadyclear that the witha viscosityof eitherzero, or a valuein therange 1018-1022 Pa s. effectiveelastic plate thicknessis lessthan the 60-70 km crustal Thesurface load was computedfrom the digitalelevation model of the thicknessin the centralAndes, and therefore presumably insufficient to centralAndes compiled by Isacks(1988) andthe lake surfaceelevation allow a sealevel signal this far inland. historywas based on our estimatesof the high shorelineage and the The directionof observedtilting is somewhatsurprising, as the most publishedlake levelcurve of Servantand Fontes (1978). easilyanticipated axis of rotationis essentiallyparallel to thePeru-Chile Theinviscid case has the advantage of requiringno knowledge of the trench (as wouldbe expectedfor either oceanloading or subduction loadinghistory, and the elastic plate thickness obtained in thiscase in a relatedtectonism). The signof the E-W componentof rotation(up to the firmupper bound for the finiteviscosity cases. We simplycompare the east)is consistentwith eitheran oceanloading signal (down on the oceanicside) or the observationthat the primary locus of active observeddeflection to that computedfrom the waterload at the highest shoreline.In the finite viscositycase we are obligedto specifya lake tectonismis on the easternmargin of the A!tiplano.Less clear is the surfaceelevation history, which we havetaken to be asymmetricand orion of the N-S component.Furthermore, it is notclear that sucha piece-wiselinear, with a slow rise from zero to maximumover the planartilt signalshould have been anticipated. The best fitting linear- interval28-17 kyr, anda rapiddecline back to zeroover the interval plus-quadraticmodel only displays--.1 m of curvatureinduced deviation 17-14kyr. Though the actualhistory is doubtlessmore complex, this from the bestfitting linear-oulymodel. If the observationsreflect a mixtureof oceanloading (which would be strongestin the west)plus simplemodel is broadlyconsistent with knownconstraints. Small tectonism(presumably strongest in the east) this planar aspect of thetilt adjustmentsto the starting and ending times of the loading interval will must representa coincidentalmatching of the components. havelittle impact on estimated viscosities, since the age of occupation of Alternatively,the lack of significantcurvature implies a very strong thehigh shoreline is nowknown. elasticplate, which is counteredby the observedresponse to the lake Duringinitial exploration ofthese models it became evident that the load. observedshorelines contained a signalthat is not directlyrelated to A surprisingresult is themagnitude of thenet tilt signal(almost 35 rebound.In fact, the residuals displayed a distinctively planar trend. As a m of tilt acrossthe extentof the basin,versus 27 m of rebound).This result,we simply postulated that the observations might be approximated rateis higherthan can plausibly be continued for evena millionyears, bythe linear combination so if it is all tectonic,it mustrepresent episodic tectonism, as mightbe AZobs-- A + Azcalc+ Sx*A(east) + Sy*A(nonh) (1) expectedfrom phase changes in thesubducting slab (Cloos, 1993). Presentobservations do notappear to allowa uniqueexplanation for where,AZob sis observed geoida! height of the high shoreline elevation, this tilting,but we can list and brieflydiscuss possible contributing referredto anarbitrary standard value of 3750m, A is an adjustmentto processesand mechanisms. The processes can be broadly categorized as thereference surface elevation, Azcalc isthe calculated deflection atthe either external(climatically driven) or internal (tectono-magmatic). relevantlocation, Sx and Sy are the (east and north) slopes ofthe planar Climaticeffects which could contribute to the observedtilting include: regressionsurface, and A(east) and h(north) are the distances east and (a) oceanicloading accompanying sea-level rise, Co) lake loading on a northfrom an arbitraryreference point located near the center of the laterallyvariable substrate, (c) asyrmneMcglacial or icesheet loading, lake. and (d) erosionalunloading of the easternflanks of the Altiplano. Tectono-magmaticeffects that may have contributedinclude: (e) Hasegawa,A., andI.S. Sack,Subduction of the Nazca plate beneath changesin the subductionprocess (convergence rate, buoyancy of slab) Peru,J. Geophys.Res. 86, 4971-4980, 1981. at thePeru-Chile trench, or (f) magmaticinflation of theAltiplano-Puna Hsu,J.T., E.M. Leonard,and J.F. Wehmiller,Aminostratigraphy of volcanic zone. Peruvianand Chilean Quaternary marine terraces, Quat. Sci. Rev. 8, It is clearthat resolutionof theseissues will be greatlybenefited by 255-262, 1989. furthermodeling results, and mostespecially by observationsof the Isacks,B.L., Upliftof thecentral Andean plateau and bending of the patternsof deflectionon the lower shorelines.Although a single Bolivianorocline, J. Geophys.Res. 93, 3211-3231, 1988. shorelinesuffices to demonstratethe existence of multiplesignals, it will James,D.E., Plate tectonic model for the evolution of theCentral Andes, Bull. Geol. Soc.Amer. 82, 3225-3346, 197!. requireseveral separate views of thevertical motions to separatethem accordingto cause.The basic lacustrinesignal is easyto separate Jordan,T.E., B.L. Isacks,R.W. Allmendinger,J.A. Brewer,V.A. becauseit hasa distinctivegeometry.. The multitude of otherpotential Ramos,and C.J. Ando,Andean tectonics related to geometryof influenceswith geometry related to subductionwill onlybe separable subductedNazca Plate, Bull. Geol.Soc. Amer. 94, 341-361, 1983. throughtheir differing temporal signatures. Jouzel,J., C. Lorius,J.R. Petit, C. Genthon,N.!. Barkov,V.M. Kotlyakov,and V.M. Petrov,Vostok ice core:A continuousisotope Acknowledgments.This researchwas supportedby the NASA temperaturerecord over the last climatic cycle, Nature 329, 403-418, 1987. GeodynamicsProgram through RTOP 465-13-04 at GSFC,grant NAG 5-1734 at ISU, and grantNAG5-1822 at UU. The cooperationof key Kono,M., Y. Fukao,and A. Yamamoto,Mountain building in the personnelat ServicioGeologico de Boliviais gratefullyacknowledged. CentralAndes, J. Geophys.Res. 94, 3891-3905,1989. The commentsof two anonymousreviewers helped clarify the Lambeck,K., Glacialrebound, sea-level change and mantleviscosity, presentationof ourideas. Quart.J. Roy.Astr. Soc. 31, 1-30, 1990. Lavenu,A., M. Fornari,and M. Sebrier,Existence de deuxnouveaux References episodeslacustres quaternaires dans l'Altiplanoperuvo-bolivien, Cah. ORSTOM Ser. Geol. 14, 103-114, 1984. Agassiz,A., Hydrographicsketch of LakeTiticaca, Proc. Amer. Acad. Lyon-Caen,H., P. Molnar,and G. Suarez,Gravity anomalies and flexure Arts Sci. 11, 283-305, 1876. of the Brazilian shield, Earth Plan, Sci. Lett. 75, 81-92, 1985. Baby,P., T. Sempere,J. Oller,L. Barrios,G. Herail,and R. Marocco, MacAyeal,D.R., Largescale ice flow over a viscousbasal sediment, Un basin en compressiond'age oligo-miocene dans le sud de Geophys.Res. 94 4071-4087,1989. I'Altiplanobolivien, Cornpt. Rend. Acad. Sci. 311, 341-347, 1990. Machare,J., and L. Ortlieb, Plio-Quaternaryvertical motionsand the Barazangi,M., andB. Isacks,Subduction of the Nazcaplate beneath subductionof theNazca Ridge, Tectonophysics205, 97-108, 1992. Peru:Evidence from spatial distribution of earthquakes,Geophys. J. Mitrovica, J.X., and W.R. Peltier, A complete formalism for the Roy.Astr. Soc. 57, 537-555,I979. inversionof post-glacialrebound data, Geoph.vs. d. Int. 104, 267- Bard,E., B. Hamelin,R.G. Fairbanks,and A. Zindler, Calibrationof 288, !991. the!4C timescale over the past 30,000 years, Nature 345, 405-410, Mitrovica,J.X., and W.R. Peltier,Constraints on mantleviscosity from 1990. relativesea level variations, Geophys.Res. Lett. 19, 1185-1188, Benjamin,M.T., N.M. Johnson,and C.W. Naeser, Recent rapid uplift in 1992. the BolivianAndes, Geology !5 680-683,1987. Nakiboglu,S.M., andK. Lambeck,A studyof the Earth'sresponse to Bills, B.G., and G.M. May, Lake Bonneville: Constraintson surfaceloading with applicationsto Lake Bonneville, Geophys. lithosphericthickness and upper mantle viscosity, J. Geophys.Res. Roy.Astr. Soc. 70, 577-620,1982. 92, 11,493-11,508,1987. Oviatt,C.G., D.R. Currey,and D. Sack,Radiocarbon chronology of Lake Bills, B.G., and D.R. Currey,Viscosity estimates for the crustand upper Bonneville,Palogeog. Paleoclim. Paleoecol. 99, 225-241, 1992. mantlefrom patterns of lacustrineshoreline deformation in the Great Payne,A.J., D.E. Sugden,and C. M. Clapperton,Modeling the gro• Basin, J. Geophys.Res. (in press),1993. anddecay of theAntarctic ice sheet,Quat. Res. 31, 119-!34, 1989. Boulton,G.S., and C.D. Clark, A highlymobile Peltier,W.R., The impulseresponse of a Maxwell Earth,Rev. Geop&vs. revealedby glaciallineations, Nature 346, 813-817,1990. SpacePhys. 12, 64%669, 1974. Bowman,i., The physiographyof the centralAndes, Amer. J. Sci. 28, Peltier, W.R., Glacial-isostaticadjustment: The inverse problem, 373-402, 1909. Geophys.d. Roy.Astr. Soc. 46, 669-705,1976. Cande,S.C., and D.V. Kent, A newgeomagnetic polarity time scale for Risacher,F., and B. Fritz, Quaternarygeochemical evolution of the tl/e late Cretaceousand Cenozoic, J. Geophys.Res. 97, 13,917- salarsof Uyuniand Coipasa,Bolivia, Chem.Geol. 90, 211431, 13.951, 1992. 1991. Cloos,M., Lithosphericbuoyancy and collisional orogenesis, Bull. Geol. Servant,M., and J.C. Fontes,Les lacs quaternairesdes hautsp!ateaux Soc.Amer. 105, 715-737, 1993. desAndes boliviennes, Cah. ORSTOMSet. Geol. !0, 9-23, 1978. Crittenden,M.D., Effectiveviscosity of the Earthfrom isostaticloading Steinmann,G., H. Hok,and A. Bistram,Zur geologicdes suedestland of PleistoceneLake Bonneville, J. Geophys.Res. 68, 5517-5530, Bolivien, Zbl. btiner. 5,1-4, 1904. 1963. Wdowinski,S., andR.J. O'Connell, Deformation of the CentralAndes Crough,S.T., Apatitefission-track dating of erosionin the eastern derivedfrom a flowmodel of subductionzones, J. Geophys.Res. 96, Andes,Bolivia, Earth Plan. Sci. Lett. 64, 396-397, 1983. 12,245-12,255, 1991. Currey,D.R., Lake Bonneville:Features of relevanceto neotectonic Wu, P., andW.R. Peltier,Glacial isostatic adjustment as a constrainton analysis,U.S. Geol. Surv. Open File Rep.,82-1070, 31 pp., 1982. deepmantle viscosity, Geophys. J. Roy. Astr. Soc. 74, 377-449, Currey,D.R., Quaternarypaleolakes in the evolutionof semidesert 1983. basins,Paleogeog. Paleoclim. Paleoecol. 76, I89-214, 1990. Dansgaard,W., S.J.Johnsen, D. Dahl-Jensen,N.S. Gundestrup,C.U. BruceG. Bills,Geodynamics Branch, NASA GoddardSpace Flight Hammer,J. Jouzel,and G. Bond,Evidence for generalinstability of Center,Greenbelt, MD 20771 pastclimate from a 250-kyrice-core record, Nature 364, 218-220, DonaldR. Curreyand Robert S. Emenger,Department of 1993. Geography,University of Utah,Salt Lake City, UT 84 ! 12 DeVries, T.J., The geologyof late Cenozoicmarine terraces in ShanakaL: de Silva,Department of Geographyand Geology, northwesternPeru, J. S. Amer.Earth Sci. I, 121-136,1988. IndianaState University, Terra Haute, IN 47809 Dewey, J.F., and S.H. Lamb, Active tectonicsof the Andes, AndreaDonnellan, Jet Propulsion Laboratory, California Institute of Tectonophysics205, 79-95, 1992. Technology,Pasadena, CA 91109 Ericksen,G.E., S.D. Vine, andR.A. Ballon, Chemicalcomposition and Karl D. Li!lquist,Department of Geographyand Geology, Drake distributionof lithium-richbrines in Salar de Uyuni and nearby University,Des Moines, IA 50311 salarsin southwesternBolivia, Energy3,355-363, 1978. BruceWorden, Seismological Laboratory, California Institute of Fairbanks,R.G., A 17,000year glacio-eustatic sea level record,Nature Technology,Pasadena, CA 91125 342, 637-4542,1989. Gilbert,G.K., Lake Bonneville, U.S. Geol. Surv. Mono. 1, •38 pp., Received:13 August 1993 Revised:14 October1993 !890. Accepted:19 November1993