Origin of the First Global Meltwater Pulse Following the Last Glacial Maximum

PALEOCEANOGRAPHY, VOL. 11, NO. 5, PAGES 563-577, OCTOBER 1996

Origin of the first global meltwater pulse following the last glacial maximum

PeterU. Clark,l Richard B. Alley, 2Lloyd D. Keigwin,3Joseph M. Licciardi,• SigfusJ.Johnsen, 4'•and Huaxiao Wang 6

Abstract. Well-datedsea level recordsshow that the glacioeustaticrise followingthe lastglacial maximumwas characterizedby two or possiblythree brief intervalsof rapid sealevel rise separating periodswith muchlower rates. Thesevery highrates of sealevel rise indicate periods of ex- ceptionallyrapid deglaciation of remainingice sheets.The LaurentideIce Sheetis commonlytar- getedas the sourceof the first, andlargest, of the meltwaterpulses (mwp-IA between-14,200 (12,200•4C years B.P.) and 13,700 years ago (11,700 •4C years B.P.)). In alloceanic records of deglaciationof the formernorthern hemisphere ice sheetsthat we review, only thosefrom the Gulfof Mexico and the Bermuda Rise show evidence oflow ;5•80 values at the time of mwp-IA, identifyingthe southernLaurentide Ice Sheetas a potentialsource for mwp-IA. We questionthis sourcefor mwp-IA, however,because (1) ice sheetmodels suggest that this sectorof the ice sheet contributedonly a fraction(<10%) of the sealevel needed for mwp-IA, (2) meltingthis sector of the ice sheetat the necessaryrate to explainmwp-IA is physicallyimplausible, and (3) ocean modelspredict a muchstronger thermohaline response to theinferred freshwater pulse out of the MississippiRiver into theNorth Atlanticthan is recorded.This leavesthe AntarcticIce Sheetas the only otherice sheetcapable of deliveringenough sea level to explainmwp-IA, butthere are currentlyno well-datedhigh-resolution records to documentthis hypothesis. These conclusions suggestthat reconstructionsof the LaurentideIce Sheetin the ICE-4G model,which are constrainedto matchthe sealevel record,may be too low for time periodsyounger than 15,000years ago.Furthermore, ;5•80 records from the Gulf of Mexico show variable fluxes of meltwater from the southernmargin of the LaurentideIce Sheetwhich can be tracedto the openingand closing of eastwarddraining glacial-lake outlets associated with surgingice sheetbehavior. Thesevariable fluxesthrough eastern outlets were apparentlysufficient to affectformation of North Atlantic DeepWater, thus underscoring the sensitivityof thisprocess to changesin freshwaterforcing.

Introduction et al., 1990a; Edwards et al., 1993; Blanchonand Shaw, 1995]. Thesevery high ratesof sealevel rise indicateperiods of excep- Well-dated sea level recordsshow that the glacioeustaticrise tionally rapid deglaciationof remainingice sheets.The source(s) followingthe last glacialmaximum (approximately 21,000 years of these rapid rates of sea level rise, or "meltwaterpulses" ago) (calendaryears are representedas years and radiocarbon [Fairbanks, 1989], however, has not been established. In the yearsare represented as•4C years B.P.) was characterized bytwo caseof the first, and largest,of the meltwaterpulses, several dif- or possiblythree brief intervalsof rapid sealevel rise separating ferent ice sheetshave been implicated. The FennoscandianIce periodswith muchlower rates (Figure 1) [Fairbanks,1989; Bard Sheetmay have beenresponsible [Birchfield et al., 1994] because it occurs"downwind" of the Nordic seaswhere large amountsof •DepartmentofGeosciences, Oregon State University, Corvallis. heat are released when active in the formation of North Atlantic 2EarthSystem Science Center and Department ofGeosciences, Penn- Deep Water (NADW). The BarentsIce Sheetmay havebeen re- sylvaniaState University, University Park. sponsiblebecause it was largely groundedbelow sea level and 3WoodsHole Oceanographic Institution, Woods Hole, Massachu- thereforesusceptible to the rapid collapsesuggested by the first setts. 4NiehlsBohr Institute, Department of Geophysics, University of Co- meltwater pulse [Lindstromand MacAyeal, 1993]. Many re- penhagen,Denmark. searchers,however, have targetedthe LaurentideIce Sheetas the 5Alsoat ScienceInstitute, Department of Geophysics, University of primary sourcefor the first meltwaterpulse for two reasons:(1) Iceland,Reykjavik. The Laurentidewas the largestice sheetand thus most likely to 6LawrenceLivermore National Laboratory, IAvermore, California. deliver sucha large volumeof meltwater,and (2) oxygeniso- topesin deep-seacores from the Gulf of Mexico and midlatitude Copyright1996 by theAmerican Geophysical Union. North Atlanticrecord an intervalof low fi•80 at the time of the Papernumber 96PA01419. first meltwaterpulse suggesting that glacialmeltwater discharged 0883-8305/96/96PA-01419512.00 down the MississippiRiver and spreadas a low-salinityplume

563 564 CLARKET AL.: ORIGIN OF THE FIRSTGLOBAL MELTWATER PULSE

acrossthe temperateNorth Atlantic [e.g., Keigwin et al., 1991; 0 Charles and Fairbanks, 1992; Fairbanks et al., 1992; Lehman and Keigwin, 1992]. This interpretationhas figured prominently Barbados in recentmodeling of the deglaciationof the global ice sheets New Guinea (ICE-4G) [Peltier, 1994], whereby the LaurentideIce Sheet is 25 constrainedto losenearly 20 m of sealevel equivalentduring the first meltwaterpulse between 15,000 and 14,000years ago. In this paper,we presentseveral lines of evidencewhich suggestthat the LaurentideIce Sheetcould not have beenthe pri- marycontributor tothis first meltwater pulse. Although •5180 re- 50 mwp-IB cordsclearly identify meltwaterflowing down the Mississippi River, our evidencesuggests that this dischargelikely only ac- counted for a fraction of the total sea level rise. Furthermore,

deep-searecords of meltwaterand icebergdischarge from the_ 75 Gulf of St. Lawrence and Hudson Strait, or areas that se•;,ed as the other primary outletsof freshwaterto the North Atlantic in mwp-IA additionto the MississippiRiver, showevidence of significantly reduceddischarge at the time of the first meltwaterpulse. Be- cause marine records near the former Fennoscandian and Barents 100 Ice Sheetsalso showno signatureof accelerateddeglaciation at the time of the first meltwaterpulse, records of deglaciationof the AntarcticIce Sheetshould be examinedfor possibleevidence of its role in thisprominent sea level event. 125 20 15 10 Sea Level Record (ka) Assuminga 105m riseof globalsea level I•etween21,000 and 6000 yearsago [Peltlet, 1994],the rateof sealevel rise would be Figure 1. Coral recordsof sealevel datedby U•'h from Barba- 7 m/1000 years if all ice sheetshad melted unitbrmly. Corals dos [Bard et al., 1990a, 1993] andNew Guinea [Edwardset ai., datedby U/Th recordingsea level rise at Barbados[Fairbanks, 1993]. Two periodsof rapid rise of sea level are identifiedas 1989; Bard et al., 1990a, 1993], New Guinea[Edwar&' eta/.,' mwp-IA and mwp-IB. 1993], and Tahiti [Bard et al., 1995], however, show that between 15,000 and 10,000 yearsago, the postglacialsea level history was punctuatedby two brief intervalsof rapid ratesof sea level rise,which Fairbanks [ 1989] referredto asmeltwater pulses northernhemisphere ice sheetswere involved,this eventshould (mwp's) IA and lB (Figure l). Accordingto the Barbadosrec- have left a clear signalin the North Atlantic Ocean,which re- ord, sea level during the first meltwater pulse (mwp-IA) rose about24 m between14,690 + 25 yearsago (12,600 + 460 ceivedmost of the meltwaterand icebergsfrom theseice sheets. yearsB.P.) and 13,730 + 100years ago (11.720 + 400inc years Here we examinerecords for evidenceof an abruptdeglaciation event from the northernhemisphere ice sheetsat the time of B.P.), with mostof this sea level rise (19 m) occurringhfter 14,230+ 100years (12,200 + incyears B.P.) [Bard et al., 1990a] mwp-IA. (Figure 1; Table l). Blanchonand Shaw [1995] arguedthat changesfrom a monospecificAcropora palmata reef framework Laurentide Ice Sheet to a deeper-watercoral reef framework constrain13.5 m of sea level rise in <290 + 50 yearsbeginning at or shortlyafter 14,230 + 100 yearsago. On the basisof this estimate,meltwater was in- The LaurentideIce Sheetwas the largestof the former ice troducedto the globaloceans during mwp-IA at an averagerate sheets. Most meltwaterand icebergsfrom the LaurentideIce of-4.5 cm/yr, correspondingto a freshwaterflux of-•0.52 Sv Sheetdrained to theNorth Atlantic through three outlets: Hudson (Sverdrup)(1 Sv = 106m 3 s-l),for about 300 years. The Fair- Strait, the Gulf of St. La'arence,and the MississippiRiver banksand Bard data give a best-estimatesea level rise of 19 m (Figure2). The oxygenisotope record in boththe Gulf of Mex- over 500 years(3.8 cm/yr), or 0.43 Sv. Thusthe freshwaterflux ico [Leventeret al., 1982] andthe BermudaRise [Keigwinet al., of mwp-IA is in the vicinity of 0.5 Sv by eithercalculation, while 1991]show a spikeof low•5•80 at the time of mwp-IA(Figure the estimateof sealevel riseranges from 13.5 to 19 m. 3b), leading several researchersto suggestthat the meltwater pulseoriginated from the southernmargin of the LaurentideIce Sheet through the MississippiRiver to the Gulf of Mexico [Keigwinet al., 1991; Fairbankset al., 1992] [seealso Emiliani, Evidence From Northern Hemisphere Ice Sheets 1976; Emiliani et al., 1978]. Extractinga meltwater flux of 0.5Sv from the 2 x 10•2 m 2 or The rapidity and magnitudeof mwp-IA strongly suggests less sectorof the LaurentideIce Sheet draining to the Gulf of abruptcollapse of someportion(s) of remainingice sheets.If the Mexico would requirethat the entire sectorbecame an ablation CLARK ET AL.: ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE 565

Table 1. Agesof SamplesConstraining the Timing of MeltwaterPulse IA

Sample Depth,am U/Th,b years (at 2 •) •4C,byears B.P. (at 2 •)

RGFI2-21-10 -70 13,730 + 100 (2) 11,720 + 400 (1)

RGF9-8-2 -89 14,230 + 100 (1) 12,200 + 75 (2)

RGF9-13-3 -94 14,690 $ 25 (2) 12,600 $ 460 (3)

aDepths are relative to thepresent sea level and can be converted to formersea level by multiplyingby the averagelong-term uplift rate of Barbados(0.34 m/1000 years) [Fa?banks,1989]. u Agesare weightedmeans if morethan one agedetermination was measured on same sample.Number of agedeterminations are shownby numbersin parentheses.Data from Fairbanks[ 1989] andBard et al. [ 1990a, 1993].

zone with an average melt rate of at least 8 m/yr of ice, an ex- ter flux of 0.03 Sv over that time, similar to estimatesmade by tremerate for any ice on Earth at presentand an implausiblyhigh Teller [1990]. rate for a large sectionof an ice sheet[cf. Andrews,1973]. Esti- We thusconclude that it is physicallyimplausible to melt the matesof the temperaturesensitivity of melting of the modem volume of ice from the Mississippisector of the LaurentideIce GreenlandIce Sheet,when scaledto the area of the Mississippi Sheetwithin the time suggestedby mwp-lA and that the volume sector,produce an annualaverage meltwater increase of 0.004- of meltwaterthat flowed down the MississippiRiver, suggested 0.007 Sv/degreewarming if snowfall does not increaseand by ice sheetmodels, was only a fraction(<10%) of the flux re- somewhatless if snowfalldoes increase [Warrick and Oerlemans, quired to explain the sea level rise during this interval [cf. An- 1990]. This suggeststhat an increaseof 0.5 Sv in meltwaterfrom drews,1976]. theMississippi drainage could be achievedtbr a l ø warmingif We next examinedrecords of meltwaterand icebergdischarge that part of the Laurentidewere at least70-125 timesmore sensi- throughHudson Strait and the Gulf of St. Lawrence,as theserep- tive to warming than the modem GreenlandIce Sheet, by a resenttwo other primary outletsof meltwaterand icebergsfrom warmingof 70ø-125ø with modemGreenland sensitivity, or by the LaurentideIce Sheet to the North Atlantic (Figure 2). Fur- somecombination of increasedwarming and •ensitivity. None of thermore,these representmore reasonableroutes for the rapid theseseems physically plausible, and they would be even less collapseof the ice sheetthan the Mississippisector, since the ice plausibleif we consideredonly a seasonalmelt rate. sheetcould discharge ice directlyto the oceanby calvingof ice- Numericalmodeling of the LaurentideIce Sheetprovides es- bergs. timates of the ice volume contained within the sector of the ice Large freshwaterdischarges originating from theseoutlets are sheetthat drainedto the MississippiRiver. Using ice sheetre- identified by oxygen isotopes and ice-rafted debris (IRD) constructionsby Hughes[1987], Teller [1990] estimatedthe vol- (Figures3c and 3d) [Andrewsand Tedesco,1992; Andrews et al.. ume of runoff to the MississippiRiver derivedfrom precipitation 1994; Bond et al., 1992, 1993; Bond and Lotti, 1995; Keigwin and ice sheet melting during the last deglaciation. Between and Jones,1995]. Thesedata, however, show evidence of only 13,000and 11,000 InC years B.P., or spanningthe interval of low or reduceddischarge of freshwaterto the North Atlantic at mwp-IA [Fairbanks,1989] (Table 1), the componentof runoff to the time of mwp-IA. Major iceberg-dischargeevents only oc- the Gulf of Mexico from glacialmeltwater alone ranged between curredout of HudsonStrait (Heinrich events) -14,500 •4C years 0.02 to 0.04 Sv, suggestingdischarges more'than an orderof B.P.and during the Younger Dryas between 10,000-11,000 14C magnitudelower than those neededto explain mwp-lA [Teller, years B.P. (Figure 3c) [Andrewsand Tedesco,1992; Hillaire- 19901. Marcel et al., 1994; Andrewset al., 1995; Bondand Lotti, 1995]; We reconstructedthe LaurentideIce Sheetduring the last de- theseare alsothe only timesof major excursionsto light oxygen glaciation(Figure 4) (J.M. Licciardiet al., Modelingthe Lauren- isotope values [Andrews et al., 1994; Hillaire-Marcel et al., tide Ice Sheetthrough the last deglaciation,submitted to Quater- 1994].At the time of mwp-IA off the coast of Nova Scotia, 15•80 nary ScienceReviews, 1996 (hereinafterreferred to as J.M. Licci- valueswere at their highest(Figure 3d) and warm-waterplank- ardi et al., submittedmanuscript, 1996)) usinga numericalmodel tonic foran•iniferafirst becamecommon, suggesting the absence that includesthe effectsof subglacialsediment deformation on of any significantmeltwater out of the Gulf of St. La;vrenceat ice flow [Clark et al., 1996; Jensonet al., 1995, 1996]. Like the time of mwp-IA [Keigwinand Jones,1995]. Teller's[1990] resultsusing ice sheetreconstructions by Hughes '[1987],ou? reconstructions alsosuggest that the sector of the ice sheetsupplying meltwater to the Gulf of Mexico containedonly a Eurasian Ice Sheets fractionof the ice volume requiredto explainmwp-IA. Between 12,000and 11,000 14C years B.P., the decrease in ice volume in Marine recordsof deglaciationof the Fennoscandianand Bar- the sectorof the ice sheetdraining to the MississippiRiver would entsSea Ice Sheets(Figure 5) showno evidencefor a majorcol- haveincreased global sea level by 2.9 m, equivalentto a meltwa- lapseof theseice sheetsduring the time of mwp-IA. Oxygen 566 CLARK ET AL.' ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE

Figure 2. Map showinglocation of palcoclimaterecords discussed in thetext andshowing reconstruction of Lau- rentideIce Sheet for 12,000 Inc years B.P. (ice-surface contour interval = 500 m; from J.M. Licciardi etal., submitted manuscript,1996). Coresare 1, EN32 PC-6; 2, KNR31 GPC5' 3, compositeof coresfrom off the Nova Scotia coast;4, SummitGreenland ice core;5, PS21295;6, HM94-34; 7, HM79 6/4; 8, Troll 3.1' 9, V23-081' and 10, SU81-18.

isotoperecords from the Norwegian and Greenland Seas (Figures parabletrends (Figure 5f), with low valuesduring the early de- 5b, 5c, 5d, and 5e) andthe easterncentral Arctic Ocean [Stein et glaciationof the Eurasianice sheetsand high valuescomparable al., 1994]show low fi180values between 15,000-13,000 to modem valuesduring the time of mwp-IA [Duplessyet al., yearsB.P. which are generallyattributed to rapid deglaciationof 1992, 1993]. somepart of the Eurasianice sheets[Jones and Keigwin, 1988; As further discussedbelow, these data are consistentwith cli- Lehmanet al., 1991; Koq and Jansen,1994], probablythe Bar- mate recordswhich suggestwarm sea surfacetemperatures and entsand Kara Sea sectors[Elverhoi et al., 1995; Polyaket al., activeformation of North AtlanticDeep Water in the Nordic seas 1995].At thetime of mwp-lA, however, 15•80 values are gener- at this time [Karpuz and Jansen, 1992; Lehmanand Keigwin, ally at or neartheir highestvalues, suggesting little meltwaterin- 1992], which would only be expectedin the absenceof signifi- put into the Norwegianand Greenland Seas at this time. Recon- cant freshwaterfluxes to these seas,particularly from immedi- structed salinities for the northeastern North Atlantic show com- ately adjacent(Eurasian) ice sheets. CLARK ET AL.' ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE 567

a b c d lO

"12

14 100 80 601 0 -1 -2 -3 0 1500 3 2.5 2

(m) •180 (%o) lithic grains/gm •180 (ø/oo)

Figure 3. The Barbadossea level recordcompared with oceanicrecords of LaurentideIce Sheetdeglaciation be- tween10,000 and 14,000 •4C years B.P. Solid horizontal line corrresponds tobeginning offastest rate of sea level rise duringmwp-IA (seeTable I for correspondingages). (a) Barbadossea level recorddated by radiocarbon [Fairbanks,1989]. (b) Oxygenisotope records from the Gulf of Mexico (solidline, coreEN32 PC-6) (oxygen isotopevalues from Leventeret al. [1982]; agemodel from Keigwin et al. [1991]) andthe BermudaRise (dashed line, core KNR31 GPC5) [from Keigwinet al., 1991]. Both recordsmeasured on planktonicforaminifera Globi- gerinoidesruber (white variety).(c) Recordof concentrationof lithic grainsper gramof sedimentfrom coreV23- 081[from Bond and Lotti, 1995]. Prominent peak at 13,800-14,000 Incyears B.P. is latter part of Heinrich event 1 andpeak from 10,400-10,000 •C years B.P. is earlier part of Heinrich event 0. (d)Results ofstacked oxygen isotoperecords measured on planktonicforaminifera sinistrally coiled (s.) Neogloboquadrinapachyderma from continentalslope of Nova Scotia[from Keigwinand Jones,1995].

North Atlantic Deep Water and Climate Change t¾omthe Mississippi River (versus other sources)on thermohaline circulation. Their model results show that a freshwater flux from the MississippiRiver of 0.5 Sv resultsin a >50% re- Formationof North AtlanticDeep Water (NADW) by thermo- ductionin NADW formationwithin 50 years,and in 150 years, halinecirculation results in significantmeridional heat transport or within the duration.of mwp-IA, NADW is nearly completely into the North Atlantic [Rooth, 1982; Broeckeret al., 1985]. shutdown (Figure 6). Geochemicaldata supportthe notion that changesin the rate Thesemodel results suggest that injectionof a flux of freshwa- and/or site of formation of NADW and resultingloss of ocean ter from the MississippiRiver equivalentto that suggestedby heat transportare associatedwith abruptclimate change in the mwp-IA shouldbe accompaniedby a majorclimatic cooling. We North Atlantic region [Boyleand Keigwin, 1987; Keigwinet al., thuscompared the Barbadossea level recordwith climaterecords 1991; Lehman and Keigwin, 1992; Oppo and Lehman, 1995]. in order to examine any relation betweenmwp-IA and climate Oceanmodels show that formationof NADW is sensitiveto any change. changesin the salinitybudget in thoseareas of deepwaterforma- All climate recordsshow that mwp-IA occurredduring the tion. Thus model studiessuggest that changesin the ratesand Bolling-Allerodwarm interval(Figures 7 and 8). We note,how- sites of formation of NADW may occur by the injection of ever,that because this warming occurred as a resultof intensified freshwater to the areas of convection and that NADW formation formationof NADW in the Nordic seas[Charles and Fairbanks, is moresensitive to high-latitudethan low-latitude meltwater dis- 1992; Lehmanand Keigwin, 1992], the climaticinfluence of this charge [Bryan, 1986; Maier-Reimer and Mikolajewicz, 1989; processwould be largelyrestricted to the Eurasianice sheetsand Wright and Stocker, 1993; Rahmstorf, 1994, 1995; Sakai and would not stronglyaffect the meltingrate of the LaurentideIce Peltier, 1995]. Sheet [Rind et al., 1986; Manabe and Stouffer, 1988]. This As summarizedabove, the only clear signal for enhanced warming apparentlydid not have a significanteffect on the meltwaterflux from northemhemisphere ice sheetsduring the Eurasian ice sheetseither, however, since there is no record of time of mwp-IA is from oxygenisotope records of meltwater enhancedmeltwater discharge to the Greenlandand Norwegian flowingdown the MississippiRiver. Usingthe Geophysical Seasduring the Bolling-Allerodinterval, and in particularduring FluidDynamics Laboratory modular ocean model, H. Wang,J.C. thetime of mwp-IA (Figure5). Evans,and R.C.. Thunell (Thermohalinecirculation and meltwa- All recordsshow a brief coolinginterval following the time of ter input: OGCM experimentsusing different surfacethermal mwp-IA, althoughthis coolingis associatedwith the OlderDryas boundaryconditions, submitted to Journalof GeophysicalRe- in the GreenlandIce coreProject (GRIP) record(Figure 7c) and search,1995 (hereinafterreferred to asH. Wanget al., submitted with the intra-Allerodcold periodin radiocarbon-datedmarine manuscript,1995)) investigatedthe effect of meltwaterinput records(Figures 8c and 8d) [Lehmanand Keigwin, 1992]. Geo- 568 CLARKET AL.: ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE 0 1000KM '••"•/• ---•0 1000KM

.•oo 200

40 ø 40 ø

120' 110ø 10tJ' 90' 80ø 70ø 120' 110ø 100' 90' 80ø 70'

h•••. 0 1000KM

•0 o

5'0ø ...... :.::•:iiiii":i!iii!

120' 110' 100' 90' 80ø 70'

Figure4. ReconstructionsofLaurentide IceSheet at(A) 13,000 •nC years B.P., (B) 12,000 •nC years B.P., and (C) 11.,000•4C years B.P. (ice-surface contour interval = 500 m) (from J.M. Licciardi etal., submitted manuscript, 1996). The sectorof the ice sheetthat drained to theMississippi River drainage is outlinedand labeled MS. These reconstructionsandthose by Hughes[1987] suggest a meltwater flux to theMississippi drainage of-0.03 Sv [see also Teller, 1990]. chemicaldata (•13C, Cd'•a) also suggest a decrease inthe rate of these periods occur at times of significantlylower meltwater productionofNADW following, mwp-IA (Figure 8b) [Keigwin et fluxesthan duringmwp-IA. al., 1991; Charlesand Fairban•s•.1992]. Oceanmodeling re- Modeling resultsby Sakai and Peltier [1995] predictthat results(Figure 6) suggest,however, that the effect of a meltwater ductionin NADW formationin responseto a freshwaterforcing flux of 0.5 Sv derivedfrom the MississippiRiver on NADW of the magnitudesuggested by mwp-IA would be significantly productionand climate would be significantlygreater than the delayedfrom the time of onsetof the freshwaterevent. Thesere- small and short res13onseidentified in the climate records. In sultssuggest therefore that the YoungerDryas cold event is the observedresponse to theirmodel prediction and is thusconsistent particular,we contrastthe modestclimate responsefollowing with a source of mwp-IA from the northern hemisphereice mwp-IA with the greatermagnitude and duration of otherperiods sheets. This model result still requires,however, observational of reduced NADW formation during the last deglaciation data that identify the northernhemisphere ice sheet(s)as being [Keigwinet al., 1991, Keigwin and Lehman, 1994;Maslin et al., responsiblefor mwp-IA which, accordingto the recordswe have 1995; Oppo and Lehman, 1995], despitethe observationthat surveyed,do not exist. CLARK ET AL ßORIGIN OF THE FIRST GLOBAL MEI,TWATER PULSE 569

a b c lO

15 I I I I 120 100 80 60 4.5 4 3.5 4.5 4 (m) fi•O(% o) fi•O(% o) d e f 10

12 13

14 15] , , I I I 3.5 2.5 fi•O(% o) 180( %o) salinity (%o)

Figure 5. The Barbadossea level recordcompared with oceanicrecords that reflectprimarily to Eurasianice sheet deglaciationbetween 10,000 and 15,000 •4C years B.P. Solidhorizontal line corresponds tobeginning of fastest rate of sealevel rise duringmwp-IA (seeTable 1 for correspondingages). (a) Barbadossea level recorddated by radiocarbon[Fairbanks, 1989]. (b) Oxygenisotope record (core PS21295-4) measured on N. pachyderma(s.) from Fram Strait [fromJones and Keigwin,1988]. (c) Oxygenisotope record (core HM94-34) measuredon N. pachyderma(s.) from the GreenlandBasin [from Koq and Jansen,1994]. (d) Oxygenisotope record (core HM79-6/4) measuredon N. pachyderma(s.) from the continentalslope of Norway [from Karpuzand Jansen,1992]. (e) Oxy- genisotope record (core Troll 3.1) measuredon benthicforaminifera Nonion labradoricum before and Cassidulina laevigataafter 10,500 •nC years B.P. from the continental shelf of Norway [from Lehman etal., 1991]. (f) Recon- structedsalinity variations (core SU81-18) off the coastof Portugal[from Duplessy et al., 1993].

Gulf of Mexico Record Loop Currentof the Gulf Streamwith a flux of-14 Sv [Schmitz, 1995], meltwaterfluxes of 0.1-0.23 Sv, with peak dischargesas high1 Sv,were required in orderto producethe observed 2.4 ø/oo Freshwater Fluxes to the Gulf of Mexico and Oxygen Isotope isotopicanomaly (compare with Figure 3b). (In makingsimilar Records arguments,Emiliani [1976] initially suggesteda freshwaterflux as high as 2.4 Sv.) The only clear signal of significantmeltwater entering the In respondingto discussionof his interpretationof large North Atlantic Oceanat the time of mwp-IA is thusfrom oxygen "floods"down the MississippiRiver, Emiliani [1976] suggested isotoperecords from the Gulf of Mexico,the BermudaRise, and that those who commented[Andrews, 1976; Farrand and Even- the Blake OuterRidge (Figure3b) [Leventeret al., 1982;Haskell son, 1976; Wrightand Stein, 1976] "revisittheir tunnelvalleys et al., 1991;Keigwin et al., 1991]. In first examiningan oxygen andspillways to look for matchingevidence" (p. 1271) from the isotoperecord from the Gulf of Mexico similarto the one dis- southernmargin of the LaurentideIce Sheetfor thesehigh fluxes. cussedhere (Figure 3b),Emiliani et al. [1978] [seealso Emiliani, We have revisitedthe glacial geologicalrecord, however, and 1976] arguedthat becausethe Gulf of Mexico is flushedby the find no evidenceto supporta sourcefrom the southernLauren- 570 CLARK ET AL.: ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE

Two developmentssince Emiliani's [1976] work, however, 16 0.0 suggesta solution to thisissue. The ill:st is theBarbados record itself (Figure 1), which now identifiesboth the magnitudeand 14 0.1 timing of sealevel rise within the contextof the Gulf of Mexico record. Althoughthe total oxygenisotope anomaly in coreEN32 PC-6from 14,000 14C to 12,000•4C years B.P. is on the order of 3%o(Figure 3b) [Leventeret al., 1982;Keigwin et al., 1991],the anomaly that occurs in direct associationwith mwp-IA is 10 0.2 -•l.5ø/oo. The important point here is thatthere are two isotopic excursionsof similarmagnitude that occurbefore mwp-IA when ice sheetmeltwater entered the global oceanat a lower rate' one at about13,000 •4C years B.P. (1.5%o) and one starting about 8 0.3 13,800•4C years B.P. (2%o) [cf. Keigwin etal., 1991]. The pres- 6 ence of these excursions at times when rates of sea level rise were low suggeststhat isotopicevents in the Gulf of Mexico were re- 4 cordedin the absenceof extremelyhigh freshwaterfluxes down the MississippiRiver. 0 50 100 150 The seconddevelopment is basedon observationsof the en- trainmentof MississippiRiver floodwatersinto the Gulf Stream Time (yr) duringthe 1993flood [Ortneret al., 1995]. Meanpeak discharge Figure6. Modelingresults of NorthAtlantic Deep Water pro- from the MississippiRiver duringAugust and September1993 duction,measured by the maximumof the meridionaloverturn- was-•0.025 Sv, or a similar flux as we estimatel¾om our ice sheet ing in theNorth Atlantic, as a functionof thetime of integration reconstructions[see also Teller, 1990]. Off the FloridaKeys and for differentmagnitude of freshwaterinput, under flux boundary Miami, Florida,the surfacesalinity was depressed by -•2%oin conditions(from H. Wang et al., submittedmanuscript, 1995). September,and remained depressed by 1%o6 weeksafter the The magnitudeof input of freshwateris labeledon eachcurve in flood. Depressedsalinity was observedas far northas Cape Sverdrupunits. Lookout,North Carolina. Ortneret al. [1995] concludedthat the followingconditions favored entrainment of the floodwaters into tideIce Sheet[e.g., Clayton and Moran, 1982;Mickelson et al., the Loop Current:(1) northwardposition of the Loop Current; 1983;Dyke and Prest,1987]. In addition,two independentnu- (2) highlystratified shelf waters; and (3) westerlywinds. mericalice sheet reconstructions [Hughes, 1987; J.M. Licciardiet A 1%oreduction in salinityis equivalent to about3% freshwa- al., submittedmanuscript, 1996] suggest that the actual meltwater ter. For this salinityreduction, therefore, Mississippi River wa- flux flowingdown the MississippiRiver was only a fractionof terswould result in a reductionof- 1ø/oo in15 •80 values of Gulf of thatrequired to explainthe sealevel rise during mwp-IA [Teller, Mexicosurface waters, if thoseflood waters had 5•80 values 1990](Figure 4). We arethus faced with an apparent conundrum equivalentto LaurentideIce Sheetmeltwater (about -30%o [Mix, betweeninterpreting the marineand terrestrialrecords with re- 1987]).The 2%o salinity reduction that occurred during the 1993 spect to the magnitudeof the freshwaterfluxes that drained peakflood could thus translate into a 2%0decrease in 5•80 val- throughthe Mississippi River. ues. The 1993 flood datathus demonstrate that an oceanographic

a b c 11

14 / 15 .... , .... , .... , .... , 'I '•-ß Iß I i 20 30 40 100 8o 6o 4o -42 -38 -34 (W/m2) (m) 5180 (ø/oo)

Figure7. Comparisonof (a) Julyinsolation anomaly at 45øNbetween 11,000 and 15,000 years ago with (b) Bar- badossea level record[from Bard et al., 1990a,1993] and (c) the GRIP oxygenisotope record (20-year means) from the Summitice coreon the GreenlandIce Sheet[from Johnsen et al., 1992]. Solid horizontalline correspondsto beginningof fastestrate of sealevel rise during mwp-IA (see Table 1 for correspondingages). CLARK ET AL.' ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE 571

a b c d lO

14 110 90 70 -1.0 -0.6 -0.2 0.2 100 50 0 -2 2 6 10 (m) fi13C(%0) (%N. pachyderma) (øC)

Figure 8. The Barbadossea level record compared with recordsthat reflect changes in the strengthof NorthAt- lanticDeep Water Production between 10,000 and 14,000 •4C years B.P. Solid horizontal line corresponds tobe- ginningof fastestrate of sealevel rise during mwp-IA (see Table I for correspondingages). (a) Barbadossea level recorddated by radiocarbon[Fairbanks, 1989]. (b) Carbonisotope record (core RC11-83) measuredon Planulina wuellerstorfifrom the SouthernOcean [from Charles and Fairbanks,1992]. (c) PercentN. pachyderma(s.) (core Troll 3.1) from continentalshelf of Norway [from Lehmanand Keigwin, 1992]. (d) ReconstructedAugust (solid line) andFebruary (dashed line) seasurface temperatures (core HM79-6/4) from continentalslope of Norway [from Karpuzand Jansen,1992].

circulation can exist which can advect a rather modest tYeshwater signalfar seawardand result in a 15•gOsignal such as that recorded in the Gulf of Mexico Pleistocene record. 6,o (%0) Deglaciation,Meltwater Routing and the Oxygen Isotope 0 -1 -2 -3 Record 10 Here we follow Kennettand Shackleton[1975], Farrand and Evenson[1976], and Leventeret al. [1982] in interpretingthe Gulf of Mexico oxygenisotope record as simplyidentifying the interplay between(1) meltwaterdischarge from the southern 12 marginof the ice sheetassociated with increasinginsolation and (2) the routingof that meltwateralong the ice sheetmargin during the lastdeglaciation. Kennett and Shackleton[1975] had no directradiocarbon age controlfor the Gulf of Mexico coresthey 14 examined,and the chronologyof the core studiedby Emiliani et al. [1975, 1978] was basedon only few conventionalbulk radio- carbonages. The chronologyof the Gulf of Mexico coreexam- ined here (EN32 PC-6) (Figure 3b) is basedon severalaccelera- tormass spectrometry (AMS) 14C ages [Broecker et al., 1988, 1989] whichallow a moreprecise comparison with the historyof meltwaterrouting along the southernice sheetmargin than dis- cussedby Leventeret al. [1982]. Broeckeret al. [1989] madea similar interpretationof the youngerpart of the Gulf of Mexico 18 isotoperecord (<11,000 •4C years B.P.) as reflecting meltwater routing along the southernice sheet margin. Keigwin et al. [1991] alsointerpreted variations in surfacewater salinities in the North Atlantic as recordinggeographic variability in meltwater 2o input. 0 10 20 30 40 We convertedthe radiocarbonchronology of this recordto calendaryears using the relation defined by Bardet al. [ 1993]to (w/m2) compareit with insolationanomalies for Julyat 45øN,a latitude Figure 9. Oxygenisotope record from the Gulf of Mexico (core whichspans the southernmargin of the LaurentideIce Sheet. In EN32 PC-6), with radiocarbonage chronologyconverted to cal- general, this comparisonshows that the rise in insolationis endaryears using relation of Bard et al. [1993], comparedwith closelymatched by an increasein meltingalong the southern July insolationanomaly at 45øNbetween 10,000 and 20,000 margin(as suggestedby the oxygenisotope record) (Figure 9) years ago. 572 CLARK ET AL.: ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE

lobe dvance

20 i i i i i i i i i i i i i 750 (km) 0

Figure 10. Oxygenisotope record from the Gulf of Mexico(core EN32 PC-6) (dasedline) [fromKeigwin et al., 1991]compared with recordof advanceand retreat of theLake Michigan Lobe of theLaurentide Ice Sheet[from Hanseland Johnson, 1992] between 9,000 and 20,000 •4C years B.P. Lowercaselabels (a- g, a•- ii) aredis- cussed in text.

untilabout 14,000 years ago (12,000 14C years B.P.), when 5180 times when meltwater from some or all of the Great Lakes basins valuessuddenly become more positive. If meltwaterrouting is was drainingto the MississippiRiver insteadof to easternoutlets alsoimportant in explainingthe Gulf of Mexicorecord [Kennett [Hanseland Mickelson,1988]. and Shackleton,1975; Farrand and Evenson,1976; Leventeret Theice margin reached its maximum extent-•19,500 incyears al., 1982; Broeckeret al. 1989; Keigwin et al., 1991], then sec- B.P. and then begana gradualretreat with severalminor oscilla- ondaryfluctuations seen in that recordas well as the abruptshift tionsuntil rapid retreat at-15,500 InC years B.P. Thiswas fol- at 14,000 yearsshould correspond closely to timeswhen eastern lowedby a significantreadvance between 15,200 and 14,100 inc outletsopened or were closed,resulting in changesin the flux of yearsB.P. (Figure 10), which is recordedelsewhere along the meltwaterflowing through the MississippiRiver. southernice sheetmargin (Port Bruce stadial; Dreimanis [ 1977]). Meltwaterflowing to the MississippiRiver was derivedfrom This readvance(labeled a on Figure10) may be responsiblefor oneof severalpossible hydrological basins [Teller, 1990, 1995]. the increasein blSOvalues seen in the Gulf of Mexicobetween The westernmostbasin, correspondingto the area eventually -•15,000and 13,700 inc years B.P. (labeled containingglacial Lake Agassiz, drained to the MississippiRiver Followingthis readvance, the icemargin began a majorretreat unlessan easternoutlet openedto the Lake Superiorbasin (labeledb) during the Mackinaw interstadial[Dreirnanis and [Teller, 1985]. The next basinto the east,corresponding to the Karrow, 1972]. Meltwater was initially directedfrom all basins basinsof Lakes Superiorand Michigan,drained to the Missis- intothe Mississippi River, but by about 13,300 inc years B.P. the sippiRiver unlessthe ice marginretreated north, thus opening southernice marginretreated far enoughto the north(labeled c) easternoutlets near or throughthe Straitsof Mackinac[Hansel to opendrainage routes to the eastfor all meltwaterin the Great and Mickelson,1988]. The next basin,corresponding to the ba- Lakesbasins [Calkin and Feenstra,1985; Hansel and Mickelson, sins of Lakes Huron, Erie, and Ontario, drainedto the west into 1988; Teller, 1995]. This retreat history seemsto be well re- the Lake Michiganbasin and then into the MississippiRiver un- cordedin theGulf of Mexico,where bl*O values initially de- less the ice margin retreatedenough to allow waters to drain creased(labeled bi), reflecting ice sheet melting, and then eastward acrossNew York to the Mohawk River or, with further abruptlyincreased (labeled cl) at the sametime as the openingof ice marginretreat, to the St. LawrenceRiver [Clark and Karrow, easternoutlets in the GreatLakes region. 1984; Calkin and Feenstra,1985]. The southernice marginin muchof the Great Lakesregion In orderto evaluatethe hypothesisthat fluctuationsin the Gulf then beganto readvanceseveral hundred kilometers during the of Mexico oxygen isotoperecord reflect changesin meltwater Port Huron stadial, reachingits maximumposition by about routing,we comparethis recordwith the recordof ice-margin 13,000inc years B.P. (labeled d) [Dreimanisand Karrow, 1972; fluctuationsreconstructed for the Lake Michigan Lobe of the Dre#nanis, 1977]. This readvanceclosed all eastern outlets LaurentideIce Sheet(Figure 10) [Hanseland Johnson,1992]. again,forcing drainage back to the MississippiRiver [Calkinand The Lake MichiganLobe record is importantbecause it identifies Feenstra, 1985; Eschmanand Karrow, 1985; Hansel and Mickel- CLARK ET AL.: ORIG1NOF THE FIRST GLOBAL MELTWATER PULSE 573 son, 1988]. This reroutingof meltwateris recordedby decreas- LaurentideIce Sheet seemsan unlikely sourceof mwp-IA for ing15•80 values in the Gulf of Mexico(labeled d•). Meltwaterin severalreasons. First, on the basisof our understandingof ice the Huron, Erie, and Ontario basinscontinued to drain west to the volume contained within the sector of the Laurentide that drained Michiganbasin until about 12,400 •4C years B.P., when ice re- to the MississippiRiver, the likely flux of meltwaterduring the treat againbegan to uncovereastern outlets [Calkin and Feenstra, time of mwp-IA was only a fraction (<10%) of that neededto 1985]. Meltwater in the Lake Michigan basin,however, contin- raise sea level at the rate suggestedby mwp-IA. Geologicalrec- ued to drain to the MississippiRiver, probablyat an increasing ords from the area of the southern Latirentide Ice Sheet indicate rate as the margin was retreatingand July insolationwas within only slightwithdrawal of the ice margin [Dykeand Prest, 1987], 80%of itsmaximum level (Figure 9). Thus•5•80 values contin- unlike the amountrequired if this ice sheetsector were responsi- ued to decreasedespite the loss of meltwaterfrom the eastern ble. Even if enoughice were presentwithin that sectorof the ice Great Lakes basins. sheet,melting it at the necessaryrate is physicallyimplausible. About12,000 •4C years B.P., or thebeginning of the Two In addition, such an acceleratedmelting rate required for the Creeks interstade [Kaiser, 1994], retreat of the ice margin Mississippiice sheetsector raises the questionof why only that (labelede) uncoveredan outletjust to the southof the Straitsof sectorwould be so sensitiveto ablation,whereas adjacent sectors Mackinac, allowing meltwater in the Lake Michigan basin to were relatively insensitive,as suggestedby the absenceof sig- drainto the east[Hansel and Mickelson,1988]. This diversionis nificant meltwaterdrainage in recordsfrom the Gulf of St. Law- contemporaneouswith and thus may explain the abruptincrease rence[Keigwin and Jones,1995] and HudsonStrait [Andrewset in8•80 values in the Gulf of Mexicofollowing the point labeled al., 1994; Hillaire-Marcel et al., 1994] (Figure 3). Furthermore, eI (Figure 10). The last ice readvancerecorded in the Lake oceanmodeling results (Figure 6) predict a much strongercli- Michiganbasin (labeled f) beganafter 11,750 •4C years B.P. matic responseto a large freshwaterflux out of the Mississippi [Kaiser, 1994], forcing meltwater in this basin to again drain River than is recordedin the North Atlanticregion (Figures 7 and brieflyto theMississippi River, possibly as late as 11,100•4C 8). Finally,there is no obviousforcing mechanism that would re- yearsB.P. [Hansel and Mickelson,1988]. This may explainthe sult in sucha brief but intenseperiod of ablationof the Missis- briefdecrease in 8•O valuesat the same time (labeled f•). Sub- sippi ice sheet sector. Insolationduring this time increased sequentretreat of the ice margin (labeledg) resultedin all melt- gradually(Figure 9), and warming associatedwith intensified waterfrom the Great Lakesbasins to drainto the east,resulting in formationof NADW in the Norwegianand GreenlandSeas is ex- increasing•80 values(labeled g•). Broeckeret al. [1989]at- pectedto stronglyaffect only the Eurasianice sheets[Rind et al., tributedthe next significant increase inGulf of Mexico•5•80 val- 1986; Manabe and Stouffer,1988], althoughthere is no evidence ues (starting at point h•) to diversionof glacial Lake Agassiz for acceleratedmelting of the Eurasianice sheetsat the time of waters from the MississippiRiver to the St. LawrenceRiver, mwp-IA (Figure 5). followed by reneweddrainage to the MississippiRiver and de- The only other ice sheetcapable of deliveringenough volume creasing•O values(labeled i•) whenice readvanced into the to explainthe sealevel risethat definesmwp-IA thereforeis the LakeSuperior basin at about 10,000 •4C years B.P. Antarctic Ice Sheet. However, there are currently no high- resolutionrecords of deglaciationof the Antarctic Ice Sheet to evaluatewhether it was responsiblefor mwp-IA. Moreover,the Discussion amountof extra ice containedin this ice sheetat the last glacial maximum,as well as the timing of its releaseto the oceanduring the last deglaciation,remains uncertain. Estimatesof the contri- Oceanic and terrestrial records show no evidence for acceler- butionto sealevel loweringby the AntarcticIce Sheetduring the ated deglaciationof any of the northernhemisphere ice sheetsat last glacial maximum range from as little as 0.5 m [Colhounet the time of mwp-IA. Our evaluationof the historyof deglacia- al., 1992] to 37 m [Nakada and Lainbeck,1988]. The small ice- tion of globalice sheetsduring mwp-IA is basedon the assump- volume estimatesmade by Colhoun et al. [1992] are basedon tion that this eventrepresents the rapidcollapse of someportion Holoceneraised beaches found at relativelylow altitudes. They of an ice sheet,resulting in the deglaciationof the equivalentof calculatedan ice thicknessbased on the assumptionthat the two to three GreenlandIce Sheetsin <500 years. We thus as- height of Holocene beachesis a function only of former ice sumethat the rapid rate of sealevel rise duringmwp-lA records thickness.Because the height of a raisedbeach is also a function either acceleratedice sheetmelting in responseto an abrupt of the rate of ice thinningprior to deglaciation[Andrews, 1970], warming event or collapseof a marine-basedportion of an ice however,their estimatesof ice thicknessare only minima;a sub- sheetby calving. stantialamount of uplift may have occurredbeneath the thinning Because the North Atlantic and Arctic Oceans received all ice load during deglaciationbefore the site becameice free, and meltwaterand icebergsfrom the former northernhemisphere ice the maximumice load may have beensignificantly greater. sheets,records from theseoceans should identify rapid ice sheet Much of the geologicaland geophysicalevidence appears to collapseat thetime of mwp-IAas an excursion to lighter •5•O support15-20 m of additionalsea level within the AntarcticIce values and/or increased IRD fluxes. In all of the records of de- Sheet at the last glacial maximum. Hugheset al. [1981] sug- glaciation of the northern hemisphereice sheetsthat we re- gestedthat the AntarcticIce Sheetmay have containedas much viewed, only those from the Gulf of Mexico, the Blake Outer as 24 m of additionalsea level during the last glacial maximum, Ridge, and the BermudaRise recordingmeltwater drainage from with -•16 m presentwithin an expandedWest Antarctic Ice Sheet, the southernLaurentide Ice Sheet down the MississippiRiver althoughthis latter estimateis likely too high [Denton et al., showany indication of a significantexcursion to lighter •5•O 1989]. Huybrechts[1990a, b] modeledan additional12-16 m sea valuesat the time of mwp-IA (Figure3). However,the southern level equivalentin a full-glacialAntarctic Ice Sheet. Tushingham 574 CLARK ET AL.' ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE and Peltlet [ 1991] constrained26 m of additionalsea level, and began14,000-14,200 years ago, or immediatelyafter initiationof Peltier [1994] constrained21.8 m. We notethat if the Antarctic mwp-IA (Figure 11). If this eventis associatedwith mwp-IA, it Ice Sheetcontained-22 m of additionalsea level [Peltier, 1994], may indicatecooling of the Antarcticcontinent as a resultof in- of which-19 m was releasedduring mwp-lA, then only 3 m of creasedice cover over the surroundingocean following ice sheet additionalice would remain from-12,000 •nC years B.P. until collapse. Alternatively,it may representan increasein isotopic sitesdescribed by Colhounet al. [1992] becameice free <7,500 fractionationduring moisturetransport to the ice sheetas the •nCyears B.P. This would allow a significantamount of upliftto sourcewaters were displacednorthward by increasedice cover occurbeneath the dynamicallythinned ice coverbefore their sites overthe surroundingocean. became ice free. Peltier [ 1988] and Tushinghamand Peltier [ 1991] arguedthat Because much of the extra ice in both the East Antarctic and deglaciationof the Antarctic Ice Sheet began well after initial West AntarcticIce Sheetswas groundedbelow sea level during deglaciationof the northernhemisphere ice sheets,and Peltier the last glaciation,these ice sheetswould have beensusceptible [1994] constrainedthe AntarcticIce Sheetas the sourceof the to rapid collapseas sea level rose following the last glacial secondmeltwater pulse (mwp-IB) (Figure 1). Geologicalrecords maximum [Hollin, 1962; Stuiver et al., 1981' Denton and of Antarcticdeglaciation reviewed above, however, may be more Hughes,1983]. The timing of deglaciationremains poorly con- consistentwith an earlierdeglaciation, similar in timingto mwp- strained,however, although many existing data can be interpreted IA, althoughbetter resolution is clearly requiredto supportthis as consistentwith, and perhapsproviding preliminary evidence hypothesis.Nevertheless, the sensitivityof the Earthmodels de- for, our hypothesisthat the sourceof mwp-IA was from the Ant- scribedby Peltier [1988, 1994] to changesin timing of Antarctic arcticIce Sheet. Denton et al. [1989] datedinitial deglaciationof Ice Sheet deglaciationand other model parameters(e.g., litho- the Ross Sea ice drainagesystem of the AntarcticIce Sheetas sphere thickness,mantle viscosity) [cf. Davis and Mitrovica, beginningaslate as 12,500-13,000 •nCyears B.P., while Licht et 1996] and assumptionsof isostaticequilibrium [cf. Mitrovica and al. [1996] dateddeglaciation in the RossSea area as beginning Davis, 1995] in explainingfar-field sealevel recordssuggest that byat least I 1,500 •nC years B.P. an evaluationof the AntarcticIce Sheetas the sourceof mwp-IA In deep-seacores from the Indian sectorof the Southern may be warranted. Ocean, Labeyrie et al. [1986] found isotopicanomalies inter- Our conclusionthat the formernorthern hemisphere ice sheets pretedas meltwaterevents occurring between 17,000 and 35,000 werenot responsiblefor mwp-IA hastwo importantimplications. yearsago. In thesame cores, but now with AMS •nC age control, First, becausePeltier [ 1994] constrainedthe LaurentideIce Sheet Labracherieet al. [1989] and Bard et al. [1990b] foundthat sea to be the sourceof mwp-lA in his ICE-4G model,we suggestthat his reconstructions of the Laurentide Ice Sheet surface elevations surfacetemperatures were nearlythe sameas modernvalues by 13,000•nC years B.P., but were then replaced by a coolperiod during the last deglaciationmay be too low for time periods between12,000 and l 1,000InC years B.P., or roughly contenq•o- younger than 15,000 yearsago. raneouswith mwp-I^ More importantto our hypothesis,Lab- Second,our interpretation thatthe •5•O record of meltwater tacherieet al. [ 1989] andBard et al. [1990b] identifieda signifi- dischargefrom the MississippiRiver to the Gulf of Mexicoiden- cantlight peak in •5•80values from one core, having an age cen- tifies the interplaybetween insolation forcing of meltingrates teredat I 1,920+ 200•4C years B.P. (800-year reservoir age cor- and the routinghistory of meltwateralong the southernice sheet rection). Becausethis peak doesnot coincidewith a changein SST, Bard et al. [1990b, p. 413] concludedthat this isotopic event could be "a responseto a pulse-likeinjection of [a] large 11 volume of ice or meltwater." Shemeshet al. [1995]examined 15•80 records of biogenicsilica from the Atlantic sector of the Southern Ocean. Their cores 12 lacked radiocarbonage control, althoughthe level of the last glacial maximum is identified from biostratigraphy.Shemesh et al. [1995] identifieda post-glacial-maximumspike-shaped de- creasein 15t80values in coresassociated with the Weddell gyre. The amplitudeof this "meltwater"spike decreases in a northeast- ACR em direction,suggesting increasing dilution with distancefrom 14 Antarctica toward the South Atlantic. On the basisof preliminary age modelsfor thesecores, Shemesh et al. [1995] concluded 15 that the meltwaterspike occurredsometime between 15,000 to 10,000tnC years B.P. and that it maybe correlative with mwp- 100 8O 6O 4O -32.5 -30.5 IA. (m) Finally,using the 15t80 of atmospheric02trapped in icecores, •180 (ø/oo) Sowersand Bender[ 1995] placedthe chronologyof the Byrd ice Figure l l. The Barbadossea level record[Bard et al., 1990a, core, West Antarctica[Johnsen et al., 1972], on the sametime- 1993]compared with the Byrd ice core 15180 record (15t80 values scale as the GreenlandIce Sheet Project (GISP2) ice core from from Johnsenet al. [1972]; age modelof ice core from Sowers Greenland.A small decrease in15•O values (-2% o) observed in and Bender[1995]). Solidhorizontal line correspondsto begin- the Byrd ice core during the last deglaciation,also seenas an ningof fastestrate of sealevel rise during mwp-IA (seeTable 1 oscillationin deuteriumprofiles in ice coresfrom EastAntarctica for correspondingages). ACR denotesAntarctic cold reversal and referredto as the Antarcticcold reversal[•Iouzel et al., 1995], [Jouzelet al., 1995]. CLARK ET AL.: ORIGIN OF THE FIRST GLOBALMELTWATER PULSE 575 margin has important implicationsregarding the sensitivityof (i.e.,Hudson River, St. LawrenceRiver, and Hudson Strait) seem NADW formationto freshwaterinjections. Haskell et al. [1991] to haveaffected NADW formation.These discharge events may and Keigwin et al. [ 1991] identifiedfour periodsof reducedflow in turnhave been controlled by surgingbehavior of the Lauren- of NADW duringthe lastdeglaciation (14,500 to 10,000 tideIce Sheet,both in termsof controllingthe flux of icebergs yearsB.P.) superimposedon a generaltrend of increasingflow, [MacAyeal,1993] and the routingoi' meltwateralong the ice and Keigwin et al. [ 1991] associatedeach period with a periodof sheetmargin [Clark, 1994],thus reinforcing an ice sheetforcing freshwaterinjection into the NorthAtlantic. l'he earliestperiod mechanismthat resulted in high-frequencyclimate change. correspondsto the low 15•80 values recorded in theNordic seas (Figure5) and Heinrichevent 1 [Keigwinand Lehman,1994], Acknowledgments. We thank E. Bard, G. Bond, S. Lehman, whereasthe most recentperiod correspondsto the Younger A. Mix, and D. Peltierfor discussions,G. Bondfor providing Dryas,which may haveits originin the diversionof drainageof dataon coreVM23-81 andT. Sowersfor providingthe Byrd ice glacialLake Agassizfrom the MississippiRiver to the St. Law- core data, and J. Andrews, C. Charles, and K. Miller for reviews renceRiver [Broeckeret al., 1989]and the releaseof icebergsto of the manuscript.This work was supportedby the NationalSci- theNorth Atlantic during Heinrich event 0 [Miller andKaufman, ence Foundation. 1990;Andrews et al., 1995]. We interpretthe causeof the two interveningperiods of reduced NADW formation to diversion of meltwater drainage References alongthe southernLaurentide Ice Sheetmargin to easternoutlets, unlike Keigwin et al. [1991], who interpretedthem as resulting from periods of increaseddischarge through the Mississippi Andrews, J.T., A geomorphologicalstudy of postglacialuplit•, with River. These two periods of reducedNADW formationjust particularreference to Arctic Canada,Inst. Br. Geogr.Spec. Publ., 2, postdateprominent peaks seen in the oxygen isotope records 156 pp., 1970. Andrews,J.T., The WisconsinLaurentide Ice Sheet:Dispersal centers, fromthe Gulf of Mexico(the first peakoccurring between b• and problemsof ratesof retreat,and climaticimplications, Arct. Alp. Res., c• and the secondpeak, labeledf• on Figure 10, equivalentto 5, 185-199, 1973. eventsb and c, respectively,given by Keigwin et al. [1991]), Andrews,J.T., Glacialsurges and flood legends, Science, 193, 1270, 1976. BermudaRise (Figure 3b), andBlake OuterRidge [Haskellet al., Andrews, J.T., and K. Tedesco, Detrital carbonate-rich sediments, 1991; Keigwin et al. 1991]. As was discussed,the expressionof northwesternLabrador Sea: Implicationsfor ice sheetdynamics and icebergraffing (Heinrich) events in the North Atlantic,Geology, 20, these15•80 peaks represents meltwater diversion from the Missis- 1087-1090, 1992. sippiRiver to easternoutlets. The first diversionoccurred during Andrews,J.T., H. Erlenkeuser,K. Tedesco,A.E. Aksu,and A.J.T. Jull,Late theMackinaw interstadial such that by-13,300 •4C years B.P., Quaternary(stage 2 and 3) meltwaterand Heinrich events, northwest all meltwater drained to the east and into the North Atlantic LabradorSea, Quat. Res., 41, 26-34, 1994. Andrews,J.T., A.E. Jennings,M. Kerwin, W. Manley, G.H. Miller, G. throughthe HudsonRiver. At the sametime, therewas a signifi- Bond, and B. MacLean, A Heinrich-like event, H-0 (DC-0): cant increasein the dischargeof icebergsthrough the Gulf of St. Source(s)for detrital carbonatein the North Atlantic during the Lawrence [Bond and Lotti, 1995; Keigwin and Jones, 1995], YoungerDryas chronozone,Paleoceanography, 10, 943-952, 1995. which would have augmentedthe freshwaterflux to the North Bard,E., B. Hamelin,R.G. Fairbanks, and A. Zindler,Calibration ofthe Atlantic. Proxy recordsindicate reducedformation of NADW timescaleover the past 30,000 yearsusing mass spectrometric U-Th agesfrom Barbadoscorals, Nature, 345, 405-410, 1990a. during this time [Haskell et al., 1991; Keigwin et al., 1991]. Bard,E., L.D. Labeyrie,J.-J. Pichon, M. Labracherie,M. Arnold,J. Duprat, Following13,000 •4C years B.P., however, the southern margin J. Moyes,and J.-C. Duplessy,The lastdeglaciation in the southernand of the Laurentide Ice Sheet readvanced and diverted meltwater northernhemispheres: A comparisonbased on oxygenisotope, sea backto the MississippiRiver, icebergdischarge from the Gulf of surfacetemperature estimates, andaccelerator •4Cdating from deep-sea sediments,in GeologicalHistory of the Polar Oceans:Arctic Versus St. Lawrence decreased,and the rate of formation of NADW in- Antarctic,edited by U. Bleil andJ. Thiede,pp. 405-415, KluwerAcad., creased. Norwell, Mass., 1990b. The next youngerperiod of reducedNADW formationwas Bard,E., M. Arnold,R.G. Fairbanks, and B. Hamelin,23øTh-234U and14C brief,occurring-•12,000-12,200 •4Cyears B.P. Thetiming of agesobtained by massspectrometry on corals,Radiocarbon, 35, 191- 199, 1993. this eventseems to correspondclosely to renewedmeltwater dis- Bard,E., B. Hamelin,M. Arnold,L. Montaggioni,G. Cabioch,A. Laurenti, chargeto the east, culminatingwith the Two Creeksinterstade G. Faure,and F. Rougerie,Sea level duringthe last 14,000 years when meltwater from the Michigan, Huron, Erie, and Ontario reconstructedby datingcorals from Tahiti (TIMS andAMS) (abstract), basinswas drainingto the North Atlantic initially throughthe 5thInt. Conf Paleoceanography,Program and Abstracts, 29, 1995. Birchfield,G.E., H, Wang, and J.J. Rich, Century-millev.nium internal HudsonRiver and then4hroughthe St. LawrenceRiver. The ef- climate oscillationsin an ocean-atmosphere-continentalice sheet fectof thiseasterly drainage on the observedreduction of NADW model,J. Geophys.Res., 99, 12,459-12,470,1994. formation,if indeedinvolved, was short-lived, perhaps as a result Blanchon,P., and J. Shaw,Reef drowningduring the last deglaciation: of the strongermode of formationof NADW at the time, com- Evidencefor catastrophicsea-level rise and ice sheet collapse, Geology, paredwith earlierevents, as suggested by Cd/Ca data[Keigwin et 23, 4-8, 1995. Bond, G., and R. Lotti, Icebergdischarges into the North Atlantic on al., 1991]. millennialtime-scales during the lastglaciation, Science, 267, 1005- This discussionreiterates arguments by Broeckeret al. [1989] 1010, 1995. andKeigwin et al. [ 1991] thatdischarge of icebergsand rerouting Bond,G., et al., Evidencefor massivedischarges of icebergsinto the of meltwateralong the southernmargin of the LaurentideIce North Atlantic oceanduring the last glacial period,Nature, 360, 245-249, 1992. Sheetmay have been importantin affectingNADW formation. Bond,G., W.S. Broecker,S. Johnsen,J. McManus,L. Labeyrie,J. Jouzel, In the above scenario,however, only significantfreshwater dis- and G. Bonani, Correlationsbetween climate recordsfrom North chargeto the North Atlanticfrom routesnorth of the Mississippi Atlanticsediments and Greenland ice, Nature,365, 143-147, 1993. 576 CLARK ET AL.: ORIGIN OF THE FIRST GLOBAL MELTWATER PULSE

Boyle,E.A., andL.D. Keigwin,North Atlantic thermohaline circulation Emiliani, C., C. Rooth,and J.J. Stipp, The late Wisconsinflood into the duringthe past20,000 years linked to high-latitudesurthce tempera- Gulf of Mexico, Earth Planet. Sci. Lett., 41, 159-162, 1978. ture, Nature, 330, 35-40, 1987. Eschman,D.F., andP.F. Karrow,Huron Basin glacial lakes: A review,in Broecker,W.S., D.M. Peteet,and D. Rind, Does the ocean-atmosphere QuaternaryEvolution of the Great Lakes,edited by P.R. Karrowand systemhave more than one stablemode of operation?,Nature, 315, P.E. Calkin,Geol. Assoc. Can. Spec.Pap., 30, 79-94, 1985. 21-26, 1985. Fairbanks,R.G., A 17,000-yearglacio-eustatic sea level record:Influ- Broeckcr,W.S., M. Andree, W. Wolfli, H. Oeschger,G. Bonani,J. Ken- enceof glacialmelting rates on the YoungerDryas event and deep- nett, and D. Peteet,The chronologyof the last deglaciation:Impli- oceancirculation, Nature, 342. 637-642, 1989. cationsto the causeof the YoungerDryas event, Paleoceanography, Fairbanks,R.G., C.D. Charles,and J.D. Wright,Origin of globalmeltwa- 3, 1-19, 1988. ter pulses,in RadiocarbonAfter Four Decades,edited by R.E. Tay- Broecker,W.S., J.P. Kennett, B.P. Flower, J.T. Teller, S. Trumbore,G. lor, A. Long, and R.S. Kra, pp. 473-500, Springer-Verlag,New Bonani, and W. Wolfli, Routing of meltwaterfrom the Laurentide York, 1992. ice sheetduring theYounger Dryas cold episode,Nature, 341, 318- Farrand,W.R., and E.B. Evenson,Glacial surgesand flood legends,Sci- 321, 1989. ence, 193, 1269-1270, 1976. Bryan, F., High-latitudesalinity effectsand interhemisphericthermo- Hansel,A.K., and W.H. Johnson,Fluctuations of the Lake Michiganlobe halinc circulations,Nature, 323, 301-304, 1986. duringthe late Wisconsinsubepisode, Sver. Geol. Unders.,81, 133- Calkin,P.E., andB.H. Feenstra,Evolution of the Erie-BasinGreat Lakes, in 144, 1992. QuaternaryEvolution of the Great Lakes,edited by P.R. Karrowand Hansel,A.K., andD.M. Mickelson,A reevaluationof thetiming and causes P.E. Calkin, Geol.Assoc. Can. Spec.Pap., 30, 149-170,1985. of high lake phasesin the Lake Michiganbasin, Quat. Res., 29, 113- Charles,C.D., and R.G. Fairbanks,Evidence from SouthernOcean sedi- 128, 1988. mentsfor the effect of North Atlantic deep-waterflux on climate, Haskell, B.J., T.C. Johnson,and W.J. Showers,Fluctuations in deep Nature, 355, 416-419, 1992. westernNorth Atlanticcirculation on the Blake OuterRidge during Clark, P.U., Unstablebehavior of the LaurentideIce Sheetover deform- the last deglaciation,Paleoceanography, 6, 21-31, 1991. ing sedimentand its implicationsfor climatechange, Quat. Res., 41, Hillaire-Marcel,C., A. de Vernal, G. Bilodeau,and G. Wu, Isotope 19-25, 1994. stratigraphy,sedimentation rates, deep circulation,and carbonate Clark, P.U., and P.F. Karrow, Late Pleistocenewater bodies in the St. eventsin the LabradorSea during the last-200 ka, Can.d. EarthSci., LawrenceLowland, New York, andregional correlations, Geol. Soc. 31, 63-89, 1994. Am. Bull., 95, 805-813, 1984. Hollin, J.T., On the glacialhistory of Antarctica,d. Glaciol.,4, 173-195, Clark, P.U., J.M. Licciardi, D.R. MacAyeal, and J.W. Jenson,Numerical 1962. reconstructionof a soft-beddedLaurentide Ice Sheetduring the last Hughes,T.J., Ice dynamicsand deglaciationmodels when ice sheetscol- glacialmaximum, Geology, 24, in press,1996. lapsed,in TheGeology of NorthAmerica, vol. K-3, NorthAmerica and Clayton,L., andS.R. Moran,Chronology of late Wisconsinanglaciation AdjacentOceans During the LastDeglaciation, edited by W.F. Rud- in middleNorth America,Quat. Sci. Rev., 1, 55-82, 1982. dimanand H.E. WrightJr., pp. 183-220,Geol. Soc. of Am., Boulder, Colhoun,E.A., M.C.G. Mabin, D.A. Adamson,and R.M. Kirk, Antarctic Colo, 1987. ice volumeand contributionto sealevel fall at 20,000 yearsBP from Hughes,T.J., G.H. Denton,B.G. Andersen,D.H. Schilling,J.L. Fastook, raisedbeaches, Nature, 358, 316-319, 1992. andC.S. Lingle,The lastgreat ice sheets:A globalview, in TheLast Davis, J.L., and J.X. Mitrovica, Glacial isostaticadjustment and the GreatIce Sheets,edited by G.H. Dentonand T.J. Hughes, pp. 263-317, anomaloustide gaugerecord of easternNorth America,Nature, 379, Wiley-Interscience,New York, 1981. 331-333, 1996. Denton, G.H., and T.J. Hughes,Milankovitch theory of ice ages:Hy- Huybrechts,P., The AntarcticIce Sheetduring the lastglacial-interglacial pothesisof ice sheetlinkage between regional insolation and global cycle: A three-dimensionalexperiment, Ann. Glaciol., 14, 115-119, climate,Quat. Res.,20, 125-144, 1983. 1990a. Denton, G.H., J.G. Bockheim, S.C. Wilson, and M. Stuiver, Late Wis- Huybrechts,P., A 3-D modelfor theAntarctic ice sheet: A sensitivitystudy consinand early Holoceneglacial history,inner Rossembayment, on theglacial-interglacial contrast, Clim. Dyn., 5, 79-92, 1990b. Antarctica,Quat. Res.,31, 151-182, 1989. Jenson,J.W., P.U. Clark, D.R. MacAyeal,C.L. Ho, andJ.C. Vela, Numeri- Dreimanis,A., Late Wisconsinglacial retreat in the GreatLakes region, cal modelingof advectivetransport of saturateddeforming sediment North America, Ann. NYAcad. Sci., 288, 70-89, 1977. beneaththe LakeMichigan Lobe, Laurentide Ice Sheet,Geomorphol- Dreimanis,A., and P.F. Karrow, Glacial historyof the Great-Lakes-St. ogy, 14, 157-166, 1995. Lawrenceregion, the classificationof the Wisconsin(an)stage, and Jenson,J.W., D.R. MacAyeal,P.U. Clark,C.L. Ho, andJ.C. Vela, Numeri- its correlatives,Int. Geol. Congr.Rep. Sess. 24th, 12, 5-15, 1972. cal modelingof subglacialsediment deformation: Implications for the Duplessy,J.-C., L. Labeyrie,M. Arnold, M. Paterne,J. Duprat, and behaviorof the Lake MichiganLobe, LaurentideIce Sheet,,d. Geo- T.C.E. van Weering,Changes in surfacesalinity of the North Atlan- phys.Res., 101, 8717-8728,1996. tic Oceanduring the last deglaciation,Nature, 358, 485-488, 1992. Johnsen,S.J., W. Dansgaard,H.B. Clausen,and C.C. Langway,Oxygen Duplessy,J.-C., E. Bard, L. Labeyrie,J. Duprat,and J. Moyes,Oxygen isotopeprofiles through the Antarctic and Greenland ice sheets, Nature, isotoperecords and salinity changesin the northeasternAtlantic 235, 429-434, 1972. Oceanduring the last 18,000 years,Paleoceanography, 8, 341-350, Johnsen,S.J., H.B. Clausen,W. Dansgaard,K. Fuhrer,N. Gundestrup,C.U. 1993. Hammer,P. Iversen,J. Jouzel,B. Stauffer,and J.P. Steffensen, Irregular Dyke, A.S., and V.K. Prest,Late Wisconsinanand Holocenehistory of glacialinterstadials in a newGreenland ice core,Nature, 359, 311-313, the LaurentideIce Sheet:G•ogr. phys.Quat., 41,237-263, 1987. 1992. Edwards,R.L., J.W. Beck, G.S. Burr, D.J. Donahue,J.M.A. Chappell, Jones,G.A., and L.D. Keigwin, Evidence from Fram Strait (78øN) for early A.L. Bloom, E.R.M. Druffel, and F.W. Taylor, A large drop in at- deglaciation,Nature, 336, 56-59, 1988. /4 12 mospheric C/ C and reduced melting in the Younger Dryas, Jouzel,J., et al., The two-stepshape and timing of the lastdeglaciation in documentedwith23øTh ages of corals, Science, 260, 962-968, 1993. Antarctica,Clim. Dyn., 11, 151-161,1995. EIverhoi,A., E.S. Andersen,T. Dokken, D. Hebbeln,R. Spielhagen,J.l. Kaiser,K.F., Two Creeksinterstade dated through dendrochronology and Svendsen,M. Sorflaten,A. R•rnes, M. Hald, andC.F. Forsberg,The AMS, Quat.Res., 42, 288-298, 1994. growthand decayof the late Weichselianice sheetin westernSval- Karpuz,N.A., andE. Jansen,A high-resolutiondiatom record of the last bard and adjacentareas based on provenancestudies of marine deglaciationfrom the southeastNorwegian Sea: Documentationof sediments,Quat. Res.,44, 303-316, 1995. rapidclimatic changes, Paleoceanography, 7, 499-520, 1992. Emiliani,C., Glacialsurges and flood legends, Science, 193, 1270-1271, Keigwin,L.D., andG.A. Jones,The marinerecord of deglaciationfrom the 1976. continentalmargin off Nova Scotia,Paleoceanography, 10, 973-985, Emiliani,C., S. Gartner,B. Lidz, K. EIridge,D.K. EIvey, T.C. Huang, 1995. J.J. Stipp, and M.F. Swanson,Palcoclimatological analysis of late Keigwin, L.D., and S.J. Lehman,Deep circulationchange linked to Quaternarycores from the northeasternGulf of Mexico, Science, Heinrichevent 1 and YoungerDryas in a middepthNorth Atlantic 189, 1083-1088, 1975. core,Paleoceanography, 9, 185-194, 1994. CLARKl•T AL.' ORIGIN OF THE FIRSTGLOBAL MELTWATER PULSE 577

Keigwin, L.D., G.A. Jones,S.J. Lehman, and E.A. Boyle, Deglacial Polyak, L., S.J. Lehman, V. Gataullin, and A.J.T. Jull, Two-step meltwater discharge,North Atlantic deep circulation,and abrupt deglaciationof the southeasternBarents Sea, Geology,23, 567-571, climatechange, J. Geophys.Res., 96, 16,811-16,826,1991. 19•5. Kennett,J.P., and N.J. Shackleton,Laurentide ice sheetmeltwater recorded Rahmstorf,S., Rapid climatetransitions in a coupledocean-atmosphere in Gulf of Mexico deep-seacores, Science, 188, 147-150, 1975. model,Nature, 372, 82-85, 1994. Koq,N., andE. Jansen,Response of the high-latitudenorthern hemisphere Rahmstorf, S., Bifurcations of the Atlantic thermohalinecirculation in to orbitalclimate forcing: Evidence from the Nordic seas, Geology, 22, responseto changesin the hydrologicalcycle, Nature, 378, 145-149, 523-526, 1994. 1995. Labeyrie,L.D., J.-J.Pichon, M. Labracherie,P. Ippolito,J. Duprat,and J.- Rind, D., D. Peteet,W.S. Broecker,A. Mclntyearse,and W.F. Ruddiman, C. Duplessy,Melting historyof Antarcticaduring the past 60,000 The impactof coldNorth Atlantic sea-surface temperatures on climate: years,Nature, 322, 701-706, 1986. Implicationsfor the YoungerDryas cooling (l 1-10 ka), Clim. Dyn., 1, Labracherie,M., L.D. Labeyrie,J. Duprat,E. Bard,M. Arnold,J.-J. Pichon, 3-33, 1986. andJ.-C. Duplessy,The lastdeglaciation in the SouthernOcean, Pa- Rooth,C., Hydrologyand ocean circulation, Prog. Oceanogr., 11, 131-149, leoceanography,4, 629-638, 1989. 1982. Lehman,S.J., and L.D. Keigwin, Suddenchanges in North Atlantic cir- Sakai,K., andW.R. Peltier,A simplemodel of the Atlanticthermohaline culationduring the lastdeglaciation, Nature, 356, 757-762, 1992. circulation:Internal and forcedvariability with paleoclimatological Lehman,S.J., G.A. Jones,L.D. Keigwin, E.S. Andersen,G. Butenko,and implications,,/.Geophys. Res., 100, 13,455-13,479,1995. S.-R. Ostmo,Initiation of Fennoscandianice sheetretreat during the Schmitz, W.J., Jr., On the interbasin-scalethermohaline circulation, Rev. lastdeglaciation, Nature, 349, 513-516,1991. Geophys.,33, 151-173, 1995. Leventer,A., D.F. Williams, and J.P. Kennett,Dynamics of the Lauren- Shemesh,A., L.H. Burckle,and J.D. Hays,Late Pleistocene oxygen isotope tide ice sheetduring the last deglaciation:Evidence from the Gulf of recordsof biogenicsilica from the Atlantic sectorof the Southern Mexico, Earth Planet. Sci. Lett., 59, 11-17, 1982. Ocean,Paleoceanography, 10, 179-196,1995. Licht, K.J., A.E. Jennings,J.T. Andrews,and K.M. Williams, Chronol- Sowers,T., andM. Bender,Climate records covering the lastdeglaciation, ogy of late Wisconsinice retreatfrom the westernRoss Sea, Antarc- Science,269, 210-214, 1995. tica, Geology,24, 223-226, 1996. Stein,R., S.-I. Nam, C. Schubert,C. Voght,D. Ftitterer,and J. Heinemeier, Lindstrom,D.R., and D.R. MacAyeal, Death of an ice sheet,Nature, The lastdeglaciation event in the easterncentral Arctic Ocean, Science, 365, 214-215, 1993. 264, 692-696, 1994. MacAyeal,D.R., Growth/purgeoscillations of the Laurentideice sheetas Stuiver,M., G.H. Denton,T.J. Hughes,and J.L. Fastook,History of the a causeof the North Atlantic'sHeinrich events, Paleoceanography, marineice sheetin West Antarcticaduring the last glaciation,in The 8, 775-784, 1993. Last GreatIce Sheets,edited by G.H. Dentonand T.J. Hughes,pp. 319- Maier-Reimer,E., and U. Mikolajewicz,Experiments with an OGCM on 436, Wiley-Interscience,New York, 1981. the causeof the YoungerDryas, in Oceanography,1988, edited by A. Teller,J.T., GlacialLake Agassizand its influenceon the GreatLakes, in Ayala-Castanares,W. Wooster,and A. Yanez-Arancibia,pp. 87-100, QuaternaryEvolution of the Great Lakes,edited by P.R. Karrowand Univ. Nac. Auton.de Mex. Press,Mexico City, 1989. P.E. Calkin,Geol. Assoc. Can. Spec. Pap., 30, 1-16, 1985. Manabe,S., and R.J. Stouffer,Two stableequilibria of a coupledocean- Teller, J.T., Volume and routingof late-glacialrunoff from the southern atmospheremodel, J. Clim., 1, 841-866, 1988. LaurentideIce Sheet,Quat. Res.,34, 12-23, 1990. Maslin, M.A., N.J. Shackleton, and U. Pflaumann, Surface water tem- Teller, J.T., History and drainageof large ice-dammedlakes along the perature,salinity, and densitychanges in the northeastAtlantic dur- LaurentideIce Sheet,Quat. Int., 28, 83-92, 1995. ing the last45,000 years:Heinrich events, deep water formation, and Tushingham,A.M., and W.R. Peltier, lce-3G: A new global model of climaticrebounds, Paleoceanography, 10, 527-544, 1995. late Pleistocenedeglaciation based upon geophysicalpredictions of Mickelson,D.M., L. Clayton,D.S. Fullerton,and H.W. Bores,The Late postglacialrelative sea level change,J. Geophys.Res, 96, 4497- Wisconsinglacial recordof the LaurentideIce Sheetin the United 4523, 1991. States,in Late QuaternaryEnvironments of the UnitedStates, edited Warrick,R., andJ. Oerlemans,Sea level rise, in ClimateChange, edited by H.E. Wright Jr., pp. 3-37, Univ. Minn. Press,Minneapolis, 1983. by J.T. Houghton,G.J. Jenkins,and J.J. Ephraums,pp. 261-281, Miller, G.H., andD.S. Kaufman,Rapid fluctuationsof the LaurentideIce IPCC Sci. Assess.,Cambridge Univ. Press,New York, 1990. Sheet at the mouth of Hudson Strait: New evidence for ocean/ice Wright, D.G., and T.F. Stocker,Younger Dryas experiments, in Ice in sheet interactions as control on the Younger Dryas, the ClimateSystem, edited by W.R. Peltier,N,4 TO ,4SISet., 112, pp. Paleoceanography,5, 907-919, 1990. 395-416, Springer-Verlag,New York, 1993. Mitrovica,J.X., andJ.L. Davis,The influenceof a finiteglaciation phase on Wright, H.E., Jr., and J. Stein, Glacial surgesand flood legends,Science, predictionsof post-glacialisostatic adjustment, Earth Planet. Sci. Lett., 193, 1268-1269, 1976. 136, 343-361, 1995. Mix, A.C., The oxygen-isotoperecord of glaciation,in The Geologyof North,4merica, vol. K-3, NorthAmerica and AdjacentOceans During R. B. Alley, Earth SystemScience Center and Departmentof Geo- theLast Deglaciation, edited by W.F. Ruddimanand H.E. WrightJr., sciences,Pennsylvania State University, University Park, PA 16802.(e- pp. 111-135,Geol. Soc.of Am., Boulder,Colo., 1987. mail: ralley•essc.psu.edu) Nakada,M., andK. Lambeck,The meltinghistory of the late Pleistocene P. U. Clark and J. M. Licciardi, Departmentof Geosciences,Oregon Antarctic ice sheet,Nature, 333, 36-40, 1988. State University,Corvallis, OR 97331. (e-mail: clarkp•ucs.orst.edu; Oppo,D.W., and S.J.Lehman, Suborbital timescale variability of North licciarj•bcc.orst.edu) AtlanticDeep Water during the past200,000 years, Paleoceanogra- S. J. Johnsen,Niehls Bohr Institute,Department of Geophysics,Uni- phy, 10, 901-910, 1995. • versityof Copenhagen,Haraldsgade 6, 2200 CopenhagenN., Denmark. Ortner, P.B., T.N. Lee, P.J. Milne, R.G. Zika, M.E. Clarke, G.P. Podesta, (e-mail:sigfus•osiris.gfy.ku.dk) P.K. Swart, P.A. Tester, L.P. Atkinson, and W.R. Johnson,Missis- L. D. Keigwin, WoodsHole OceanographicInstitution, Woods Hole, sippiRiver flood watersthat reached the Gulf Stream,J. Geophys. MA 02543. (e-mail:Ikeigwin•cliff. whoi.edu) Res., 100, 13,595-13,601, 1995. H. Wang, AtmosphericScience Division, LawrenceLivermore Na- Peltier,W.R., Lithosphericthickness, Antarctic deglaciation history., and tionalLaboratory, P.O. Box 808, L-256, Livermore,CA 94550. (e-mail: oceanbasin discrctization effects in a globalmodel of postglacialsea huaxiao•extreme.llnl.gov) level change:A summaryof somesources of nonuniqueness,Quat. Res.,29, 93-112, 1988. (ReceivedJuly 26, 1995' revisedMay 2, 1996; Peltier,W.R., Ice agepaleotopography, Science, 265, 195-201,1994. acceptedMay 2, 1996.)