ChronologyofQuaternaryglaciationsinEastAfricaTimothyM.Shanahan*,MarekZ - PDF document

ChronologyofQuaternaryglaciationsinEastAfricaTimothyM.Shanahan*,MarekZredaDepartmentofHydrologyandWaterResources,UniversityofArizona,Tucson,AZ85721,USAReceived5August1999;accepted31January2000AnewglacialchronologyforequatorialEastAfricaisdevelopedusinginsitucosmogenicClmeasuredin122bouldersfrommorainesonMountKenyaandKilimanjaro.TheoldestdepositssampledonKilimanjaroyieldaClageof360calendarkyr(allClagesareincalendaryears,cal.kyrorcal.yr).OnMountKenya,theoldestmorainesgiveagesof355^420kyr(LikiI)and255^285kyr(Teleki).GiventheuncertaintyinourClages,theLikiImorainemaycorrespondtoeithermarineisotopestage10or12,whereastheTelekimorainecorrelateswithstage8.Thereisnoevidenceforstage6oneithermountain.TheLikiIImorainesonMt.KenyaandmorainesoftheFourthGlaciationonKilimanjarogiveagesof28þ3kyrand20þ1kyr,respectively.Theyrepresentthelastglacialmaximum(LGM)andcorrelatewithstage2ofthemarineisotoperecord.AseriesofsmallermorainesabovetheLGMdepositsrecordseveralreadvancesthatoccurredduringthelateglacial.OnMt.Kenya,thesedepositsdateto14.6þ1.2kyr(LikiIIA),10.2þ0.5kyr(LikiIII),8.6þ0.2kyr(LikiIIIA)and200yr(Lewis);thecorrespondingdepositsonKilimanjarohavemeanagesof17.3þ2.9kyr(FourthGlaciation^Saddle),15.8þ2.5kyr(LittleGlaciation^Saddle),and13.8þ2.3kyr(FourthGlaciation^Kibo).Thesedataindicatethattheclimateofthetropicswasextremelyvariableattheendofthelastglacialcycle.ß2000ElsevierScienceB.V.Allrightsreserved.Keywords:Quaternary;glaciation;EastAfrica;chronology1.IntroductionTropicalEastAfricaisoneofthreeplacesontheequatorwithdirectglacial^geomorphologicevidenceofformer,expandedglaciers;theothertwoareinSouthAmericaandNewGuinea.Theageandextentofmorainesdepositedbytheseglaciersprovidearecordofthemagnitudeandtimingofclimatechangeinthetropics.OnMountKenyaandKilimanjaro,theseglacialdepositshavebeendescribedindetail[1,2].However,theconstructionofanaccurateglacialchronologyforthesemountainshasbeenhinderedbydi¤cultiesintheabsolutedatingofthedeposits.Thesedi¤-cultiesareprimarilyduetothepaucityofdatablematerialandthelimitedtimerangeofapplicabil-ityofavailablemethods.Inthisstudy,weusedcosmogenicCltodeterminesurfaceexposureagesofmorainesonMt.KenyaandKilimanjaro(Fig.1)andtoreconstructthetimingandextentofglaciationsintropicalEastAfrica.Develop-mentofareliableglacialchronologywillaidinoure¡ortstocorrelatetheterrestrialrecordofglaciationsinEastAfricawithotherrecordsof0012-821X/00/$^seefrontmatterß2000ElsevierScienceB.V.Allrightsreserved.PII:S0012-821X(00)00029-7 *Correspondingauthor.Tel.:+1-520-621-4072;Fax:+1-520-621-1422;E-mail:shanahan@hwr.arizona.eduEPSL538927-3-00EarthandPlanetaryScienceLetters177(2000)23^42 www.elsevier.com/locate/epsl paleoclimate,suchaspollen,seasedimentsandicecores.2.Priorwork2.1.Mt.KenyaGlacialdepositsonMt.Kenyaarepresentfrom4600mtolessthan2900m.Figs.2a^darecom-prehensivemapsoftheseglacialdeposits,modi-¢edfrompreviousstudies[1,3].Sevenstagesofglacialmoraineshavebeenidenti¢ed(fromoldesttoyoungest):Gorges,LakeEllis,NaroMoru,Teleki,Liki(I,II,III),TyndallandLewis[1].Thechronologyfortheoldestdeposits(Gorges,LakeEllis,NaroMoruandTeleki)isbasedonpaleomagneticdataandstratigraphicrelation-ships[1].AgeestimatesfortheLiki(I,II,III)depositshavebeenhinderedbythelackofasso-ciateddatablematerialandproblemswithobtain-ingsu¤cientorganicmaterialfromcorebottoms[4,5].Thechronologyforneoglacialdepositsisknownfromhistoricalrecords(Lewis)andlichen-ometry(Tyndall)[1].2.2.KilimanjaroDownieandWilkinson[2]mappedtheglacialdepositsonKilimanjaroanddevelopedarough Fig.1.LocationsofKilimanjaroandMt.KenyainEast Fig.2.(a)MorainesonMt.Kenya.Solidlinesindicatemor-ainesmappedfromaerialphotographs.Boxesindicateloca-tionofstudyareasmagni¢edinb^d.MapsofstudyareasadaptedfromMahaney[1]withmorainesandsampleloca-tions:(b)lowerTelekiValley;(c)upperTelekiValley;(d)GorgesValley.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 chronologyfortheoldestglaciationsbasedonvolcanicrocksassociatedwithglacialdeposits.Theirmappingwaslatercon¢rmedusingaerialphotographs[3].FourmajorPleistoceneglacia-tionsareidenti¢ed(fromoldesttoyoungest):theFirst,theSecond,theThirdandtheFourth(Main).Ofthese,onlytheMainGlaciationisas-sociatedwithasystemofmoraines.TheabsenceofvolcanicsassociatedwiththeMainGlaciationhasleftitsageinquestion.TwoHoloceneglacia-tionswerealsoidenti¢ed:theRecentandtheLit-tle,bothofwhichformmorainesinsidethoseoftheMainGlaciation.Nochronologicalcontrolisavailablefortheyoungerdeposits.Mapsofthedepositsbasedupon¢eldobservations,aerialphotographsandpreviousstudies[2,3]areshowninFig.3a^d.3.Methods3.1.SamplingandanalyticalproceduresSamplingonMt.Kenyawasconductedineast^westtrendingvalleys,whichwerereported[1]tohavethemostfavorabledepositsforsurfaceex-posuredating(Fig.2).OnKilimanjaro,samplingconcentratedonthesouthernslopesofMawenziandontheSaddlebetweenKiboandMawenzi(Fig.3).Theprimarycriteriausedinselectionofsampleswerebouldersize,positiononthemor-ainecrestandsurfacecharacter.Tominimizethee¡ectoflandformerosionandsnowcoverontheexposurehistoryofthesamples,wecollectedfromthetopsofthelargestboulderslocatedalongmorainecrests.Boulderswerealsoselectedonthebasisofsurfacepreservationandglacialchar-acter(shapeandthepresenceofstriaeorpolish).Whole-rocksamplesweregroundtoasizefrac-tion0.25^1.00mmandleachedin5%HNOremovemeteoricClandanysecondarycarbonates Fig.3.(a)MorainesonKilimanjaro.Solidlinesindicatemorainesmappedfromaerialphotographs.Boxesindicatelocationofstudyareasmagni¢edinb^d.Mapsofstudyareaswithmorainesandsamplelocations:(b)eastslopesofKiboPeak;(c)theSaddle;(d)southslopesofMawenziEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 Table1DataforClagesofoldglaciationsSampleIDCl/ClCl)(ppm)(Cl/(gyr))(kyr)OldestGlaciation,SouthMawenziMP-816640þ13041.0þ2.145117.0þ3.0MP-8220600þ80026.4þ0.638355.0þ22.0MP-836460þ26022.4þ2.23674.0þ3.0MP-8414360þ29024.2þ3.740200.0þ5.0MP-858550þ17045.1þ0.140205.0þ5.025390þ59025.5þ0.632663.0þ37.0Coe¤cientofvariationLandformageGorgesValley:LikiIGV-303500þ120670.0þ63.398432.0þ26.0GV-312490þ801177.5þ80.4169329.0þ16.0GV-322620þ901423.1þ89.1202383.0þ22.0Coe¤cientofvariationLandformageTelekiValley:TelekiTV-171810þ70768.8þ11.5124243þ132820þ90837.2þ36.6129531þ35TV-191940þ70667.5þ25.4111255þ13TV-202180þ48652.7þ28.2109300þ10TV-211641þ49927.8þ58.5141231þ10TV-225070þ14058.4þ5.543134þ5TV-237200þ28053.5þ3.540200þ10Coe¤cientofvariationLandformageGorgesValley:NaroMorutillGV-74503þ22497þ1.412133þ2GV-75476þ15930þ45.119437þ1GV-761278þ51115þ6.86936þ2GV-77666þ4.21136þ56.123854þ0.3GV-781029þ39639þ38.916371þ3GV-791043þ43257þ10.810046þ2GV-80723þ311106þ58.622960þ3GV-811172þ511171þ81.1251101þ5Coe¤cientofvariationLandformage55þ23TelekiValley:LikiIITV-71520.0þ60806þ15.7201114.0þ5TV-81770.0þ60831þ0.2210135.0þ6TV-9746.0þ301039þ10.221562.0þ3TV-101027.0þ371123þ0.123887.0þ4TV-111007.0þ20211þ0.58642.0þ0.9TV-12367.0þ141695þ36.030433.0þ1TV-13608.0þ241744þ0.732256.0þ2TV-14239.0þ8943.8þ3.917119.2þ0.7TV-15283.0þ81139þ15.920923.0þ0.7TV-161294.0þ39373þ5.712171.0þ2Coe¤cientofvariationLandformage64þ4020kyrroundedtonearest1kyr.Sampleexcludedfromconsideration.Ageestimatedusingerosionmodel.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 fromporespaces.ClwasextractedbydissolvinginamixtureofhotHFandHNOandprecipi-tatedasAgClforClanalysisbyacceleratormassspectrometry[6]atPRIMELab,PurdueUniversity.Majorelementsweremeasuredbypromptgammaspectroscopy,UandThbyneu-tronactivationanalysisandtotalClbyspeci¢cionelectrode[7].3.2.PrinciplesProductionofCloccursviainteractionsbe-tweencosmic-rayneutronsandmuonswithCaandClinlandformsexposedattheEarth'ssurface.Becausethedepthdependenceofthepro-ductionanddecayratesforClareknown,thetimeofdepositionofthesurfacecanbedeter-mined.Theprinciplesofcosmogenicdatingoflandformshavebeendescribedelsewhere[8,9].Theutilityofcosmogenicdatingtechniquesfordevelopingglacialchronologieshasbeendemon-stratedbynumerousworkers[10^19].Calculationsoftheelevationandlatitudede-pendenceofClproductionrateswereperformedaccordingtoLal[9].CorrectionsfortopographicshieldingweremadeusingtheformulationofZre-daandPhillips[20].TheproductionrateofduetothermalandepithermalneutronabsorptionClwascalculatedusingthemethodoutlinedinLiuetal.[21].AgecalculationswereperformedusingthecomputerprogramCHLOE[22].WeusedtheClproductionparametersdeter-minedbyPhillipsetal.[23]:73.3þ4.9atoms(gCa),154þ10atomsCl(gK)and586þ40fastneutrons(gair).Thoughotherworkershavedetermineddi¡erentproduc-tionrates[24,25],thereisnoindicationthattheyaremoreaccuratethanthoseofPhillipsetal.[23].Randomerrorsassociatedwiththecalibrationshouldbesmallerthan10%[23]andoverallran-domerrorsformoraineagesshouldnotexceed15%[17].Uncertaintiesrelatedtotemporalvaria-bilityinproductionratesarelesswellunderstood.Becausethesevariationsaresigni¢cantatlow-lat-itudesites,thesee¡ectsarediscussedingreaterdetailbelow.Analyticalerrorisprimarilyafunctionofun-certaintiesinthemeasurementsofCl/ClandoftotalCl.TheerrorassociatedwithAMSmeasure-mentswasonaverage4.9%,thoughsomesampleshaderrorsashighas28.6%oraslowas0.6%.UncertaintyinmeasurementsoftotalClrangedfrom0.0%(basedonthreeidenticalmeasure-ments)to16.7%withameanof4.6%.ThetotalanalyticalerrorfortheageofeachsamplewasdeterminedusingMonteCarlosimulation.TenthousandagesweregeneratedassumingnormaldistributionsofpossibletotalClandCl/Clval-ues.Individualsamplestandarddeviationswerethencalculatedbasedupontherangeofpossibleagesproducedbythesimulation.ErrorsreportedinTables1and2re£ectonlytheMonteCarlo-deriveduncertainties.3.3.CorrectionsfortemporalvariabilityincosmicVariationsinthe£uxofcosmicradiationresultinchangesincosmogenicproductionratesovertime.Thesevariationsareafunctionof(i)changesintheintensityofgalacticcosmicrays,(ii)changesinsolaractivityand(iii)changesinthestrengthofthegeomagnetic¢eld.Becausevar-iationsinthe¢rsttwofactorsaretoolongandtooshort,respectively,tosigni¢cantlya¡ectpro-ductionratesfortimescalesofinterest(ca.10^700kyr),weconsideronlythee¡ectofvariationsinthegeomagnetic¢eldstrengthonproductionrates.Cosmicraysconsistprimarilyofchargedpro-tonsandalphaparticles.Thosewithsu¤cienten-ergiesareabletopenetratethegeomagnetic¢eldandproducecosmogenicnuclides.Anyvariationsinthestrengthofthegeomagnetic¢eldovertimeresultinchangesinthecosmic-ray£uxandchangesinproductionrates.Becausethemagnetic¢eldhasagreaterin£uenceonthecosmic-ray£uxattheequatorthanatthepoles,thee¡ectofvariationsinthe¢eldstrengthonproductionratesislargeratlowlatitudes.Productionratesdeter-minedatmiddleandhighlatitudes,wherethee¡ectofthegeomagnetic¢eldisnegligible[23],needtobemodi¢edtoaccountforvariationsinthemagnetic¢eld,especiallyforapplicationsatlowlatitudes.Recentpaleomagneticrecords[26]indicatethatvariationsinthegeomagnetic¢eldEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 Table2DataforClagesofyoungglaciationsandstromatolitesSampleIDCl/ClClAge(kyr))(ppm)(Cl/(gyr))0mmkyr2mmkyrLikiII,GorgesValley,Mt.KenyaGV-33425þ2121.0þ45.711032þ228þ2GV-34751þ3535.0þ4.54424þ123þ1GV-35293þ1313.0þ17.218927þ123þ1GV-362830þ140140.0þ0.24528þ130þ1GV-37316þ1212.0þ6.215228þ124þ1GV-381354þ5555.0þ2.44632þ131þ1Landformage28þ3MainGlaciation,Mawenzi,KilimanjaroMP-87841þ3859.8þ6.24320þ120þ18870þ34021.0þ3.534103þ5119þ51071þ5168.4þ1.24528þ128þ1474þ19103.7þ1.35116.4þ0.716.1þ0.7MP-911270þ5033.3þ3.83719.7þ0.819.8þ0.8MP-921230þ3440.5þ0.74221þ121þ1Landformage20þ1MainGlaciation,Saddle,KilimanjaroMP-94450þ70194.7þ0.63117.5þ2.916.7þ2.9MP-95500þ21150.1þ0.92416.7þ0.716.1þ0.7MP-96569þ25143.5þ10.82219.1þ0.918.5þ0.9MP-97405þ16181.8þ0.72815.8þ0.715.2þ0.7MP-98428þ29156.7þ3.32415.4þ1.114.9þ1.1MP-99587þ25179.1þ6.52823þ122þ1MP-100347þ14192.8þ5.03113.6þ0.613.2þ0.6Landformage17.3þ2.9LittleGlaciation,Saddle,KilimanjaroMP-101904þ4172.4þ0.47115.7þ0.715.7þ0.7MP-102689þ33101.5þ5.57415.9þ0.815.7þ0.8MP-103415þ19138.2þ8.97113.5þ0.713.1þ0.7MP-104377þ16208.3þ14.98315.7þ0.715.0þ0.7MP-106905þ3388.1þ0.97219.0þ0.718.8þ0.7MP-107470þ22188.5þ3.38018.6þ0.917.8þ0.9MP-108471þ20158.7þ0.97716.2þ0.715.7þ0.7MP-109605þ28116.4þ8.66717.7þ0.917.2þ0.9MP-110365þ26222.0þ5.19414.4þ1.113.9þ1.1MP-111512þ14100.0þ0.06712.7þ0.412.4þ0.4MP-112490þ140183.5þ10.67819.2þ5.818.4þ5.8MP-113298þ10203.2þ5.58611.5þ0.411.1þ0.4Landformage15.8þ2.5FourthGlaciation,KiboPeak,Kilimanjaro777þ35175.7þ13.911718.5þ0.917.7þ0.91866þ5583.1þ2.89428þ128þ11482þ4878.7þ2.68922þ122þ1KB-119603þ19120.8þ3.79412.0þ0.411.7þ0.4KB-1201166þ5360.1þ1.87814.8þ0.714.7þ0.7KB-121596þ26176.4þ3.09716.8þ0.816.1þ0.8KB-122445þ17178.9þ3.39712.1þ0.511.8þ0.5Landformage13.8þ2.3EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 Table2(SampleIDCl/ClClAge(kyr))(ppm)(Cl/(gyr))0mmkyr2mmkyrLikiIIA,GorgesValley,Mt.KenyaGV-39170þ111040.4þ45.815513.9þ0.912.7þ0.9GV-40191þ11891.9þ62.614015.5þ0.914.1þ0.9Landformage14.6þ1.2LikiIIIboulders,GorgesValley,Mt.KenyaGV-691171þ3932.8þ0.05910.8þ0.410.8þ0.4349þ15560.0þ12.716218.4þ0.816.8þ1GV-711343þ5040.5þ0.16314.5þ0.614.6þ0.6GV-723170þ11010.8þ0.54213.7þ0.513.9þ0.53360þ12012.2þ0.14216.6þ0.616.9þ0.6Landformage14.1þ0.6LikiIII,TelekiValley,Mt.KenyaTV-45222þ12447.7þ17.312810.4þ0.69.9þ0.6TV-46363þ16192.9þ5.19010.9þ0.510.6þ0.5TV-47139þ91044þ29.91639.5þ0.68.9þ0.6TV-48286þ14211.0þ2.59110.5þ0.510.1þ0.5227þ12469.8þ22.512013.1þ0.712.3þ0.7TV-501680þ7017.4þ1.7509.9þ0.410.0þ0.4TV-511662þ5918.6þ1.65110.1þ0.410.2þ0.4Landformage10.2þ0.5LikiIIIA,TelekiValley,Mt.KenyaTV-52907þ3838.3þ0.45310.4þ0.510.5þ0.5TV-531350þ6020.2þ0.3538.4þ0.48.4þ0.4TV-54661þ2733.6þ1.6595.9þ0.36.0þ0.3TV-55250þ11238.3þ9.78111.2þ0.510.9þ0.5GV-56167þ7663.7þ26.21498.8þ0.48.4þ0.4TV-57175þ7651.8þ20.614311.9þ0.511.2þ0.5TV-58806þ4033.0þ0.1458.6þ0.58.6þ0.5Landformage8.6þ0.2Lewis,TelekiValley,Mt.KenyaTV-5952þ5180.7þ3.42990.10þ0.010.10þ0.01TV-6074þ4179.8þ1.62680.33þ0.020.33þ0.02TV-61166þ7132.7þ10.02271.2þ0.051.2þ0.05TV-6255þ5209.5þ12.34080.20þ0.020.20þ0.02TV-6373þ699.8þ9.32310.20þ0.020.20þ0.02Landformage0.21þ0.21LateGlacialdeposits,Mt.KenyaTV-6460þ6196.0þ5.53670.15þ0.020.15þ0.02TV-65431þ2122.9þ2.72410.63þ0.030.63þ0.03TV-6657þ528.8þ3.92310.03þ0.000.03þ0.00TV-6745þ655.4þ3.02570.00þ0.000.00þ0.00PointLenana,striatedbedrock,Mt.KenyaPL-68114þ51140.2þ63.73374.1þ0.24.0þ0.2Gorgeslandform,GorgesValley,Mt.Kenya178þ8446.1þ1.456.722þ120þ1GV-27159þ851.6þ1.221.85.9þ0.35.9þ0.3GV-2874þ10361.9þ4.150.17.6þ1.17.3þ1.1GV-2983þ8675.1þ52.381.26.3þ0.66.0þ0.6Landformage6.6þ0.9EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 strengthshouldaverageoutovertimeslongerthanabout50kyr.Toaccountforthee¡ectoftemporalchangesinthegeomagnetic¢eldoncosmogenicnuclideproduction,productionrateswerescaledusingageomagneticcorrectionfactor.Thisfactorwasbasedonthetheorythateachgeomagneticlati-tudehasanassociatede¡ectiveverticalcuto¡rigidity.Onlythosecosmic-rayparticleswithen-ergiesgreaterthanthisrigiditycanpassthroughthemagnetic¢eldandentertheEarth'satmos-phere.Thecuto¡rigidity()canbecalculatedforanylatitudeusingthefollowingrelation:=14.9()cos,whereisthegeomag-neticlatitudeandistherelativestrengthofthegeomagneticdipole[27].Fortimeswhenthegeomagnetic¢eldwasreduced(anewe¡ectiverigidityandacorrespondinggeo-magneticlatitudecanbecalculated.Then,anad-justedproductionrateforthesitecanbedeter-minedusingthescalingfactorsofLal[9].Forsituationsofincreasedmagnetic¢eldstrength1.0),itisimpossibletocalculateanef-fectivelatitudebecausemagnetic¢eldstrengthismaximizedattheequator,where=1.0.Inthiscase,weusedtheapproximation:,whereisthecorrectedproductionrateandistheproductionratecorrespondingcorrespondingThe200kyrrelativemagneticintensityrecord(SINT-200)[26]wasusedasarecordofpastmag-netic¢eldvariations.SINT-200isacompositestackof17marinesedimentrecordsofpaleoin-tensitycollectedfromaroundtheworld.Althoughpaleomagneticdatafromsourceslikelava£owsandarcheomagneticstudiesarepreferredtosedi-mentrecordsbecausetheyrecordabsoluteinten-sities,di¤cultieswithdevelopmentofcontinuousrecordsandwithchronologicalcontrollimittheusefulnessofthesedata.Previousstudies(e.g.[29])haveshownthatabsoluterecordsaregener-allyingoodagreementwithmarinesedimentre-cords.Furthermore,strongcorrelationsbetweenwidelyseparatedmarinerecordsaswellasbe-tweenmarinerecordsandBeinseasediments[30]indicatethatmarinerecordsaccuratelyrecordchangesinthemagnetic¢eld.Theratioistherelativestrengthofthegeomagneticdipole.Athighlatitudes,bothareessentiallyconstantwithtimebecausevariationsinthe¢eldarenegligible.Forlowlat-itudes,changesinthe¢eldarelargeandmustbecalculatedusingarecordofpaleo-magnetic¢eldvariations.Becausechangesinthe¢eldarecyclicoverlongtimes,thelong-termaverageofthepaleomagneticrecordwasusedasanapproximationfor.Theparametereachsamplewascalculatedbyintegratingthemagnetic¢eldrecordofSINT-200overthemeanuncorrectedlandformage.Then,wasusedtocalculateanewproductionrate,sam-pleagesandanadjustedlandformage.Theseval-ueswereusedtoproduceanewandthepro-cedurewasiterateduntiltheindividualsampleagesconverged.Correctedagesusingthisproce-durearereportedinTables1and2. Table2(SampleIDCl/ClClAge(kyr))(ppm)(Cl/(gyr))0mmkyr2mmkyrAgesofstromatolitesLakeTauca,BoliviaCH-472400þ10065.8þ1.918614.6þ0.614.8þ1LakeMagadi,KenyaKE-1-ML267þ1741.8þ3.51912.5þ0.812.6þ1KE-3-ML509þ2522.5þ1.31912.9þ0.713.1þ1KE-4-ML726þ3416.2þ1.42112.2þ0.612.4þ1475þ4420.2þ0.2229.2þ0.99.3þ1Stromatoliteage12.5þ0.4,Bedrocksampleexcludedfromconsideration.20kyrroundedtonearest1kyr.Sampleoutlierexcludedfromconsideration.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 3.4.TestofgeomagneticcorrectionToverifyourmethodofgeomagneticcorrec-tion,wecollectedsamplesfromindependentlydatedstromatolitesattwolow-latitudelocations:paleolakeMagadiinequatorialEastAfrica(2³S)andpaleolakeTaucainBolivia(19³S).Thestro-matolitesfromLakeMagadiformanalmostcon-tinuousbeltapproximately65mabovethecur-rentlakelevel,delineatingtheformerextentofthelake.Thewesternsideofthelakeisboundedbyalargefaultscarponwhichthestromatolitebeltiswell-preservedandcaneasilybesampled.Weidenti¢edandsampledfourinsitustromato-liteconcretions.Previousstudiesofthesestroma-tolitesusingCyieldedages(convertedtocalen-daryears(cal.yr)usingStuiveretal.[31])rangingfrom7826þ240to14782þ890cal.yr,withthemajorityofsamplesbetween11000and13000cal.yr[32].Oneadditionalsamplewascollectedfromalarge(1mdiameter)algalbiohermonashorelineofpaleolakeTauca.ThissampleprovidesacheckofourpaleomagneticcorrectionforaslightlyhigherlatitudethanthatoftheAfricansamples.PreviousstudiesusingChaveyieldedagesrang-ingfrom14744þ453to16518þ825cal.yrwithameanageof15420þ764cal.yr[33,34].Fig.4showsthedistributionsofCstromato-liteages(cal.kyr),uncorrectedClagesandcor-Clages.ThecorrectedClagesagreeatthe1levelwithCagesforbothLakeMagadiandpaleolakeTauca.Uncorrectedcosmogenicagessigni¢cantlyunderestimatethetrueageofthestromatolites.Theseresultssupportouras-sumptionthatacorrectionforvariationsinthegeomagnetic¢eldisnecessaryandvalidateourmethodofadjustinglow-latitudeproductionUncertaintiesassociatedwiththesemagneticcorrectionsofproductionratesaredi¤culttoevaluateforseveralreasons.Withtheexceptionofthestromatolitesreportedinthisstudy,fewotherattemptshavebeenmadetoquantifythee¡ectsofchangesinthemagnetic¢eldstrengthatlowlatitudesusingsamplesofknownage[35,36].Furthermore,littleisknownabouttherelationshipbetweenchangesinmagnetic¢eld Fig.4.ComparisonbetweenpreviouslymeasuredCageson(A)stromatolitesfrompaleolakeMagadi[32]and(B)stromatolitesfrompaleolakeTauca(allCagesconvertedtocal.yrusingStuiveretal.[31]),withClagescorrected(¢lledcircles)anduncorrected(emptycircles)forthee¡ectofchangesinthegeomagnetic¢eldstrength.Individualagesareshownasarangeduetouncertaintiesintheconver-sionbetweenradiocarbonandcal.yr.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 strengthandcosmogenicisotopeproductionrates,particularlyforequatorialregionswhereinthepast¢eldstrengthswerehigherthananywheretoday.Errorsassociatedwiththesecalculationscannotbeassesseduntilmoreworkinthisareahasbeencompleted.Additionaluncertaintiesarisefromtheinaccuraciesofexistingpaleointensityrecords.BecausetheSINT-200recordisacom-positeof17marinesedimentcores,ithaserrorsassociatedwiththeanalyticalandchronologicaluncertaintiesfromindividualcoresaswellasun-certaintiesrelatedtothestackingoftherecords.Despitetheseuncertainties,theSINT-200recordgenerallyshowsgoodconsistencybetweenthecompositerecordandeachindividualrecord(meancorrelationcoe¤cientbetweenSINT-200andindividualrecords=0.6þ0.16[26])aswellaswithrecordsofcosmogenicisotopeproduction[30]andotherrecordsofmagnetic¢eldstrengthstrength3.5.E¡ectoferosiononClagesAdditionaluncertaintyinsurfaceexposureagesmayresultfrompost-depositionalprocesses.Ero-sionofthesoilmatrixanddegradationofbouldersurfacesmayresultinapparentsurfaceexposureagesdi¡erentfromthetrueageofthelandform.Dependingontheproductionmechanismsandtheerosionrate,erosionofthebouldersurfaceresultsinapparentagesthatareeitherolderoryoungerthanthetrueage.Forspallogenicnu-clides,e.g.Al,productionratesdecreaseexponentiallywithdepth.Thisresultsinapparentagesthatalwaysunderestimatethetrueageofthesurface.FornuclideswhicharealsoproducedbyneutronactivationofCl(e.g.Cl),agesmayei-therunder-oroverestimatethetrueage.Thisisbecausethedepthpro¢leforproductionbyneu-tronactivationinitiallyincreaseswithdepthbe-forefollowinganexponentialdecreasewithdepth.Fig.5illustratesthetheoreticaldistributionofapparentagesfortwoboulderchemistriesasafunctionofsurfaceerosionrate.ForsampleswithlowCl,Clproductioniscontrolledbyspal-lationandthereforedecreasesexponentiallywithdepth.ForhighClsamples,inwhichneutronactivationisasigni¢cantproductionmechanism,ages¢rstincreasewithincreasingerosionandthendecreasewithadditionalerosion.Foryounglandforms,bouldererosionisasigni¢cantprocessa¡ectingapparentboulderage.Forolderland-forms,erosionofmorainematrixismoresigni¢-cantandwilltendtodominateovertherelativelyslowerosionofthebouldersurfaces.Toillustratethee¡ectofbouldererosionontheapparentageofyoungsamples,wereportinTable2agesforzeroerosionandforanerosionrateof2mmAlthoughnoindependentinformationonero-sionratesofbouldersurfacesisavailablefortheselocalities,webelievethatamaximumerosionrateof2mmkyrisreasonablebasedupon¢eldevidenceandtheresultsofstudiesfromsimilarclimates.Erosionratesof0.2mmkyrwerepre-viouslyestimated[17]forglaciallydepositedbouldersintheWindRiverRange,WY,USA,usingtheratioofCltoBeinthesurfaces.Furthermore,Gosseandco-authors[13]arguedthatbouldererosionratesmustbelowinaridregionsbaseduponobservationsofglacialstria-tionsandpolishonbouldersdeposited20000yrago.Signi¢cantlyhighererosionrates(2^8mm)reportedforbedrockinAfrica[39,40]andBrazil[41]maynotbeapplicabletooursam-ples.Bedrockerosionratesmaybehigherthanthoseofglacialdepositsbecausetheyhavelongerexposurestoweatheringprocesses.Incontrast,theprocessofglacialerosionpreservesonlythe Fig.5.ApparentcosmogenicagesasafunctionofbouldererosionrateforsampleswithhighCl(solid)andlowCl(dashed).HighClsamplehassigni¢cantClproductionfromneutronactivationofCl,whilelowClsamplehasmostlyspallogenicEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 hardest,mostresistantrocksaslargeboulders.Surfacespre-weatheredpriortoglacialpluckinganderosionaremoresusceptibletodisintegrationduringtransportbytheglacier.Furthermore,bysamplingonlythelargest,freshestlookingbould-ers,wepreferentiallyselectthebouldersthataremostresistanttoweatheringprocesses.Erosionofsoilmatrixalsoresultsinapparentboulderagesthatdi¡erfromthetruelandformage.Bouldersdepositedonthelandformsurfaceinitiallyremaintherewhereasbouldersdepositedatsomedepthinthelandformaregraduallyex-posedwiththeremovalof¢nesedimentandtheprogressiveloweringofthelandformsurface.Thisresultsinanassemblageofbouldersontheland-formsurfacewithadistributionofexposurehis-toriesandapparentagesthatrangefromslightlyoldertoyoungerthantheageofthesurface.Byselectingthelargestboulders(1minmostcases),locatedonorneartothemorainecrest,weminimizethein£uenceofgradualexposureonapparentboulderages.Foryounglandforms30kyr),weassumethatcarefulsamplingmakesthee¡ectofsoilerosionnegligiblewhencomparedwithothersourcesofuncertainty.Thisassumptionissupportedbythetypicallynar-rowdistributionofapparentboulderagesforyounglandforms.Forolderlandforms,thedistributionofbould-eragesindicatesthatthee¡ectofsoilerosionissigni¢cant.Apreviousstudy[42]usedMonteCarlosimulationofthisprocessbygeneratingasetofbouldersrandomlydistributedinagradu-allyerodinglandform.Theyfoundthatthecoef-¢cientofvariationforthesimulatedboulderagesiscorrelatedwiththeerosiondepth.Therefore,thedistributionofapparentboulderagesandthemeanapparentboulderagemaybeusedtodetermineboththetrueageandthesoilerosionrateforthelandform.Inthisstudy,thisMonteCarlosimulationwasexpandedtoincludethee¡ectsoferosionofbouldersurfaces.Tenthousandbouldersweregeneratedwithchemistriesrandomlyselectedfromtheknownchemicalcompositionofbould-ersatthesite.Theinitialdepthofburialfortheindividualboulderswasselectedfromauniformdistributionofdepthsbetween,theactualheightoftheboulderinthe¢eld,tosomedepthwhichisequaltothemaximumerosiondepth(foraparticularerosionrate)subtractingtheboulderheight.Weassumedthatacrossthelandform,soilerosionratesareconstant,eventhoughinitialboulderburialdepthsarenot.Eachboulderwasalsorandomlyassignedabouldererosionratefromalognormaldistributionwithvaluesrangingfromzerotoanassignedmaximumerosionrate).Thisdistributionwasusedontheas-sumptionthatbouldersaremorelikelytohavebouldererosionratesthatareclosertozerothantheyaretohaveerosionratesnearthemax-imum.Thesimulationwasrunforallpossiblecombinationsofbetweenzeroandgcm(ca.0^2mmkyr)andbetweenzeroand2000gcm(ca.10m).Asmentionedabove,thevalueforisbaseduponpreviouslymeasuredbouldererosionratesinaridregions[17].Themaximumsoilerosiondepthusedinthemodelwaslargeenoughtoin-cludeallpossiblematrixerosionratesthatwouldproducetheobserveddistributionofages.Foreachcombinationof,adistribu-tionofageswasdeterminedandacoe¤cientofvariationwascalculated.Theresultisadistribu-tionofallpossiblelandformages,dependent Fig.6.Resultsofthesoil^bouldererosionmodelfortheTel-ekimoraine.Linesofconstantapparentage/trueagearedashed.Solidlinesrepresentconstantvaluesforcoe¤cientofvariation.TheTelekimorainesamplesgaveacoe¤cientofvariation0.24,whichcorrespondstoapparentage/trueageof0.805^0.895.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 uponthecoe¤cientofvariation,thesoilerosionrateandthebouldererosionrate.AcontourplotoftheresultsfortheTelekimoraineisshowninFig.6.Solidlinesrepresentthecoe¤cientofvar-iationanddashedlinesrepresentvaluesoftheratioofthetrueagetoapparentage(orconstantageratio).Theintersectionsofthecoe¤cientofvariationlinefortheTelekisamples(0.247)andthelinesofconstantageratioyieldarangeoftrueagesforallcombinationsofsoilandboulderero-sionrates.Foraparticularcoe¤cientofvaria-tion,thereisasmallrangeofsoilerosionratesassociatedwiththepossiblerangeoftrueagesforthelandform.Foroldlandforms,wereporttheageastherangeofpossibletrueages,andesti-matethesoilerosionrateusingtherangeofallpossiblebouldererosionratescorrespondingtoexperimentalcoe¤cientofvariation.Becausethismodelassumesthatsoilerosionrateisconstantintimeanduniformacrossthesamemoraine,deviationsfromtheseconditionsmayleadtoincorrectageestimates.Acarefulgeomorphologicalexaminationofthedepositsiscriticaltosuccessfulapplicationofthemodel.Samplingshouldavoidareaswithevidenceofnon-steady-stateprocessessuchasmeltwaterinci-sionordepositionofbouldersbysuccessivegla-cialadvances.Furthermore,themodelshouldnotbeappliedtodepositsthatmayhaveundergonedi¡erentialerosionalanddepositionalprocessesduringtheirhistory.Insuchcases,theobservedcoe¤cientofvariationmaybetoolargetobeexplainedbysteady-stateerosionandthereforecannotbemodeledusingthetechniquesdescribedabove(forexample,theLikiIImoraineinTeleki3.6.PriorexposureWecollectedfoursamplesofbouldersandbed-rockfromthepresentmarginofLewisGlacierand¢vesamplesfromtheLewismoraineinupperTelekiValleytoinvestigatethee¡ectofpriorex-posureonsurfaceexposureages.Historicalre- Fig.7.SummaryofapparentClagesfor(A)oldand(B)younglandformsonMt.KenyaandKilimanjaro.Filledcirclesrepresentagesforzeroerosion,un¢lledcirclesrepre-sentagescalculatedwith2mmkyrerosionrates.Foroldsamplesanalyzedwiththesoilerosionmodel,wereportarangeofages(solidlines).Foryoungsamples,wereportthemean(solidline)þ1(dashedlines)asthemostprobableEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 cordsfortheLewismoraineindicatethatitwasdepositedduringtheLittleIceAge[1].Boulderscollectedfromthepresenticemarginarethusex-pectedtohaveazeroexposureage.Withtheex-ceptionofTV97-61whichgaveanageof1200yr,theaverageageoftheLewismorainewas210þ90yr.Theicemarginalsohadoneoutlier,TV97-65,withanageof630yr;otherbouldersyieldedanageof60þ80yr.Eveninthecaseofananom-alouslyoldsample,suchasTV97-61,priorexpo-sureaddsonly1000yr,withinthetypicalanalyt-icaluncertaintyformostPleistocenesamples.Thee¡ectofpriorexposureonsampleagesissigni¢-cantforveryyoungsamples(tenstohundredsofyr)andlimitsourabilitytocalculatelandformagesinthistimerange.Foroldersamples(thou-sandsofyr),priorexposureshouldnotposeasigni¢cantproblem.4.ResultsThecalculatedsurfaceexposureagesforthebouldersusedinthisstudyaresummarizedinTables1and2andFig.7.Fullanalyticalresults,withindividualsampleproductionratesandchemistry,areavailableasanEPSLonlineBack-grounddataset(AppendixA).4.1.Mt.Kenya,TelekiValleyWeidenti¢edandsampledthreedistinctglaciallandformsinTelekiValley:theTeleki,theLikiIIIandtheLewismorainesof[1].Inaddition,wecollectedsamplesfromaLikiIImorainemappedby[1]andfromglaciallystriatedbedrockonPointLenana.OursamplinglocationsforthesemorainesareindicatedonFig.2.TheTelekimoraineispreservedasabroad,gentlyslopingridgewithseveralwell-roundedbouldersupto2.0mindiameter.Mostbouldersshowindicationsofgranularweatheringandspal-ling.TheageoftheTelekimorainewaspreviouslyestimatedat150^300kyr,baseduponthehighpercentageofclastsfromthecentralvolcanicplugpresentintheTelekideposits[1].Wecol-lectedsevensamplesfrombouldersontheleftlateralportionofthemorainenearthemeteoro-logicalstation.Usingthesoil^bouldererosionmodel,wecalculateaprobableagerangefortheTelekiglaciationof255^285kyr,inagreementwithpreviousestimates[1].Thisresultcorre-spondstoacoe¤cientofvariationof0.247andasoilerosionrateof15.1^17.8mmkyrSeveralcloselyspacedmoraineridgesat4000m,designatedtheLikiIIImorainesofMahaney[1],arelitteredwithbouldersupto1.0mindi-ameter.Bouldersurfacesarefreshandshownoevidenceoferosion.Previousradiocarbondatingofcoresfromglaciallydammedlakeshaspro-videdminimumagesforLikiIIIdeglaciationrangingfromca.8.5cal.kyrtoca.15cal.kyr[5](CagesconvertedtocalendaragesusingTaylor[43]).Surfaceexposureagesgiveanaver-agelandformageof10.2þ0.5kyrfortheLikiIIImoraines.SamplescollectedfromtheLikiIIIAmorainegiveanaverageageof9.9þ1.5kyr.ThewiderangeofagesfortheLikiIIIAcom-paredwiththeLikiIIImorainemayindicatethaterosionorpriorexposurea¡ectedboulderages.WiththeexceptionoftheanomalouslyyoungTV-54sample(5.9kyr),boulderagesfromthislandformfallintotwoagegroups:11.2þ0.8kyrand8.6þ0.2kyr.WesuggestthatperhapstheLikiIIIAlandformisacompositeofbouldersdepositedduringtheLikiIIIepisodeandbouldersdepositedduringasecondglacialad-vanceat8.6kyr.Furtherstudiesofthedepositsareneededtoaddressthisproblem.ThesurfacemappedbyMahaney[1]asLikiIIinTelekiValleydisplaysnoclearmoraineform(notrecognizableonaerialphotographs)andisconfusedbyoutcropsofbedrockobservedinthe¢eldandonaerialphotographs.PreviousstudiesonburiedpaleosolsassociatedwiththesedepositsindicatethattheLikiIIglaciationmayhavebe-gunaround30cal.kyr[4].Wecollected10sam-plesfromwell-rounded,butheavilyweatheredbouldersupto2.0mindiameter.Thesegivearangeofsurfaceexposureagesfrom19.0kyrto130kyr(Fig.7A).Thecoe¤cientofvariationandmeanapparentagesforthesebouldersare0.60 http//:www.elsevier.nl/locate/epsl;mirrorsite:http//:www.elsevier.com/locate/epslEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 and64kyr,respectively.However,suchhighvar-iabilitycannotbereproducedbytheconstantratesoilerosionmodel.Thisresultcanbeexplainedbyeitherthein£uenceofnon-uniformerosionalprocessesorthedepositionofboulderswithsig-ni¢cantpriorexposure.Possiblemechanismsin-cludemodi¢cationbymeltwaterincisionorre-workingbysubsequentglacialor£uvial^glacialprocesses.Theseprocessescanalsoexplainthelackofawell-de¢nedlandformshapeandthepresenceofbedrockoutcroppings.WeconcludeonlythattheageoftheLikiIIsurfacemustliebetween10.2kyr(LikiIIIintheuppervalley)and255^285kyr(Teleki).PolishedandstriatedbedrockonPointLenana(thethirdhighestsummitofMt.Kenya,4985m)indicatethatice£owedtotheeastandsoutheastdownGorgesandHobleyValleys[1].Thestriatedsurfacegivesanexposureageof4.1kyr.Becausethebedrockisstillcoveredbyloosematerialinmanyplaces,thisdateshouldbeconsideredaminimumdeglaciationageforPointLenana.4.2.Mt.Kenya,GorgesValleyOntheeasternsideofMt.Kenya,inGorgesValley(Fig.2d),weidenti¢edandsampledtwodistinctLikiIandLikiIImoraines[1].WealsocollectedsamplesfromtheLikiIIAmoraineloops,theNaroMorutill,theGorgesmoraineandfromLikiIIIbouldersandglaciallyscouredbedrock.TheLikiImorainehasawell-preservedshapewithsteepslopesthatcanbetracedeasilyinthe¢eldandonaerialphotographsinmanyvalleysabove3200m.Threeverylarge(2.0^3.0mdiam-eter)boulderswereidenti¢edontheLikiImor-ainecrestandsampled.Nopreviouschronologi-calcontrolisavailablefortheLikiIlandform.Usingthesoil^bouldererosionmodel,wecalcu-lateasoilerosionrateof5.2^6.7mmkyrestimateanageofbetween420and355kyrforthelandform.Thecoe¤cientofvariationforthesethreeboulders,0.135,issurprisinglysmallforalandformofsuchantiquity.Theerosionratesarelessthanhalfthoseestimatedforthewesternsideofthemountain.Thismaybeex-pected,giventhewell-preservedformoftheLikiImorainecomparedwiththesubduedtopogra-phyoftheTelekimoraine.Observedvegetationpatternscorrespondingtomoistandsemi-aridmi-croclimatesontheeastandwestsidesofthemountain,respectively,lendfurthersupporttothesecalculatederosionrates.TheLikiIImoraineinGorgesValley[1]ispreservedasasharp,well-de¢nedridgejustinsidetheLikiImoraineridge.Bouldersapproximately0.5^1.0mindiameterarefoundonthemorainecrest.Theyexhibitlimitedsignsoferosion.Weproposeanageof28þ3kyrforthelandformbasedonanaverageofthezeroerosionboulderages.Thisageagreeswellwithpreviousestimatesof30kyrfrompaleosolsonacorrelativemoraineinTelekiValley.LikiIIAmorainesarepreservedasseveralsmallloops,justinsidethelimitsofthemassiveLikiIImoraine.Theyareextremelybouldery,havelittletopographicreliefandareverywell-preserved.Samplesoftwobouldersyieldanaver-ageexposureageof14.6þ1.2kyr.AdditionalsampleswerecollectedfromlargebouldersandpolishedbedrocksurfacesjustbeyondtheLikiIIIlimitsinUpperGorgesValley.Thesesurfacesareabout5kmfromtheLikiIIAmorainesandgiveanaverageexposureageof13.0þ2kyr.Thetwobedrocksamplesgiveaslightlyolderageof17.4þ1.1kyrwhichweattributetoinsu¤cientremovalofpreviouslyexposedbedrockmaterial.Theagesoftheerraticsshouldrecordthemini-mumtimeofdeglaciationforthesurfaceonwhichtheyarelocated.Fromtheageoftheter-minalLikiIIAmoraine(14.6kyr),theageoftheglacialerratics(13.0kyr)andthedistancebe-tweenthetwo,wecalculateanaveragerateofglacierretreat(3myr)duringthelastdeglaci-ation.TheNaroMorutillispreservedasanisolateddepositcontaininglargeboulders(1.0^2.0mdi-ameter).Basedonpaleomagneticinformation,Mahaney[1]interpretedthistillastheremnantofaveryoldglaciation(730kyr).Exposureagesforeightbouldersrangefrom33kyrto101kyr.LiketheLikiIIdepositsinTelekiValley,thisdistributionofagesissurprisingforaland-forminthisagerange,especiallyinlightofitsapparentlyexcellentpreservation.InterpretationEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 ofthetillisfurthercomplicatedbyitsstrati-graphicrelationshiptotheLikiIandLikiIIde-posits.TheNaroMorutillislocatedonthetopofaridge,tensofmetersabovebothoftheseunits.Itseemsdi¤cultforicetohavedepositedtillonsuchahighridgewithoutobliteratingtheolderandsmallerLikiImoraine.Additional¢eldinves-tigationsareneededtoprovideamoreaccurateinterpretationofthislandformintermsofitsstratigraphyandchronology.TheGorgesmoraineslocatedat2850mhavebeeninterpretedastheoldestandmostextensiveglaciallandformpreservedonMt.Kenya[1].Baseduponlithologiccharacteristicsandremnantmagnetization,thesemorainesareconsideredtobeolderthan2.4Myr[1].However,thesharpreliefandsteepslopesoftheselandformsseemunlikelyforamoraineofsuchage.Wesampledtheonlyfourboulderswecould¢ndonthecrestofthelandform.Thesebouldershaverelativelyfreshsurfacesbutaresmall(lessthan0.5mindiameter).Thelandformyieldsanaverageexpo-sureageof6.5þ0.9kyr.Baseduponitsstrati-graphicposition,thislandformcannotbeofgla-cialoriginifthisageiscorrect.However,thepotentialforpriorsoilcoverofthesebouldersishighbecausetheyaresosmall.A2.5mdeepsoilpitdugintothecrestofthelandformshoweddepositsofwell-sorted¢nesandandsiltwithnogravel,cobbleorboulderfractionsandlittletonodevelopmentofsoilhorizons.Theseresultsfur-thersupportthehypothesisthattheGorgesland-formisnotglacial.4.3.Kilimanjaro,MawenziPeakOnthesouthwestslopesofMawenziPeakandthesoutheastedgeoftheSaddle,morainesas-cribedtotheFourthandtheLittleGlaciationswereidenti¢edandsampled.BoulderswerealsosampledfrominfrontoftheMainGlaciationmorainesandfromthepreviouslyreportedlimitsoftheThirdGlaciation(ca.3000m).MoraineandsamplelocationsareindicatedinFig.3b^d.TheThirdGlaciationisconsideredthemostextensiveglaciationonKilimanjaro,baseduponreportsofboulderbedsat4000mandglacialstriaeat3000m.Basedontheageofvolcanicclastspresentintheboulderbed,theglaciationisbelievedtohaveoccurredbetween120and150kyr[2].Weidenti¢edandsampledanumberofroundedbouldersthatweattributedtotheThirdGlaciationbasedupontheirpositionbe-yondtheFourthGlaciationlimits.Wealsocol-lectedaboulderandbedrockpairfrom3000m,thepreviouslyestimatedlimitoftheThirdGlaci-ation.Thebouldersrangedbetween0.5and3.0mdiameterandexhibitedsignsofintenseweatheringanderosion.Thebedrocksamplegivesanin¢niteageof660kyrandisassumedtohavepriorexposure.AswiththeLikiIIdeposits,thehighcoe¤cientofvariationfortheboulderages(0.49)indicatesthatextensivesoilerosionhasoccurredandcannotbemodeledaccurately.Wetentativelychoosethemaximumboulderage(360kyr)asaminimumageforthelandform.SincethisageisbeyondpreviousestimatesfortheThirdGlacia-tion,itispossiblethatwemayhavesampledbouldersfromanolderepisodesuchastheFirst(ca.500kyr)ortheSecond(360^240kyr).OnthesouthernslopesofMawenzi,morainesoftheMainGlaciationareupto30mhighandseveralkilometerslong.Numerouswell-roundedandwell-preservedbouldersupto3.0mindiam-eterlitterthecrestsofthemoraines.Previousat-temptstodatethesedepositshavefailedbecauseofthelackofdatablematerial[44].However,anestimateof10^70kyrhasbeenbasedonweath-eringandrelativeposition[2].Fiveboulderscol-lectedfromthesesurfacesyieldanaveragesurfaceexposureageof21þ4kyr,indicatingthatthesedepositswereemplacedatthelastglacialmaxi-mum(LGM).Althoughthestandarddeviationislarge,itisgreatlya¡ectedbyoneyoung(MP-90,16.4kyr)andoneold(MP-89,28kyr)outlier.Usingonlytheotherthreebouldersgivesasimilarmeanage(20kyr),butwithagreatlyreducedstandarddeviation(0.7kyr).AcomplexofmorainessouthwestofMawenziPeakhavebeenascribedtotheFourthandtheLittleGlaciations.BasedupontheirproximitytothemorainesoftheMainGlaciation,theageofthesedepositshasbeenpreviouslyestimatedat8^10kyr[2].Usingaerialphotographsandlimited¢eldinvestigations,thesemoraineswereinter-pretedasajuxtapositionoflandformsproducedEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 bythetwoglacialepisodes.ThetwooutermostlateralmoraineswereascribedtotheFourthGla-ciation.AseriesofsmallerrecessionalmorainesinsidethesewereattributedtotheLittleGlacia-tion.Allofthesurfacesareboulderyandap-pearedyoungandwell-preserved.WecollectedsamplesfromtheleftlateralmoraineandfromseveraloftherecessionalmorainesoftheLittleGlaciation.Thebouldersgiveasurprisinglylargespreadinsurfaceexposureages,consideringtheagesofthelandforms.Theaverageexposureagefortheoutermoraineis16.3þ1.9kyr.Theinnermorainesyieldanaverageageof15.8þ2.5kyr.Theseagesarestatisticallyindistinguishablebe-causeofthelargespreadintheboulderagesfromtheindividuallandforms.4.4.Kilimanjaro,KiboPeakOntheeasternslopesofKiboPeak,twodis-tinctmoraineloopswereidenti¢edinthe¢eldandonaerialphotographs.Bouldersappearedun-weathered,angularandrangefrom1.0to2.0mindiameter.Inplaces,boulderswereobserveddirectlyonstriatedbedrocksurfaces.Wecollectedfourbouldersamplesandthreestriatedbedrocksamples.Thebouldersyieldanaverageageof13.8þ2.3kyrandthebedrocksamplesgiveanolderageof23þ5kyr.Previousstudies[19,45]haveshownthatbedrocksurfacesarelikelytogiveolderapparentagesbecauseofpriorexpo-suretocosmicrays.Theboulderagesarethere-foreconsideredamorereliableestimateoftheageofthesurface.4.5.Kilimanjaro,SaddleWecollectedtwosamplesfromlargeboulderspresentontheSaddlebetweenMawenziandKiboPeaks.The¢rstboulder,whichwasapproxi-mately1.5mdiameterandonlyslightlyweath-ered,gaveanageof38kyr.Thesecondboulderwaslarger(3.0m)butheavilyweathered.Itgaveasigni¢cantlyyoungerageof6.9kyr.Thisresultwasdiscountedbecauseoftheevidenceofspallingoftheboulder.MajorelementanalysisofbothbouldersindicatesthattheywerederivedfromMawenziPeak.Weproposethatthisbouldertrainwasformedbyice£owingfromMawenziPeakbefore40kyrwhenthecoalescedKiboandMawenziicemassesbrokeupandtheSaddlewasdeglaciated.4.6.CalculatedsoilerosionratesforglacialThenewlydevelopedsoil^bouldererosionmod-elpermitstheestimationofsoilerosionratesforglaciallandforms.Table3displaysthesoilero-sionratescalculatedfortheLikiIandtheTelekimorainesonMt.Kenya.Theolder,butsteeper,LikiImoraineyieldssoilerosionratesthataretwicethoseoftheyounger,subduedTelekimor-aine.Theseresultsmayre£ectthewetterclimateinTelekiValley(Telekimoraine)onthewesternslopesofthemountaincomparedwiththearid,easternGorgesValley(LikiI).Thesepreliminaryestimateso¡erintriguingpossibilitiesforfutureapplicationsofcosmogenicisotopestotheratesoflandscapeevolution.5.SummaryandconclusionsNewcosmogenicClagesfortheglacialde-positsofMt.KenyaandKilimanjaropermitthedevelopmentofanimprovedchronologyforgla-ciationsinequatorialEastAfricaduringtheQua-ternary.Theseresultsprovideimportantnewin-formationaboutthetimingofclimatechangeinequatorialregionswhereterrestrialdataarescarce.Therevisedcompositechronologyofgla-ciationsfrombothmountainsisdisplayedinFig.ThemostextensiveglaciationinequatorialEastAfricaoccurredbefore360kyrago,baseduponcosmogenicagesofheavilyweathereddepositsonKilimanjaro.Althoughbedrockgivesanageof Table3ErosionratesofmorainecrestsMoraineAgeCoe¤cientofSoilerosionrate(kyr)variation(mmkyrTeleki254^2840.24715.1^17.8LikiI353^4190.1355.2^6.7EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 660kyr,itisnotclearthatthisbedrockagerep-resentstheageofaglacialepisodebecauseofcomplicationswitherosionandpriorexposure.TheoldestdepositssampledonMt.Kenya(355^420kyr)indicatethattheoldestglaciationsonbothmountainsmaybecorrelative.Thisagerangemaycorrespondtoeitheroxygenisotopestage(OIS)10or12dependingupontherateofboulderandsoilmatrixerosion.TheageoftheTelekideposits(255^285kyr)indicatesthattheycorrelatewithmarineOIS8.Interestingly,theelevationoftheTelekimoraineisapproximately500mlowerthantheolderLikiIdepositsontheeasternslopesofMt.Kenya.EitherthecorrelativeLikiIdepositsinTelekiValleywereatlowerelevationsonthewesternsideofMt.Kenyathanontheeasternside,andwerenotpreserved,orpaleoclimatechangesman-ifestedthemselvesdi¡erentlyontheeasternandwesternslopesofMt.Kenya.Fewterrestrialpaleoclimaterecordsextendasfarbackas400kyr.Forthisreasonalone,therecognitionofglacialdepositsinthetropicscor-relativewithOIS8and10isasigni¢cant¢nding.Combinedwiththeapparentlackofdepositscor-relativewithOIS6,thisresultisespeciallyintri-guing.Theabsenceofstage6depositsindicatesthateithertheywerenotdepositedatallorthattheywereobliteratedbymoreextensive,youngerglacialadvances.Thisisincontradictiontowhathastypicallybeeninferredabouttherelativeex-tentofstage6fromglacialdepositsathighlat-itudes[46^48].Furthermore,Oinmarinesedi-ments(aproxyforsealevelandicevolume)[49]inicecores(aproxyfortemperature)[50]bothindicatethatOIS6waslargerthanOIS8.However,therecordofOincalcitefromDevilsHole,NV,USA[51],appearstoagreewellwiththedatafromEastAfricanmoraines:OIS10islarge,OIS8isslightlysmallerandOIS6issigni¢cantlysmallerthanbothoftheseepi-sodes.Thesigni¢canceoftheseresultsisunclearinparticularbecauseofthelackofotherproxyrecordsforthistimeperiod.However,theymayindicatethatthelow-latitudeclimateresponseatthistimebecamedecoupledfromthatathigherlatitudes,aresultwhichmaybecriticaltoevalu-atingtherelativeimportanceofhigh-andlow-latitudesitesintheforcingofmajorclimateshifts.Clagesindicatethataseriesofyoungeradvancesoccurredat28þ3,20þ1,between15.8þ2.5and13.8þ2.8,10.4þ0.6andpossiblyat8.6þ0.5kyr.ThisrecordofglaciationsagreesstronglywiththeresultsofotherstudiesinEastAfrica.Reconstructionsoftemperatureandpre-cipitationforthelast40kyrusingpollenrecordsindicatethatAfricawascoldanddrybetweenca.35and15kyr[52].Sedimentaryanddiatomre-cordscon¢rmthatlakelevelswerelowduringthisinterval,withthedriestperiodoccurringataround21kyr[53],thetimewhenthemorainesoftheMainGlaciationweredepositedonKili-manjaro.Theendofthiscold,dryperiodwaspunctuatedbyasharpincreaseintemperatureandprecipitationtopresentvaluesataround15kyr[52].ThisageisinagreementwiththetimeofglacialretreatfromtheLGMglaciallimits(15.8þ2.5to13.8þ2.8),withintheuncertaintiesinboththepollenandglacialchronologies.Priortothisabruptclimateshift,thepollenrecordischaracterizedby3^4³C£uctuationsintempera-turearoundameanwhichwas4þ2³Clowerthantoday[52].Pollendataalsoindicatelargevariabilityinprecipitationduringthistime,whichiscorroboratedbythelakelevelrecordof£uctu-ationsbetweendryandwetperiodsatca.14^18 Fig.8.CompilationofcosmogenicClagesforEastAfricanglaciations.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 kyr[54,64].ThesevariationsintemperatureandprecipitationareconsistentwiththerecordofmultipleglacialadvancesandretreatsjustinsidetheLGMicelimit.Followingaperiodofhighlakelevels,anepi-sodeofstrongaridityisrecordedacrossEastAfricaat13^11cal.kyr[32,54^57],whichhasbeencorrelatedwiththeEuropeanYoungerDryascoldinterval.TheLikiIIIreadvance(10.2þ0.5)intheuppervalleysofMt.Kenyaap-pearstopost-datethisclimateepisode.However,correlationwiththeYoungerDryaseventcannotberuledoutuntilthesystematicerrorsinvolvedinourgeomagneticcorrectionarebetterunderstood.The¢nalglacialevent,at8.6þ0.5kyr,maycorrespondtoanearlyHolocenecoldeventre-cordedelsewherebetween8.4and8.0kyr[58,59].Ifaccurate,thisagewouldindicatethatglaciersonMt.Kenyarespondtocenturyscaleclimaticshifts.However,thisageisuncertainandstrongerevidenceisneededtocon¢rmthishy-pothesis.Theresultsofthisstudyprovideauniquere-cordofclimatechangeintropicalEastAfrica.Fewotherlong-termrecordsofclimatechangeinAfricahavebeenidenti¢ed.Mostoftherecordswhichextendfurtherthan100kyrarebaseduponpollenandmineralogicalstudiesofmarinecores[60^63].Theserecordsclearlyindicatethattheglacialperiodscorrespondingtoeven-numberedmarineOISswerecoolanddry,whereasthein-terglacialperiodswerewarmandhumid.Numer-ousshort-termrecordsofclimatederivedfromlakelevels[56]andfrompollen[52,64]con¢rmtheseresultsforthelast50kyr.Thenewchronol-ogyofglacial£uctuationsinequatorialEastAfri-calendsadditionalsupporttothevariabilityoftropicalclimatesduringtheQuaternaryandhintsatthepossibilityoflinksbetweenlow-andhigh-latitudeclimatechanges.AcknowledgementsWethankS.PorterandT.Swansonfortheirroleinthepreparationoftheresearchproposal,andT.Swansonandhisstudentsfortheircom-panyduringthe¢eldworkonMt.Kenya.Wethankallofthemfordiscussions.WealsothankDavidElmore,PankajSharmaandtheaccelera-torsta¡atPurdueUniversityformeasuringtheCl/Clinthesamples.WearegratefulforthecriticalreviewsandhelpfulsuggestionsprovidedbyFredPhillipsandErikBrown.ThisworkwassupportedbytheNationalScienceFoundation(researchGrantEAR-9632277toM.Z.andop-erationalGrantEAR-9809983toDavidElmore,PurdueUniversity),andbytheDavidandLucilePackardFoundation(FellowshipinScienceandEngineering951832toM.Z.)..).References[1]W.C.Mahaney,IceontheEquator:QuaternaryGeologyofMountKenya,EastAfrica,WmCaxton,1990,398pp.[2]C.DownieandP.Wilkinson,TheGeologyofKiliman-jaro,UniversityofShe¤eld,She¤eld,1972,253pp.[3]S.Hastenrath,TheGlaciersofEquatorialAfrica,Reidel,1984,353pp.[4]R.W.Barendregt,W.C.Mahaney,Paleomagnetismofse-lectedQuaternarysedimentsonMTKenya,EastAfricaareconnaissancestudy,J.Afr.EarthSci.7(1988)219^225.[5]W.C.Mahaney,HoloceneglaciationsandpaleoclimateofMountKenyaandotherEastAfricamountains,Quat.Sci.Rev.7(1988)211^225.[6]D.Elmore,F.M.Phillips,Acceleratormassspectrometryformeasurementoflong-livedradioisotopes,Science236(1987)543^550.[7]P.J.Aruscavage,E.Y.Campbell,Anion-selectiveelec-trodemethodfordeterminationofchlorineingeologicalmaterials,Talanta30(1983)745^749.[8]T.E.Cerling,H.Craig,Geomorphologyandin-situcos-mogenicisotopes,Annu.Rev.EarthPlanet.Sci.22(1994)[9]D.Lal,Cosmicraylabelingoferosionsurfacesinsitunuclideproductionratesanderosionmodels,EarthPlan-et.Sci.Lett.104(1991)424^439.[10]P.R.Bierman,K.A.Marsella,C.Patterson,P.T.Davis,M.Ca¡ee,Mid-Pleistocenecosmogenicminimum-agelim-itsforpre-WisconsinanglacialsurfacesinsouthwesternMinnesotaandsouthernBa¤nislandamultiplenuclideapproach,Geomorphology27(1^2)(1999)25^39.[11]E.J.Brook,M.D.Kurz,J.R.P.Ackert,ChronologyofTaylorGlacieradvancesinArenaValley,Antarctica,us-inginsitucosmogenicHeandBe,Quat.Res.39(1993)[12]E.J.Brook,E.T.Brown,M.D.Kurz,J.R.P.Ackert,G.M.Raisbeck,F.Yiou,Constraintsonage,erosion,andupliftofNeogeneglacialdepositsintheTransantarcticMoun-tainsdeterminedfrominsitucosmogenicBeandGeology23(12)(1995)1063^1066.EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 [13]J.C.Gosse,J.Klein,E.B.Evenson,B.Lawn,R.Middle-ton,Beryllium-10datingofthedurationandretreatofthelastPinedaleglacialsequence,Science268(1995)1329^[14]J.C.Gosse,E.B.Evenson,J.Klein,B.Lawn,R.Middle-ton,PrecisecosmogenicBemeasurementsinwesternNorthAmericaSupportforaglobalYoungerDryascool-ingevent,Geology23(1995)877^880.[15]F.M.Phillips,M.G.Zreda,S.S.Smith,D.Elmore,P.W.Kubik,P.Sharma,Acosmogenicchlorine-36chronologyforglacialdepositsatBloodyCanyon,EasternSierraNe-vada,California,Science248(1990)1529^1532.[16]F.M.Phillips,M.G.Zreda,M.A.Plummer,L.V.Benson,D.Elmore,P.Sharma,Chronologyfor£uctuationsinLatePleistoceneSierraNevadaglaciersandlakes,Science274(5288)(1996)749^751.[17]F.M.Phillips,M.G.Zreda,J.C.Gosse,J.Klein,E.B.Evenson,R.D.Hall,O.A.Chadwick,P.Sharma,Cosmo-ClandBeagesofQuaternaryglacialand£uvialdepositsoftheWindRiverRange,Wyoming,Geol.Soc.Am.Bull.109(11)(1997)1453^1463.[18]M.G.Zreda,F.M.Phillips,InsightsintoalpinemorainedevelopmentfromcosmogenicClbuildupdating,Geo-morphology14(2)(1995)149^156.[19]M.Zreda,J.England,F.Phillips,D.Elmore,P.Sharma,UnblockingoftheNaresStraitbyGreenlandandElles-mereice-sheetretreat10000yearsago,Nature398(1999)[20]M.G.ZredaandF.M.Phillips,Surfaceexposuredatingbycosmogenicchlorine-36accumulation,in:C.Beck(Ed.),DatinginExposedandSurfaceContexts,Univer-sityofNewMexicoPress,1994,pp.161^183.[21]B.Liu,F.M.Phillips,J.T.Fabryka-Martin,M.M.Fowler,R.S.Biddle,CosmogenicClaccumulationinunstablelandforms.1.E¡ectsofthethermalneutrondis-tribution,WaterResour.Res.30(11)(1994)3115^3125.[22]F.M.Phillips,M.A.Plummer,CHLOEAprogramforinterpretingin-situcosmogenicnuclidedataforsurfaceexposuredatinganderosionstudies,Radiocarbon38(1996)98.[23]F.M.Phillips,M.G.Zreda,M.R.Flinsch,D.Elmore,P.Sharma,AreevaluationofcosmogenicClproductionratesinterrestrialrocks,Geophys.Res.Lett.23(9)(1996)949^952.[24]T.Swanson,DeterminationoflproductionratesfromthedeglaciationhistoryofWhidbeyandFidalgoIslands,Washington,Radiocarbon38(1996)172.[25]J.O.Stone,G.L.Allan,L.K.Fi¢eld,R.G.Cresswell,Cos-mogenicchlorine-36fromcalciumspallation,Geochim.Cosmochim.Acta60(1996)679^692.[26]Y.Guyodo,J.P.Valet,Relativevariationsingeomagneticintensityfromsedimentaryrecords:Thepast200000years,EarthPlanet.Sci.Lett.143(1^4)(1996)23^36.[27]D.LalandB.Peters,Cosmicrayproducedradioactivityonearth,in:K.Sitte(Ed.),EncyclopediaofPhysics:CosmicRaysII,EncyclopediaofPhysics46/2,Springer-Verlag,Berlin,1967,pp.551^612.[28]W.Elsaesser,E.P.Ney,J.R.Winckler,Cosmic-rayinten-sityandgeomagnetism,Nature178(1956)1226^1227.[29]E.Tric,J.-P.Valet,P.Tucholka,M.Paterne,L.Labeyrie,F.Guichard,L.Tauxe,Paleointensityofthegeomagnetic¢eldduringthelast80000years,J.Geophys.Res.97(1991)9337^9351.[30]M.Frank,B.Schwarz,S.Baumann,P.W.Kubik,M.Suter,A.Mangini,A200kyrrecordofcosmogenicradio-nuclideproductionrateandgeomagnetic¢eldintensityfromBe-10ingloballystackeddeep-seasediments,EarthPlanet.Sci.Lett.149(1^4)(1997)121^129.[31]M.Stuiver,P.Reimer,E.Bard,G.Burr,K.Hughen,B.Kromer,G.McCormac,J.V.d.Plicht,INTCAL98radio-carbonagecalibration,24000^0calBP,Radiocarbon40(3)(1998)1041^1083.[32]C.Hillaire-Marcel,O.Carro,J.Casanova,CandTh/UdatingofPleistoceneandHolocenestromatolitesfromEastAfricanpaleolakes,Quat.Res.25(1986)312^329.[33]M.Servant,M.Fournier,J.Argollo,S.Servantvildary,F.Sylvestre,D.Wirrmann,J.P.Ybert,Thelastglacialinter-glacialtransitioninthesouthtropicalAndes(Bolivia)basedoncomparisonsoflacustrineandglacial£uctua-tions,ComptesRendusAcad.Sci.-SerieII320(8)(1995)729^736.[34]J.D.Clayton,C.M.Clapperton,Broadsynchronyofalate-glacialglacieradvanceandthehighstandofpalaeo-lakeTaucaintheBolivianAltiplano,J.Quat.Sci.12(3)(1997)169^182.[35]M.D.Kurz,D.Colodner,T.W.Trull,R.B.Moore,K.O'Brien,Cosmicrayexposuredatingwithinsitupro-ducedcosmogenicHe:Resultsfromyounglava£ows,EarthPlanet.Sci.Lett.97(1990)177^189.[36]P.Sarda,T.Staudacher,C.J.Allegre,A.Lecomte,Cos-mogenicneonandheliumatReunionmeasurementoferosionrate,EarthPlanet.Sci.Lett.119(1993)405^[37]J.E.T.Channell,D.A.Hodell,B.Lehman,Relativegeo-magneticpaleointensityandOatODPSite983(Gar-darDrift,NorthAtlantic)since350ka,EarthPlanet.Sci.Lett.153(1997)103^118.[38]J.Brassart,E.Tric,J.P.Valet,E.Herrero-Bevera,Abso-lutepaleointensitybetween60and400kafromtheKo-halamountain(Hawaii),EarthPlanet.Sci.Lett.148(1997)141^156.[39]R.Braucher,F.Colin,E.Brown,D.Bourles,O.Bamba,G.Raisbeck,F.Yiou,J.Koud,Africanlateritedynamicsusinginsitu-producedBe-10,Geochim.Cosmochim.Acta62(9)(1998)1501^1507.[40]H.Cockburn,M.Seidl,M.Summer¢eld,Qualifyingde-nudationratesoninselbergsinthecentralNambinDesertusinginsitu-producedcosmogenicBe-10andAl-26,Geol-ogy27(5)(1999)399^402.[41]R.Braucher,D.Bourles,F.Colin,E.Brown,B.Bou-lange,Brazilianlateritedynamicsusinginsitu-producedBe-10,EarthPlanet.Sci.Lett.163(1^4)(1998)197^205.[42]M.G.Zreda,F.M.Phillips,D.Elmore,Cosmogenicaccumulationinunstablelandforms,2.SimulationsandEPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42 measurementsonerodingmoraines,WaterResour.Res.30(1994)3127^3136.[43]R.E.Taylor,M.Stuiver,P.J.Reimer,Developmentandextensionofthecalibrationoftheradiocarbontimescalearcheologicalapplications,Quat.Sci.Rev.15(1996)655^[44]H.Osmaston,Glaciers,glaciationandequilibriumlinealtitudesonKilmanjaro,in:W.C.Mahaney(Ed.),Qua-ternaryandEnvironmentalResearchonEastAfricanMountains,A.A.Balkema,Rotterdam,1989,p.483.[45]J.P.Briner,T.W.Swanson,UsinginheritedcosmogenicCltoconstrainglacialerosionratesoftheCordilleranicesheet,Geology26(1998)3^6.[46]T.Nilsson,ThePleistocene:GeologyandLifeintheQua-ternary,FerdinandEnkeVerlag,Stuttgart,1983,651pp.[47]J.Ehlers,QuaternaryandGlacialGeology,Wiley,Chi-chester,1996,578pp.[48]R.F.Flint,GlacialandQuaternaryGeology,JohnWileyandSons,NewYork,1971,892pp.[49]N.J.Shackleton,Oxygenisotopes,icevolumeandsea-level,Quat.Sci.Rev.6(3^4)(1987)183^190.[50]J.R.Petit,J.Jouzel,D.Raynaud,N.I.Barkov,J.M.Bar-nola,I.Basile,M.Bender,J.Chappellaz,M.Davis,G.Delaygue,M.Delmotte,V.M.Kotlyatov,M.Legrand,V.Y.Lipenkov,C.Lorius,L.Pepin,C.Ritz,Climateandatmospherichistoryofthepast420000yearsfromtheVostokicecore,Antarctica,Nature399(1999)429^[51]I.J.Winograd,J.M.Landwehr,K.R.Ludwig,T.B.Co-plen,A.C.R.AC,Durationandstructureofthepastfourinterglaciations,Quat.Res.48(2)(1997)141^154.[52]R.Bonne¢lle,J.C.Roeland,J.Guiot,Temperatureandrainfallestimatesforthepast40000yearsinequatorialAfrica,Nature346(1990)347^349.[53]F.Gasse,V.Lee,M.Massault,J.C.Fontes,Water-level£uctuationsofLakeTanganyikainphasewithoceanicchangesduringthelastglaciationanddeglaciation,Na-ture342(6245)(1989)57^59.[54]R.Bonne¢lle,G.Riollet,G.Buchet,M.Icole,R.Lafont,M.Arnold,D.Jolly,Glacial/interglacialrecordfromin-tertropicalAfrica,highresolutionpollenandcarbondataatRusaka,Burundi,Quat.Sci.Rev.14(1995)917^936.[55]M.Servant,J.Maley,B.Turcq,M.L.Absy,P.Brenac,M.Fournier,M.P.Ledru,TropicalforestchangesduringthelateQuaternaryinAfricanandSouthAmericanlow-lands,Glob.Planet.Change7(1993)25^40.[56]N.Roberts,M.Taieb,P.Barker,B.Damnati,M.Icole,D.Williamson,TimingoftheYoungerDryaseventinEastAfricafromlake-levelchanges,Nature366(6451)(1993)146^148.[57]F.A.Street-PerrottandN.Roberts,Fluctuationsinclosed-basinlakesasanindicatorofpastatmosphericcirculationpatterns,in:A.Street-Perrott,M.BeranandR.Ratcli¡e(Eds.),VariationsintheGlobalWaterBudg-et,D.Reidel,Boston,1983,pp.331^345.[58]R.B.Alley,P.A.Mayewski,T.Sowers,M.Stuiver,K.C.Taylor,P.U.Clark,Holoceneclimaticinstability:aprom-inent,widespreadevent8200yrago,Geology25(6)(1997)483^486.[59]S.R.O'Brien,P.A.Mayewski,L.D.Meeker,M.S.Twick-ler,S.I.Whitlow,ComplexityofHoloceneclimateasre-constructedfromaGreenlandicecore,Science270(1995)1962^1964.[60]S.Jahns,M.Huls,M.Sarnthein,VegetationandclimatehistoryofwestequatorialAfricabasedonamarinepollenrecordo¡Liberia(siteGIK16776)coveringthelast400000years,Rev.Palaeobot.Palynol.102(3^4)(1998)[61]F.X.Gingele,P.M.Muller,R.R.Schneider,Orbitalforc-ingoffreshwaterinputintheZaireFanarea-claymineralevidencefromthelast200kyr,Palaeogeogr.Palaeoclima-tol.Palaeoecol.138(1^4)(1998)17^26.[62]T.C.Partrige,WarmingphasesinSouthernAfricaduringthelast150000yearsanoverview,Palaeogeogr.Palaeo-climatol.Palaeoecol.101(1993)237^244.[63]H.Hooghiemstra,H.Stalling,C.O.C.Agwu,L.M.Du-pont,VegetationalandclimaticchangesatthenorthernfringeoftheSahara250000^5000yearsBP:evidencefrom4marinepollenrecordslocatedbetweenPortugalandtheCanaryIslands,Rev.Palaeobot.Palynol.74(1992)1^53.[64]A.Vincens,F.Chalie,R.Bonne¢lle,J.Guilot,J.-J.Tier-celin,Pollen-derivedrainfallandtemperatureestimatesfromLakeTanganyikaandtheirimplicationfortheLatePleistocenewaterlevels,Quat.Res.40(1993)343^EPSL538927-3-00T.M.Shanahan,M.Zreda/EarthandPlanetaryScienceLetters177(2000)23^42