climate variability and ice sheet dynamics during the last three glaciations

pachy-derma [s] δ18O purple, E>150 μm lithic grain concentration black, F detrital carbonate DC; orange, G Icelandic glass percentage IG; blue, H hematite-stained grain percentage HSG; r

Earth and Planetary Science Letters 406 (2014) 198–212

Contents lists available atScienceDirect

www.elsevier.com/locate/epsl

glaciations

aAtmosphere and Ocean Research Institute, University of Tokyo, 277-8564, Japan

bBraeheads Institute, Scotland EH40 3DH, UK

cDepartment of Geological Sciences, University of Florida, 32611, USA

dDepartment of Earth Sciences, University of Cambridge, CB2 3EQ, UK

eDivision of Earth and Ocean Sciences, Duke University, 27708, USA

a r t i c l e i n f o a b s t r a c t

Article history:

Received 30 November 2013

Received in revised form 24 August 2014

Accepted 3 September 2014

Available online 6 October 2014

Editor: J Lynch-Stieglitz

Keywords:

insolation

IRD

Heinrich Event

MIS 6

MIS 8

AMOC

A composite NorthAtlantic recordfromDSDP Site 609and IODP Site U1308spansthe past300,000 years and shows that variability withinthe penultimateglaciation differed substantially from thatof thesurroundingtwoglaciations.Hematite-stainedgrainsexhibitsimilarrepetitivedown-corevariations withintheMarineIsotopeStage(MIS)8and4–2intervals,butlittlecyclicvariabilitywithintheMIS 6 section There isalsonopetrologic evidence,intermsof detrital carbonate-rich(Heinrich)layers,for surgingoftheLaurentideIceSheetthroughtheHudsonStraitduringMIS 6.Rather,veryhighbackground concentrationoficeberg-rafteddebris(IRD)indicatesnearcontinuousglacialmeltwaterinputthatlikely increased thermohaline disruption sensitivity to relativelyweak forcingevents, suchas expanded sea ice overdeepwaterformationsites Altered(sub)tropical precipitationpatterns andAntarctic warming during highorbitalprecessionand low 65◦Nsummer insolationappear related tohigh abundanceof Icelandicglassshardsandsouthwardseaiceexpansion.DifferingEuropeanandNorthAmericanicesheet configurations,perhapsaidedbylargervariationsineccentricityleadingtocoolersummers,mayhave contributedtotherelativestabilityoftheLaurentideIceSheetintheHudsonStraitregionduringMIS 6

©2014TheAuthors.PublishedbyElsevierB.V.ThisisanopenaccessarticleundertheCCBYlicense

1 Introduction

Climate wasrelatively unstable during thelast glaciation,

Ma-rineIsotopeStages(MIS)4–2(e.g.,Dansgaardetal.,1993).Global

variationandinter-hemisphericcouplingwere linkedto

reorgani-zationsinNorthAtlanticthermohalinecirculationthatareinturn

relatedprimarilytoicesheetinstabilityonmultiplescales

Hydro-graphic changesresulting fromfreshwater forcing during Hudson

Strait (HS) Heinrich Events (Heinrich, 1988; Bond et al., 1992;

Hemming,2004) decreasedNorthAtlanticDeepwater(NADW)

pro-duction,reducing northwardheatflowandwarmingtheSouthern

Hemisphere (Crowley, 1992) Similarly, instability and freshwater

forcing on millennial scales appear to play a role in Dansgaard–

Oeschger(D–O)variability(e.g.,Menvieletal.,2014)

* Corresponding author.

E-mail address:obrochta@aori.u-tokyo.ac.jp (S.P Obrochta).

1 Deceased.

Pioneering sedimentological studies, such as those conducted

atclassicDSDPSite609fromthecentralsub-polarNorthAtlantic Ocean (Bond et al., 1992, 1993; Bond and Lotti, 1995), provided thefirstevidencethatthehigh-amplitude,rapidclimateshiftsfirst observedforthelastglaciationinGreenlandicecoresprobably af-fected a much wider area This work substantially advanced our understanding of Northern Hemisphere ice sheet instability and the origin of rapid climate change, as well as stimulated a vast field ofinterdisciplinary research However,mostobservations on this importantsubject are limited to thismost recentglaciation, mainly due to the shallow depths below seafloor of last-glacial sedimentarysequencesandhencetheiraccessibility.Muchremains

to be learnedregarding its nature andforcing by further under-standing the longer-termbehavior ofice-sheet oceaninteractions underdiffering boundaryconditionsduringmultipleglacial “real-izations”

The few recordsoficeberg-rafteddebris (IRD) variabilityfrom earliericeagestendtoindicatesignificantdifferencesbetweenthe lastandpenultimateglaciation(MIS 6).Forexample,a 500-ky IRD recordfromtheFeniDrift(ODP980)intheeasternNorthAtlantic http://dx.doi.org/10.1016/j.epsl.2014.09.004

0012-821X/©2014 The Authors Published by Elsevier B.V This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/3.0/ ).

Fig 1 SiteU1308/609 (49◦53N, 24◦14W; Expedition303 Scientists, 2006 ) is represented by a red star Other sites discussed in the text are: 1) the North GRIP ice core (75◦6N, 42◦20W; Andersenet al., 2004 ), 2) M23414 (55◦32N, 20◦17W; Didiéand Bauch, 2000 ), 3) ODP Site 980 (55◦29N, 14◦42W; Oppoet al., 2006 ), 4) V23-81 (54◦15N, 16◦50W; Bondet al., 1992 ), 5) Site U1304 (53◦3N, 33◦32W; Expedition303 Scientists, 2006 ), 6), Sites U1302 and U1303 (50◦10N, 45◦34W; Expedition

303 Scientists, 2006 ), 7) V28-82 (49◦27N, 22◦16W; Bondet al., 1992 ), 8) MD95-2040 (40◦35N, 9◦52W; de Abreuet al., 2003 ), 9) MD01-2443 (37◦53N, 10◦11W) and MD01-2444 (37◦34N, 10◦09W) ( Martrat et al., 2007 ), 10) MD03-2707 (2◦30N, 9◦24W; Weldeabet al., 2007 ), 11) CDH-86 (0◦20N, 44◦13W; Naceet al., 2014 ), 12) El Condor Cave (5◦56S, 77◦18W) and Cueva del Diamante (5◦44S, 77◦30W) ( Cheng et al., 2013 ), 13) Pacupahuain Cave (11◦14S, 75◦24W; Kanneret al., 2012 ), 14) Bahia state (10◦10S, 40◦50W; Wanget al., 2004 ), 15) TN057-14PC (51◦59S, 4◦31E; Andersonet al., 2009 ), 16) Hulu Cave (32◦30N, 119◦10E; Wanget al., 2001 ), 17) Sanbao Cave (31◦40N, 110◦26E; Wanget al., 2008 ) and Linzhu Cave (31◦31N, 110◦19E; Chenget al., 2009 ), 18) Yangkou Cave (29◦2N, 107◦11E; Liet al., 2014 ), 19) GeoB 10053-7 (8◦41S, 112◦52E; Mohtadiet al., 2011 ), 20) Lynch’s Crater (17◦22S, 145◦42E; Mulleret al., 2008 ), and 21) the Epica Dome C ice core (75◦06S, 123◦21E; Jouzel

et al., 2007 ).

(McManusetal., 1999) indicatesthatIRDvariabilityduringMIS 6

wastheleastamongthe pastfiveglaciations(CrowleyandHyde,

2008) Farthersouth,Iberian MargincoreMD95-2040alsoshows

similar results(de Abreu etal., 2003), andrecent proxies forHS

HeinrichEventssuggestnosurgingoftheLaurentideIceSheet(LIS)

(HodellandCurtis, 2008; Hodelletal.,2008; Ji etal., 2009; Stein

etal.,2009; Naafsetal.,2011; Channelletal.,2012; Channelland

Hodell,2013; Naafs etal.,2013).However, other records,suchas

that of coreM23124 fromthe Rockall Plateau (Didié andBauch,

2000),exhibitrelativelyincreasedIRDconcentrationduringMIS 6,

raisingthepossibilityofalteredcirculationpatterns

To further explore the potential effects of differing boundary

conditionsduring previousglaciations,wereportadetailed,

high-resolutionlithicrecordfromIODPSiteU1308(Fig 1),a

reoccupa-tionofSite 609,that includestheprevious twoglaciations,MIS 6

and8.ToallowdirectcomparisontotheMIS4–2petrologictracer

records from this same location (Bond et al., 1999) (Fig 2), we

employ similar methods (Bond et al., 1997) and determine the

abundance of hematite-stained quartz and feldspar grains (HSG)

primarilyoriginatingfromtheGulf ofSt Lawrence;dolomite-rich

detritalcarbonate(DC)sourcedfromtheHudsonStrait;and

non-weathered volcanic glass derived from Iceland (IG) These data

supplement bulk lithic content data and allow us to infer lithic

grainprovenanceanddeliverymechanism(i.e.,calvedicebergs

ver-susseaice).Additionalpetrologicinformationfromearliericeages

will allow us to assess the validity of Heinrich-like layer

prox-ies beyond the last glaciation A detailed record from MIS 6 in

particularwillhighlightotherdetrital constituentstowhichthese

proxiesmaybesensitive,whilealsopotentiallyidentifying

impor-tant drivers of climate change in the apparent absence of major

LIS surging Given the lack of an ice-core record prior to MIS 5

onGreenland, thenature ofHSG variabilityinmarine cores

dur-ingglacialintervalsbeyondthelastoneisimportanttodetermine whethersimilarconditionsexisted

2 Background

2.1 Study area

Sites609andU1308wereoccupied20yearsapartatthesame location on the upper-middle eastern flank of the Mid-Atlantic Ridge, approximately 250 km south of the Charlie Gibbs Frac-tureZone,betweenthemodern NorthAtlanticsubpolarand sub-tropical gyres (49◦53N,24◦14W,3871 mbsl;Fig 1). During the

last glaciation, the Polar Front was displaced southward of its modern location, creating steep SST gradients in this area (e.g.,

CLIMAP Project Members, 1976) and resulting in a concentrated zone of iceberg melting and IRD deposition, i.e., the “IRD Belt”

ofRuddiman (1977).Site609recordedpronouncedsurface hydro-graphicchangesthatwereinterpretedtoreflectstadial–interstadial swingsinthepositionofthePolarFront(Bondetal.,1993).Thus, Sites 609/U1308 are strategically positioned to record changing climate conditions in the North Atlantic Basin and surrounding continental regions The rationale for drilling Site U1308 was to recoverademonstrablycompletestratigraphicsectionusing mod-erncoringmethodstoreplacetheremainingmaterialfromtheSite

609coresthatisnowinpoorcondition(Expedition303 Scientists,

2006)

2.2 Last glacial variability at Site 609

Results from classic DSDP Site 609 (Fig 2) were particularly important to the development of the current paradigm view of millennial climate variability This site preserveda record of DC

200 S.P Obrochta et al / Earth and Planetary Science Letters 406 (2014) 198–212

Fig 2 Site609 record from MIS 4–2 ( Bond et al., 1999 ) with radiocarbon dates (red triangles) and GICC05 model-extended tie points (red squares) ( Obrochta et al., 2012 ) A) North GRIP ice coreδ18 O ( Andersen et al., 2004 ),B) N pachyderma [s]percentage (green), C) Site U1308 benthicC wuellerstorfi δ13 C (gray; Hodellet al., 2008 ),D), N

pachy-derma [s] δ18O (purple), E)>150 μm lithic grain concentration (black), F) detrital carbonate (DC; orange), G) Icelandic glass percentage (IG; blue), H) hematite-stained grain percentage (HSG; red), and I) EPICA Dome C ice coreδD ( Jouzel et al., 2007 ) using the AICC2012 chronology ( Veres et al., 2013 ) DC, IG, and HSG are percentages of 63–150 μm lithic fraction Orange vertical bars denote Heinrich Events, and blue bars non-Heinrich cold stadials (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

IRDdepositedduringHeinrichEventsH1toH6(Bondetal.,1992),

and results from Site 609 were the first to demonstrate a link

betweenNorthAtlanticseasurface temperature(SST) and

Green-landairtemperature(Bondetal.,1993),leadingtotheobservation

that progressively cooler Greenland insterstadials were bundled

between Heinrich Events (“Bond Cycle”) Results from Site 609

furtherdemonstratedthateachoftheseHeinrichEventswas pre-ceded by increased lithic flux withhighproportions of HSGand fresh,unweatheredIG(BondandLotti,1995; Bondetal.,1999)

AtSite609,peaklithicgrainconcentrationsoccurwithin inter-vals ofhighabundanceoftheplankticpolarforaminiferN pachy-derma [s], indicating cold surface conditions As lithic content

Fig 3 Site609 and Hole U1308A were aligned to the U1308 post-cruise, modified composite splice using sediment lightness(L∗ records.L∗was calculated from the core photograph for Site 609 (blue) and from the U1308 shipboard line-scanner images for the modified splice (red) and U1308A (black) The age model used in this study is based on alignment of U1308 benthicC wuellerstorfi δ18O (purple) to the LR04 benthic isotope stack (green; Lisieckiand Raymo, 2005 ) from MIS 9 to late MIS 5 (top) The younger interval uses the latest Site 609 age model ( Obrochta et al., 2012 ) consisting of recalibrated radiocarbon dates (triangles) and GICC05 ages (squares) Modified U1308 splice, benthicδ18 O, LR04 age tie points (red circles), and 609 to U1308 depth match points are from Hodellet al (2008) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

increases, planktic δ18O values begin to decrease primarily due

to the effects of isotopically light meltwater (but also

poten-tially due to either contamination by fine DC, as pointed out

sug-gestedbyHillaire-Marcelandde Vernal (2008)).Concurrently,

ben-thicforaminifer δ13C exhibitsdepletedvalues,whichisconsistent

withreducedNADWformationcompensatedby encroachmentof

southern-sourcedwatersintotheNorthAtlanticbasin

The Site 609 HSG record was central to the theory that a

“1500-yr” pacing underlies andsets the D–OEvent tempo(Bond

etal.,1997) andtriggered muchdiscussion ofperiodicsuborbital

climatechange Subsequent work to modernize the Site 609age

model (using the Marine09 radiocarbon calibration and GICC05

model-extendedchronology;Fig 2A,2B)showedthatHSGatthis

siteisnotcharacterizedbyasingle1500-yrcyclebutismore

con-sistent withdistinct 1000and 2000-yr components(Obrochta et

al.,2012)

3 Materials and methods

3.1 Stratigraphic correlation

The uppermost ∼8 m of Site 609 (representing the last

∼115 ka)hasbeensynchronizedtotheSiteU1308modified

com-positesplice(Hodelletal.,2008),asshowninFig 3.Wecalculated

colorreflectance(e.g., L∗a∗b∗; Ortizetal., 1999) valuesfromred,

green, andblue (RGB)channel data extracted from the Site 609

corephotograph(Ruddimanetal.,1987) andtheSiteU1308

ship-boardcore line-scanner images (Expedition303 Scientists, 2006)

along the modified composite splice (0–21.56 mcd), as well as

fromCoreU1308-2Hovertheinterval 11.94to18.37 mbsf,which

is not in the Site U1308 composite splice but preserves a

sedi-mentaryrecordspanningearlyMIS 7tolateMIS 9.Theshipboard

spectrophotometercolorreflectance data (2-cmresolution)

corre-late well with the line-scanner color reflectance data used here

(L∗R2=0.95; Supplemental Fig 1) The much higher resolution

(0.01 cm)line-scanner reflectancedatawere requiredtoprecisely alignCoreU1308A-2Htothemodifiedcompositesplice

3.2 Age model

The Site 609 age control points (Fig 2) were transferred to Site U1308(Fig 3), incorporatingrecentrefinements(Obrochta et al., 2012) These improvements are 1) inclusion of eleven addi-tional radiocarbon datesthat were previously uncalibrated (Bond

etal.,1993; Elliotetal.,1998),2)usageoftheMarine09 radiocar-boncalibrationcurve(Reimeretal.,2009) withaconstant 405 yr reservoircorrection,and3)optimizedcorrelationofthe%N pachy-derma [s] ties(Bond etal., 1999) tothe virtually-complete North GRIP (NGRIP) icecorewiththe GICC05 model-extended chronol-ogy (Rasmussen et al., 2006; Svensson et al., 2008; Wolff et al.,

2010) Our age model uses radiocarbon dates from 13 to 34 ka, tiestoNGRIPfrom35to76 ka,andcorrelationoftheU1308 ben-thic δ18O recordtotheLR04 δ18O stack(LisieckiandRaymo,2005) from86to325 ka.TheLR04tiepointsusedherearefromHodell

etal (2008) The resulting average sedimentation rates are 7.1 cm/ky for MIS 4–2,4.4 cm/ky forMIS 6, and7.3 cm/kyfor MIS 8.At 1-cm sampling,thisproduces anaverageresolutionof ∼180, 230, and

150 yr,respectively.Theresultsofspectralanalysisare,ofcourse, highlydependentonthequalityoftheagemodels

3.3 Grain abundance analysis

The abundance of HSG, IG, DC, N pachyderma [s], and bulk lithics (Fig 4) was determined from U1308 material within the MIS-6and8intervals.Toperformthesecountsthemodifiedsplice was sampled by taking half-round slices (∼10 cc) at a 1-cm in-terval from8.84 mcdto 11.55 mcd,corresponding toMIS 6 Core U1308A-2Hwassampledwith5 ccscoopsa1-cmresolutionfrom 12.88 mbsfto 17.62 mbsf(14.94–20.05 mcd).Bulklithic,DC,and foraminifer counts were performed at an average 2-cm interval

202 S.P Obrochta et al / Earth and Planetary Science Letters 406 (2014) 198–212

Fig 4 Photomicrographsof the primary sediment types used in the study From grain-mount slides (63–150 μm) are A) a hematite-stained quartz grain and B) an Icelandic glass shard, and from loose sediment (>150 μm) are C) detrital carbonate, andD) N pachyderma [s].

on loose sediment from the >150 μm fraction using a reflected

light microscope HSG and IG counts were performed at a 1-cm

intervalonthe63–150 μm size fractionusinggrain mountslides

Accuratecounts ofHSG in theMIS-6 interval required carbonate

digestiondue tothe increased abundance offoraminifers

There-fore63–150 μm DCcountsarelimitedtotheMIS-8interval

Because sample weights are not available for Core U1308-2H,

grainconcentrationforkeyintervalsisestimatedbasedonthe

vol-umeandweightofsamplesfromcorrespondingcompositedepths

in the modified splice These are ∼20-cc, half-round continuous

2-cmslices

Countsofthe63–150 μm fractionwere performedwitha

pet-rographicmicroscope,asdescribedbyBondetal (1997).The

sam-plewasilluminatedwithanexternalfiberopticlightsourcewitha

whitereflector placedoverthe microscope’scondenser Adjusting

the height of the reflector relative to the stage allows for a

po-sitionto “be found thatcreates astrong impressionof relief and

brings into striking view details of surface textures and coatings

ongrains”(Bondetal.,1997)

3.4 Spectral analysis

SpectralanalysisofHSGwasperformedby multitapermethod

(MTM) usingthe software package SSA-MTM Toolkit (Ghilet al.,

2002) HSG was first interpolated to an even 240-yr and 150-yr

stepforMIS 6and8,respectively.MTManalysisofMIS 8HSGwas

limitedtotheintervalfrom300to260 ka

3.5 Elemental analysis

A total of 64 individual DC grainsfrom MIS 8 and4–2 were

measured forMg/Ca and Sr/Caratios to determine whetherthey

appearderived fromthe samepopulation andtherefore likely of

similarprovenance.GrainsfromtheMIS 8intervalweremeasured

by laser ablation inductively coupled plasma mass spectrometry

(LA-ICP-MS)attheUniversityofTokyoAtmosphereandOcean Re-search Institute.Grains fromthe MIS4–2intervalwere measured

byinductivelycoupledplasmaatomicemissionspectrometry (ICP-AES)attheUniversityofCambridgeDepartmentofEarthSciences

3.5.1 ICP-AES

Five individual DC grains were picked from each of the H1, H2, H4,and H5intervals in the U1308 splicefor analysisat the University of Cambridge, Department of Earth Sciences using a Vista ICP-AES For H1,H2, H4, and H5,grains were picked from samples U1308C-1H-1W 92–94 cm (0.92 mcd), U1308C-1H-1W 130–132 cm (1.30 mcd),U1308C-1H-2W116–118 cm (2.66 mcd), and 1308E-1H-2W 70–72 cm (3.52 mcd), respectively Prior to analysis, sampleswere dilutedin0.1MHNO3 toobtainaMg con-centration of ∼0.3 ppm Calibration of calcium, magnesium, and strontiumwas performedwithmixedstandard solutions(Greaves

etal.,2005)

3.5.2 LA-ICP-MS

Individual DC grains (Supplemental Fig 2) were picked from each of the three MIS-8 DC deposition events (labeled 8.1, 8.2,

us-ing a Resonetics Resolution M-50 ablation system attached to a ThermoElementXRICP-MS.Forthe8.1event(∼243 ka)15grains were picked from sample U1308A-2H-4W 8–9 cm (13.18 mbsf) For the 8.2 event (∼249 ka), 9 grains were picked from sam-ple U1308A-2H-4W 54–55 cm (13.64), 3 grains from 57–58 cm (13.67 mbsf), and4 grainsfrom 58–59 cm (13.68 mbsf).For the 8.3Event(∼263 ka), 13grainswerepicked fromU1308A-2H-5W 4–5 cm(14.64 mbsf)

Laser energy was 50 mJ with a spot beam Beam width was setto155 μm formostanalysesbutreducedto112 μm forgrains lessthan∼500 μm.Beamwidthwaskeptconstantduring individ-ual runs Continuous ICP-MS measurements were made foreach run using triple detectionand low resolution settings, and mass

offset was determined at the start of each day of analyses Up

to10individual grainsofsimilar sizewere analyzed during each

run withconstant beamsize Priorto grain ablation,background

valuesweremeasured for ∼3 min,andstandardsJCp-1and

NIST-614 were measured along a line for 5 min each at a speed of

5 μm/s Grains were held stationaryand ablatedfor 120 s, with

60 sofequilibriumtime betweeneach grainmeasurement

Mea-suredintensities werebackground subtractedanddrift corrected

The mean intensity of 24Mg, 43Ca, and 88Sr was calculated for

each grain after discarding the initial and final ∼2 s When the

laser fullybroke through a grain, the corresponding values were

discarded.The meanablationtime fromwhichintensitywas

cal-culatedis96 switha rangeof46to116 s.NIST-614wasusedto

calibratetheresultspresentedhere.CalibrationwithJCp-1yielded

similarresults

One grain with moderate Mg/Ca ratios (0.42 mol/mol) was

selected from the 8.2 DC event from sample U1308A-2H-4W

57–58 cm for higher resolution measurements Twenty-nine

in-dividualmeasurements were performedacross thegrain’s surface

witha24 μm spotbeamusingthesameparametersasdescribed

above

3.6 Stable isotopes

Stable isotopes were measured on the planktic foraminifer

N pachyderma [s] following the method of Hodell et al (2008)

Specimenswerepickedfromthe >212-μm sizefractionata2-cm

spacingwithin the MIS-6 and8 intervalsof theU1308 modified

splice Analyseswere performed usinga Finnigan MAT 252mass

spectrometerattheUniversity ofFlorida,Departmentof

Geologi-calSciences

4 Results

NewgrainabundancedataandplankticN pachyderma [s] δ18O

fromSite U1308areshowninFig 5togetherwiththepreviously

publishedbenthicC wuellerstorfi δ13C record(Hodell etal.,2008)

Note that (with the exception of lithic grain concentration) the

vertical axes of individual panels presenting the same proxy are

identicallyscaled inordertohighlightthedifferingamplitudesof

variability during each glaciation, and the horizontal axes cover

an identicaltime range(80 ky) Elemental analyses of individual

DCgrainsareshowninFig 6.Data reportedhereareavailableat

http://dx.doi.org/10.1594/PANGAEA.834640

4.1 Iceberg and sea ice rafted debris

Withthe exception of IG, the lithic grains atSite U1308,

in-cludingHSG,areinterpretedtoprimarilyreflecticeberg-rafted

de-bris, i.e., IRD While the primary source for HSG was the Gulf

ofSt Lawrence during the last glaciation (Bond andLotti, 1995;

Bondetal.,1999),HSGwas likelyderived froma widerareaand

transported by sea ice during the Holocene and perhaps other

long interstadials (Bond et al., 1997, 1999, 2001) Regarding IG,

Kuhs et al (2014) concluded that >150 μm shards deposited at

Site U1304 (53◦03N, 33◦32W) were likely iceberg-rafted based

ongenerallylargesize,geochemicalheterogeneity,andassociation

withhighconcentrationsofotherlithicgraintypes.AtSiteU1308,

however,we observealessconsistentrelationshipbetweenfresh,

non-weatheredandtransparentIGcontentandtotallithic

concen-tration,particularlyinthe63–150 μm fraction(Fig 5)

North Atlantic sea ice distribution is sensitive to changes in

windstress,whichisinturnaffectedbythepresenceand

configu-rationoftheLIS(ManabeandBroccoli,1985; Okaetal.,2012).We

thereforefeelitislikelythatbothicebergsandseaicecontributed

IGtoSiteU1308,andtherelativeimportanceofeachprocessmay

varywithdifferingoceanographicconditionsbetweenthediscrete glacialintervalsstudied,aswellaswithinthesameglaciation.For example,HSGandIGcovariedwithinMIS 4–2(Fig 2G, 2H), sug-gesting a similar delivery mechanism, but IG is decoupled from other lithicgrain typesthroughoutmuchofMIS 8 and6(Fig 5) Within the MIS-6interval in particular, peak IG content is more consistentlyassociated withhightotalforaminifer abundance, es-pecially N pachyderma [s], which is associated with sea ice and brine (e.g., Hillaire-Marcel and de Vernal, 2008, and references therein) IG could be entrained as atmosphericfallout and deliv-ered to the site as the sea-ice margin expanded south of Ice-land Coastal sea ice formation can directly incorporate seafloor sediment, and in this case the rafted shards could exhibit geo-chemical signatures associated with multiple Icelandic volcanic provinces

4.2 MIS 8

Results from MIS 8 (Fig 5B) are comparable to those of the last glaciation (Fig 2; Section 2.2), with abrupt, high-amplitude variations in all measured proxies Each interval of increased

N pachyderma [s]abundanceisassociatedwithelevatedIRD con-tent.Prior toTerminationIII (∼243 ka),lithic grainandN pachy-derma [s]abundanceisconsistentlypositivelycorrelated(r=0.64,

p <0.0001) Benthic δ13C exhibits high-amplitude fluctuations (−0.65hto 1.28h),with a meanof 0.26 ±0.36 (1σ), and con-sistently tracks both lithic (r= −0.41, p <0.0001)and N pachy-derma [s] abundance (r= −0.65, p <0.0001) N pachyderma [s]

comprises as little as ∼1% to as much as 95% of the plank-tic foraminifer assemblage, andlate-glacial(<275 ka) changes in abundanceare rapid,occurringinlessthan 1000 yr.At∼249 ka,

N pachyderma [s] abundance increases from ∼30% to 90% in

∼850 yr,thenreturnsto30%in ∼1000 yr

Four distinct, IRD events occur within the latter portion of MIS 8at ∼275,263,249,and243 ka (8.4–8.1events),theyounger threeofwhicharecharacterizedby highDCcontent.Light plank-tic δ18O excursions up to a 1.3h magnitudefollow theonset of glacial IRDevents.Estimates of lithicgrain concentration suggest

anincrease of ∼1000 to 2500 grains/g abovethenear-zero(<20) ambientlevelsduringthe8.3,and8.1DCevents,andtheseare di-rectlyprecededby“precursor”increasesinsiliceous IRD,HSGand

IG,similarto thelastglacialHS HeinrichEvents (H1,H2,H4,and H5; Fig 2E–H) (Bond and Lotti, 1995; Bond et al., 1999) While onsetofthe8.4and8.3IRDeventscoincides withhighN pachy-derma [s]abundance,lithicgrainsreachover99%ofthe >150 μm fraction and planktic foraminifers are virtually absent N pachy-derma [s] abundance decreases during a brief reduction in IRD input earlyinthe 8.4event.The 8.1DCeventandH1,which oc-curduring glacial terminations,arecharacterized byintermediate

tolowN pachyderma [s]abundance

Mg/Ca andSr/Ca of individual MIS-8 DC grains from the 8.1, 8.2, and 8.3 DC events suggest they are from the same popula-tion as those fromthe last glacial HS Heinrich Events (Fig 6A) ThehighMgcontent,uptoa1:1ratiowithCa,isconsistentwith thepresenceofdolomiteandaHudsonStraitsource.Narrow-beam LA-ICP-MSanalysisacrossthesurfaceofasinglegrainfromthe8.2 eventrevealshighlyvariableMgcontent(Fig 6B),likelyindicative

ofheterogeneousdolomitization,perhapsacrossmicrofronts OnedistinctionbetweenMIS 8(Fig 5B)andMIS4–2(Fig 2F)is that,during theformer, 63–150 μm-sizedDCremainsanelevated lithiccomponentforanextendedtime(upto ∼2500 yr)following surface anddeeprecovery(lowN pachyderma [s] abundanceand highbenthicδ13C).Thisis observedfollowingthe8.1and8.3DC events but not during 8.2or any of the last glacial HS Heinrich

204 S.P Obrochta et al / Earth and Planetary Science Letters 406 (2014) 198–212

Fig 5 Newresults for A) MIS 6 from the U1308 splice and B) MIS 8 from Hole U1308A: plankticN pachyderma [s] δ18 O (purple; U1308 splice);>150 μm lithic percent (black, open circles) and concentration (red);N pachyderma [s]percent of total planktic foraminifers (green) for samples with>200 individual foraminifers (open circles) and>100 individuals (filled circles);>150 μm DC percent (orange, open circles) and 63–150 μm percent (gray; MIS 8 only); 63–150 μm percent (blue) with concentration (red, open circles; MIS 6) and percent (black, open circles; MIS 8) from the>150 μm IG; 63–150 μm HSG percent (red).>150 μm data are percent of total fraction, and 63–150 μm data are percent of lithic fraction Blue vertical bars:>50% lithic grains; light green bars: IG is∼15% of the 63–150 μm fraction; light orange bars:>10% DC BenthicC wuellerstorfi δ13 C (gray) from the U1308 splice are from Hodellet al (2008) (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Events Bioturbation is unlikely to havetransported thismaterial

therequired ∼18 cm upcore

TheMIS 8 HSGrecord exhibitsa range inabundancefrom1%

to 17%, witha mean and standard deviation of 6.2 and 2.8,

re-spectively.Withtheexceptionofdirectlyprecedingthe8.1and8.3

DCevents,HSG generallyexhibitslow abundanceduring the

lat-ter portion of MIS 8, from ∼270 ka, perhaps due to dilution by

DC.Priortothis,duringthemid-glacial interval,from300–260 ka, HSGexhibitsenhancedamplitudevariabilityandidentifiable max-imaandminima,withmidpointsspacedonaverage1400±340 yr apart(Fig 7).A broadspectralpeakspansthemillennialbandfrom 1/1690to1/1360 yr,centeredat1/1500 yr

Fig 6 A)Cross plot of Mg/Ca and Sr/Ca for individual DC picked from MIS 4–2

HS Heinrich layers in the U1308 splice (filled symbols) and from the MIS 8

DC-rich layers in Hole U1308A (open symbols) MIS 4–2 and 8 grains where analyzed

by ICP-AES and LA-ICP-MS, respectively B) Mg/Ca ratios from 25-μm spot

LA-ICP-MS measurements across the surface of an individual grain from the 8.2 event

(1308A-2H-4W 57–58 cm) Large indention in grain center is from ablation for data

used in (A) See Supplemental Fig 2 for photomicrographs of all grains.

4.3 MIS 6

Lithicresults from MIS 6 (Fig 5A) exhibit pronounced

differ-ences from those of MIS 8 and the last glaciation While there

arelightplanktic δ18O excursionsofsimilarmagnitudeasthe

sur-roundingglaciations,thesearenotassociatedwithintervals

domi-natedbylithicgrains.Plankticforaminiferabundanceremainshigh

throughoutthe entire MIS 6 glaciation, and the maximum lithic

grain abundance is less than 75% of the >150 μm fraction The

concentrationoflithicgrains,however,isnearly continuously

ele-vated,withbackgroundlevelsrarelybelow1000 grains/g

Intervalsofpeaklithic concentration donotcorrespond to

in-tervals of particularly high lithic percentage During the earliest

(∼175 ka) of four distinct MIS-6 lithic deposition events, lithic

contentcomprisesonly15%ofthe >150 μm fractionbutexceeds

3500grains/g,a similarconcentrationasduringthelastglacialHS

Heinrich Events.The eventfrom 161to 156 ka issimilarly

char-Fig 7 A)U1308A MIS 8 HSG from 260 to 300 ka showing cycle minima, maxima, and midpoints B) Histogram of elapsed time between cycle midpoints (∼1400±

340 yr mean) C) MTM power spectrum showing a wide, 99%-confident spectral peak centered at a 1/1500 yr frequency (dashed line) Shaded region denotes the range of 99% confidence (1/1690–1/1360 yr).

acterized bylow lithicpercentage(<25%)andhighconcentration (∼4000 grains/g).Themaximumconcentrationof >6500 grains/

occurredat ∼151 ka,duringanextendedintervaloflithic deposi-tion, lastingfor ∼15 ky andcentered on 146 ka,butlithics only comprise at most ∼50% of the >150 μm fraction at this time While H11 exhibitsthe highestpercentage oflithic grains (74%),

it is characterized by relatively low concentration that averages

∼3500 grains/g

Therealsoappearstobe lesssurface-deepcouplingatthissite duringMIS 6.WhilethecorrelationofN pachyderma [s]abundance with lithic percentage (r=0.64, p <0.0001) and with concen-tration (r=0.64, p <0.0001) is significant and comparable to the other glaciations, benthic δ13C is not highly correlated with any of these parameters Correlation coefficients are: r= −0.32

(p <0.001) forN pachyderma [s], r= −0.21 (p <0.05) forlithic percentage, and r= −0.34 for lithic concentration (p <0.0001) Benthic δ13C exhibits, on average, lighter (0.11h) andless vari-able(0.24;1σ)valuesthanduringMIS 8(0.26±0.36h)and4–2 (0.52±0.30h),withnoextendedintervalsheavierthan0.6h.For comparison, thelastglacial interstadialsatthislocation exhibited valuestypicallyexceeding0.95h(Fig 2C)

Thecoldphasesat ∼184 ka and ∼140 ka arenotdistinguished

by high total lithic percentage or concentration Lithic grains in theseintervalsareprimarily IG,withboth theolderandyounger intervalscharacterizedbyincreasedrelativeabundance(proportion

oflithicfraction)of63–150 μm IG.Thecoldperiodat ∼140 ka is alsomarkedbyhighIGconcentration(∼1000 IGgrains/g).Periods

of increased IG occur at ∼20 ky intervals, i.e., the cold event at

∼162 ka isalsomarkedbyhighIGabundanceandconcentration

N pachyderma [s]variations suggestthat MIS 6 was character-ized by two distinct phases(e.g., Margari etal., 2010) N pachy-derma [s] is relatively abundant throughout the entire glaciation (15–98%); during the early glacial from ∼190–155 ka, N pachy-derma [s] abundance averages ∼20% during warm periods and increasesto ∼65% duringcooler phases Subsequentto ∼155 ka, therecordexhibitsabaselineshifttohigherabundance,indicating

206 S.P Obrochta et al / Earth and Planetary Science Letters 406 (2014) 198–212

Fig 8 Lithiccontent proxy data from the U1308 splice section compared to petrologic count data for A) MIS 4–2 from Site 609, B) MIS 6 from the U1308 splice, and C) MIS 8 from U1308A Proxy data are Si/Sr ratio (red), bulkδ18 O (blue), magnetic grain size (κARM/ κ; brown), Ca/Sr (gray), and XRD-based dolomite percentage ( Ji et al., 2009 ; black) Grain count data are lithic concentration (black) and percent (green) and DC percent (orange) For B and C, non-IG lithic concentration and abundance (gray) are also shown All abundance data are percent of total>150 μm fraction Lithic percent is unavailable for MIS 4–2, and lithic concentration is unavailable for MIS 8 Orange vertical bars show high DC events; blue vertical bars show siliceous IRD events (For interpretation of the references to color in this figure legend, the reader is referred to the web version

of this article.)

generally colder conditions with more severe cold events

Maxi-mumabundancereaches∼95% butdecreasestoonly∼50% during

intervening warmings, which is the same relative magnitude as

duringtheearlyglacial(andsubstantiallylessthanthatofMIS 8)

In addition, both warm-to-cold and cold-to-warm transitions are

lessabruptthan during the surroundingtwo glaciations,

exhibit-ingatypically4000-yrduration.Thisfour-foldincreaserelativeto

MIS 8istoolargetobeentirelyexplainedbylowersedimentation

rate/samplingresolution

TheMIS 6HSGrecordexhibitsasimilarrange,mean,and

stan-dard deviation (1–15%, 6.3 and 2.5, respectively) as the MIS 8

record However, it does not exhibit clear, visually apparent

“cy-cles”(thoughitdoescontainone 99%-confidencespectralpeak at

1/950 yr;SupplementalFig 3)

5 Discussion

5.1 Heinrich events prior to the last glacial

Hudson Strait-derived material is distinguishable from other

sourcesofNorthAtlanticsedimentbya Paleozoicdolomitic

lime-stonecomponentwithhighly depleted δ18O (∼−5h;Hodell and

Curtis, 2008), relatively low Sr content (Hodell et al., 2008), and

a distinctive biomarker signature (Naafs et al., 2011) The

par-ticular nature of this material has allowed the development of

a numberofproxies forHS Heinrich-likeevents(Hodelland

Cur-tis, 2008; Hodell et al., 2008; Ji et al., 2009; Stein et al., 2009;

Naafsetal.,2011; Channelletal.,2012; ChannellandHodell,2013;

Naafs et al., 2013) Results generally indicated that HS

Heinrich-likelayersareprominentfeaturesofall“100-kyworld”glaciations

(backtoMIS16),exceptforMIS 6

Until now, these proxies were only calibrated for the last

glaciation Our results both extend the directly-calibrated

inter-val through MIS 8 andconfirm the evidence for virtual absence

ofsand-sized DCinthe MIS-6interval atSite U1308 (Fig 8) DC

concatenated over all three glaciations exhibits high correlation withbulk δ18O (r=0.64,p <0.001;n=594),split-coreXRFCa/Sr

(r=0.59, p <0.001;n =594), andmagneticgrain size(κARM / κ;

r=0.47, p <0.001;n =594).Lastglacial HSHeinrich Events are all expressed aspeaksinCa/Sr exceeding150,asare the 8.1–8.3

DCevents.However, magneticgrain size(κARM / κ) andbulk δ18O arenotstrictlyaproxyforHudsonStrait-derivedDC.Both κARM/ κ

and bulk δ18O are highly correlated to total lithic percent dur-ing MIS 6 (r= −0.66, p <0.001 and r= −0.84, p <0.001 re-spectively),andeachexhibitsa significantresponsetoIRDevents withinintervalslackingdolomiteandDC

We also show that individual DC grainsdeposited during the MIS 8 DC events exhibit an identical geochemical signature to those from thelast glacial HS Heinrich Events, indicating similar provenance.ThisimpliesthatthesethreeDCIRDevents(8.3,8.2, and8.1) representsurgingoftheLIS intheHudsonStrait region, constitutingtrueHSHeinrichEventsat∼263,∼249,and∼243 ka Similar to the last glaciation,8.1 and8.3, the larger ofthe three

intermsoflithicgrainconcentration(2500and1600 grains/g, re-spectively) aredirectly precededby “precursor”IRDevents(Bond andLotti,1995; Bond etal., 1999),withHSG,IG,andbulklithics all peaking directlybefore the HS Heinrich Event Thisis consis-tent withan externaltrigger forHeinrich Events Conversely, the so-called H11event during TII (∼130 ka) doesnot appearto be associatedwitheitherprecursoreventsorwithlarge-scalesurging fromtheHSregion,andhasverylowCa/Sranddolomite content

atSiteU1308(Fig 8B)

5.2 Nature of variability during the past three glaciations 5.2.1 Implications of HSG variations

TheHSGrecordsreportedherearethefirstpublishedafterthe pioneeringinvestigationsofBondandLotti (1995)andBondetal (1997, 1999, 2001),aswell asthefirst reportedfortime periods priortothelastinterglacial(MIS 5).TheMIS 8HSGresults(Figs.7

and5B)aresimilartothoseofthelastglaciation(Fig 2H)inthat

bothrecordsexhibitbroadlycomparabledown-corevariabilityand

pacing,particularlywithinmid-glacialintervalsofintermediateice

volume, that become less pronounced as each glacial maxima is

approached(∼250–240 ka; ∼30–20 ka).The MIS4–2HSGrecord

exhibitsameanelapsedtimebetweencyclemidpointsof∼1570±

700 yr,andthemostsignificantspectralfrequencyisabroadpeak

centeredat1/1390 yr(Obrochtaetal.,2012;SupplementalFig 4)

ThisresemblesMIS 8results inboth cyclelength (∼1400 ±340)

andfrequency(1/1500 yr)

Cycle lengths within the radiocarbon- and GICC05-dated

por-tionoftheMIS4–2HSGrecordarebimodallydistributedand

pri-marilycomprised of∼1000 and ∼2000 yr components(Obrochta

etal., 2012) This is less apparent in MIS 8, which lacks precise

millennial-scale age control The relatively invariable

sedimenta-tionratesinherentlyproducedbyourorbital scaleagemodel,for

whichnoalternativeiscurrentlyavailable,maybeunrealisticdue

tooccurrenceofHSHeinrichEvents,particularlyduringlateMIS 8

AlthoughtheapparentcyclicityobservedinHSGiswithinthemid

to early glacial interval that lacks major lithic deposition events,

statistical durations and frequencies should be interpreted with

caution.WenotethatthepacingoftheMIS 8and4–2HSGrecords

isgenerallysimilar

The MIS 6HSG record (Fig 5), on the other hand, shares

lit-tlesimilaritywiththoseofthesurroundingglaciations Relatively

high-amplitudedepositionaleventsarespacedatmuchlonger

in-tervals,approximately4 to 8 ky,which issimilar in scaleto the

typically longer-duration cold events reported from the Iberian

MarginduringMIS6(Margarietal.,2010).Also,unliketheMIS 8

and 4–2 HSG records, there is no significant spectral peak at

a frequency that corresponds to this mean pacing

Mathemati-cally,thereisone99%-confidencespectralpeakatanapproximate

1/950-yrfrequency (SupplementalFig 3).While thisis thesame

frequency at which the stacked Holocene HSG record is

coher-entwithcosmogenic14C and10Be production(Bond etal.,2001;

Obrochtaetal., 2012), itcould simplyreflectthe240-yrsampling

step(4×240 =960)inanoisyrecord

HSG variability was originally suggested as both a precursor

to HS Heinrich Events and to be a pacemaker for D–O Events

(Bond andLotti,1995;Bond etal., 1997, 1999) While increasing

HSGconsistentlyprecedesallHSHeinrichEventsincluding8.1and

8.3, thelinkagebetweenice-rafted HSG andD–OEvents through

freshwaterforcingistenuous.Nosystematicrelationshipexists

be-tweenHSGpeaksandGreenlandclimate.Evenwiththeagemodel

usedhere, themostpreciseavailableforthissiteduring MIS4–2,

thereisnoconsistentphaserelationshipbetweenHSGandNGRIP

icecore δ18O.The associationwas primarilybased onacommon

1470-yrperiod,butrecentwork showsthat littleevidence exists

foractual 1500-yrintervalsofclimatevariabilityineither

Green-landrecords(Ditlevsenetal.,2007) ortheHSGrecordfrom

well-dated intervals ofSite 609 (Obrochta etal., 2012) Regardless of

forcing,ifHSGvariabilityisinsomewaycausallylinked toNorth

Atlanticregionalclimatechange,anypotential forcingwouldhave

beenlessofafactorduringMIS 6thanduringMIS 8and4–2

5.2.2 Global connections to North Atlantic variability

Detectable sea level increase at far-field locations indicates

a measurably large ice mass loss of up to 15 m sle (sea-level

equivalent) during the HS Heinrich Events of the last glaciation

(Yokoyamaetal.,2001a,2001b) Earlymodelingresultssuggested

that such a large amount of freshwater would have greatly

af-fected North Atlantic hydrography (Manabe and Stouffer, 1988),

decreasing North Atlantic deepwater (NADW) formation This is

consistentwiththestrongcouplingbetweenIRDevents,cold

sur-face conditions, and depleted benthic δ13C observed at this site

duringMIS4–2(Fig 9B–E).Consistentwith“bipolarsee-saw”

the-ory,asproposed byCrowley (1992),northwardheat flowwas re-duced, resulting in Antarctic warming (Blunier and Brook, 2001; EPICACommunityMembers,2006; Veresetal.,2013;e.g.,EDC δD;

Fig 9I)

Furtherobservational evidence forthisinterhemispheric phas-ingisprovidedbyIberianMarginrecordsthatexhibitbenthic δ18O similar to Antarctic icecores (Shackleton et al., 2000), decreased grainsizealongdeepwater routesthat suggestsslowercirculation (McCaveetal., 1995; Hall andMcCave, 2000), andincreased14C reservoir ages atupwelling sites that indicate slower meridional overturning circulation(Kubotaetal., 2014) Southward displace-ment of circulationbelts, mostlikely due to cooling andsea ice expansion(e.g.,Dentonetal.,2010),isindicatedbyaltered precip-itationpatternsinSouthAmerica(Wangetal.,2004; Kanneretal., 2012; Chengetal.,2013; Naceetal.,2014),Africa(Weldeabetal.,

2007),Asia(Wangetal.,2001),andAustralia(Mulleretal., 2008; Mohtadi etal.,2011),aswellasAntarcticupwelling(Andersonet al.,2009) anddeep-oceanventilation(AhnandBrook,2007,2008) Resultspresentedhereindicateasimilarseriesofeventslikely occurredduringMIS 8.ConsideringthattheSiteU1308agemodel for MIS 8 is orbital in scale and cannot resolve the expected changesinsedimentationrateassociatedwithHSHeinrichEvents, Antarcticairtemperatureincrease(Jouzeletal.,2007; Bazinetal.,

2013) corresponds in general to the timing of high IRD deposi-tion,highN pachyderma [s]abundance,anddepletedbenthic δ13C (Fig 9C, 9D, 9E, 9I) This suggeststhat the hydrographicchanges during the MIS-8HS Heinrich Eventsmayhave beencomparable

tothosethatoccurredduring thelast glaciation.WhileEastAsian monsoonintensity(Fig 9H)appearscontrolled primarilyby high-amplitudeNorthernHemisphereinsolationchanges(Fig 9G),brief lowintensityevents(high δ18O)alsocorrespondtointervalsof en-hancedDCdeposition

During MIS 6, however, these same proxies indicate a nearly continuous,low-levelinputofglacialmeltwater,a greaterinfluence

of orbital precession, andperhaps muted D–O variability NADW production appears to have been generally reduced (or shoaled) and preconditioned for Heinrich-like reorganization that was ul-timately triggered by comparatively weak forcing (e.g., due to a lower disruption threshold, Margari etal., 2010) Benthic δ13C is

onaveragemoredepletedanddoesnotreachsustained,high val-ues typicalof interstadials atthissite(∼0.95h), andwarm/cool oscillations are lower amplitude and less abrupt There are very highbackgroundconcentrationsofsiliceous IRD(∼1000 grains/g) that rarely drop to the near-zero levels characteristic of MIS 8 and 4–2 interstadials, and the highestconcentrations occur dur-ing high 65◦N north insolation (Fig 9D, 9G), suggesting a

con-nection betweenicesheet instabilityandhighsummerinsolation (Timmermannetal.,2010) MIS 6IRDresultsare consistentwith modelsimulations that implyhighbackgroundIRDwithlow rel-ative abundance is associated withweaker D–O cycles (Marshall andKoutnik,2006)

Thethreeintervalsoflowinsolation(highprecession)at ∼184,

∼162, and ∼140 ka are characterized by high IG and N pachy-derma [s]abundance(Fig 9B,9F,9G).(Onlyat ∼162 ka wasthere

anappreciableincreaseintheconcentrationofnon-IGlithicgrains, though IGstill accountsfora largepercentageofthe 63–150 μm fraction andone third ofthe total >150 μm lithic grain concen-tration.) Corresponding to each increase in IG, speleothem δ18O from China (31◦N) indicates weakening of the East Asian

mon-soon (Wang etal., 2008; Fig 9H), and renewed speleothem and travertine growth in northeastern Brazil (10◦S) reflects increased

precipitation(Wangetal.,2004),bothofwhichareconsistentwith

asouthwarddisplacementoftheITCZ

If IG reaching Site U1308 is indeed primarily sea-ice rafted, then seaiceexpansion over deepwater formationsites mayhave been the final trigger for reorganizations during times of low