Comparison of ice-rafted debris and physical properties in ODP Site 1094 (South Atlantic) with the Vostok ice core over the last four climatic cycles

Comparison of ice-rafted debris and physical properties inODP Site 1094 (South Atlantic) with the Vostok ice core over

the last four climatic cycles

Sharon L. Kanfoush a;b;�, David A. Hodell a, Christopher D. Charles c,Thomas R. Janecek d, Frank R. Rack e

a Department of Geological Sciences, University of Florida, Gainesville, FL 32611, USAb Department of Geology, Utica College of Syracuse University, Utica, NY 13502, USA

c Scripps Institution of Oceanography, University of California, San Diego, CA 92093, USAd Antarctic Marine Geology Research Facility, Florida State University, Tallahassee, FL 32016, USA

e Joint Oceanographic Institutions, Washington DC 20036-2102, USA

Received 28 July 2000; accepted 5 December 2001

Abstract

Visual counts of ice-rafted debris (IRD), foraminifera, and radiolaria were made for V1500 samples in Site 1094spanning the last four climatic cycles (marine isotope stages 1^11). Most, but not all, of the IRD variability iscaptured by whole-core physical properties including magnetic susceptibility and Q-ray attenuation bulk density.Glacial periods are marked by high IRD abundance and millennial-scale variability, which may reflect instability ofice shelves in the Weddell Sea region. Each interglacial period exhibits low IRD and high foraminiferal abundanceduring the early part of the interglacial, indicating relatively warm sea-surface temperatures and reduced influence ofsea ice. IRD increases and foraminiferal abundances decrease during the latter part of each interglacial, indicating areturn to more glacial-like conditions. Glacial terminations I and V are each characterized by a step-wise reduction inice-rafting punctuated by a brief pulse in IRD delivery and reversal in N

18O. The coarse fraction of the sediment isdominated by ash and radiolaria, and the relative abundance of these components is remarkably similar to theconcentration of Naþ in Vostok. Each of these variables is believed to be controlled mainly by sea-ice cover, therebyproviding a means for sediment^ice core correlation. ; 2002 Elsevier Science B.V. All rights reserved.

Keywords: South Atlantic; late Pleistocene; ice-rafted debris; Antarctic Ice Sheet; Vostok ice core; paleoclimate; Milankovitch;suborbital variability

1. Introduction

There is much interest and speculation abouthow stable the Antarctic Ice Sheet (AIS) hasbeen in the past, what factors control its stability,and how the AIS will respond to anthropogenicgreenhouse warming in the future (Alley and

0031-0182 / 02 / $ ^ see front matter ; 2002 Elsevier Science B.V. All rights reserved.PII: S 0 0 3 1 - 0 1 8 2 ( 0 1 ) 0 0 5 0 2 - 8

* Corresponding author. Tel. : +1-315-792-3134;Fax: +1-315-792-3831.

E-mail address: [email protected](S.L. Kanfoush).

PALAEO 2854 30-5-02

Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349

www.elsevier.com/locate/palaeo

Whillans, 1991; MacAyeal, 1992; Bindschadler,1998; Bindschadler et al., 1998; Scherer et al.,1998). Because much of the West Antarctic IceSheet (WAIS) is grounded below sea level, it isthe most susceptible to perturbations in sealevel and climate (Denton et al., 1986; Andersonand Thomas, 1991; Bindschadler, 1998). Here westudied downcore changes in concentration of ice-rafted detritus (IRD) at Site 1094 in the southeastAtlantic to decipher the history of two di¡erentcomponents of the cryosphere: ice sheets dis-charging into the Weddell Sea region of Antarc-tica (Kanfoush et al., 2000) and sea ice (Fig. 1).The AIS expanded at the Last Glacial Maxi-

mum, with most of the change concentratedin West Antarctica (Bindschadler, 1998). TheWAIS was about two-thirds larger than todayand grounded ice shelves in marginal seas aroundAntarctica extended to sea level which was V120m lower than today. Kanfoush et al. (2000) docu-mented millennial-scale pulses in IRD delivery tothe South Atlantic during the last glaciation thatrepresent episodes of instability of grounded iceshelves in the Weddell Sea on suborbital time-

scales. Variability of IRD has also been docu-mented for the Holocene from the South Atlanticincluding an abrupt late-Holocene increase in ice-rafting and decrease in sea-surface temperatures(SST) at V5.5 ka (Hodell et al., 2001). In thisstudy, we extend the high-resolution record ofice-rafting at Site 1094 from marine isotopic stage(MIS) 1 through 11 in order to examine the na-ture of ice-sheet and sea-ice variability over thelast four climatic cycles.The longest, most continuous record of Antarc-

tic climate variability is the Vostok ice core, whichextends through the last 420 kyr and captures var-iability on millennial timescales (Petit et al., 1999).Here we present a South Atlantic deep-sea recordthat has a length equaling and resolution ap-proaching, or in some intervals exceeding, thatof the Vostok ice core. The utility of such a rec-ord is twofold. First, a high-resolution IRD rec-ord spanning several climatic cycles provides anopportunity to understand the interactions be-tween variability in the suborbital and Milanko-vitch bands. Second, correlation of the marine-sediment and ice-core records will enable compar-ison of the timing and magnitude of IRD variabil-ity in the South Atlantic with changes in atmo-spheric variables recorded in Vostok. Our study issimilar to that of Bond et al. (1993) who pre-sented a correlation between North Atlantic coresand the Greenland ice cores which revealed therelationship between IRD events (Heinrich events^ Heinrich, 1988) and temperature changes overGreenland during the last glaciation (Dansgaard^Oeschger cycles ^ Dansgaard et al., 1993). Simi-larly, correlation of the high-resolution South At-lantic IRD record from Site 1094 will help deter-mine if Antarctic ice-discharge events were linkedto external climatic factors.

2. Materials and methods

The material analyzed in this study comes froma 14-m piston core (TTN057-13-PC4) taken at53.2‡S 05.1‡E in 2850 m of water depth andODP Site 1094 taken at the same location in2818 m of water depth each in a small sedimen-tary basin near Bouvet Island (Shipboard Scien-

Fig. 1. Site map of the southeast Atlantic indicating the loca-tion of Site 1094 relative to modern positions of importantoceanographic boundaries, including the Polar Frontal Zone,and the winter sea-ice margin. Modi¢ed after ¢gure F1 in(Shipboard Scienti¢c Party, 1999).

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349330

ti¢c Party, 1999). The site is located in the bio-genic silica belt of the Southern Ocean south ofthe present-day position of the Polar FrontalZone (PFZ) and about 2‡ north of the averagelimit of winter sea ice (Zwally et al., 1983; Gloer-son et al., 1993). The piston core was sampled at3- or 5-cm intervals, corresponding to a temporalresolution of approximately 50^150 years duringthe interglacial periods and approximately 300^500 years during the glacial periods. In order toextend the record through the last four climaticcycles, the piston core was spliced to ODP Core1094C-2. This core then was appended to the

shipboard-generated composite record that wasproduced from the same site (Site 1094) duringLeg 177 (Shipboard Scienti¢c Party, 1999). TheODP cores were sampled at 5-cm intervals, yield-ing similar temporal resolution to that of the pis-ton core.Ten AMS 14C dates on monospeci¢c samples of

Neogloboquadrina pachyderma (sin.) from ninestratigraphic horizons in TTN057-13 providechronological control to a depth of 9.7 m belowsea£oor (mbsf) (to V38 kyr; explained in Kan-foush et al., 2000). All ages presented subse-quently are in calendar years. Using planktonic

Table 1Composite stratigraphy analyzed from Site 1094

Core Section and interval Depth Age Latitude Longitude Water depth

(mbsf) (mcd) (ka) (‡S) (‡E) (m)

TTN057-13-PC4 X 0^2 cm 0.0 3.0 0.0 53.2 5.1 2850IV 141^143 cm 9.7 12.7 37.8

ODP 1094C-2H 2W 60^62 cm 11.0 12.7 37.8 53.2 5.1 28187W 70^72 cm 18.7 20.5 88.0

ODP 1094 splice A 3H3W 95^97 cm 18.1 20.5 88.0 53.2 5.1 2818A 8H4W 40^42 cm 65.6 68.6 345.0

Fig. 2. (a) Age^depth model for the Site 1094 stratigraphy derived from piston core TTN057-13, ODP core 1094C-2 and theshipboard splice. Ages were determined by AMS 14C to 37 ka and beyond this by correlation of the Site 1094 N

18O record to theSPECMAP record (Martinson et al., 1987). Squares indicate ages determined from radiocarbon or N

18O correlation points be-tween which ages were interpolated. Dashed lines indicate core splice positions, and that at 37 ka also denotes the limit for AMS14C control. Notice in this and all subsequent ¢gures depths in piston core TTN057-13 were adjusted downward by 3 m, placingthe core top at 3 mcd, in order to keep depths for the composite stratigraphy consistent with the Site 1094 shipboard splice mcdscale. Ages in this and all subsequent ¢gures are in calendar years. (b) (Left) The reference stratigraphy of SPECMAP (Martin-son et al., 1987). (Right) Correlation of the Site 1094 planktonic N

18O record to SPECMAP.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 331

foraminiferal oxygen isotopic records, TTN057-13-PC4 and 1094C were spliced at a depth of9.7 mbsf in the piston core and depths in thepiston core were adjusted downward by 3 m (Ta-ble 1). This is equivalent to a depth of 12.7 mcomposite depth (mcd) in Core 1094C-2, whichis not part of the shipboard splice. The downwardadjustment of depths in the piston core is neces-sary to produce a record that is consistent withthe depth assignments (i.e., mcd) in ODP site1094. To the base of 1094C (at 20.5 mcd), the1094 composite was then spliced extending therecord to 73.7 mcd. Below the TTN057-13-PC4/1094C splice position (12.7 mcd), the chronologyis constructed by correlating the oxygen isotopicrecord to SPECMAP (Imbrie et al., 1984; Mar-tinson et al., 1987) and intervening ages were in-terpolated assuming linear sedimentation rates be-tween tie points (Fig. 2).Downcore IRD concentrations were visually

determined on the 150-Wm^2-mm size fraction(Allen and Warnke, 1991). At least 500 grainswere counted and included quartz, volcanic ash,¢ne-grained volcanics, coarse-crystalline rockfragments, metamorphics, radiolarians and fora-minifera. Relative percent within the 150-Wm frac-tion (%), number of grains/g of total dry sedi-ment, mg/g of total dry sediment and accumula-tion rate (mg/cm2/kyr) were calculated for the in-dividual components (Allen and Warnke, 1991).Abundance of planktonic foraminifera at Site

1094 was variable, resulting in a discontinuousoxygen isotope record, including a gap in latestage 11 and 10 and also in mid stage 5. Oxygenisotopic measurements of the planktonic foramin-ifer Neogloboquadrina pachyderma (sin.) weremade at the University of Florida using a Finni-gan MAT252 mass spectrometer coupled to anautomated carbonate preparation system (KielIII). Isotopic ratios are reported in standard N

Fig. 3. (a) Weight percent of the s 150-Wm sediment fraction, relative percent each within the coarse (s 150-Wm) sediment frac-tion of (b) total lithics, (c) ash, (d) quartz and (e) foraminifera, and (f) N18O of the planktonic foraminifer Neogloboquadrina pa-chyderma (sin.) all versus mcd in the sediment record at Site 1094.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349332

notation (N18O) relative to V-PDB (Coplen, 1996).Analytical precision, estimated by the standarddeviation (1c) of repeated analyses of a powderedcarbonate standard (NBS-19), was P 0.06x forN18O.Shipboard measurements of magnetic suscepti-

bility (MS) and Q-ray attenuation (GRA) bulkdensity were made at 2-cm sample intervals usinga GEOTEK multisensor core logger on whole-core sections recovered from Site 1094 (Weber etal., 1997; Shipboard Scienti¢c Party, 1999). Theseparameters were measured on both the pistoncore and the Leg 177 material. Consequently,MS and GRA bulk density data are from theidentical composite stratigraphy utilized for IRDanalysis. Natural Q-radiation (NGR) was collectedonly on Site 1094 cores, however, therefore theshipboard splice above 12.7 mcd was correlatedto the piston core using oxygen isotope stratigra-phy. Quantitative estimates of di¡use spectral re-£ectance were generated on split core sections at4^6-cm sample intervals using the Oregon StateUniversity Split Core Analysis Track (Mix et al.,1992; Shipboard Scienti¢c Party, 1999).

3. Results

IRD at Site 1094 is composed predominantly ofash (88.3% of the lithic fraction and 600 grains/gon average) and quartz (10.9% and 50 grains/g onaverage) (Fig. 3). The strong dominance of ashwithin the total lithic assemblage results in nearlyidentical records of ash and total lithics. Conse-quently, hereafter, only plots of quartz and totallithics are shown. IRD levels are generally highduring the glacial periods (1200 grains/g onaverage) and low during the interglacial periods(300 grains/g on average) with a high degree ofvariability within each of these periods. Expressedas a relative percent of the coarse fraction, thetotal lithics were roughly equivalent during eachof the four glacial periods (reaching s 80% ineach case). However, expressed as a concentration(Fig. 4), peak values of the last glacial period(5000^10 000 grains/g) were considerably higherthan those of MIS 6, 8 and 10 (2000^4000grains/g).

The whole-core sediment physical propertiesmeasurements, MS, NGR and GRA bulk density,are each high during the glacial periods (averag-ing 38.1U1035 SI, 4.8 cps and 1.25 g/cm3, respec-tively) and low during the interglacials (averaging10.2U1035 SI, 1.9 cps and 1.20 g/cm3) (Fig. 6).The highest levels (nearing 200U1035 SI, 11.5 cpsand 1.5 g/cm3) occurred during the last glacialperiod. Both MS and GRA bulk density exhibita degree of variability within each marine isotopicstage similar to that of the visually determinedIRD, capturing many of the prominent featuresdisplayed by that record. The NGR record exhib-its a prominent saw-toothed pattern whereby val-ues increase gradually throughout each glacial pe-riod and then decrease abruptly at glacial termi-nations. The percent re£ectance record is low dur-ing glacial periods and high during interglacialperiods, paralleling the N18O record. It also exhib-its a saw-toothed pattern, although the relation-ship is inverted relative to that of the NGR. Thegeneral saw-toothed structure displayed by bothNGR and percent re£ectance has numerous mil-lennial-scale oscillations superimposed upon thispattern.

3.1. Glacial periods

Throughout the last glacial period, discrete epi-sodes of IRD deposition recur on millennial time-scales (V6^10 kyr; Figs. 4 and 6; Kanfoush etal., 2000). None lasts for more than a few thou-sand years, and the concentration typically re-turns to baseline values before increasing again.Peaks are composed predominantly of ash andquartz, with ash abundances roughly an orderof magnitude greater than those of quartz (sur-passing 5000 grains/g and 400 grains/g, respec-tively), and the two primary constituents closelyparallel each other throughout. As in the IRDrecord of the last glacial period, MS and GRAbulk density also contain discrete peaks recurringon millennial timescales. Values of these whole-core physical properties generally do return tobaseline values before increasing again but, unlikethe IRD record, the baseline is substantially ele-vated above the values of the contiguous intergla-cial periods (MIS 5 and the Holocene). In all pa-

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 333

rameters, higher-frequency oscillations are alsosuperimposed upon the V6^10-kyr events.Discrete peaks in IRD, dominantly ash and

quartz, occurred during MIS 6, MIS 8 and MIS10 with a temporal spacing (V3^12, V2^9 andV2^8 kyr, respectively) similar to that observedfor the last glacial period (Kanfoush et al., 2000).Unlike the last glaciation, however, the magnitudeof the IRD peak concentrations in the three olderglaciations is smaller (V2000^3000 grains/g, 3000^4000 grains/g and 1500^3000 grains/g). The rec-ords of MS and GRA bulk density do not parallelthe IRD record as closely as they did in the re-cords of the last glaciation. In both parameters,the peaks are often much greater than those foundin the IRD record, particularly at V145^155 ka.As in the last glacial period, higher-frequency os-cillations are superimposed upon the V3^12-kyr,V2^9-kyr and V2^8-kyr events displayed in allmeasured sediment parameters.

3.2. Interglacial periods

Within each of the past ¢ve interglacial periodsthere is an early interval during which IRD valuesapproach zero then rise abruptly. The early Ho-locene (Figs. 4 and 7) is marked by a period inwhich foraminiferal abundances are high (s 45%)and IRD is extremely low (V5.4 grains/g on aver-age). At V5.5 ka, percent IRD increases abruptly(to s 60%). With the exception of two briefevents (atV1500 andV600 yr BP), foraminiferalabundances accompanying this late-Holocene in-crease in IRD are very low (V8.2%). Highestpercent IRD occurs at V4.7 ka and there is adecrease in IRD following this although levelsremain higher than those of the early Holocene.Expressed as a concentration, the large increase inIRD (to s 140 grains/g) occurs later, beginning atV4 ka and peaking at V3 ka. Unlike the lastglacial period, in which ash dominated IRD de-

Fig. 4. (a) Weight percent s 150 Wm, (b^e) concentration (grains/g dry sediment) each of total lithics, ash, quartz and foraminif-era, and (f) N18O of the planktonic foraminifer Neogloboquadrina pachyderma (sin.) all versus mcd in the sediment record at Site1094.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349334

position at the site, the composition of the IRD inthe late Holocene is dominantly quartz. MS,NGR and GRA bulk density are each low duringthe early Holocene (averaging 0.0U1035 SI, 1.2cps and 1.150 g/cm3, respectively) and show anincrease in the latter part of the Holocene (aver-aging 9.9U1035 SI, 3.2 cps and 1.163 g/cm3). Sim-ilar to the record of foraminiferal abundances,high percent re£ectance marks the early Holocene(27.5% on average) and is followed by lower per-cent re£ectance values in the late Holocene (23.9%on average). Unlike the IRD and whole-corephysical properties, the N

18O signal is relativelyconstant throughout the Holocene.The patterns of IRD and foraminiferal abun-

dance during MIS 5 are similar to those observedfor the Holocene. IRD values are near zero earlyin substage 5e (the Eemian) and increase to great-er values at V120 ka. High foraminiferal abun-dances (s 60%) in the early Eemian are replaced

by generally lower foraminiferal abundances(6 40%) throughout the remainder of the in-terglacial. MIS 5 is marked by large £uctuationsin IRD abundances, with total lithics peaking ats 80% and s 1600 grains/g. High IRD and lowforaminiferal abundances occur in substages 5dand 5b. A general reversal of these conditionsoccurs in substages 5c and 5a, although levelsdo not return to those observed in the early Ee-mian. The composition of the IRD in the latterpart of MIS 5 is dominated by ash. As in theHolocene, MS, NGR and GRA bulk densityduring MIS 5 each increase (to V7U1035 SI,V6 cps and V1.18 g/cm3, respectively) followinglow Eemian values (V31U1035 SI, V0.5 cpsand V1.19 g/cm3). The opposite pattern (decreas-ing from s 50% to V25%) is displayed by per-cent re£ectance.MIS 7 shows two substages centered on V240

ka and V215 ka, of high foraminiferal abundan-

Fig. 5. (a) Time series in the spliced TTN057-13/1094C-2/1094 shipboard splice of total lithic concentration (grains/g), magneticsusceptibility (1035 SI), GRA bulk density (g/cm3), N18O of the planktonic foraminifer Neogloboquadrina pachyderma (sin.) andpercent re£ectance of the blue (450^550-nm) band. (b) Time series of NGR in the Leg 177 Site 1094 shipboard splice correlatedto the composite splice used for IRD analysis.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 335

ces (s 55% in each case) and essentially no IRD.As in MIS 5, the record is marked by periods ofhigh IRD occurring during the stadials (s 1500grains/g in each case). Ash is the dominant com-ponent of the IRD throughout. From values ap-proaching zero, MS reaches values of 95U1035 SIand 53U1035 SI during the stadials. Similarly,GRA bulk density values rise from V1.17 g/cm3

to greater than 1.25 g/cm3 during stadials ; NGRvalues rise from 6 0 cps to s 4 cps. From valuesgreater than 30% during MIS 7, percent re£ec-tance drops to less than 20% during stadials. Sim-ilar to MIS 7, two pronounced substages occur inMIS 9. Each is marked by an abrupt increase inIRD following intervals of low IRD. The earliestpart of MIS 9 is essentially ‘IRD-free’. UnlikeMIS 7, however, percent re£ectance values remain

relatively constant throughout, whereas MS andGRA bulk density increase from 20U1035 SI tos 35U1035 SI and from about 1.23 to 1.3 g/cm3.The early part of MIS 11 was marked by a

period of high foraminiferal abundances(s 50%) and very little IRD. Although punc-tuated at V405 ka by a brief reversal in plank-tonic N

18O and a decline in foraminiferal abun-dances to near zero, IRD remains low throughoutthe early part of MIS 11. The planktonic N

18Ominimum of MIS 11 remained heavier than thoseobserved in MIS 9 and 5e (s 2.9x versus 6 2.7and V2.7x, respectively). Ash is the dominantcomponent of the IRD throughout MIS 11. Fromconcentrations near zero in the earliest part of thestage, ash begins to gradually increase at V395ka. Quartz displays an abrupt, large-magnitude

Fig. 6. Comparison of lithics and quartz concentrations and relative percent foraminifera in the s 150-Wm size fraction withineach of the past four glacial periods: (a) MIS 2^4, (b) MIS 6 (c) MIS 8 and (d) MIS 10. Ash strongly dominates the record ineach period, making up s 90% of the total lithic concentration, and is therefore not shown. Quartz closely parallels ash and to-tal lithics throughout each. IRD peaks are associated with high foraminiferal abundances during MIS 2^4 and MIS 6 exceptwhere indicated with an ‘x’. In MIS 8 and 10, this association is not as clearly identi¢able due to the widely varying magnitudesof peaks in foraminiferal abundance.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349336

(V150 grains/g) peak at V383 ka. From a mini-mum of 6 0 cps at V407 ka, NGR values in-crease to 3.7 cps at V390 ka and from less than25% at 405 ka, percent re£ectance rapidly in-creases to s 50% by 400 ka. MS begins atV395 ka to gradually increase from values ap-proaching zero and reaches values of s 15U1035 SI by the end of the interglacial period. Sim-ilarly, GRA bulk density values rise from 6 1.20g/cm3 to s 1.25 g/cm3 at V400 ka and continueto £uctuate near that value throughout the re-mainder of MIS 11.

3.3. Glacial terminations

From high glacial concentrations, IRD de-creases across each of the glacial terminations.This decrease occurs in a somewhat gradual‘step-wise’ manner in some cases (terminations I,III and V) and abruptly in others (terminations II

and IV). In all cases, however, the decrease inIRD occurs early during the termination. AsN18O values decline across termination I, they dis-play a brief reversal toward heavier values beforeagain declining (Fig. 8). Within this N18O reversal,there is a prominent IRD peak. Only terminationV contains a similar IRD peak centered within areversal in N

18O. The IRD peak within termina-tion V is strongly dominated by quartz.

4. Discussion

4.1. Whole-core physical properties as proxies ofIRD

Analysis of IRD is a time-consuming proce-dure. Consequently, several researchers havesought, with success, to employ other sedimentparameters as a proxy of IRD in North Atlantic

Fig. 7. (Top) Comparison of lithics and quartz concentrations and N18O of the planktonic foraminifer Neogloboquadrina pachy-

derma (sin.) in each of the past ¢ve interglacial periods. (Bottom) High relative percent foraminifera accompanies IRD values ap-proaching zero during the early part of each interglacial period (indicated by shading). Following these times of apparent hypsi-thermal conditions, IRD delivery to Site 1094 resumes and foraminiferal abundances decline suggesting occurrence of aneoglaciation characterized the latter part of each interglacial period. Notice the resumption of ice-rafting during MIS 5 beginsprior to the end of the substage 5e (the Eemian) as de¢ned by N

18O values that remain low and constant relative to the neoglaci-ations of MIS 7 and 9. Similarly, IRD returns to Site 1094 during the Holocene at a time when N

18O values are unchanged.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 337

sediments (Robinson et al., 1995; Stoner et al.,1995, 1996; Chi and Mienert, 1996). An opportu-nity exists here to compare directly IRD concen-trations with other sediment parameters in Site1094 to assess their utility as IRD proxies in theSouth Atlantic. This also includes weight percentof the coarse sediment fraction (wt% s 63 Wmand wt% s 150 Wm), which is routinely measuredduring sample processing.In general, the glacial^interglacial pattern dis-

played by MS and GRA bulk density is similar tothat for IRD (Fig. 5). This implies these whole-core physical properties are controlled by terrige-nous in£ux, largely via ice-rafting (Shipboard Sci-enti¢c Party, 1999). Discrepancies between thevisually counted IRD and the whole-core physicalproperties may be attributed in part to grain-sizedi¡erences resulting from the selected upper andlower sieve limits (2 mm and 150 Wm, respectively)utilized for IRD analysis. Speci¢cally, the sub-stantially elevated glacial baseline values of thewhole-core physical properties may re£ect thepresence of terrigenous ¢nes omitted in the visualIRD cps. Examination of the weight percent ofthe various size fractions of the sediment supports

this conclusion (Fig. 9). Because of the greaterweight of the lithics as compared to the biogeniccomponents, the weight percent of the s 150-Wmfraction provides a good approximation of theIRD cps. However, inclusion of ¢ner sediments(the weight percent 63^150^Wm fraction) ap-proaches more closely the structure of the MSand GRA bulk density records, speci¢cally theelevated baseline of glacial periods (Fig. 6).Both the start of glacial periods and glacial ter-

minations are recorded as an abrupt change in therecords of IRD, MS and GRA bulk density (Fig.5). Each of the past three glacial terminations isalso abrupt in the records of N18O, percent re£ec-tance and NGR. However, the transition frominterglacial to glacial conditions in these parame-ters is recorded as a gradual change. NGR is de-rived from decay of natural radioisotopes, includ-ing potassium, uranium and thorium, and re£ectsin£ux of terrigenous materials containing theseradioisotopes. The strong anti-correlation be-tween the prominent saw-toothed patterns ofNGR and percent re£ectance, therefore, impliesthat gradual decreases in percent re£ectance resultfrom either an increase in terrigenous in£ux or a

Fig. 8. Comparison of N18O and IRD (total lithics and quartz) concentrations across the past ¢ve glacial terminations. Termina-tions I, III and IV (and perhaps II) are punctuated by a reversal to heavier N

18O values suggesting a brief return toward moreglacial conditions. Termination V also is punctuated by a reversal to heavier N18O values, but it is of smaller magnitude than inthe other terminations. In terminations I and V, this N18O reversal is accompanied by an in£ux of IRD (marked with arrows).

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349338

decrease in biogenic input (Shipboard Scienti¢cParty, 1999). All other factors being equal, how-ever, both processes would impact the other indi-cators of terrigenous in£ux as well. Their lack ofanti-correlation with percent re£ectance, there-fore, implies that the NGR signal is dominatedby ¢ne-grained sediments delivered to the sitevia ice-rafting, focusing by deep currents, advec-tion within the nephaloid layer and/or eoliantransport. Alternatively, the NGR signal may re-£ect dilution by biosiliceous sediment (diatomproductivity).Discrete IRD peaks observed within each of the

past four glaciations begin and end abruptly andare very brief, on the order of a few thousand toeven a few hundred years in duration. Even with

the high sedimentation rates at Site 1094 (averag-ingV40 cm/yr andV18 cm/yr during interglacialand glacial periods, respectively), the ice-raftingevents are recorded within only a few centimeterssediment thickness. This, in conjunction withslightly o¡set sampling depths and di¡ering sam-pling intervals for the di¡erent sediment measure-ments, may explain discrepancies between the mil-lennial-scale events present within the IRD andthe whole-core physical properties records. Forexample, the records of MS and GRA bulk den-sity, while also containing discrete millennial-scalepeaks, display some peak magnitudes much high-er than peaks within the IRD record, particularlyat V145^155 ka (Fig. 5).The sediment parameter that most closely ap-

Fig. 9. (a) Comparison of total lithic concentration and magnetic susceptibility during the last glaciation (MIS 2^4) demonstratesthat MS provides a good approximation of visually determined IRD. (b) Comparison of the s 63-Wm sediment fraction withmagnetic susceptibility reveals elevated baselines of MS and GRA bulk density (not shown) relative to the IRD cps. This may beattributed to in£uence of the ¢ne size fraction that is omitted in the arbitrarily selected upper and lower sieve limits used in theIRD analysis.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 339

proximates the IRD concentration is weight per-cent s 150 Wm, displaying most (but not all) ofthe discrete IRD peaks (Figs. 5 and 9). Althoughthe majority of IRD peaks is present within theweight percent s 150-Wm signal, the magnitude ofsome of the events di¡ers from the IRD cps. Be-cause larger lithic grains typically weigh more,their presence is re£ected by greater values ofweight percent s 150 Wm. Within IRD cps, how-ever, each lithic grain is counted once regardlessof size and weight. This problem is exacerbated bythe selection of upper and lower size limits forIRD analysis. Nonetheless, because the othersediment parameters incorporate ¢ne sedimentsomitted in the IRD cps, weight percent s 150Wm provides the best approximation of IRD atSite 1094.

4.2. Glacial-to-interglacial variability

IRD in sediments at Site 1094 is composedoverwhelmingly of ash and quartz. Geochemicalanalysis of IRD in nearby sediment cores suggeststhe ash in South Atlantic sediments is derivedprincipally from volcanic eruptions in the SouthSandwich Islands in the Scotia Arc, West Antarc-tica, with minor contribution from nearby BouvetIsland (Cooke and Hays, 1982; Smith et al.,1983). Carried to the east by prevailing winds,ash then settles on ice shelves in the WeddellSea and on seasonal sea ice, which acts as a con-veyor of ash to the South Atlantic Ocean duringits seasonal advance and retreat. The quartz andother non-ash lithics are sourced from Antarcticaby calving of icebergs (Smith et al., 1983; Labey-rie et al., 1986). The majority of icebergs originatefrom ice shelves such as those found in the Wed-dell Sea where debris-rich basal ice can be pro-duced by melting and freezing at the base of theseice shelves (Warnke, 1970; Hulbe, 1997).If IRD transport were exclusively by icebergs,

the relative proportions of the primary compo-nents, ash and quartz, would be the samethroughout the South Atlantic. However, ash isthe dominant component in Site 1094 at 53‡S,whereas quartz is the dominant component incore TTN057-21 at 41‡S. Site 1094 is situatedsouth of the polar front where cool SST permits

sea-ice rafting. In contrast, core TTN057-21 at41‡S is situated north of both the present-daypolar front and the glacial position of the polarfront where the comparably warm SST inhibitssea-ice rafting (Brathauer and Abelmann, 1999;Kanfoush et al., 2000). Consequently, althoughash is transported both by icebergs and by seaice, sea-ice rafting appears to be an importantmechanism in the South Atlantic.Strong glacial^interglacial distinctions occur in

the record of IRD delivery to Site 1094, with veryhigh IRD levels during the glacial periods andvery low levels during the interglacial periods(Figs. 6 and 7). The increase in IRD during gla-cials may be due to increased production of ice-bergs, increased survival of icebergs and sea ice,or dilution by biogenic material.During the last glaciation when sea level was

V120 m lower than today, peripheral ice shelvesaround Antarctica were more extensive asgrounding lines advanced seaward over the con-tinental shelf (Denton and Hughes, 1983; Bind-schadler et al., 1998). Expansion of the WAISduring the glacial period increased the proportionof marine-based ice, and attendant delivery ofIRD to the South Atlantic. Additionally, isostaticdepression of the continental shelf beneath thelarge glacial WAIS could render it susceptible torelative sea-level rise and subsequent buoyancyincrease and unpinning along its margins (Ander-son and Thomas, 1991).In addition to the higher calving rates associ-

ated with increased ice volume in the Weddell Searegion, cooler glacial SST would enhance the sur-vival and long-distance transport of debris-ladenicebergs and sea ice to the core site (Matsumoto,1996). Increased survivability associated with coolglacial SST may have resulted in an increase inthe relative percent IRD within the sediments,even without an increase in the rate of icebergproduction (Keany et al., 1976).Sea-ice coverage and frontal boundary posi-

tions control biosiliceous sedimentation to SouthAtlantic sediments (Burckle et al., 1982; Westfalland Fenner, 1991; Charles et al., 1991; Asmus etal., 1999) and, consequently, IRD content. Duringcool intervals, the polar front migrated northwardand sea ice expanded over the site resulting in a

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349340

decline in biosiliceous productivity (Crosta et al.,1998; Brathauer and Abelmann, 1999). As a re-sult, the relative percent IRD within the sedimentsincreased, even without an increase in their rate ofdelivery to the site. However, IRD expressed asan accumulation rate corrects for dilution andshows increases during cool intervals similar tothat observed in the concentration data.

4.3. Suborbital variability

In addition to well-de¢ned glacial^interglacialcycles, suborbital variability is clearly present inthe records of IRD and whole-core physical prop-erties measurements at Site 1094. For the last gla-cial period, these prominent events have been cor-related with IRD layers in cores situated north ofthe PFZ and have been interpreted as evidence ofabrupt episodes of instability of the AIS in theWeddell Sea region (Kanfoush et al., 2000). Ex-amination of the past four glaciations revealsthat, in addition to generally high IRD, each ischaracterized by discrete pulses in IRD delivery toSite 1094 that occurred on millennial timescales(Fig. 6). With few exceptions, the abrupt IRDevents are accompanied by brief episodes ofhigh foraminiferal abundances originating fromiceberg-induced upwelling or high SST (Fig. 6)(Labeyrie et al., 1986; Grobe and Mackensen,1992).Based on our chronology derived from SPEC-

MAP tie points, the temporal spacing of discretepeaks in IRD during MIS 6, MIS 8 and MIS 11appears to be V3^12, V2^9 and V2^8 kyr, re-spectively. However, two factors make it di⁄cultto ascertain if this di¡ers signi¢cantly from thatobserved for the last glacial period (V6^10 kyr;Kanfoush et al., 2000). First, the IRD £uctuationsduring MIS 6 and 8 are less distinct making itdi⁄cult to identify individual events and, second,the chronology is coarse. We can, however, ascer-tain if the relationship of these events with othersuborbital changes within the same record is con-sistent for di¡erent glacial periods. For example,do the millennial-scale IRD peaks ‘march’through the changing climatic conditions of MIS2, 3 and 4 as well as similar (unnamed) changeswithin MIS 6? Whereas MIS 3 is marked by sev-

eral rapid N18O £uctuations, the variability in MIS

2 and 4 is comparably reduced. A similar patternoccurs in estimated temperatures at Vostok dur-ing the last glaciation. Millennial-scale IRDevents do occur during each of the three stages,but their magnitude and frequency is greatest dur-ing MIS 3. This is consistent with McManus et al.(1999) who argue for relative climatic stability inthe northern hemisphere during full glacial peri-ods and interglacial periods, and increased vari-ability in the intervening climate states. As in thelast glaciation, IRD variability occurs throughoutMIS 6. However, unlike the last glaciation, IRDevents of greater magnitude and frequency do notoccur during the interval containing numerousrapid N

18O £uctuations.Suborbital variability within the interglacial pe-

riods exists with less regularity than the glacialperiods (Fig. 7). Within the glacial periods SSTwas near freezing year-round and we interpretIRD, therefore, as a signal of iceberg production.During interglacial periods, Site 1094 is south ofthe Polar Front and north of the winter sea-icemargin. Fluctuations in the positions of theseboundaries would a¡ect iceberg and sea-ice sur-vivability. Therefore, the controls on IRD duringinterglacial periods may be di¡erent from thoseduring glacial periods. Each of the last ¢ve inter-glacial periods begins similarly, marked by anIRD-free interval with abundant foraminifera oc-curring at a time when temperature over Antarc-tica is high (Figs. 4 and 7). Following this, eachinterglacial is characterized by a return to moreglacial conditions, or a ‘neoglaciation’ in whichice-rafting resumes abruptly and £uctuates irreg-ularly while foraminiferal abundances decline(Kirby et al., 1998; Brach¢eld and Banerjee,2000; Yoon, 2000).The nature of the hypsithermal, however, dif-

fers among the interglacial periods. Previous stud-ies have postulated that MIS 11 was the warmestinterglacial period of the late Pleistocene (Oppo etal., 1990; Hodell, 1993). In contrast, the plank-tonic N

18O minimum of MIS 11 at Site 1094 re-mained heavier than minimum values observed inMIS 9 and 5e which suggests, all other factorsbeing equal, that SST in the southeast Atlanticwere not signi¢cantly warmer during MIS 11

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 341

compared to the subsequent interglacial stages(Hodell et al., 2000). Although not unique withrespect to the magnitude of its SST, the MIS 11hypsithermal exhibited relatively IRD-free condi-tions that appear to have persisted for a sub-stantially longer period of time than observed inthe later interglacial periods. Our age model, de-rived from correlation to SPECMAP, is admit-tedly coarse, particularly during MIS 11 (Fig. 2).However, based upon this chronology, indicatorsof ice-rafting to the southeast Atlantic appear tohave remained near zero for more than 30 kyr,compared to V5^15 kyr in the hypsithermalintervals of subsequent interglacial periods. Sucha particularly IRD-free interval implies dimin-ished calving or size of the ice sheet or both(McManus et al., 1999). The duration of sus-tained IRD-free conditions at Site 1094 may bedue to orbital con¢guration. Eccentricity at MIS11 was low and the amplitude of the precessioncycle was damped with the result that fewer coldsubstages punctuated the interglacial (Hodell etal., 2000).The nature of the neoglaciation also di¡ers

among the interglacial periods. Although thetwo youngest interglacial periods (MIS 5 and 1)and MIS 9 and 11 were marked by a resumptionof ice-rafting, delivery of IRD remained low rel-ative to those of the glacial periods. In contrast,the neoglaciation of MIS 7 exhibits IRD concen-trations approaching values typical of full glacialperiods. The N

18O during MIS 7 increased con-comitantly with the IRD. Again, this di¡erencebetween the interglacial periods is likely a re£ec-tion of the di¡erent orbital con¢gurations of each(Fig. 10). Precession is large during MIS 7 withthe result that the magnitude of cold substages ismuch stronger.Millennial-scale variability on termination I is

characterized by a reversal to cooler temperaturesor the so-called Antarctic Cold Reversal (ACR)(Jouzel et al., 1987). In the Site 1094 record, theACR is marked by greater N

18O values and abrief, but prominent pulse in IRD delivery tothe core site (Fig. 8) (Kanfoush et al., 2000).IRD within the ACR is comprised of two rapidpeaks in quartz, the younger of which is accom-panied by a single large peak in ash. Termination

V is the only other glacial termination punctuatedby a pronounced quartz peak. The rate of IRDdecrease across the terminations di¡ers also, witha gradual, step-wise decrease across terminations Iand V and an abrupt decrease across terminationsII and IV. In each case, however, concentrationsof IRD decrease very early in the terminationswith values approaching zero well prior to theend of each glacial termination.Estimates of SST derived from diatoms in Site

1094 sediments indicate that substantial warmingof South Atlantic surface waters and reduction ofsea-ice occurred early in terminations I and V(Kunz-Pirung and Gersonde, this volume; She-mesh et al., submitted). Consequently, high biosi-liceous productivity may have led to dilution ofIRD delivered to Site 1094 sediments. In addition,SST increase early in the termination may havedecreased survivability and transport distances ofdebris-laden icebergs resulting in a decrease inIRD delivery to Site 1094 early in the termination.The brief reversal to cooler SST associated withthe ACR may have caused an expansion of sea iceand increased survivability of icebergs, both ofwhich may have increased IRD concentrationsin sediments at Site 1094. If SST increase is re-sponsible for the reduction in IRD delivery to Site1094 during glacial terminations, the gradual,step-wise IRD decrease across terminations Iand V implies SST increase occurred in a gradual,step-wise manner. In contrast, the abrupt IRDdecrease early in terminations II and IV suggestsSST increase was abrupt and began early in theterminations.

4.4. Marine sediment^ice correlation

We propose that Site 1094 at 53‡S can be cor-related to the Vostok ice core using sodium con-centration in Vostok and percent ash in SouthAtlantic sediments because both parameters arein£uenced by the extent of sea-ice cover in theSouth Atlantic Ocean. Atmospheric circulation,driven by the ocean^Antarctica temperature gra-dient, is most vigorous during the austral winterwhen sea ice is at its maximum extent. This resultsin e⁄cient transport of this marine aerosol to theSouth Pole (Petit et al., 1999). In Site 1094, ash

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349342

and radiolarians strongly dominate the coarsesediment fraction and both are related to sea-icecover at Site 1094. Ash is transported in part bysea ice, and sea-ice cover inhibits radiolarian pro-ductivity (Keany et al., 1976; Smith et al., 1983;Gersonde and Zielinski, 2000). Consequently, assea-ice cover increases, percent ash increases andradiolarian abundance falls. We elected to usepercent ash to correlate the Site 1094 record toVostok, but could have used percent radiolaria asit is essentially the inverse of the percent ash sig-nal (Fig. 11).The correlation yields close agreement between

changes in Vostok-inferred temperature and N18O

of the marine record on both orbital and millen-nial timescales (Fig. 11). On glacial^interglacialtimescales, regional atmospheric temperature con-trols changes in ice-rafting via its in£uence onSST and, hence, survivorship and transport dis-tances of sea ice and icebergs. Reconstructions ofSST based upon South Atlantic diatoms and ra-diolarians re£ect the regional atmospheric temper-ature control on SST, showing much similaritywith temperature recorded at Vostok (Pichon etal., 1992; Brathauer and Abelmann, 1999). Ingeneral (but not always), IRD delivery to Site1094 is high when inferred temperature at Vostokis cool (less than V33.5 to 34.0‡C). IRD often

Fig. 10. (a) Comparison of IRD at Site 1094 and SPECMAP. (b) July solar insolation at 65‡N. (c) The Earth’s orbital parame-ters throughout the period spanned by the Site 1094 IRD record (440 000 years). (Modi¢ed after Hodell et al., 2000.) This ¢gureillustrates that the past four glacial periods occurred under di¡erent orbital con¢gurations and di¡erent insolation values. Simi-larly, the past ¢ve interglacial periods and glacial terminations occurred under di¡erent orbital con¢gurations and di¡erent insola-tion values as well.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 343

drops precipitously, sometimes approaching zero,at times when temperatures over Antarctica warmto greater than this. Such declines in ice-raftingoccur during the two most prominent warm inter-vals within MIS 7, much of MIS 5, the brief warminterval at V14.5 ka and the early Holocene.Dominated by quartz, the IRD of the late Ho-

locene di¡ers in composition relative to that ofthe previous interglacials and the majority of therecord (which is dominated by ash). TerminationV is punctuated by a brief but large peak inquartz and is the only other instance in whichquartz is dominant. One possible explanation isa di¡erence in provenance. However, for the Ho-locene, the correlation to Vostok suggests thismay result from SST too warm to allow signi¢-

cant sea-ice rafting of ash. Similarly, estimates ofSST derived from diatoms in Site 1094 sedimentsindicate that substantial warming of South Atlan-tic surface waters and reduction of sea ice oc-curred in termination V (Kunz-Pirung, this vol-ume).Resumption of IRD delivery to Site 1094 in the

late Holocene occurs when inferred temperatureat Vostok remains warm (sV0‡C) and N

18O re-mains relatively unchanged. Ice-rafting to Site1094 resumes during MIS 5 even before the endof the Eemian following an essentially IRD-freeinterval. Similar to the Holocene, the ¢rst neogla-cial IRD peak during the Eemian occurs whileinferred temperature at Vostok is V0‡C. How-ever, it does follow a rapid drop in temperature

Fig. 11. (a) Correlation of percent ash in the marine sediment record from Site 1094 and sodium concentration in the Vostok icecore. We propose both are controlled by areal extent of sea ice, and therefore provided a means for correlation between theSouth Atlantic sediments and Antarctic ice at resolution su⁄cient to examine the interaction of ice discharge and atmospheric cli-mate on millennial scales. (b) The resulting comparison of N18O and lithic concentration at Site 1094 with inferred atmospherictemperature in Vostok on the GT4 Vostok chronology (Petit et al., 1999).

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349344

from the early Eemian maximum (s 3‡C). DuringMIS 11, IRD delivery increases while N

18O re-mains relatively unchanged, however temperatureat Vostok has decreased substantially by this time(from more than +2‡C to less than 34‡C). InMIS 7 and 9, the neoglaciation corresponds tosubstantially heavier N18O values and cooler Vos-tok temperatures (less than V33.5 to 34.0‡C)that likely enhanced long-distance transport ofsea ice.The record of IRD concentration exhibits

broad similarities with the dust record in Vostok(Fig. 12). There is similarity on timescales some-what intermediate between orbital and suborbital.Glacial^interglacial patterns are clearly evident ineach and some (but not all) of the millennial-scaleevents match. High dust concentrations duringglacial periods re£ect continental aridity, dust mo-bilization and transport, whereas the presence oflarger particles indicates more turbulent atmo-

spheric circulation in the high southern latitudes(Petit et al., 1999). Petit et al. (1999) suggestedthat peaks in dust in Vostok correspond to peri-ods of steepened winter ocean^Antarctica temper-ature gradient resulting from expansion of sea icein the South Atlantic. The general correspondenceof percent ash with dust in Vostok may be ex-plained by sea-ice expansion because ash is rafted,in part, by sea ice (Keany et al., 1976; Smith etal., 1983). Quartz, which is not dominantly raftedby sea ice, also exhibits a similar broad-scale cor-respondence with dust. This may be explained byincreased iceberg survivorship and transport dis-tances at times when SST are cool enough to per-mit sea-ice expansion. The MS and GRA bulkdensities, which approximate the IRD record,show a corresponding degree of similarity withthe dust record.Comparison of NGR and log Vostok dust re-

veals striking similarity (Fig. 12). The prominent

Fig. 12. Comparisons of (a) total lithic concentration at Site 1094 (solid line) and dust in Vostok (dashed line). (b) NGR at Site1094 (solid line) is smoothed with a ¢ve-point moving average and compared with log dust in Vostok (dashed line), both on theGT4 chronology (Petit et al., 1999).

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 345

saw-toothed pattern displayed by NGR is presentin the dust record and many of the millennial-scale peaks are represented in both. NGR re£ectsin£ux of uranium- and thorium-bearing terrige-nous materials, including dust, much of which issourced from the Patagonia region of SouthAmerica (Grousset et al., 1992; Basile et al.,1997). The NGR signal, therefore, is dominatedby ¢ne-grained terrigenous particles largely of eo-lian origin whereas the other proxies of terrige-nous in£ux to Site 1094 (IRD, MS and GRAbulk density) re£ect terrigenous material trans-ported to the site via ice-rafting. Other processesthat may also contribute terrigenous ¢nes to Site1094 include eolian transport, focusing by deepcurrents, advection within the nephaloid layer,or a combination of these.

5. Conclusions

(1) The whole-core measurements of sedimentphysical properties, MS and GRA bulk density,displayed much (but not all) of the variabilitypresent within the visually determined IRD recordfor both glacial and interglacial periods. Conse-quently, for Site 1094, and likely others situatedsouth of the present-day PFZ where sedimenta-tion is controlled by similar processes, these pa-rameters are to a limited extent a proxy of IRD.Weight percent s 150 yielded the best estimate ofIRD. The NGR signal is not a proxy for IRD butrather exhibited strong similarity with the recordof dust concentration in Vostok on millennialtimescales.(2) Suborbital variability in ice-rafting at Site

1094 characterizes glacial periods and interglacialperiods of the late Pleistocene. Millennial-scalepulses in IRD delivery to the South Atlantic,interpreted as episodes of instability of theAIS, occurred during each of the last four glacialperiods (MIS 2^4, MIS 6, MIS 8 and MIS 10).Variability within each of the last ¢ve interglacialperiods (the Holocene, MIS 5, MIS 7, MIS 9 andMIS 11) includes an early hypsithermal periodcharacterized by low IRD, MS and GRA bulkdensity and high foraminiferal abundance andpercent re£ectance. A reversal of these conditions

characterized the latter parts of the interglacialperiods and is interpreted as a return to moreglacial conditions, a neoglaciation.(3) Suborbital variability in ice-rafting at Site

1094 di¡ers among glacial periods and among in-terglacial periods. Thus, the nature (magnitude,frequency, perhaps even composition) of IRDpeaks is to some extent dependent upon orbitalcon¢guration. Within the limitations of our chro-nology, the duration of hypsithermal conditionsappears to have been greater for MIS 11 thanother interglacial periods of the late Pleistocene.The neoglaciations of MIS 7 and 9 were associ-ated with a substantial increase in N

18O; however,the neoglaciation of MIS 5 began prior to the endof the Eemian and was accompanied by only amodest increase in N

18O. Oxygen isotopic valuesduring the late Holocene neoglaciation are un-changed. Precession is large during MIS 7 withthe result that the magnitude of cold substageswas much larger and ice-rafting more abundantthan in other interglacial periods of the past420 000 years.(4) We postulate percent ash in the marine sedi-

ment record from Site 1094 and sodium concen-tration in the Vostok ice core are both controlledby areal extent of sea ice. They, therefore, pro-vided a means for correlation between the SouthAtlantic sediments and Antarctic ice at resolutionsu⁄cient to examine the phasing of marine andatmospheric changes on millennial scales. NGRexhibits similarities with dust concentration inVostok. One possible explanation is that it is at-tributed to periods of vigorous atmospheric circu-lation resulting from steepened ocean^Antarcticathermal gradients associated with sea-ice expan-sion. Peaks in IRD also show broad-scale (ontimescales somewhat intermediate between Milan-kovitch and suborbital) similarities with Vostokdust and likely re£ect enhanced sea-ice and ice-berg survivability in cooler SST at times of sea-ice expansion.

Acknowledgements

Samples for this study were provided by theOcean Drilling Program on Leg 177 and the

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349346

preceding site survey cruise with sponsorship fromNSF. This research was supported by U.S. ScienceSupport Program Grant F000849 (D.A.H. andS.L.K.), NSF Grant OCE-9907036 (D.A.H. andC.D.C.), a 1999 Geological Society of Americagrant for graduate student research (S.L.K.) and aUtica College of Syracuse University Dean’s FundSmall Research Grant (S.L.K.). Physical proper-ties data used were generated shipboard by Leg177 scientists and on-shore at Lamont-DohertyEarth Observatory by J. Ortiz.

References

Allen, C.P., Warnke, D.A., 1991. History of ice-rafting at Leg114 Sites, Subantarctic South Atlantic. In: Ciesielski, P.F.,Kristo¡erson, Y. (Eds.), Proc. ODP Sci. Results 114, 599^607.

Alley, R.B., Whillans, I.M., 1991. Changes in the West Ant-arctic Ice Sheet. Science 254, 959^963.

Anderson, J.B., Thomas, M.A., 1991. Marine ice-sheet decou-pling as a mechanism for rapid, episodic sea-level change:The record of such events and their in£uence on sedimenta-tion. Sediment. Geol. 70, 87^104.

Asmus, T., Frank, M., Koschmieder, C., Frank, N., Gersonde,R., Kuhn, G., Mangini, A., 1999. Variations in biogenicparticle £ux in the southern Atlantic section of the subant-arctic zone during the late Quaternary: Evidence from sedi-mentary 231Paex and 230Thex. Mar. Geol. 159, 63^78.

Basile, I., Grousset, F.E., Revel, M., Petit, J.R., Biscaye, P.E.,Barkov, N.I., 1997. Patagonian origin of glacial dust depos-ited in East Antarctica (Vostok and Dome C) during glacialstages 2, 4 and 6. Earth Planet. Sci. Lett. 146, 573^589.

Bindschadler, R., 1998. Future of the West Antarctic IceSheet. Science 282, 428^429.

Bindschadler, R., Alley, R.B., Anderson, J., Shipp, S., Borns,H., Fastook, J., Jacobs, S., Raymond, C.F., Shuman, C.A.,1998. What is happening to the West Antarctic Ice Sheet?EOS Trans. Am. Geophys. Union 257, 264^265.

Bond, G., Broecker, W., Johnsen, S., McManus, J., Labeyrie,L., Jouzel, J., Bonani, G., 1993. Correlations between cli-mate records from the North Atlantic sediments and Green-land ice. Nature 365, 143^147.

Brach¢eld, S.A., Banerjee, S.K., 2000. Rock-magnetic carriersof century-scale susceptibility cycles in glacial-marine sedi-ments from the Palmer Deep, Antarctic Peninsula. EarthPlanet. Sci. Lett. 167, 443^455.

Brathauer, U., Abelmann, A., 1999. Late Quaternary varia-tions in sea-surface temperatures and their relationship toorbital forcing recorded in the Southern Ocean (Atlanticsector). Paleoceanography 14, 15^148.

Burckle, L.H., Robinson, D., Cooke, D., 1982. Reappraisalof sea-ice distribution in Atlantic and Paci¢c sectors of

the Southern Ocean at 18,000 yr BP. Nature 299, 435^437.

Charles, C.D., Froelich, P.N., Zibello, M.A., Mortlock, R.A.,Morley, J.J., 1991. Biogenic opal in Southern Ocean sedi-ments over the last 450,000 years: Implications for surfacewater chemistry and circulation. Paleoceanography 6, 697^728.

Chi, J., Mienert, J., 1996. Linking physical property records ofQuaternary sediments to Heinrich events. Mar. Geol. 131,57^73.

Cooke, D.W., Hays, J.D., 1982. Estimates of Antarctic Oceanseasonal sea-ice cover during glacial intervals. In: Craddock,C. (Ed.), Proceedings of the 3rd Symposium on AntarcticGeology and Geophysics. Univesity of Wisconsin Press,Madison, WI, pp. 1017^1025.

Coplen, T.B., 1996. More uncertainty than necessary. Pale-oceanography 11, 369^370.

Crosta, X., Pichon, J.-J., Burckle, L.H., 1998. Application ofmodern analogue technique to marine Antarctic diatoms:Reconstruction of maximum sea-ice extent at the last glacialmaximum. Paleoceanography 13, 284^297.

Dansgaard, W., Johnson, S.J., Clausen, H.B., Dahl-Jensen, D.,Gundestrup, N., Hammer, C.U., Hvidberg, C.S., Ste¡ensen,J.P., Sveinbjornsdottir, A.E., Jouzel, J., Bond, G., 1993.Evidence for general instability of past climate from a 250kyr ice core record. Nature 364, 218^220.

Denton, G.H., Hughes, T.J., 1983. Milankovitch theory of iceages: Hypothesis of ice-sheet linkage between regional inso-lation and global climate. Quat. Res. 20, 125^144.

Denton, G.H., Hughes, T.J., Karlen, W., 1986. Globalice-sheet system interlocked by sealevel. Quat. Res. 26, 3^26.

Gersonde, R., Zielinski, U., 2000. The reconstruction of lateQuaternary Antarctic sea-ice distribution ^ The use of dia-toms as a proxy for sea-ice. Palaeogeogr., Palaeoclimatol.,Palaeoecol. 162 (3-4), 263^286.

Gloerson, P., Campbell, W.J., Cavalieri, D.J., Comiso, J.C.,Parkinson, C.L., Zwally, H.J., 1993. Arctic and AntarcticSea Ice, 1978^1987: Satellite Passive Microwave Observa-tions and Analysis. NASA, Washington, DC.

Grobe, H., Mackensen, A., 1992. Late Quaternary climaticcycles as recorded in sediments from the Antarctic continen-tal margin. In: Kennett, J.P., Warnke, D. (Eds.), The Ant-arctic Paleoenvironment: A Perspective on Global Change.AGU Antarctic Research Series, Washington, DC, pp. 349^376.

Grousset, F.E., Biscaye, P.E., Revel, M., Petit, J.R., Pye, K.,Joussaume, S., Jouzel, J., 1992. Antarctic (Dome C) ice-coredust at 18 ky BP: Isotopic constraints on origins. EarthPlanet. Sci. Lett. 111, 175^182.

Heinrich, H., 1988. Origin and consequences of cyclic ice raft-ing in the northeast Atlantic Ocean during the past 130,000years. Quat. Res. 29, 142^152.

Hodell, D.A., 1993. Late Pleistocene paleoceanography ofthe South Atlantic sector of the Southern Ocean: OceanDrilling Program Hole 704A. Paleoceanography 8, 47^67.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 347

Hodell, D.A., Charles, C.D., Ninnemann, U.S., 2000. Com-parison of interglacial stages in the South Atlantic sectorof the southern ocean for the past 450 kyr: Implicationsfor marine isotope stage (MIS) 11. Glob. Planet. Change24, 7^26.

Hodell, D., Shemesh, A., Crosta, X., Kanfoush, S., Charles,C., Guilderson, T., 2001. Initiation of neoglaciation in theSouthern Hemisphere at V5500 yr BP. Quat. Res. 56, 191^198.

Hulbe, C.L., 1997. An ice shelf mechanism for Heinrich layerproduction. Paleoceanography 12, 711^717.

Imbrie, J., Hays, J.D., Martinson, D.G., McIntyre, A., Mix,A.C., Morley, J.J., Pisias, N.G., Prell, W.L., Shackleton,N.J., 1984. The orbital theory of Pleistocene climate: Sup-port from a revised chronology of the marine N

18O record.In: Berger, A., Imbrie, J., Hays, J., Kukla, G., Saltzman, B.(Eds.), Milankovitch and Climate, Part I. Reidel, Dordrecht,pp. 269^305.

Jouzel, J., Lorius, C., Petit, J.R., Genthon, C., Barkov, N.I.,Kotyakov, V.M., Petrov, V.M., 1987. Vostok ice core: Acontinuous isotope temperature record over the last climaticcycle (160,000 years). Nature 329, 403^408.

Kanfoush, S.L., Hodell, D.A., Charles, C.D., Guilderson,T.P., Mortyn, P.G., Ninnemann, U.S., 2000. Millennial-scale instability of the Antarctic ice sheet during the lastglaciation. Science 288, 1815^1819.

Keany, J., Ledbetter, M., Watkins, N., Huang, T.-C., 1976.Diachronous deposition of ice-rafted debris in sub-Antarcticsediments. Geol. Soc. Am. Bull. 87, 873^882.

Kirby, M.E., Domack, E.W., McClennen, C.E., 1998. Mag-netic stratigraphy and sedimentology of Holocene glacialmarine deposits in the Palmer Deep, Bellingshausen Sea,Antarctica: Implications for climate change? Mar. Geol.152, 247^259.

Kunz-Pirrung, M., Gersonde, R., 2002. Mid-Brunhes century-scale diatom sea surface temperature and sea ice recordsfrom the Atlantic sector of the Southern Ocean (ODP Leg177, Sites 1093, 1094 and core PS 2089-2). Palaeogeogr.Palaeoclim. Palaeoecol. S0031-0182(01)00501-6.

Labeyrie, L.D., Pichon, J.J., Labracherie, M., Ippolito, P.,Duprat, J., Duplessy, J.C., 1986. Melting history of Antarc-tica during the past 60,000 years. Nature 322, 701^706.

MacAyeal, D.R., 1992. Irregular oscillations of the West Ant-arctic ice sheet. Nature 359, 29^32.

Martinson, D.G., Pisias, N.G., Hays, J.D., Imbrie, J.,Moore, T.C., Jr., Shackleton, N.J., 1987. Age dating andthe orbital theory of the ice ages: Development of a high-resolution 0 to 300,000-year chronostratigraphy. Quat. Res.27, 1^29.

Matsumoto, K., 1996. An iceberg drift and decay model tocompute the ice-rafted debris and iceberg meltwater £ux:Application to the interglacial North Atlantic. Paleoceanog-raphy 11, 729^742.

McManus, J.F., Oppo, D.W., Cullen, J.L., 1999. A 0.5 millionyear record of millennial-scale climate variability in theNorth Atlantic. Science 283, 971^975.

Mix, A.C., Rugh, W., Pisias, N.G., Veirs, S., Leg 138 Ship-

board Sedimentologists (Hagelberg, T., Hovan, S., Kemp,A., Leinen, M., Levitan, M., Ravelo, C.), Leg 138 Scienti¢cParty, 1992. Color re£ectance spectroscopy: A tool for rapidcharacterization of deep-sea sediments. In: Mayer, L., Pi-sias, N., Janecek, T. (Eds.), Proc. ODP Init. Rep. 138,pp. 67^77.

Oppo, D.W., Fairbanks, R.G., Gordon, A.L., Shackleton,N.J., 1990. Late Pleistocene Southern Ocean N

13C variabili-ty. Paleoceanography 5, 43^54.

Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N.I., Barnola,J.-M., Basille, I., Bender, M., Chappellaz, Davis, M., De-laygue, G., Delmotte, M., Kotlyakov, Legrand, M., Lipen-kov, V.Y., Lorius, C., Pepin, L., Ritz, C., Saltzman, E.,Stievenard, M., 1999. Climate and atmospheric history ofthe past 420,000 years from the Vostok ice core, Antarctica.Nature 399, 429^436.

Pichon, J.-J., Labeyrie, L.D., Bareille, G., Labracherie, M.,Duprat, J., Jouzel, J., 1992. Surface water temperaturechanges in the high latitudes of the southern hemisphereover the last glacial-interglacial cycle. Paleoceanography 7,289^318.

Robinson, S.G., Maslin, M.A., McCave, I.N., 1995. Magneticsusceptibility variations in upper Pleistocene deep-sea sedi-ments of the NE Atlantic: Implications for ice rafting andpaleocirculation at the last glacial maximum. Paleoceanog-raphy 10, 221^250.

Scherer, R.P., Aldahan, A., Tulaczyk, S., Possnert, G., Engel-hardt, H., Kamb, B., 1998. Pleistocene collapse of the WestAntarctic Ice Sheet. Science 281, 82^85.

Shemesh, A., Hodell, D.A., Crosta, X., Kanfoush, S.L.,Charles, C., Guilderson, T., in press. The sequence of eventsduring the last deglaciation in Southern Ocean sedimentsand Antarctic ice cores. Paleoceanography.

Shipboard Scienti¢c Party, 1999. Leg 177 summary: SouthernOcean paleoceanography. In: Gersonde, R., Hodell, D.A.,Blum, P. (Eds.), Proc. ODP Init. Rep. 177, 1^67.

Smith, D.G., Ledbetter, M.T., Ciesielski, P.F., 1983. Ice-raftedvolcanic ash in the South Atlantic sector of the SouthernOcean during the last 100,000 years. Mar. Geol. 53, 291^312.

Stoner, J.S., Channell, J.E.T., Hillaire-Marcel, C., 1995. Mag-netic properties of deep-sea sediments o¡ southwest Green-land: Evidence for major di¡erences between the last twodeglaciations. Geology 23, 241^244.

Stoner, J.S., Channell, J.E.T., Hillaire-Marcel, C., 1996. Themagnetic signature of rapidly deposited detrital layers fromthe deep Labrador Sea: Relationship to North AtlanticHeinrich layers. Paleoceanography 11, 309^325.

Warnke, D.A., 1970. Glacial erosion, ice rafting, and glacial-marine sediments: Antarctica and the Southern Ocean. Am.J. Sci. 269, 276^294.

Weber, M.E., Niessen, F., Kuhn, G., Wiedicke, M., 1997. Cal-ibration and application of marine sedimentary physicalproperties using a multisensor core logger. Mar. Geol. 136,151^172.

Westfall, F., Fenner, J., 1991. Pliocene-Holocene polar frontzone in the South Atlantic: Changes in its position and sedi-

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349348

ment-accumulation rates from holes 699A, 701ZC and 701B. In: Ciesielski, P.F., Kristo¡ersen, Y. (Eds.), Proc. ODPSci. Results 114, 609^646.

Yoon, H.I., 2000. Glaciomarine sedimentation and its paleo-ceanographic implications along the fjord margins in theSouth Shetland Islands, Antarctica during the last 6000

years. Palaeogeogr. Palaeoclimatol. Palaeoecol. 157, 189^211.

Zwally, H.J., Comiso, J.C., Parkinson, C.L., Campbell, W.J.,Carsey, F.D., Gloerson, P., 1983. Antarctic Sea Ice, 1973^1976: Satellite Passive Microwave Observations. NASA,Washington, DC.

PALAEO 2854 30-5-02

S.L. Kanfoush et al. / Palaeogeography, Palaeoclimatology, Palaeoecology 182 (2002) 329^349 349