Last glacial sea surface temperatures and sea-ice extent in the Southern Ocean (Atlantic-Indian sector): A multiproxy approach



[1] On the basis of the quantitative study of diatom, radiolarian, and planktic foraminiferal assemblages, we estimated summer sea surface temperature (SSST) and sea-ice extent at 50 sediment core localities in the Atlantic and western Indian sector of the Southern Ocean to reconstruct the last glacial environment at the GLAMAP (18 to 15 ka, equal to 21,500 to 18,000 calendar (cal) years BP) and EPILOG (19.5 to 16.0 ka, equal to 23,000 to 19,000 cal years BP) time slices. Stratigraphic identification of the time slices was accomplished by a combination of AMS 14C measurements, benthic isotope, and siliceous microfossil abundance records. While the SSST estimates reveal greater surface water cooling than reconstructed by CLIMAP [1981], reaching a maximum in the area of the present Subantarctic Zone, the sea-ice reconstruction indicates that CLIMAP overestimated the expansion of the Antarctic sea-ice field, especially for austral summer. During winter the sea-ice field was expanded by 60–70% compared to the present. Last glacial summer sea surface isotherms indicate a northward shift of the zonal bands of the Antarctic Circumpolar Current and a relative expansion of the cold water realm south of the Subantarctic Front by ∼5° in latitude. This coincides with a northward displacement of the zone of enhanced biogenic silica deposition and iceberg occurrence. As a result of northward expansion of Antarctic cold waters and a relatively small displacement of the Subtropical Front, thermal gradients were steepened during the last glacial in the area of the present Subtropical Front. The northward displacement of Antarctic cold waters and the related deflection of Southern Ocean waters along the eastern boundary of South America may have resulted in a weakened “cold water route” across the Drake Passage. In contrast, the transport of warm and salty surface water from the Indian into the Atlantic Ocean via the “warm water route” was not blocked allowing continuous but reduced import of heat into the South Atlantic.

1. Introduction

[2] About 25 years after the first reconstruction of Southern Ocean surface water temperature and sea-ice extent during the Last Glacial Maximum (LGM) published by Hays et al. [1976], we present a new investigation of this climatic end-member of late Pleistocene climatic variability from an oceanic area that is critical for global climate development. The pioneering study of Hays et al. [1976] was based on 34 cores from the Atlantic and the western Indian sectors of the Southern Ocean between 35° S and 55°S and 80°E and 50°W. Surface water temperatures were estimated using the transfer function technique developed by Imbrie and Kipp [1971], the so-called Imbrie and Kipp Method (IKM), applying a radiolarian-based paleoecological equation proposed by Lozano and Hays [1976] to estimate austral summer (February) and winter (August) temperature values with a standard error of 1.5 and 1.4°C, respectively. The LGM sea-ice boundary was reconstructed by mapping the lithological boundary between diatom-rich and diatom-poor sediments. Hays et al. [1976] based their stratigraphy for the definition of the LGM, set at 18 ka, on the abundance pattern of the radiolarian Cycladophora davisiana, which was calibrated in four cores recovered from the Subantarctic and Subtropical Zone with the oxygen isotope stratigraphy obtained from planktic (3 cores) and benthic (1 core) foraminiferal records. Additionally, three 14C measurements, all Holocene in age, were used to improve the age models of the cores. The stratigraphic determination of the LGM and the temperature reconstruction proposed by Hays et al. [1976] was used by CLIMAP [1976, 1981] to estimate circumantarctic temperature and sea-ice distribution as a part of the first global ocean LGM reconstruction. CLIMAP [1976, 1981] placed the austral winter and summer sea-ice edge at the faunally identified 0°C winter and summer isotherm, respectively. Later, Cooke and Hays [1982] presented a revised estimation of the LGM summer and winter sea-ice extent, considering additional parameters such as changes in sedimentation rates and the quantification of ice-rafted debris. This idea was supported by Burckle et al. [1982] and Burckle [1983], who proposed that the lithological boundary between silty diatomaceous clay and diatom ooze more appropriately identifies the spring/summer sea-ice limits. Alternative attempts to reconstruct past sea-ice cover rely on the distribution of siliceous microfossils (diatoms) in the sediment record. Crosta et al. [1998a, 1998b] used the Modern Analog Technique (MAT), established by Hutson [1980], applied to diatom assemblages for quantitative reconstruction of seasonal summer and winter sea-ice distribution (months/year) at the LGM. The stratigraphic assignment of the LGM used by Crosta et al. [1998a, 1998b] was based on CLIMAP [1976, 1981] age determinations.

[3] In this study we present a new LGM reconstruction of the Atlantic and western Indian sector of the Southern Ocean applying IKM or MAT on siliceous (diatoms, radiolarians) and calcareous (planktic foraminifers) microfossil assemblages for estimation of surface water temperatures, and diatom indicator species for the description of the sea-ice extent. The dating of sediments documenting the last glacial period was based on a combination of benthic and planktic foraminiferal oxygen isotope records and abundance fluctuation records of Cycladophora davisiana and diatoms, such as Eucampia antarctica. Additional age control for these records has been achieved with accelerator mass spectrometry (AMS) 14C age determinations to identify two LGM time slices proposed by the GLAMAP-2000 working group [Sarnthein et al., 2003] and the international EPILOG working group [Mix et al., 2001], respectively. While the GLAMAP time slice covers the “last benthic oxygen isotope maximum” (LIM), the EPILOG time slice coincides with a maximum sea level-low stand at ∼135 m below present, which lasted about 3000 to 4000 years according to Yokoyama et al. [2000]. Although we provide data for both the GLAMAP and EPILOG chronozone Level 1, defined at 18 to 15 14C ka (equal to 21,500 to 18,000 cal years BP) and 19.5 to 16.0 14C ka (equal to 23,000 to 19,000 cal years BP), respectively, our presentation and discussion here focuses on the data sets obtained from the international EPILOG Last Glacial Maximum (ELGM) time slice (for explanation, see second paragraph in section 4).

[4] In addition to surface water temperature and sea-ice estimations at 50 locations, we also document sedimentation rates and the distribution of biogenic silica (BSi) and ice rafted debris (IRD) at the ELGM from most of the studied cores to augment our paleoceanographic interpretation of the LGM environment. Additionally, we consider the ELGM surface water temperature results from 11 cores presented by Niebler et al. [2003] to connect our results with the adjacent Subtropical Zone of the South Atlantic.

[5] The combination of different proxies and the accurate identification of the selected time slices provides a new quality to last glacial reconstructions in the Southern Ocean. This will help to better understand what happened during the last glacial in the Southern high-latitudes, a region that acts as an amplifier of global climate change through feedback mechanisms driven by changes in albedo, ocean/atmosphere exchange rates, physical parameters of ocean surface waters, water mass structure and formation, and biological productivity. Together with the data from other GLAMAP working areas, our study allows the reconstruction of last glacial conditions in the entire Atlantic Ocean and takes an important step toward a new comprehensive reconstruction of LGM conditions on a global scale, as promoted by EPILOG. Such reconstructions are essential before further progress in modeling and understanding of driving mechanisms of ice age climate change can be made.

2. Material and Methods

[6] Samples representing the last glacial period have been examined from 53 cores recovered during expeditions with R/V Polarstern in the Atlantic and western Indian sector of the Antarctic Zone between 40°S and 70°S, and 45°E and 60°W (Figure 1, Table 1) using gravity or piston coring devices. Samples have mostly been taken at a spacing of around 10 cm. Out of the 53 LGM intervals, a total of 50 (Table 1, core numbers 1–50) were analyzed for the estimation of LGM surface water temperatures and sea-ice extent using siliceous and calcareous microfossil assemblages. Sample preparation and microscopic counting procedures followed methods and conventions described by Gersonde and Zielinski [2000] for diatoms, Abelmann [1988] and Abelmann et al. [1999] for radiolarians, and Niebler and Gersonde [1998] for planktic foraminifers.

Figure 1.

Location of cores used for LGM reconstructions and modern surface water isotherms in the Atlantic and western Indian sector of the Southern Ocean according to Gouretski et al. [1999]. The Gouretski et al. [1999] data set represents the most advanced hydrographic data for the Southern Ocean and reflects annual mean temperatures at 10 m water depth. Core numbers in italics indicate locations where only LGM biogenic silica (BSi) values are available. For location coordinates, see Table 1.

Table 1. Designation, Geographic Coordinates, and Water Depth of Cores Used for LGM Reconstructiona
NumberCoreLatitudeLongitudeWater Depth, m

[7] Surface water temperature values were estimated using statistical methods, such as the Imbrie and Kipp Method (IKM; Imbrie and Kipp [1971]) and the Modern Analog Technique (MAT; Hutson [1980]). The basic assumption of both methods is that modern spatial variability of microfossil assemblages in surface sediment samples deposited at known environmental conditions serves as a proxy for past environmental variability documented down-core. The IKM, which was the primary method used for CLIMAP [1976, 1981] reconstructions, resolves microfossil assemblages preserved in surface sediment samples via a factor analysis. The resulting varimax factors are calibrated in terms of surface water hydrographic parameters, such as temperature, using stepwise multiple regression analysis that is then applied to down-core assemblages for estimation of past hydrographic parameters. For further theoretical and mathematical background of the IKM we refer the reader to Imbrie and Kipp [1971], Jöreskog et al. [1976], and Malmgren and Haq [1982]. The MAT matches down-core with surface sediment assemblages that have similar species composition using a similarity index. The resulting estimates are based on an average of the measurements for the subset of modern samples, weighted by their similarity index, with a calculation of the standard error for each estimate.

[8] Documentation of the surface sediment reference data sets and paleoceanographic equations applied here can be found in the work by Zielinski et al. [1998] for diatoms, Abelmann et al. [1999] for radiolarians, and Niebler and Gersonde [1998] for planktic foraminifers (for a summary, see Table 2). In contrast to previous LGM reconstructions in the Southern Ocean based on statistical techniques, the surface sediment samples included in the present reference data sets have been recovered with a multicorer or minicorer device, allowing undisturbed sampling of the sediment surface. Given that the biogenic particle flux to the seafloor in the Southern Ocean is restricted to austral summer, also in areas not affected by ice cover [Abelmann and Gersonde, 1991; Gersonde and Zielinski, 2000; Fischer et al., 2002], only summer sea surface temperatures (SSST) have been estimated. The SSSTs of the hydrographic reference data sets represent values measured at 10 m below sea surface and were retrieved from Olbers et al. [1992] and Levitus and Boyer [1994]. In cores from the Subantarctic and northern Antarctic zones, we used IKM for SSST estimation applying equations based on regional diatom and radiolarian reference data sets, following the suggestion of Le [1992] that regional data sets exhibit better statistical results using the IKM than other techniques. However, we used the MAT on planktic foraminiferal assemblages from cores recovered in the Subantarctic Zone, applying a larger reference data set that includes 186 surface samples from the Atlantic and Indian sectors of the Southern Ocean (partly published in the work of Niebler and Gersonde [1998]). The temperature estimates taken from Niebler et al. [2003] to connect the Southern Ocean data to the Subtropical realm have been calculated using IKM. Here we present the SSST values instead of the annual mean values given by Niebler et al. [2003]. Both the IKM and MAT calculations were done with the software package of Sieger et al. [1999]. IKM or MAT derived SSST estimates cannot be accomplished in sequences where the microfossil assemblage composition has been altered by dissolution, an effect that precludes an appropriate comparison with the assemblages from the surface sediment reference data set. To overcome such problems in cores from the Weddell Sea, we applied to one core (PS2575-5) a method that allows temperature reconstruction based on the ratio of morphologically different valve types (terminal and intercalary valves) of the coarsely silicified diatom Eucampia antarctica. This Eucampia-Index Method (EIM), proposed by R. Gersonde and B. Censarek (personal communication, 2002), is restricted to cold waters below 4°C (Table 2).

Table 2. Summary of Statistical Methods and Equations, as Well as the Standard Deviation of Estimates and the Temperature Range of the Reference Data Setsa
Application AreaFossil GroupMethodEquationStandard Deviation of Estimate, °CTemperature Range of Reference Data Set, °CReference
  • a

    Equation designations indicate fossil or species group (F, foraminifers, R, radiolarians, D, diatoms, E, Eucampia)/number of reference samples/number of taxa or taxa groups/number of factors; lg or ln indicates logarithmic conversion of species abundance data used to compensate the dominance of single taxa.

South Atlantic (<40°S)
 Pl. foraminifers, >150 μmIKMF271/24/4ln±1.210–27Pflaumann et al. [1996]Nieber and Gersonde [1998]Niebler et al. [2002]
Subantarctic zone (>3500 m water depth)
 Pl. foraminifers, >125 μmMAT/IKMF186/26/5±1.30–28Nieber and Gersonde [1998]
Subantarctic-Antarctic zone
 radiolariaIKMR55/23/4±1.2−1–21Abelmann et al. [1999]
 diatomsIKMD93/29lg/3±0.66−1.8–21Zielinski and Gersonde [1997]Zielinski et al. [1998]
Antarctic zone
 diatomsEucampia-Ind.EI56/2±1.1−1.8–5R. Gersonde and B. Censarek (personal communication, 2002)

[9] The sea-ice extent was derived from the presence of sea-ice indicator diatoms. Following the suggestion of Gersonde and Zielinski [2000], we used the abundance pattern of F. curta and F. cylindrus, which were combined into a F. curta/cylindrus group, as tracers of the winter sea-ice edge and of waters influenced by the spring melt back. A relative abundance of the F. curta/cylindrus group higher than 3% of the total assemblage is considered as a qualitative threshold between the average presence of winter sea-ice and year-round open waters. Values between 3 and 1% of the total are considered to monitor the maximum winter sea-ice edge. The proximity of the summer sea-ice limit was deduced through the enhanced presence of Fragilariopsis obliquecostata, a taxon restricted to very cold waters (<−1°C). This species is relatively thickly silicified and thus not much affected by opal dissolution and it remains a valuable ice cover tracer even in conditions of low sedimentation rates and enhanced opal dissolution [Gersonde and Zielinski, 2000].

[10] The proportions of biogenic silica (BSi) in the sediment, which have been used for delineation of sea-ice extent [e.g., Hays et al., 1976], but also reflect the export production of siliceous microorganisms [Frank et al., 2000], were measured on freeze-dried and pulverized sub-samples. The samples were washed and treated with 10-% acetic acid for sea salt and carbonate removal. For the determination of BSi (calculated for bound water content of 10%), an automated leaching method was applied [Müller and Schneider, 1993]. In addition, BSi contents were estimated from the maximal peak height of the broad opal hump in x-ray diffractograms, calibrated with standard samples [Eisma and van der Gaast, 1971; Hempel and Bohrmann, 1990; Leg 177 Shipboard Scientific Party, 1999].

[11] The amount of gravel clasts >2 mm were counted from 1cm thick x-radiographs from the sediment cores as a parameter of ice-rafted detritus (IRD) [Grobe, 1987]. The IRD distribution provides hints on the distribution of sea-ice and icebergs originating from continental ice sheets. For mapping purposes, the contouring of isolines of the various paleoenvironmental parameters was made by linear interpolation between the values obtained at neighboring sites.

3. Stratigraphic Identification of LGM Time Slices

[12] The establishment of accurate stratigraphic age models for late Pleistocene sediments deposited south of the Subantarctic Zone is complicated by the scarcity or lack of biogenic carbonate, especially during glacial intervals, and hence by the lack of continuous benthic and planktic foraminiferal stable isotope records that can be correlated with the standard isotope stratigraphic records (e.g., SPECMAP; Martinson et al. [1987]). In addition, the lack of sufficient foraminifers makes it difficult to obtain AMS 14C datable samples. Another major problem arises from the highly variable sedimentation rates. In the sea ice covered areas sedimentation rates are typically high during interglacials, and low, sometimes with extremely condensed sections, during glacials [Frank et al., 1996; Gersonde and Zielinski, 2000], which produces unusual signatures of stratigraphic parameters that are not easy to correlate to standard records. To identify the GLAMAP and EPILOG time slices as accurately as possible, we calibrated the abundance fluctuations of siliceous microfossils such as the radiolarian Cycladophora davisiana and the diatom Eucampia antarctica with benthic and planktic oxygen isotope records from a subset of eight cores dated by means of 53 AMS 14C measurements of organic carbon extracted from planktic foraminifers, or from the humic acid fraction in diatomaceous ooze samples that did not allow for extraction of sufficient amounts of foraminiferal carbonate (Figure 2, Table 3). Comparison of AMS 14C dates obtained from organic carbon extracted from planktic foraminifers and the humic acid fraction of the bulk sediment (C. Bianchi and R. Gersonde, personal communication, 2002) demonstrates the applicability of using the humic acid fraction for accurate 14C dating of diatomaceous ooze from latest Pleistocene and Holocene sediment cores recovered in the Southern Ocean. In addition to C. davisiana and E. antarctica, we also considered abundance fluctuations of other diatom taxa, such as Chaetoceros spp., Fragilariopsis kerguelensis, Azpeitia tabularis and Fragilariopsis curta, age-calibrated in cores dated by means of isotope records and AMS 14C measurements. The stratigraphic value of the latter taxa at least for regional intercorrelation of Southern Ocean sediment cores has been documented by Kunz-Pirrung et al. [2002] and Bianchi and Gersonde [2002].

Figure 2.

Correlation of benthic (Cibicidoides spp.) and planktic (Neogloboquadrina pachyderma) oxygen isotope records dated by means of AMS 14C dating (arrows indicate level of 14C samples; for sample depth, organic carbon source, and AMS 14C dating results, see Table 3) with Cycladophora davisiana and Eucampia antarctica abundance fluctuations in the time window between 5 and 30 ka BP (reservoir corrected 14C age) in cores from the Subantarctic, Polar Front, and Antarctic zones. The benthic oxygen isotope records for cores PS2082-1 and PS2495-3, PS2498-1, and PS2499-5 are from Mackensen et al. [1994, 2001], respectively. The C. davisiana records for cores PS1778-5, PS2082-1, and PS2498-1 are from Brathauer et al. [2001]. For core PS1649-2 the abundance pattern of the summer sea-ice indicator Fragilariopsis obliquecostata is also shown. The shadow bar represents the EPILOG LGM time slice (19.5–16 ka), the open rectangle the GLAMAP LGM time slice (18–15 ka).

Table 3. Sample Depth, Laboratory IDsa
CoreDepth, cmLaboratory IDLaboratory IDC Source14C Age, ka BPError, kyReservoir Age, kyReservoir Corrected 14C Age, ka BPData Source
  • a

    Designation of samples: HB indicates samples graphitized at University of Bremen; AAR samples were dated at Institut for Fysik og Astronomy, Aarhus University, Denmark; and KIA samples at Leibniz Labor für Altersbestimmung und Isotopenforschung, Christian-Albrechts-Universität, Kiel, Germany, organic carbon source (N.p. sin, N. pachyderma sinistral; G.b., Globigerina bulloides; humic acid, humic acid fraction of bulk sample), AMS 14C values and measurement errors, applied reservoir age and reservoir corrected AMS 14C age. The definition of the reservoir age is according to Bard [1998].

  • b

    Age determinations of duplicated samples have been averaged for age model construction.

 95–100 KIA15510humic acid7.320±0.0500.6756.645C. Bianchi and R. Gersonde (personal communication, 2002)
 266–278 KIA15513N.p.sin8.215±0.0450.6757.540 
 316–321 KIA15511humic acid8.930±0.0600.6758.255 
 441–445 KIA15512humic acid9.820±0.0700.6759.145 
 460–464 KIA15514N.p.sin10.270±0.0600.6759.595 
 533–538 KIA15515N.p.sin10.530±0.0550.6759.855 
 748–752 KIA14946humic acid11.670+0.08/−0.070.67510.995 
 905–909 KIA14942humic acid13.070±0.0800.67512.395 
 1009–1013 KIA14943humic acid13.700+0.120/−0.1100.67513.025 
 1086–1072 KIA14944humic acid17.320±0.2300.67516.645 
 1208–1212 KIA14958humic acid26.300+0.330/−0.3200.67525.625 
 53–55 AAR-1536N.p.sin9.540±1100.7758.765this study
 77–79 AAR-1537N.p.sin10.510±1400.7759.735 
 141.5–143.5 AAR-1538N.p.sin12.850±1500.77512.075 
 249–253 KIA14949humic acid7.240±0.0500.8106.430C. Bianchi and R. Gersonde (personal communication, 2002)
 460–464 KIA14950humic acid9.040±0.0800.8108.230 
 550–554 KIA14955humic acid9.960±0.0800.8109.150 
 675–679 KIA14951humic acid11.130±0.1200.81010.320 
 720–724 KIA14956humic acid11.700±0.1000.81010.890 
 766–770 KIA14952humic acid12.660±0.1100.81011.850 
 826–830b KIA14953humic acid15.750+0.190/−0.1800.81014.940 
 826–830b KIA14953humic acid16.050±0.1100.81015.240 
 896–900 KIA14957humic acid26.230+0.590/−0.5500.81025.420 
 971–975b KIA14954humic acid29.140+0.950/−0.8500.81028.330 
 971–975b KIA14954humic acid28.910+0.460/−0.4400.81028.100 
 211HB776KIA10019N.p.sin7.110±0.0900.8006.310van der Beek et al. [2002]
 300–302.5 AAR-1533N.p.sin10.440±0.1400.8009.640 
 370.5–372.5 AAR-1534N.p.sin23.670±0.2200.80022.870this study
 20bHB671KIA7249G.b.8.060±0.0500.5107.550this study
 21HB247KIA3258N.p.sin5.240±0.0500.5654.675this study
 11 KIA1866G.b.5.390±0.0300.6004.790this study
 31 KIA1865G.b.10.440±0.0500.6009.840 
 95 KIA7960humic acid16.620±0.090.76015.860Diekmann et al. [2000]
 335 KIA7961humic acid23.200±0.2100.76022.440 
 415 KIA7962humic acid23.790±0.2700.76023.030 

[13] The GLAMAP (18–15 ka) time slice can be clearly identified by the occurrence of the LIM in cores that allow the establishment of continuous oxygen isotope records on benthic foraminifers. In two cores (PS2495-3, PS2498-1; Figure 2) the LIM could be dated using the AMS 14C method, the results of which show that the stratigraphic application of the LIM represents a useful time marker for correlation of northern and southern hemisphere sediment cores. The central portion of the GLAMAP time slice, specifically the LIM, is correlated with the onset and culmination of an abundance peak of C. davisiana, labeled b1 according to the C. davisiana stratigraphy nomenclature [Morley et al., 1982; Brathauer et al., 2001], and with a maximum and subsequent decline of E. antarctica (Figure 2).

[14] The EPILOG (19.5–16 ka) time slice covers, in its lower portion, an interval with slightly depleted benthic δ18O values followed by LIM values. This is correlated with the upper portion of the C. davisiana b2 abundance peak, which has its maximum between 25 and 20 ka (∼29,000–23,500 cal years BP), a C. davisiana abundance minimum, and the lower portion of the C. davisiana b1 abundance peak (Figure 2). E. antarctica generally exhibits a maximum abundance peak between 19.5 and 16 ka.

4. Results

[15] Summer sea surface temperature (SSST) records of the past 5–30 ka obtained from planktic foraminifers, radiolarians and diatoms are available from Subantarctic core PS2498-1, which has been dated on the basis of a continuous benthic oxygen isotope record and 9 AMS 14C age determinations (Figure 3). This core represents the only example allowing such intercomparison of SSSTs. Although the radiolarian SSSTs result in somewhat warmer temperatures than either diatom or foraminiferal estimates in the glacial period, the three records show a similar trend documenting the last glacial temperature low, as well as the Antarctic Cold Reversal (ACR) around 13–12 ka and the Southern Ocean Holocene climatic optimum (SOHCO) at around 10–9 ka. The record shows that the GLAMAP time slice falls within the onset of the transition from glacial to interglacial temperatures in the Southern Ocean, while the EPILOG time slice covers a more stable cold temperature period. A similar pattern is documented in the oxygen isotope record of the Byrd ice core, dated on the basis of correlation of its atmospheric methane record with that of the Greenland Ice Sheet Project 2 (GISP2) [Blunier and Brook, 2001]. However, there is some indication that even EPILOG slice does not document the coldest last glacial conditions in the Southern Ocean, but that the coldest period occurred pre-EPILOG, between 25 and 20 ka BP (∼29,000 to 23,000 cal years BP), at least in the Atlantic sector of the Southern Ocean (Figure 3). This period is characterized by a distinct abundance peak of C. davisiana (Figure 2), labeled b2 according to Brathauer et al. [2001]. This early climate response highlights the lead of Southern high-latitude climate change on the global scale at glacial/interglacial transitions.

Figure 3.

Summer sea surface temperature records obtained from planktic foraminiferal (F), radiolarian- (R), and diatom (D)-based transfer functions and the benthic oxygen isotope record [from Mackensen et al., 2001] in the 5–30 ka interval of core PS2498-1, compared with the oxygen isotope record of the Byrd ice core [Blunier and Brook, 2001] and the time window of the GLAMAP and EPILOG time slices. The age model of core PS2498-1 is based on AMS 14C dates (arrows; for data, see Table 3); ages are in ka BP. Isotopic data from Byrd ice core are on GISP2 timescale. SOHCO, Southern Ocean Holocene climatic optimum; SOLGM, Southern Ocean last glacial maximum; ACR, Antarctic cold reversal.

[16] The SSST estimates, relative abundance of diatom sea-ice indicators, age assignment, and sedimentation rates obtained from samples in the sections out of 50 sediment cores assigned to the GLAMAP and EPILOG time slices are presented in Table 4 (available at All derived parameters have been averaged using all samples assigned to the GLAMAP and the EPILOG time slices, respectively, for each core section. In cases where multiple SSST estimates derived from different microfossil groups were available from the same core section, these values have also been averaged as indicated in Table 5. Although GLAMAP is located in the early portion of significant temperature increase in the Southern Ocean, the averaged SSSTs derived for GLAMAP are almost identical or only slightly warmer than their respective estimates for the EPILOG interval with differences of a few tenths up to a maximum of around 1°C (PS2563-2) (Table 4). Similarly only minor differences occur between the amounts of diatom sea-ice indicators in the two time slices, except for core PS1786-1, which besides slightly higher SSST for the EPILOG time slice shows higher sea-ice indicator values during the GLAMAP interval. However, these values must be viewed with caution considering that they were obtained from a last glacial section with low sedimentation rates that hindered a very accurate age determination (Table 4). Considering that the minor differences between EPILOG and GLAMAP related values range in the statistical error of the applied methods (Table 2), and that the EPILOG time slice follows an international agreement for the construction of a global last glacial time slice [Mix et al., 2001], we focus here on the presentation and interpretation of our EPILOG data.

Table 5. Summary of Averaged Paleoenvironmental Data Obtained for the EPILOG (19.5–16 ka) Time Slice and Derivation of SSST Values Used for LGM Mappinga
CoreSSST, °CSediment Rate MeanBsi Mean, %IRD Mean(Maximum)F.c. + F.c. Mean, %F.obl. Mean, %SSST-D Mean, °CSSST-R Mean, °CSSST-F Mean, °CFinal SSST DerivationSSST, °CDelta SSST, °C
  • a

    See Figures 456. Modern SSST (°C at 10 m water depth) at core location according to Olbers et al. [1992] and Levitus and Boyer [1994]. Bsi, biogenic silica; IRD, ice rafted debris, mean values and maximum values. F.c. + F.c., sea-ice indicators Fragilariopsis curta and F. cylindrus; F.obl., sea-ice indicator Fragilariopsis obliquecostata; SSST-D, diatom-based SSST; SSST-R, radiolarian-based SSST; SSST-F, foraminiferal-based SSST; average SSST values have been rounded to one digit; Delta SSST, difference between modern and average LGM SSST rounded to one digit. Values from GeoB cores are from Niebler et al. [2003].

  • b

    Value additionally taken from GLAMAP time.

PS1433-16.183.2 D+R/23.9−2.3
PS1444-10.518.3  4.11b0.69b0.1b1.1 D0.1−0.4
PS1649-2−0.041.7534.3(14)7.681.29−0.1  D−0.1−0.1
PS1651-10.442.1 10.12(21)8.40b2.40b0.0b0.5b D+R/20.3−0.2
PS1652-20.242.9 10.1(25)18.291.1−0.5  D−0.5−0.8
PS1654-24.568.9529.85(21)  D1.4−3.1
PS1754-17.311.1201.00(2)    2.9F2.9−4.4
PS1756-55.0351.7630.47(5)1.840.21.63.8 D+R/22.7−2.3
PS1765-32.936.2573.47(12)4.680.420.8  D0.8−2.1
PS1768-81.477.0415.81(20)7.850.60.50.9 D+R/20.7−0.8
PS1775-42.036.758  D0.4−1.7
PS1777-64.9254.3 0.67(4)1.560.12.0  D2.0−3.0
PS1778-54.8118.0531.33(6) D+R/21.8−3.0
PS1779-23.974.4 3.36(6)2.450.11.20.8 D+R/21.0−3.0
PS1780-52.8611.8 0.67(3)3.710.01.1  D1.1−1.7
PS1782-51.16 361.89(5)3.650.001.4  D1.40.2
PS1783-50.854.2 1.61(4) D+R/20.7−0.1
PS1786-11.404.1151.36(3)  D1.1−0.3
PS2082-111.3514.1340.15(1)0.340.14.86.6 R6.6−4.8
PS2089-11.361.8 D+R/20.4−1.0
PS2090-11.366.8421.67(3)5.760.80.4  D0.4−0.9
PS2102-20.557.0  D0.0−0.5
PS2104-23.60      1.9 R1.9−1.7
PS2250-512.104.340.33(1) D+R/22.7−9.4
PS2271-53.10 38    2.7 R2.7−0.4
PS2276-40.899.3 1.42(4)3.360.00.5  D0.5−0.4
PS2278-30.6611.1431.54(5)2.760.10.6  D0.60.0
PS2280-40.2211.9593.25(6)2.650.10.8  D0.80.6
PS2305-61.406.7 0.42(1)7.900.30.2  D0.2−1.2
PS2307-10.58  D0.0−0.6
PS2319-10.8214.570.84(3)  D0.5−0.3
PS2489-210.693.35     4.7F4.7−6.0
PS2491-39.32   0.760.003.25.0 R5.0−4.3
PS2492-211.421.1  0.300.004.16.5 R6.5−4.9
PS2493-111.427.5 0.15(1)0.410.04.06.6 R6.6−4.8
PS2495-314.723.31     5.4F5.4−9.3
PS2498-111.118.8120.42(3) R6.2−4.9
PS2499-55.5450.6420.25(4)0.670.02.7  D2.7−2.8
PS2502-24.378.3291.57(5)  D1.2−3.1
PS2515-34.3523.270.22(3)  D0.8−3.5
PS2561-216.40<1  0.840.0010.9  D10.9−5.5
PS2563-29.38  D4.1−5.3
PS2564-39.03<14 0.330.004.3  D4.3−4.7
PS2567-26.904.9644.11(14)0.480.003.93.7 D+R/23.8−3.1
PS2575-50.40<111   −0.8  D-0.8−1.2
PS2603-31.514.934 3.660.70.9  D0.9−0.6
PS2606-63.094.1730.4(2)3.640.50.5  D0.5−2.6
PS2608-13.094.0 7.11(15)  D0.1−3.0
PS2610-33.743.1  D0.6−3.1
PS2821-115.323.89    12.1 R12.1−3.2
PS1575-10.15 4         
PS1648-1  3         
PS1821-60.17 7         
GeoB1306-120.17       17.6F17.6−2.6
GeoB1309-222.85       21.8F21.8−1.1
GeoB1312-222.83       23.3F23.30.5
GeoB2004-219.47       19.0F19.0−0.5
GeoB2016-121.47       19.7F19.7−1.8
GeoB2019-119.02       15.2F15.2−3.8
GeoB2021-518.68       14.0F14.0−4.7
GeoB2819-123.69       21.6F21.6−2.1
GeoB3603-220.24       18.1F18.1−2.2
GeoB3808-623.83       21.1F21.1−2.7
GeoB3813-323.2       20.4F20.4−2.9

[17] In EPILOG LGM (ELGM) sequences obtained from cores recovered in the modern Antarctic Zone (AZ, south of the Polar Front) and Polar Front Zone (PFZ, between the Polar Front and the Subantarctic Front), the SSST estimates based on diatom and radiolarian transfer functions result in coherent values within the range of statistical error. This supports the reliability of the resulting SSST data, which have been averaged to consider both, the diatom and the radiolarian-based results. However, in cores from the modern Subantarctic Zone (SAZ, between the Subantarctic Front and Subtropical Front) the radiolarian-based SSST generally give values that are up to 2°C above the diatom-based estimates (Tables 4 and 5). This mismatch is interpreted to result (1) from a bias of the diatom-based estimates toward colder values due to selective dissolution of diatom assemblages in sediment with low biogenic opal content leading to a relative increase of the coarsely silicified colder-water diatom Fragilariopsis kerguelensis [Zielinski and Gersonde, 1997], and/or (2) from the dominant occurrence of Chaetoceros spp., a diatom group not considered in the transfer function [see Zielinski et al., 1998] and thus strongly reducing the remaining portion of the considered diatom signal leading to questionable estimations. Relatively low SSST values also result from foraminiferal-based estimates. This can also be attributed to selective dissolution that leads to relative increase of the thickly calcified cold water taxon Neogloboquadrina pachyderma. Consequently, when available, we preferred using the radiolarian-based estimates for the mapping of sea surface temperatures in the northern zone of the ACC (Table 5).

[18] The SSSTs derived at locations in the modern AZ generally display values around and below 2°C, and thus are up to 3°C colder than modern values (Figures 4 and 5a). The Weddell Sea “cold water tongue,” which at present extends into the western Enderby Basin around 20°E and between 55 and 60°S (Figure 1) is expanded by approximately 5° in latitude to the north and up to 10° in longitude to the east, indicating enhanced formation of cold surface water in the Weddell Sea area. In the modern PFZ, ELGM SSSTs range between 2 and 4°C, thus 3 to 4°C colder than present. Strongest SSST cooling during the ELGM is recorded in sediments gathered from the modern SAZ, where the SSST decreased in general by 4–5°C. In contrast, in the adjacent modern Subtropical Zone (STZ), the foraminiferal-based estimates indicate only a minor decrease in SSST ranging between 1 and 2°. This is also true for a location south of the African Cape (GeoB3603-2), which monitors the temperatures in the realm of the so-called “warm water route” at the Agulhas Current Retroflection, transporting heat from the Indian into the Atlantic sector by ring shedding [Ruijter, 1982; Gordon et al., 1992]. Here the temperatures of the ELGM declined by only around 2°C compared to modern conditions. This may indicate a reduction of warm surface water import, but does not point to a shutdown of the “warm water route” during the ELGM. A strong decline in SSST was observed in core PS2250-5, located in the northward extending tongue of the Subantarctic Front, east of the Argentinean coast (Figure 1). Although only recording a single location, this core provides hints of water mass distribution in an area governed by the interplay of the warm southward flowing Brazil Current and the so-called “cold water route” that imports heat through the Drake Passage by intermediate waters from the Pacific, replacing deep outflow from the Atlantic [Rintoul, 1991]. Both diatom and radiolarian-based estimates, indicate strongly decreased ELGM SSSTs (more than 9°C), which may point to less southward penetration of the warm Brazil Current and reduced import of surface waters with temperature properties of the present PFZ via the “cold water route” (Figures 4 and 5a).

Figure 4.

Austral summer sea surface temperatures (°C) obtained from the EPILOG time slice. For generation of values, see Tables 4 and 5. Values in parentheses represent foraminiferal-based estimates not considered because of dissolution-biased results. Distribution of Antarctic Ice Sheet during the LGM according to J. B. Anderson et al. [2002].

Figure 5.

(a) Austral summer sea surface temperature anomaly (modern/EPILOG time slice) and summer sea-ice extent; area with SSST anomalies >4°C is shaded. Location of modern oceanic fronts (MPF, modern Polar Front; MSAF, modern Subantarctic Front; MSTF, modern Subtropical Front) according to Peterson and Stramma [1991] and estimated location of oceanic fronts during EPILOG time slice (EPF, EPILOG Polar Front; ESAF, EPILOG Subantarctic Front; ESTF, EPILOG Subtropical Front). Upward pointing arrows indicate range of northward displacement of oceanic fronts during EPILOG time slice. (b) Austral summer sea surface temperature anomaly (modern/LGM) and summer sea-ice extent from CLIMAP [1981].

[19] Assuming that the modern relationship between SSST and the location of the oceanic frontal systems can be applied to the ELGM, the Polar Front, the Subantarctic Front and the Subtropical Front in the central Atlantic sector would have shifted to the North during the ELGM by ∼4°, 5°, and 2–3°in latitude, respectively, compared to their present location. This pattern points to a relative expansion of the AZ and the PFZ during the LGM.

[20] The estimated average winter sea-ice edge during the ELGM was located at around 50°S, thus close to or slightly south of the modern Polar Front. The maximum sea-ice boundary, defined by the minimum occurrence of 1% of the diatom sea-ice indicators F. curta and F. cylindrus, was 2–3° in latitude north of the average winter ice edge, which is in the modern PFZ. This indicates an average northward displacement of the winter sea ice edge by 5 or more degrees in latitude (Figure 6). Assuming that the ELGM average winter sea-ice edge straddled the Drake Passage in its central northern portion, the winter sea-ice coverage increased in the study area between 45°E and 70°W from a modern value of 7.7 × 106 km2 to 12.8 × 106 km2 in the ELGM, which represents an increase of approximately 65%. The maximum winter sea-ice distribution increased from 9.8 to 13.7 × 106 km2.

Figure 6.

Austral winter sea-ice edge average and maximum extent (numbers at core locations indicate percent amount of sea-ice indicators F. curta and F. cylindrus; see Tables 4 and 5), summer sea-ice extent, maximum extent of IRD deposition (for values, see Table 5), and distribution of BSi (for values, see Table 5) at the EPILOG LGM. Modern sea-ice boundaries according to Naval Oceanography Command Detachment [1985]. Upward pointing arrows indicate range of northward displacement of maximum and average winter sea-ice during EPILOG time slice.

[21] While we have a well-established set of data for the reconstruction of winter sea-ice, estimating summer sea-ice extent is more difficult. In southerly located cores, such as PS2319-1, PS1649-2, and PS1652-2, the abundance of the cold water diatom F. obliquecostata, which preferentially dwells in waters at or close to freezing (<−1°C) is found at values between 1 and 2% of the total; however, the 3% value that has been identified to signal a summer sea-ice edge in the vicinity of the core location [Gersonde and Zielinski, 2000] is not reached. The observed occurrence of F. obliquecostata may thus indicate only sporadic expansion of summer sea-ice into the surroundings of the Bouvet Island (Figures 5a and 6). However, F. obliquecostata values above 3%, combined with low sedimentation rates have been observed in sediments of PS1649-2, but only in an interval assigned to a pre-EPILOG age between 25 and 20 ka (∼29,000 to 23,000 cal years BP) (Figure 2). This points to a more extended summer sea-ice edge during a period before the selected time slice.

5. Discussion

[22] While the average winter sea-ice edge obtained for the ELGM is at a more southerly location compared with previous estimates, our maximum sea-ice extent compares well with earlier reconstructions. With the exception of CLIMAP [1981], which shows a sea-ice limit close to 45°S, the majority of reconstructions indicate that LGM winter sea-ice reached a zone around 50°S, i.e., a few degrees in latitude south of the EPILOG Polar Front (Figures 5a and 7). This shows that coverage of winter sea-ice during the LGM was expanded in the Atlantic Southern Ocean by around 60–70% compared to modern conditions.

Figure 7.

Comparison of LGM winter ice edge reconstructions presented by Hays et al. [1976], CLIMAP [1981], Cooke and Hays [1982], Crosta et al. [1998a, 1998b], and our study. Modern sea-ice boundaries according to Naval Oceanography Command Detachment [1985]. Upward pointing arrows indicate range of northward displacement of maximum and average winter sea-ice during EPILOG time slice.

[23] Comparison of the ELGM SSSTs with those presented by CLIMAP [1981] shows that both reconstructions result in the strongest LGM cooling in the area of the modern SAZ, with temperature decreases of more than 4°C. However, while CLIMAP [1981] shows a more patchy distribution of areas of maximum cooling, the EPILOG reconstruction points to a broader zone of enhanced cooling, and thus to generally colder summer sea surface conditions than estimated by CLIMAP [1981] (Figure 5). Strongest LGM cooling was reconstructed by CLIMAP [1981] from cores located in the Argentine Basin. We did not select cores for our reconstruction from this area because this region is noted for the widespread occurrence of giant mud waves, indicating significant reworking of sediments [Flood et al., 1993], which complicates accurate dating of the LGM time slice and may lead to biased paleoenvironmental reconstructions.

[24] CLIMAP [1981] reconstructed a summer sea-ice edge reaching 50°S in the Atlantic and western Indian sector (Figure 5b). This would indicate an extremely expanded permanent sea-ice cover and little sea-ice seasonality because CLIMAP placed the winter sea-ice edge just north of the summer sea-ice boundary (Figures 5b and 7). Such an interpretation is neither supported by our data nor by the results of Crosta et al. [1998a, 1998b]. The latter authors speculate that the location of the LGM summer sea-ice edge was similar to its modern position. Indeed, due to widespread opal dissolution in the Weddell Sea, the reconstruction of the summer sea-ice edge and of temperatures based on the siliceous microfossil record as well as estimates of the LGM summer sea-ice extent in the Weddell Sea area, can only be based on second and third-order evidence. The SSST obtained from core PS2575-5 and the distribution of the SSST pattern in the southeastern Atlantic and southwestern Indian sector points to enhanced advection of cold waters from the southern Weddell Sea. Together with increased abundances of cold water diatoms preferentially dwelling close to the freezing point of ocean waters in cores recovered in the vicinity of Bouvet Island, this permits us to speculate that the summer sea-ice during the ELGM was more expanded than today and may have reached sporadically as far north as the region of Bouvet Island in the eastern portion of the Atlantic sector. Less northward extent of the summer sea-ice occurred in the western Atlantic sector where only a single core located close to the South Orkney Islands (PS2319-1) indicates an increase of the cold water diatom F. obliquecostata (Table 4). In fact, given the colder atmospheric [e.g., Petit et al., 1999] and sea surface temperatures, the diminished temperature of the LGM Antarctic deep-waters due to reduced import of warm North Atlantic Deep Water and formation of cold intermediate and deep waters in the Southern Ocean [Mackensen et al., 2001], which would not allow heat advection to the sea surface and subsequent sea-ice melting as at present [Gordon, 1981], make the scenario of a LGM summer sea-ice field that has an extent comparable to the modern conditions, as suggested by Crosta et al. [1998a, 1998b] and discussed by Moore et al. [2000], rather unlikely. The indication for a summer sea-ice field extending close to the Bouvet Island area found in core PS1649-2 for the pre-EPILOG period between 25 and 20 ka BP, an interval also documented to represent coldest conditions in the Subantarctic realm (Figure 3), is supported by findings from another core from this area [Gersonde and Zielinski, 2000] that shows temporal differences in the summer sea-ice extent during the last glacial. While enlarged perennial sea-ice cover and reduced sea-ice seasonality occurred between 25 and 20 ka in the Weddell Sea sector, during the following 5 ka the summer sea-ice field was reduced, although not by as much as at present, which may have resulted in a magnitude of sea-ice seasonality similar to present conditions.

[25] Our observation of significantly changing environmental conditions during the period assigned to the last glacial (sensu lato MIS 2) points to the need of accurate dating of any time slice assigned to represent LGM conditions, to prevent data from different periods and documenting different environmental conditions from being averaged into reconstructions of the LGM environment. Indeed, there is indication that CLIMAP [1981] used unequal stratigraphic assignments to identify the LGM sample depth in the set of cores used for reconstruction of the LGM Southern Ocean. The CLIMAP [1981] strategy for the LGM definition was outlined by Hays et al. [1976]. The presented data show that the C. davisiana abundance peak used to identify the LGM sample represents a peak that in some cores can be related to the b1, and in other cores to the b2 abundance maximum. As a consequence, CLIMAP [1981] combined paleoceanographic estimates from periods that represent different paleoenvironmental conditions, separated in time by as much as 5000 years. In such CLIMAP cores where no b1 and b2 peak could be distinguished, the center of the C. davisiana maximum was assigned to represent the LGM, which according to our results may be correlated to the C. davisiana maximum b2 that occurred during the cold period between 25 and 20 ka. Later LGM reconstructions based on CLIMAP cores, such as the sea-ice reconstructions presented by Crosta et al. [1998a, 1998b], might be similarly biased because these authors claim to have applied the CLIMAP stratigraphy for LGM identification. Thus it can be speculated that differences between data sets and resulting interpretations based on CLIMAP LGM definitions and our EPILOG LGM may also be caused by stratigraphic uncertainties resulting from the definition of the CLIMAP LGM.

[26] The ELGM SSST estimates indicate a northward expansion of Southern Ocean cold surface waters south of the Subantarctic Front compared to modern conditions. While ELGM summer sea surface isotherms that can be related to the temperature regime at the modern Polar and Subantarctic fronts shifted to the north by 4–5° in latitude compared to modern conditions, leading to an expansion of the ELGM Polar Front zone, the ELGM Subtropical Front displacement was minor (Figures 5a and 8). As a consequence of northward expansion of Southern Ocean cold waters and of minor changes of ELGM SSST in the southern subtropical realm, the surface water temperature gradients steepened during the ELGM around the Subtropical Front and the Subantarctic zone compared to modern conditions. Such steepening of hydrographic gradients between the Antarctic cold waters and the subtropical warm waters should have had an impact on the velocity of zonal water transport in the northern realm of the ELGM ACC. The gradient change should also have affected atmospheric circulation, e.g., a northward shift of the westerly winds, as proposed by Sigman and Boyle [2000] in a glacial Southern Ocean model.

Figure 8.

Modern SST [from Gouretski et al., 1999] and EPILOG LGM SSST variation on a meridional transect at 15°W. Also indicated are the locations of the oceanic fronts and ACC zones at modern conditions (according to Peterson and Stramma [1991]) and as estimated for the ELGM (prefix M, indicates modern; prefix E, EPILOG LGM related front or zone; PF, Polar Front; SAF, Subantarctic Front; STF, Subtropical Front; AZ, Antarctic zone; PFZ, Polar Front zone; SAZ, Subantarctic zone), the maximum and average winter sea-ice extent (modern and ELGM), the maximum summer sea-ice extent (modern and ELGM), the zone of maximum BSi deposition today (from W. Geibert et al., personal communication, 2001) and at the ELGM.

[27] The ELGM ice rafted debris (IRD) boundary, which monitors the extent of iceberg occurrence, coincides with the ELGM 6°C-isotherm, located in the EPILOG SAZ (Figures 4 and 6). This pattern is consistent with the observation and modeling of modern iceberg distribution in the Atlantic sector, which shows that icebergs spread out from the Antarctic Peninsula region as far north as the southern realm of the modern SAZ [Gladstone et al., 2001], and reveals a similar relationship between surface water temperature and iceberg distribution at modern and LGM conditions. Thus imports of freshwater through iceberg melt and related effects on water mass structure and production [e.g., Jenkins, 1999] has to be considered in ELGM scenarios as far north as 45°S in the eastern Atlantic sector of the Southern Ocean.

[28] An uniform relationship during modern and ELGM conditions has also been encountered between the location of the zone of maximum biogenic silica (BSi) deposition in the Atlantic sector, and the distribution of the sea-ice field and ACC hydrography. During modern and ELGM times, the zone of enhanced BSi export, which leads to massive BSi accumulation at the seafloor (opal belt) [Quéguiner and Brzezinski, 2002], is located in a zone between around the average winter sea-ice edge and the southern Polar Front zone (Figures 6 and 8). As a consequence of the expansion of the winter sea-ice field and the northward displacement of the ACC zones, the ELGM opal belt was shifted to the north by approximately 5° in latitude (Figure 8). Such a shift has also been reported from time series studies using export productivity proxies from a latitudinal sediment core transect to occur not only during the last glacial but also during the penultimate glacial maximum [Frank et al., 2000]. The uniform relationship between SSST and sea-ice distribution, and enhanced deposition of BSi during modern and ELGM conditions, indicates that these parameters set conditions for the geochemical and biological factors controlling BSi accumulation in the sediment record, independent of climate related differences in water mass structure and supply of micronutrients, such as iron. It can be speculated that the zone of enhanced silica in Antarctic surface waters bounded to the north approximately by the Polar Front [e.g., Treguer and van Bennekom, 1991; Fischer et al., 2002] was expanded during the LGM, accompanying the northward displacement of the Polar Front.

[29] Unrelated to biological processes in the last glacial Southern Ocean, yet controversial with respect to potential mechanisms for reducing glacial atmospheric CO2 [R. F. Anderson et al., 2002], the EPILOG LGM reconstruction shows distinct changes in physical parameters in the Southern Ocean that potentially enhance CO2 draw down. The Antarctic surface water temperatures were lower than present, and the cold water sphere bounded to the north by the Subantarctic Front was expanded by ∼5° in latitude. These changes alone would have increased the Southern Ocean carbon uptake capacity in an area, which according to ocean-climate modeling, is crucial for carbon dioxide draw-down via isopycnal transport [Caldeira and Duffy, 2000]. The enlargement of the winter sea-ice field (increased by ∼60–70% compared to present), although to a lesser extent than proposed by CLIMAP [1981], would increase the area where the air-sea exchange of CO2 is reduced, as is observed during modern austral winter in the ice covered Southern Ocean [Bakker et al., 1997]. On the basis of an atmosphere-ocean box-model study in which sea-ice area is prescribed, Stephens and Keeling [2000] have suggested that as much as 80% of the glacial-interglacial CO2 difference was caused by the reduction of outgassing due to an enlarged glacial winter sea-ice field. However, the large impact of sea ice, proposed by Stephens and Keeling [2000] has recently been questioned by Morales Marqueda and Rahmsdorf [2002]. They argue that a sea-ice concentration of close to 100% as assumed by Stephens and Keeling [2000] is unlikely due to disturbances by wind and ocean circulation, and conclude on the basis of a coupled sea-ice-upper ocean model that winter sea-ice expansion could account for no more than 15–50% of the glacial carbon dioxide decrease. In any case, the modeling results highlight the potential of Antarctic sea-ice as a major accelerator in climate driving processes.

[30] The northward shift of the surface water isotherms and related frontal zones of the ACC during the ELGM, as well as the low temperatures in the western Argentine Basin documented by only one core but supported by the CLIMAP [1981] reconstruction, indicates that transport of waters having the properties of the PFZ waters across the Drake Passage was reduced during the ELGM. This scenario, which implies a northward deflection of cold Southern Ocean waters in the eastern Pacific along the Chilean and Peruvian coast, similar to the modern deflection of Subantarctic waters, is corroborated by foraminiferal-based SST estimates and faunal analyses along the eastern boundary of South America and in the eastern equatorial Pacific recently presented by Feldberg and Mix [2002]. They show that the cooling in the LGM equatorial Pacific was related to increased northward advection of polar waters along the eastern boundary of South America. Such northward deflection of a broad portion of the LGM ACC would weaken heat transport from the Pacific into the Atlantic Ocean via the “cold water route,” claimed to be an important part of the climate-driving global thermohaline circulation [Rintoul, 1991; Drijfhout et al., 1996].

[31] In contrast to the indications of a weakened “cold water route,” the foraminifer-based SSST reconstruction south of the African Cape and the adjacent area to the west, which shows little cooling, does not point to a strong reduction of the import of warm surface waters from the Indian Ocean via the “warm water route.” Similar results were obtained by CLIMAP [1981] (Figure 5b). In addition calcareous nannofossil records also indicate a continuous shedding of warm surface waters from the Indian into the Atlantic Ocean during the late Pleistocene [Flores et al., 1999]. Such transport may have prevented stronger cooling in the South Atlantic subtropical gyre [Niebler et al., 2003] resulting in the steepening of thermal gradients in the area of the present Subtropical front during the ELGM (Figure 5a). However, clay mineralogical studies [Diekmann et al., 2003] as well as strontium isotopes in detrital material [Goldstein et al., 1999] show that bottom water transport around the African Cape was distinctly reduced during the last glacial.

6. Conclusions

[32] Summer sea surface temperature and sea-ice estimates reveal that during the ELGM (19.5 to 16.0 ka, equal to 23,000 to 19,000 cal years BP) the surface waters in the Atlantic and western Indian sectors of the Southern Ocean were distinctly colder than at present and the winter sea-ice field was expanded by 60–70%. Strongest cooling occurred in the area of the present Subantarctic zone and resulted in an expansion of the Antarctic cold waters south of the Subantarctic Front by ∼5° in latitude, while the location and surface temperature of the South Atlantic subtropical gyre was not significantly changed. As a result, the thermal gradient between cold Antarctic waters and the subtropical realm was steeper than today. Compared to the reconstruction by CLIMAP [1981], our estimates indicate greater cooling of the Southern Ocean SSST during ELGM. However, we find that CLIMAP overestimated the expansion of the Antarctic sea-ice field, especially for austral summer. While the reconstruction of the winter sea-ice is feasible, on the basis of the record of diatom sea-ice indicators, the reconstruction of the summer sea-ice in the Weddell Sea area is more problematic. Evidence exists that the summer sea-ice experienced a strong expansion in the pre-EPILOG time period, between ∼25 and 20 ka, reaching the area of the present average sea-ice limit in the eastern Atlantic sector. This expansion led to a reduction in sea-ice seasonality. However, during the EPILOG LGM the summer sea-ice field was more reduced. The SSST and diatom cold water indicators point to sporadic expansions into the area of the present winter sea-ice edge; however, the average boundary was somewhere in the Weddell Sea area north of the present average summer sea-ice edge. To better understand the magnitude of the segregating effect of glacial sea-ice on the air-sea carbon dioxide exchange, new methods for a more accurate reconstruction of the glacial summer sea-ice field are urgently needed. Such methods, which may be based on a combination of paleo sea-ice modeling and SSST prescription from paleoceanographic reconstructions, should allow better determination of the distribution of the perennial ice cover as well as the sea-ice seasonality, both of which impact sea-air gas exchange, biological productivity, atmospheric circulation and water mass formation.

[33] Our data also provide hints as to the operation of the “cold” and “warm water” routes that are both crucial components of the global thermohaline circulation. While we suggest that the “cold water route” across the Drake Passage was weakened, the SSST data do not point to a strong reduction in the transfer of warm surface water from the Indian Ocean into the South Atlantic via the “warm water route.” However, for a more accurate reconstruction of the operation of the “cold water route” more data are needed from the southeastern Pacific and the southwestern South Atlantic.

[34] Sea-ice and SSST reconstructions based on a well established stratigraphic identification of the ELGM are also needed from the Pacific and Indian sectors to better understand the behavior of the LGM Southern Ocean on a circumantarctic scale, and to provide more accurate data for future numerical modeling of the LGM world.


[35] We thank Hans Schrader and Larry C. Peterson for review of the paper and useful comments. Discussions with U. Pflaumann, M. Sarnthein, M. Weinelt, and R. Spielhagen were also helpful. We also thank U. Bock, R. Cordelair, I. Klappstein, and T. Pollak for technical assistance. This work was generously supported by the Deutsche Forschungsgemeinschaft Sonderforschungsbereich 261: “Der Südatlantik im Spätquartär: Rekonstruktion von Stoffhaushalt und Stromsystemen.”