Because trace metals are supplied to sediments via a number of different delivery mechanisms, a suite of trace metals is required to differentiate the processes that produce reducing conditions in pore waters. Here we follow the definition of Berner (1981), and classify sediments whose pore waters contain no measurable dissolved oxygen or dissolved sulfide as suboxic, and those whose pore waters contain measurable dissolved sulfide as anoxic. Onset of suboxic conditions occurs between the reduction potentials of Mn and NO3- reduction (Froelich et al., 1979). Silver and Cd are thought to be delivered by sinking biogenic particles (Hendy and Pedersen, 2005; Wagner et al., 2013) and preserved as their respective sulfides where trace quantities of dissolved sulfide are present (Crusius and Thomson, 2003; Rosenthal et al., 1995b). Hence, enhanced preservation of these metals indicates suboxic conditions caused by a high flux of organic matter (Hendy and Pedersen, 2005). Cadmium also displays a strong tendency toward authigenic precipitation in sediments where reducing conditions occur close to the sediment–water interface (Calvert and Pedersen, 1993; Pedersen et al., 1989; Rosenthal et al., 1995b). In contrast, authigenic Ag2S precipitation appears to be negligible except in severely anoxic sediments where sulfate reduction is occurring (McKay and Pedersen, 2008). Elevated Re concentrations also indicate pore water suboxia. Rhenium exists in seawater as the conservative perrhenate anion (ReO4-) (Koide et al., 1986). It diffuses into sediments along concentration gradients and precipitates directly from seawater under reducing conditions, possibly as ReO2 (Crusius et al., 1996), although the probability for this reaction to proceed spontaneously in typical seawater has recently been questioned (Helz and Dolor, 2012; and references therein). Due to its low crustal concentrations, nearly all sedimentary Re is authigenically precipitated (Crusius et al., 1996). Hence, increases in Re accumulation without corresponding increases in Ag and/or Cd indicate that low bottom water O2 concentrations were likely present, and this deficit was primarily responsible for suboxia in underlying sediments.

Molybdenum enrichment occurs under anoxic conditions and requires measurable sulfide levels to precipitate. In seawater Mo is present as the conservative molybdate anion (MoO42-). Under oxic conditions, Mo associates with Mn-oxyhydroxides (Calvert and Pedersen, 1993; Zheng et al., 2000) which produces low-level enrichments unrelated to reductive processes. When redox conditions become favorable to Mn-oxyhydroxide reduction, adsorbed Mo can be released to pore waters (Crusius et al., 1996), decreasing preserved solid-phase Mo. Significant enrichments of Mo (up to tens of ppm in non-euxinic settings (Scott and Lyons, 2012)) above lithogenic background are controlled by pore water H2S concentrations. At sulfide concentrations below ∼11 µM, progressive substitution of O for S atoms occurs in molybdate, leading to the formation of multiple thiomolybdate species (MoOxS4-x2-) that co-precipitate as a labile Mo-Fe-S phase (Erickson and Helz, 2000; Helz et al., 1996). The substitution reaction is kinetically slow, but above the threshold value of 11 µM, a “geochemical switch” (Helz et al., 1996) is flipped and complete conversion to MoS42- (tetrathiomolybdate) occurs. Downcore measurements of Mo and H2S in Santa Barbara Basin sediments provide empirical support for this theoretical “geochemical switch” (Zheng et al., 2000). Tetrathiomolybdate is highly particle-reactive and adsorbs to pyrite, after which it is retained in sediments as a Mo-Fe-S cubane compound (Bostick et al., 2003). Above sulfide concentrations of ∼100 µM, direct precipitation of a discrete Mo-sulfide phase is hypothesized to occur (Zheng et al., 2000). Similar to Re, elevated Mo concentrations not coupled with increased Ag and/or Cd concentrations imply that low bottom water O2 concentrations caused anoxic conditions in pore waters.

Because Ag and Cd delivery to sediments is associated with biogenic particles, and Cd precipitates in sediments where oxygen penetrates 0.5 cm or less (Morford and Emerson, 1999), enrichment of Ag and Cd is assumed to be approximately contemporaneous with the modelled sediment age. Under reducing conditions, Re and Mo precipitation begin within ∼1 cm of the sediment–water interface (Crusius et al., 1996; Morford et al., 2009), minimizing any offset with the sediment age. A quantitative relationship has yet to be shown between sedimentary trace metal accumulation and export production or bottom water ventilation due to the combined effects of biogenic particle remineralization, sedimentation rate, dilution by other phases, and post-depositional re-oxidation of trace metals. For this reason, trace metals can provide a qualitative determination of the relative importance of export production vs. bottom water ventilation by comparing Ag and Cd (metals associated with biodetritus) enrichments with Re and Mo (authigenic metals) enrichments. For this study, when Ag and/or Cd concentrations increase with increasing Re and/or Mo concentrations, we interpret this pattern as high export production that depletes sedimentary oxidants. When Ag and/or Cd concentrations decrease with increasing Re and/or Mo concentrations, we interpret this pattern as a relative decrease in export production and concomitant decrease in bottom water oxygenation. However it is important to note that export production and bottom water oxygenation may operate synergistically to produce contemporaneous trace metal enrichments.

Core TN057-13-4PC was recovered from a small sedimentary basin northeast of Bouvet Island, at the same location as ODP Site 1094 (Kanfoush et al., 2002), so the two cores can be directly compared. Presently, TN057-13-4PC is located slightly to the south of the Southern ACC Front (Orsi et al., 1995) and ∼2∘ north of the average winter sea ice edge (Kanfoush et al., 2002). Primary productivity in this region is high (Moore and Abbott, 2000) and dominated by the heavily-silicified diatom Fragilariopsis kerguelensis (Abelmann et al., 2006). Sediments are composed primarily of diatomaceous ooze and diatom mud (Shipboard Scientific Party, 1999b). Site RC13-254 is located southwest of Meteor Rise in the Cape Basin, on the northern edge of the Antarctic Polar Frontal Zone (APFZ), and within the Permanently Open Ocean Zone. Productivity within the APFZ is still high, but transitions from diatom-dominated in the south to coccolithophorid-dominated in the north (Honjo et al., 2000). Consequently, sediments in the upper portion of the core are composed primarily of CaCO3 and transition to diatomaceous ooze during deglaciation (Charles et al., 1991).

Today both cores are overlain by Lower Circumpolar Deep Water (LCDW) (Fig. 2), a cold, saline, recirculating water mass of the Antarctic Circumpolar Current (ACC). Modern LCDW is a mixture of North Atlantic Deep Water (NADW), Antarctic Shelf Waters, and recirculated Pacific and Indian Deep Waters (PDW and IDW) that is exported to the three major ocean basins through Deep Western Boundary Currents (Talley, 2008). LCDW is ventilated from the north by penetration of NADW into the ACC, and from the south by upward mixing of Antarctic Bottom Water (Orsi et al., 1999). (Note, however, that bottom waters overlying TN057-13-4PC receive less influence from NADW than sites further to the north (Shipboard Scientific Party, 1999a), such as RC13-254/ODP 1091.) Once LCDW enters the Indian and Pacific basins, it upwells diapycnally to form the major volume of PDW and IDW. These subsequently re-enter the Atlantic basin through Drake Passage, mix with ACC waters, and upwell isopycnally to form modern Upper Circumpolar Deep Water (UCDW) (Talley, 2008), a nutrient-rich and slightly lower-oxygen water mass. Thus in the modern ACC, waters are well ventilated but the vertical water column structure can be characterized as having a lower-oxygen water mass overlying a higher-oxygen water mass (Fig. 2b).

Cores were sampled at millennial- to submillennial-scale resolution from the core tops to ∼30 ka. Concentrations of Ag, Cd, Re, and Mo were measured at the University of Michigan Keck Environmental Geochemistry Laboratory (KEGL) by isotope dilution inductively coupled plasma-mass spectrometry (ID ICP-MS) using a Finnigan Element mass spectrometer. Prior to analysis, samples were spiked and subjected to total microwave digestion in Teflon vessels using an acid cocktail of trace metal grade HF, HCl, and HNO3 (Crusius et al., 1996). An aliquot was removed from digested samples for Cd and Ag analysis and passed through a column of Dowex 1X-8 resin to remove interfering isotopes. Working spike concentrations and ratios of 109Ag/107Ag, 111Cd/113Cd, 185Re/187Re, and 95Mo/97Mo were calibrated using natural element ICP standards (Ricca Chemical Company). Errors are reported as one-sigma standard deviations for each run based on repeated analysis of the sediment standard MESS-3.

Weight percent organic carbon (Corg) was determined by paired analysis of acidified and unacidified bulk sediment samples using a Costech ECS4010 elemental analyzer. Relative SD is 3 % based on duplicate analysis of ∼10 % of samples.

Published age models were used for both cores. The initial age model for core RC13-254 (Charles et al., 1991) was based on oxygen isotope chronostratigraphy. More recently, calibrated 14C dates have been incorporated that indicate that the core top is ∼7 kyr younger than originally estimated, and the age model for the upper ∼190 cm of RC13-254 has been modified accordingly (R. F. Anderson, personal communication, 2011; Anderson et al., 1998; Kumar et al., 1995). The age of the core top (0–1 cm) was set to 3500 years before present. A constant sedimentation rate was assumed between age-depth points provided by R. F. Anderson (personal communication). The age model for core TN057-13-4PC is based on a series of 14C dates (Shemesh et al., 2002) that were recalibrated using the program CALIB 6.0 and the Marine09 calibration, with an assumed 800-year reservoir effect. A constant sedimentation rate was assumed between age control points. All dates are shown in calendar kyr.

Oxidative burndown occurs when previously deposited sediments become re-exposed to oxygen, remobilizing reduced trace metal phases. It is more likely to occur in slowly accumulating sediments and during times of rapid environmental change, such as during a glacial–interglacial transition (Tribovillard et al., 2006). Once remobilized, trace metals can either diffuse downward into the sediment column if concentration gradients decrease in that direction, or diffuse upward out of sediments into the water column. In the former case, evidence of oxidative burndown usually presents as a sharp peak below a redox front. The effect has been observed downcore for Re in organic carbon-rich turbidites (Crusius and Thomson, 2003) and for Cd and U in Subantarctic sediments (Rosenthal et al., 1995a) directly below the glacial–interglacial transition.

Burndown is likely to have affected the upper ∼200 cm of TN057-13-4PC and the upper ∼60–70 cm of RC13-254. Evidence for oxidative burndown is clearest for RC13-254. An abrupt transition from very low to glacial-level trace metal concentrations occurs between ∼60 and 70 cm core depth (∼13–13.5 ka), coinciding with decreases in pore water NO3- and increases in pore water Mn concentrations at a nearby core (King et al., 2000). No evidence for oxidative burndown is apparent further downcore as at no other point do all redox-sensitive trace metals decrease abruptly and coherently to crustal concentrations. Muted trace metal enrichments in TN057-13-4PC from the core top to ∼200 cm (∼5.5 ka) suggest the onset of oxidative burndown during the late Holocene. However, NO3- and Mn pore water profiles (King et al., 2000) indicate that the present oxic-suboxic transition occurs ∼20–30 cm core depth (∼0.7–1 ka), implying that a slowdown in trace metal delivery rather than re-oxidation may exert primary control over trace metal enrichments in the upper portion of the core.

However, available proxy data fail to give an unambiguous determination of the existence and timing of burndown in TN057-13-4PC near the glacial–interglacial transition. The sharp Ag and Cd peaks observed in TN057-13-4PC between ∼820 and 880 cm core depth (∼21.6–26.9 ka) indicate possible burndown. If true, co-occurring Re peaks appear to have migrated further downcore. However, U fluxes increase slightly from ∼832 to 884 cm (∼22.1–26.5 ka) (Anderson et al., 2009), suggesting that burndown was minimal or affected U in a similar – though less severe – manner to Ag and Cd. Changes in Corg flux can control the onset of burndown (Crusius and Thomson, 2000) and may help to diagnose when burndown began. Sedimentary Corg concentrations in TN057-13-4PC decline from a maximum of about 0.5 % to about 0.3 % from 30 ka to the end of MIS 3 (∼22 ka), coincident with the large Ag, Cd, and Re peaks. Subsequently, agreement among Ag, Cd, Re, and Corg peaks between ∼760 and 820 cm (∼16.2–21.6 ka) suggests a true increase in export productivity. Sedimentary Corg concentrations in the Antarctic Zone therefore increase going into the LGM and a decrease in the intensity of organic matter respiration cannot be responsible for re-oxygenation of sediments. Rather, burndown appears to have occurred sometime during late MIS 3. Therefore, with the evidence at hand, we will assume that oxidative burndown has had a minimal effect on trace metal concentrations near the glacial–interglacial transition.

Trace metal concentrations in Holocene-age sediments reported here for the Atlantic sector Southern Ocean are low, similar to those measured for Pacific sector core top sediments both within and outside the opal belt (Wagner et al., 2013). These data suggest that marine chemistry unfavorable to trace metal accumulation exists throughout the modern Southern Ocean. In contrast, trace metal concentrations at both sites are significantly higher downcore (Figs. 3a–d, 4a–d). Although both RC13-254 and TN057-13-4PC are situated in the open ocean, trace metal concentrations in deglacial- and glacial-age sediments are similar to concentrations found at modern continental margin sites (e.g., McKay and Pedersen, 2008; Morford and Emerson, 1999). These continental margin sites are often located in oxygen minimum zones such as the eastern Pacific Ocean where dissolved oxygen concentrations are much lower than the ∼200–225 µM typical of modern LCDW (Fig. 2b).

At RC13-254 (Cape Basin, modern APFZ), the obvious difference in trace metal enrichments between late MIS 2 and the late Holocene (Fig. 3a–d) indicates a significant change in deep Southern Ocean chemistry ∼3.6 km water depth. Prior to ∼13.5 ka, the record presents a strong case for pore water suboxia near the sediment–water interface. During MIS 3, mostly elevated Ag (79–273 ppb) and Cd (0.22–1.26 ppm) concentrations coupled with a trend of relatively stable Re concentrations (average 16.3 ppb) that do not correspond to changes in Ag and Cd concentrations suggest that low bottom water O2 concentrations acted synergistically with variations in primary productivity to produce suboxic pore waters. Transitioning into the Last Glacial Maximum (∼19.1–21.1 ka; ∼181–206 cm), decreasing Ag and Cd and increasing Re concentrations (maximum 38.5 ppb) suggest that low bottom water O2 concentrations became more important in controlling sedimentary redox conditions. Subsequent deglacial increases in Ag and Cd concentrations that correspond to Re concentrations from ∼13.5–19.1 ka (∼70–181 cm) imply that suboxic pore waters resulted primarily from high organic matter flux to sediments. After ∼13.5 ka (above ∼70 cm), Ag, Cd, and Re concentrations greatly decrease to near or below crustal concentrations, probably due to oxidative burndown (see Sect. 4.1). The Mo record is dissimilar to the other trace metals and suggests little, if any, authigenic enrichment (Fig. 3d). A possible exception occurs ∼25.4 ka (∼309 cm) when a corresponding increase in Cd concentrations is observed.

Glacial- and deglacial-age Ag, Cd, and Re enrichments (Fig. 3a–c) do not likely result from lithogenic input. Estimates of terrigenous content do not exceed 35 % (Latimer and Filippelli, 2001, Fig. 3g) and Ag, Cd, and Re concentrations are nearly always greater than bulk continental crust, a purely lithogenic phase (Fig. 3a–c). Furthermore, Corg concentrations (Fig. 3e) correspond well to Ag and Cd concentrations, substantiating their interpretation as a record of export production.

The pattern of trace metal enrichments at TN057-13-4PC (modern Antarctic Zone) is dissimilar to that at RC13-254. The overall record (Fig. 4a–d) suggests that bottom waters at the site have been well ventilated for approximately the last 25 kyr, with organic matter respiration mostly responsible for periods of sedimentary suboxia. Silver, Cd, and Re show large enrichments prior to ∼16 ka (below ∼760 cm) (Fig. 4a–c). Rhenium concentrations as high as 65 ppb (Fig. 4c) between ∼25.2 and 30.6 ka (∼860–921 cm) suggest that low bottom water O2 concentrations caused suboxic pore waters during this time. Rhenium concentrations then decrease as Ag and Cd concentrations increase to maxima of 589 ppb and 1.13 ppm, respectively, between ∼23.4 and 25.2 ka (∼840–860 cm). These trends suggest increasing export production and bottom water O2 concentrations. From ∼16.2–23.4 ka (∼760–840 cm) Re peaks show reasonable agreement with Ag and Cd, indicating a change to suboxia driven by high export production. From ∼11–16.2 ka (∼560–760 cm), relatively low Ag, Cd, and Re concentrations suggest improved bottom water ventilation at the site. Long-term increases in Ag, Cd, Re, and Corg concentrations between ∼5.5 and 11 ka (∼200–560 cm) suggest that trace metal enrichments responded to pore water suboxia driven by high export production, but coeval opal fluxes decline. Well-oxygenated sediments similar to modern conditions appeared ∼5.5 ka (∼200 cm), coincident with the onset of the Neoglacial (Hodell et al., 2001) and increased sea ice presence at the site (Stuut et al., 2004, Fig. 4h). Burndown and/or a slowdown in trace metal delivery appears to be responsible for low trace metal concentrations after ∼5.5 ka (see Sect. 4.1). Similar to site RC13-254, Mo concentrations are low throughout the core (Fig. 4d), thus indicating no authigenic enrichment.

Estimates of terrigenous content from Ti-normalization (Fig. 4g) suggest that increased lithogenic input probably contributed to high total Ag, Cd, and Re concentrations during early MIS 2 and MIS 3. However, similar to site RC13-254, estimates of terrigenous content are no more than 55 % and therefore cannot be the only factor contributing to high trace metal concentrations. Sedimentary Corg concentrations (Fig. 4e) also correspond reasonably well to Ag and Cd concentrations in TN057-13-4PC. However, as noted in Sect. 4.1, it is impossible to unequivocally state that oxidative burndown near the glacial–interglacial transition has not affected the record. Additionally, increasing terrigenous input might dilute biogenic phases.

In summary, Ag, Cd, and Re enrichments in sediments from the modern APFZ (RC13-254) and the Antarctic Zone (TN057-13-4PC) reveal millennial-scale changes in both productivity and bottom water ventilation. These changes indicate a significant shift in Southern Ocean water chemistry from the last glacial period and deglaciation to the late Holocene, when sedimentary trace metal accumulation is nearly absent. Additionally, the two cores are separated by only ∼5∘ of latitude, and so the different trace metal records demonstrate glacial/deglacial spatial heterogeneity in the Southern Ocean.

Constraining the timing, location, and intensity of bottom water ventilation changes can help to unravel the mechanism of glacial–interglacial atmospheric CO2 variations. Rhenium is a valuable tool for assessing bottom water ventilation changes because it precipitates directly from seawater under suboxic conditions and is not cycled with biogenic particles (Colodner et al., 1993; Crusius et al., 1996). The radiocarbon (14C) content of deep ocean waters is an alternate, more common measure of water mass ventilation that has been applied to the last glacial period and deglaciation (e.g., Broecker and Barker, 2007; Marchitto et al., 2007; Robinson et al., 2005). By comparing these two measures of bottom water ventilation, a consistent picture emerges in the Atlantic. Rhenium concentrations from RC13-254, Δ14C apparent ventilation ages from Subantarctic core MD07-3076 (44.074∘ S, 14.208∘ W, 3770 m) (Skinner et al., 2010), and Δ14C apparent ventilation ages from northeast Atlantic core MD99-2334K (37.80∘ N, 10.17∘ W, 3146 m) (Skinner et al., 2014) were binned into 500 year intervals to establish a common age scale. Both Δ14C apparent ventilation age records are positively correlated with Re concentrations at this temporal resolution (Fig. 5). Most prominently, elevated Re concentrations from ∼19 to 20 ka correspond to the highest measured benthic-atmospheric Δ14C offsets of nearly 4 kyr (Fig. 6), thus supporting the conclusion of Skinner et al. (2010) that the carbon sequestration capacity of the Southern Ocean was enhanced at this time due to slower ventilation rates at the depth of modern LCDW. Overall, oxygen depletion appears to have intensified in the deep Atlantic as global climate moved into the full glacial conditions of the LGM, became less intense during deglaciation, and then Southern Ocean oxygen concentrations reset to more modern conditions sometime after ∼13.5 ka.

A lower-resolution Re record from PS2489-2 (42.873∘ S, 8.973∘ E, 3700 m) (Martínez-Garcia et al., 2009, Fig. 6b) corroborates the pattern from RC13-254 of lower Re concentrations/lower ventilation ages during MIS 3 and higher Re concentrations/higher ventilation ages during MIS 2. Rhenium data (this study) also support the prediction (Sarnthein et al., 2013) that oxygen-depleted conditions similar to the modern Eastern Pacific oxygen minimum zone (see Sect. 4.2) occurred in the Atlantic sector of the Southern Ocean at the LGM. Furthermore, recent deep-sea coral ΔΔ14C measurements from Drake Passage (Burke and Robinson, 2012) tentatively support better ventilation at depths corresponding to modern UCDW compared to LCDW during the LGM.

Δ14C (MD07-3076) and trace metal (RC13-254) results therefore suggest that ocean circulation changes may have inhibited glacial LCDW from mixing with better ventilated waters and/or that vertical stratification increased, similar to conclusions drawn from δ18O and δ13C tracer data (Lund et al., 2011). Other observations further support these conclusions: (1) Glacial North Atlantic Intermediate Water (GNAIW) penetrated only to ∼30∘ S (Curry and Oppo, 2005), reducing ventilation from the north. Lund et al. (2011) also suggest that shoaling of modern North Atlantic Deep Water to GNAIW reduced mixing by bottom topography. (2) Indian Ocean δ15N measurements (Francois et al., 1997) suggest that increased Southern Ocean vertical stratification was responsible for lowered bottom water O2 concentrations in glacial LCDW. One plausible mechanism to increase Southern Ocean vertical stratification invokes increased sea ice extent and cooling of the open ocean water column as a driver for increased brine formation (Keeling and Stephens, 2001).

The sum of the data therefore points to glacial LCDW as the storage reservoir of glacial CO2. Because modern LCDW forms the bulk of Indian and Pacific Deep Waters (Talley, 2008), a low 14C signature during the last glacial period is also expected for deep Pacific and Indian Ocean sites downstream of glacial South Atlantic LCDW. In agreement with this prediction, apparent ventilation ages ∼3600 years older than modern values (and ∼2500 years older than those reported by Skinner et al., 2010) were measured at Chatham Rise (Sikes et al., 2000; Vandergoes et al., 2013), where the Deep Western Boundary Current flows into the Southwest Pacific (Carter and McCave, 1997). Additionally, increased LGM benthic-planktonic age differences have been observed in the deep North Pacific (Galbraith et al., 2007). Targeting glacial LCDW and its downstream deep water products (glacial IDW and PDW, as well as deep Atlantic waters; Orsi et al., 1999) allows for a radiocarbon-depleted Southern Ocean source to be diffused into a large volume of deep ocean water, thereby circumventing the need for a smaller, extremely depleted 14C reservoir (Broecker and Barker, 2007).

Based on glacial age benthic-planktonic age differences from the western Pacific that are similar to modern values, it has been argued that an isolated, 14C-depleted abyssal reservoir could only have existed below 2.8 km water depth (Broecker et al., 2008). This extremely “old” reservoir would then have dissipated by mixing into the global ocean at the deglaciation (Broecker and Barker, 2007). However, Burke and Robinson (2012) point out that the age of such a reservoir would need to be less extreme if a portion of the trapped CO2 is allowed to vent directly to the atmosphere, perhaps through increased Southern Ocean upwelling (Anderson et al., 2009). Using the volume and average depth of each major ocean basin, a simple calculation estimates that the total volume of the Atlantic, Indian, and Pacific Oceans below 2.8 km water depth is ∼25 % of global ocean volume. Because modern LCDW feeds all three basins, it is reasonable to suppose that glacial LCDW acted as the isolated reservoir, dispersing radiocarbon-depleted water throughout the deep ocean. Solid evidence for a globally distributed, “old” oceanic carbon reservoir is so far lacking. However, benthic-planktonic age differences often assume constant surface reservoir ages. If this assumption is not valid, and surface water reservoir ages are high, the “true” age of deep waters may not be apparent. Furthermore, CaCO3 is not well preserved in the deep Pacific, and suitable samples from depths below 2.8 km are difficult to obtain (Broecker et al., 2007).

Presently, LCDW bathes site TN057-13-4PC (Antarctic Zone); however, the lack of strong Re enrichments (Fig. 4c; Table S1) suggests that a different, well-ventilated water mass (perhaps a glacial equivalent of UCDW) existed at 2.8 km. The contrast with ventilation records from RC13-254 (Cape Basin) showing older, more O2-depleted glacial LCDW suggests that the modern water mass arrangement in which a lower-O2 component of CDW overlies a higher-O2 component may have been reversed during the last glacial period. The ventilation source at TN057-13-4PC is unclear, however, and the paucity of cores from all sectors of the Antarctic Zone leaves the glacial vertical water column structure an open question.

Most other trace metal records available from the Southern Ocean utilize Cd or authigenic U as suboxic indicators, although interpreting bulk sediment U is not straightforward because its accumulation is influenced by a combination of bulk sedimentation rate/sediment focusing, organic carbon rain rate, and bottom water O2 concentration (Frank et al., 2000). Higher Cd and U accumulation rates/concentrations in Subantarctic sediments during the last glacial period have been primarily interpreted to record suboxic conditions caused by a northward shift of the ACC high productivity zone (Anderson et al., 1998; Bareille et al., 1998; Chase et al., 2001; Frank et al., 2000; Kumar et al., 1995; Rosenthal et al., 1995a). Elevated glacial Ag, Cd, and Corg concentrations (Fig. 3a, b, e) concur with previous work by demonstrating strong export production north of the modern APF. In comparison, Antarctic Zone sediments from the Indian and Atlantic sectors give inconsistent results. In the Atlantic sector, some studies show minima in bulk sediment U concentrations and Corg and Sibio (opal) fluxes at the LGM before increasing during deglaciation (Anderson et al., 1998; Frank et al., 2000), when the high productivity zone shifted southward again. Other studies from the Indian sector demonstrate modest increases in bulk sediment Cd and U accumulation rates during the LGM similar to deglacial accumulation rates (Rosenthal et al., 1995a). Organic carbon accumulation rates also show a modest increase during the LGM, but afterward increase rapidly during the early deglaciation, suggesting that preservation issues could have influenced proxy accumulation. South Atlantic trace metal concentrations (this study) follow the results of Anderson et al. (1998) most closely by decreasing toward the LGM and increasing through the deglacial and Holocene (Fig. 4a–d).

A simple explanation of trace metal enrichment patterns is complicated by Cd and U accumulation trends that imply decreased CDW oxygen concentrations (Bareille et al., 1998; Francois et al., 1997; Rosenthal et al., 1995a) in addition to glacial–interglacial shifts of the high productivity zone. Increasing Re concentrations coupled with decreasing Ag and Cd concentrations during the LGM (Fig. 3a–c) support this conclusion, and suggest that export production and bottom water O2 concentrations both influenced trace metal accumulation north of the modern APF. However, available data do not provide a clear indication of oxygenation levels in the glacial Antarctic Zone, and how they may have changed over time. Further studies employing a suite of redox proxies are needed from all three sectors of the Southern Ocean to sufficiently characterize glacial deep water ventilation.