A stagnation event in the deep South Atlantic during the last interglacial period

See allHide authors and affiliations

Science  19 Dec 2014:
Vol. 346, Issue 6216, pp. 1514-1517
DOI: 10.1126/science.1256620


During the last interglacial period, global temperatures were ~2°C warmer than at present and sea level was 6 to 8 meters higher. Southern Ocean sediments reveal a spike in authigenic uranium 127,000 years ago, within the last interglacial, reflecting decreased oxygenation of deep water by Antarctic Bottom Water (AABW). Unlike ice age reductions in AABW, the interglacial stagnation event appears decoupled from open ocean conditions and may have resulted from coastal freshening due to mass loss from the Antarctic ice sheet. AABW reduction coincided with increased North Atlantic Deep Water (NADW) formation, and the subsequent reinvigoration in AABW coincided with reduced NADW formation. Thus, alternation of deep water formation between the Antarctic and the North Atlantic, believed to characterize ice ages, apparently also occurs in warm climates.

A brief hiccup in deep ocean circulation

During the last interglacial period, Antarctic Bottom Water (AABW) formation slowed markedly. This densest ocean water sinks to the bottom of the sea, and its production helps to flush the oceans and eventually to recycle the carbon dioxide (CO2) that forms from sinking organic matter back into the atmosphere. If the AABW production rate decreases, then CO2 accumulates at depth, potentially causing a corresponding drop in atmospheric CO2 concentration. Hayes et al. found evidence, in the form of a uranium spike, in deep sea sediments that such a slowdown in AABW formation occurred ∼127,000 years ago, which may have caused the atmospheric CO2 minimum observed at that time.

Science, this issue p. 1514

The circulation and biological productivity of the Southern Ocean may help regulate atmospheric CO2 over millennial to glacial-interglacial time scales (1, 2). Evidence underpinning this view comes largely from study of climate oscillations during ice ages. Here, we report evidence for rapid changes in Southern Ocean circulation during the last interglacial period that is particularly relevant to today’s warming climate. Authigenic uranium (aU), a redox-sensitive trace-element proxy for the oxygen content in sediment pore waters (3), reveals a millennial-scale reduction in the ventilation of the deep Southern Ocean around 127,000 years ago, when global temperatures were ~2°C warmer than they are today (4, 5).

Most sediment pore water oxygen concentrations are controlled primarily by the balance between diffusive supply from bottom water and consumption by respiration of sedimentary organic matter. We find evidence in samples from Ocean Drilling Program (ODP) Site 1094 (53.2°S 5.1°E, 2807 m water depth, Fig. 1) for changes in deep water oxygen supply by combining aU measurements with controls on the supply of organic matter to the sediments. In this core, the 230Th-normalized biogenic opal flux records changes in the vertical rain of organic matter (2, 6). Lateral supply of organic matter by sediment redistribution, or focusing, is assessed by calculating the focusing factor, Ψ (7). For age control during the period surrounding Termination II (T-II), a new sea-surface temperature record from Site 1094, based on the marine lipid index, TEX86L, was correlated to the Vostok ice core δD (7, 8) (Fig. 2A, fig. S1, and table S1).

Fig. 1 Vertical section along 5°E in the South Atlantic putting ODP Site 1094 into hydrographic context.

Color-mapped oxygen concentrations, with salinity contours overlain, as drawn from the 2009 World Ocean Atlas (30, 31). ODP Site 1094 is marked with a white circle. The inset map indicates the plotted section (red line) and the locations of other cores mentioned in the text: MD03-2664, white triangle; ODP Site 1089, white diamond; and PS2561-2, white square. Figure created with Ocean Data View (

Fig. 2

Records from ODP Site 1094 spanning the penultimate glacial period to the last interglacial. (A) TEX86L sea-surface temperature (SST) aligned with δD of the Vostok ice core (8, 32). (B) Bulk (square wave) and 230Th-normalized (blue line) mass accumulation rates and 230Th-normalized biogenic opal flux (green line). (C) Focusing factor, the intensity of lateral supply of sediments. (D) Authigenic uranium (aU) accumulation rate calculated as an instantaneous function of the aU concentrations (discrete points) and as interval averages (square wave). (E) aU concentrations from Site 1094 and mean size of sortable silt from Agulhas basin core PS2651-2 (7, 13). Age model tie points for Site 1094 are marked with inverted triangles. Marine isotopestages (MIS) and Termination II (T-II) are roughly indicated and are consistent with available foraminiferal δ18O data (25). Gray shading highlights T-II, and yellow shading highlights the authigenic uranium event.

The penultimate glacial period [marine isotope stage (MIS) 6] at Site 1094 is characterized by low export production (Fig. 2B), typical of the Antarctic South Atlantic during ice ages (9). Although substantial sediment focusing occurred during MIS 6 (Ψ = 4, Fig. 2C), these glacial age sediments are low in biogenic opal, our proxy of organic carbon export. Given the absence of an enhanced supply of organic carbon to the sea floor (either vertically or laterally), the aU enrichment during glacial MIS 6 (Fig. 2E) must be due to low bottom water oxygen concentrations, consistent with sluggish ventilation of the Southern Ocean during glacial periods (10). This scenario is corroborated by glacial-age aU enrichments observed in other sediment records from the Atlantic (9, 11) and Indian (12) sectors of the Southern Ocean.

During the deglacial period (T-II), export production increased rapidly, with 230Th-normalized opal fluxes increasing roughly 10-fold (Fig. 2B). Accounting for concurrent changes in sediment focusing, although aU concentrations decreased (Fig. 2E), the mass accumulation rate (MAR) of aU increased from MIS 6 into the earliest part of T-II (Fig. 2D), consistent with an increased load of organic matter to the sea floor enhancing aU deposition. During the course of T-II, export production remained high or even increased, whereas aU MAR gradually decreased (Fig. 2D). We therefore infer that the gradual deglacial decline in aU MAR was driven by increased oxygenation in deep water as a consequence of increased Southern Ocean overturning during T-II.

Against this backdrop of MIS 6–to–T-II changes that are consistent with previous findings, we observed an entirely unexpected feature in the interglacial period MIS 5e: a 3000- to 4000-year spike in aU at 127 ± 1 ka (ka, thousand years ago) (7). Similar peaks in both aU concentration and aU MAR (Fig. 2, D and E; 125 to 129 ka) indicate that the rapid aU enrichment cannot be primarily explained by changes in dilution. Furthermore, the opal flux and focusing factor records (Fig. 2, B and C) show that vertical and lateral supplies of organic matter decreased from T-II to early MIS 5e. This combination of increased aU and reduced organic carbon sedimentation requires that the supply of oxygen to the bottom waters at Site 1094 went through a minimum around 127,000 years ago.

Today, Site 1094 is bathed by circumpolar deep water (CDW) which, in turn, is composed of contributions, mixed together in the Antarctic Circumpolar Current, from North Atlantic Deep Water (NADW), Indo-Pacific Deep Water (IPDW), and Antarctic Bottom Water (AABW) (Fig. 1). Both NADW and AABW supply oxygen to global deep waters. The last interglacial aU peak at Site 1094 is most simply explained by a reduction in AABW, partly because its formation region is most proximal to Site 1094. This is supported by a record from the abyssal Agulhas basin (PS2561-2, 41.9°S 28.5°E, 4465 m water depth), which contains a notable minimum in sortable silt size during MIS 5e at the time of the aU peak at Site 1094 (Fig. 2E), interpreted as a slow-down of AABW flow (13).

During glacial periods in general and especially during the Antarctic cold intervals preceding Heinrich events, there is evidence for reduced Antarctic productivity, a higher degree of nitrate consumption, and lower atmospheric CO2, which taken together suggest reduced Antarctic ventilation of the interior and thus greater deep ocean storage of regenerated carbon (2, 12, 14). aU enrichment during ice ages, such as is evident in the data reported here (Fig. 2), supports this interpretation. Previous authors have postulated a mechanistic link between Antarctic overturning and the strength and position of the global wind belts during glacial periods (15). This hypothesis has been strengthened by the observations that Antarctic productivity changes were coeval with changes in the position of the Intertropical Convergence Zone (ITCZ) (16) and in the strength of the Asian Monsoon (17).

In contrast, the aU spike within MIS 5e occurs in the context of little coherent change in Antarctic productivity (Fig. 2). Neither does the aU event correlate with possible far-field drivers, such as ITCZ migration (16, 17). Given these findings, the mechanism for the MIS 5e event was most likely fundamentally different than the process that reduced Antarctic deep ventilation during ice ages.

Following peak temperatures in Antarctica, global sea-level rose by ~20 m in the period 128 to 125 ka (18, 19) (Fig. 3A). Substantial melting at the margins of the Antarctic ice sheet at this time (20) may have freshened Antarctic coastal waters, interfering with the AABW formation that occurs on the Antarctic shelves today. This freshening may also have reached adequately far from shore to reduce deep ocean convection in the Weddell Sea (21). Admittedly, we have no direct evidence for Antarctic melting as the cause of the AABW reduction, and, as described below, other possibilities exist.

Fig. 3 Seesawing deep circulation events during the last interglacial period compared to changes in global sea level.

(A) Estimated record of global sea level (19) and (C) its rate of change (19). Heavy lines mark the median projections of data-based probability density distribution, with dotted lines representing the 16th and 84th percentiles. (B) Benthic foraminiferal carbon isotopic composition from MD03-2664 (24), ODP Site 1089 (25), and ODP Site 1094. (D) Authigenic uranium mass accumulation rate (aU MAR) from Site 1094.

Coinciding with the interglacial peak aU accumulation at Site 1094, the brief drop and rebound in atmospheric CO2 (Fig. 4, B and C, 127 ka) is consistent with increased deep ocean storage of regenerated carbon due to a reduction in AABW formation. Under interglacial conditions, a box model suggests that a shutoff of AABW formation should cause a CO2 decline of ~33 parts per million (ppm) due to increased regenerated carbon storage and related deep ocean CaCO3 dissolution (22). However, a partial offset is expected from the CO2 rise driven by warming of the deep ocean due to the loss of AABW. Given modern deep ocean conditions, AABW shutdown would translate to an ocean interior warming of 1° to 2°C, as NADW becomes the sole ventilator of the abyss, yielding a temperature-driven CO2 rise of 10 to 20 ppm (23). Taking the difference yields a net CO2 decline of 13 to 23 ppm, indistinguishable from the range in the reconstructions of the decline, 13 to 24 ppm (Fig. 4B).

Fig. 4

Records from ODP Site 1094 in the context of atmospheric composition and North Atlantic circulation. (A) TEX86L SST aligned with δD of the Vostok ice core (8, 32). (B) Atmospheric CO2 from the EPICA Dome C (33, 34) and Vostok (32) ice cores, the spread in which indicates the uncertainty associated with these measurements. (C) aU accumulation rate. (D) The carbon isotopic composition of benthic foraminifera from MD03-2664, with three-point running means in black, (24), ODP Site 1089 (25), and ODP Site 1094.

The counteracting effect of deep ocean warming may explain why the mid-interglacial AABW shutdown discovered here coincided with a weaker CO2 decline than has been proposed for ice age reductions in AABW formation, of 30 to 40 ppm or more (22). During glacial times, reduced Antarctic ventilation may have coincided with climate cooling that prevented deep ocean warming, yielding a greater net CO2 decline. The 3000- to 4000-year duration of the AABW shutdown during MIS 5e also may have been too short for ocean calcium carbonate dynamics to express the full CO2 effect that applies to extended glacial periods.

The period 124 to 125 ka, immediately following the aU event at ODP Site 1094, is characterized by a reduction in NADW formation (24), as recorded by benthic foraminiferal δ13C in core MD03-2664 (57.4°N, 48.6°W, 3442 m water depth) in the North Atlantic (Fig. 4D). The isotopic signature of this event propagated into the Southern Ocean. Within age model uncertainties, we observe a synchronous decrease in benthic δ13C (Cibicidoides wuellerstorfi) of ODP Sites 1089 (25) and 1094 (Fig. 4D). This demonstrates that the reduction of AABW recorded by the aU peak at ODP Site 1094 preceded the reduction in NADW. The phasing suggests that the millennial “bipolar seesaw” in deep water formation between the Antarctic and North Atlantic reconstructed for the last ice age (26) also occurs during warm interglacials.

The climate sensitivity of the wind belts has been proposed as the central connection driving millennial-scale North Atlantic–Antarctic seesaw of the last ice age (15). However, when the aU peak at Site 1094 ended, we are aware of no evidence for an atmospheric teleconnection between AABW and NADW formation. Rather, the coincidence of the reinvigoration of AABW with the reduction in NADW is best explained by changes in deep ocean density (26). With the loss of deep water formation in one hemisphere, the density of the deep ocean declines, leading to a restart or reinvigoration of deep water formation in the other hemisphere on millennial time scales. The reduction in AABW may have encouraged a brief increase in NADW formation at ~127 ka (24) and thus spurred additional circum–North Atlantic ice loss, which has been proposed to drive a well-documented reduction in NADW around 125 ka (24). The NADW reduction, in turn, may have driven the reinvigoration of the AABW that is recorded by the end of the aU peak at 125 ka.

As to the cause of AABW reduction at ~127 ka, rather than being driven by Antarctic melting, it is possible that this event also originated from a deep ocean density teleconnection, triggered by the interglacial restart of NADW formation. However, the δ13C data suggest that NADW formation at ~127 ka was not stronger than in the Holocene, which is characterized by deep water formation in both the North Atlantic and the Antarctic. Conversely, Antarctic temperature and sea-level rise were anomalously high before and during the AABW reduction, favoring the Antarctic melting trigger.

Although the South Atlantic stagnation occurred early during the last interglacial, there may be an analogy between this event and the reduction in Southern Ocean ventilation that has occurred over the past 40 years, during a much later stage in the current interglacial. Since 1976, deep convection in the Weddell Sea, as identified by polynya formation, has been absent (27), possibly due to anthropogenic warming and an associated freshening of the surface Southern Ocean (28). Such a tendency would be reinforced by coastal freshwater input should Antarctic ice loss accelerate in the coming decades and centuries. In this context, although there is great concern and substantial investigation regarding the potential for reduced NADW formation in the future (29), a marked reduction in AABW formation upon warming would not be unprecedented, in light of the previous interglacial. Moreover, the evidence that the bipolar seesaw in deep water formation (26) also applies to interglacial climate suggests that a global warming–induced reduction in deep water formation in either the North Atlantic or the Southern Ocean would lead to an increase in deep water formation in the other region.

Supplementary Materials

Materials and Methods

Figs. S1 and S2

Tables S1 to S4

References (3551)

References and Notes

  1. See supplementary materials on Science Online.
  2. Acknowledgments: We acknowledge the Integrated Ocean Drilling Program for providing samples. Data not previously reported are archived at the U.S. National Climate Data Center ( and Pangaea ( and can be accessed in tables S3 and S4. Financial support was provided by the Comer Science and Education Foundation (CTH/RFA), the W.O. Crosby Fellowship (C.T.H.), Swiss National Science Foundation grants PZ00P2_142424 (AM-G) and PP00P2_144811 (SLJ), and ETH grant ETH-04 11-1 (SLJ/APH).
View Abstract

Navigate This Article