The geostrophic shear associated with the meridional overturning circulation is reflected in the difference in density between the eastern and western margins of the ocean basin. Here we examine how the density difference across 30°S in the upper 2 km of the Atlantic Ocean (and thus the magnitude of the shear associated with the overturning circulation) has changed between the last glacial maximum and the present. We use oxygen isotope measurements on benthic foraminifera to reconstruct density. Today, the density in upper and intermediate waters along the eastern margin in the South Atlantic is greater than along the western margin, reflecting the vertical shear associated with the northward flow of surface and intermediate waters and the southward flowing North Atlantic Deep Waters below. The greater density along the eastern margin is reflected in the higher δ18O values for surface sediment benthic foraminifera than those found on the western margin for the upper 2 km. For the last glacial maximum the available data indicate that the eastern margin foraminifera had similar δ18O to those on the western margin between 1 and 2 km and that the gradient was reversed relative to today with the higher δ18O values in the western margin benthic foraminifera above 1 km. If this reversal in benthic foraminifera δ18O gradient reflects a reversal in seawater density gradient, these data are not consistent with a vigorous but shallower overturning cell in which surface waters entering the Atlantic basin are balanced by the southward export of Glacial North Atlantic Intermediate Water.
Today there is a relatively strong meridional overturning circulation in the Atlantic Ocean basin, reflecting the export of North Atlantic Deep water and the compensating inflow of surface and intermediate waters. Most data-based reconstructions [e.g., Ganachaud and Wunsch, 2000; Roemmich and Wunsch, 1985; Schmitz, 1995; Schmitz and McCartney, 1993; Talley et al., 2003] of this overturning cell show that the strength of the shallow inflow of thermocline and intermediate waters is relatively constant throughout the subtropics and tropics from the tip of Africa in the south (30°S) to just south of the polar seas in the north (Figure 1). The crossover between northward flowing upper waters and southward flowing North Atlantic Deepwater occurs at approximately 1 km in both the North and South Atlantic. It is still unclear whether a similar surface to deep overturning cell in the Atlantic, along with the associated northward transport of heat, existed during the last glacial maximum (LGM).
Deepwater tracer distributions inferred from chemical and isotopic measurements on the tests of benthic foraminifera suggest a strong stratification in the North Atlantic Ocean, with a low nutrient, high δ13C water mass (often called Glacial North Atlantic Intermediate Water, GNAIW) occupying depths down to about 2 km depth, and a high nutrient, low δ13C water mass underneath [e.g., Boyle and Keigwin, 1987; Curry and Oppo, 2005; Duplessy et al., 1988; Oppo and Lehman, 1993; Sarnthein et al., 1994]. These data are generally interpreted as supporting a shallower overturning circulation in the glacial Atlantic, similar to that observed in some models of LGM ocean circulation. Curry and Oppo  argue that a vigorous circulation (with respect to the strength of the vertical mixing) is required in order to maintain the sharp property gradient between the shallower and deeper water masses and the large bathymetric gradient in δ13C below 2 km.
The deep water below 2 km had higher nutrient concentrations than today, and almost certainly contained predominantly waters with a southern (Antarctic) source [e.g., Marchitto et al., 2002]. However, it is clear from many tracers that there was also a contribution of low nutrient waters from the north to the deep (below 2 km) Atlantic. Boyle and Keigwin  reconstructed Cd in the deep North Atlantic and showed that these waters were more nutrient depleted than average ocean values, arguing against a complete shut-down in North Atlantic deepwater production. More recently, Rickaby et al.  find that reconstructed glacial Cd concentrations were lower in the western basin than the eastern basin in the North Atlantic below 2.5 km water depth, supporting a North Atlantic source of deep waters. The reconstructed seawater δ13C are also higher in the North Atlantic than in the South Atlantic during the LGM, suggesting continued contribution of deep waters from the north [e.g., Broecker, 2002; Matsumoto and Lynch-Stieglitz, 1999]. Lea and Boyle  found that the barium/calcium ratios in benthic foraminifera demand continued contributions of low nutrient waters from the north as well. Radiocarbon reconstructions for the deep North Atlantic show older ventilation ages than today [Broecker et al., 1990; Keigwin, 2004; Keigwin and Schlegel, 2002], but still younger than waters in the South Atlantic [Goldstein et al., 2001] suggesting a continued contribution of high radiocarbon water to deep waters in the North Atlantic.
While the processes controlling the formation rates and properties of the upper (GNAIW) and lower deep waters in the North Atlantic were probably different for each water mass and are not well understood at this time, it is clear that the renewal of these water masses did not cease entirely during the last ice age (perhaps except during Heinrich event 1 [McManus et al., 2004]). The waters above 2 km may have even been renewed faster than they are today. Today, deepwater formation involves the cooling and sinking of surface waters in the North Atlantic, and is therefore associated with a strong northward transport of heat. It is not clear whether the LGM Atlantic had an overturning cell involving the northward transport of surface waters (and thus heat), or whether the ventilation was accomplished differently.
Yu et al.  find the Pa-231/Th-230 ratio in the LGM sediments in the Atlantic is lower than the production ratio, implying relatively short residence time for waters in the LGM Atlantic. McManus et al.  come to a similar conclusion based on a high-resolution Pa/Th record from the North Atlantic. This data could reflect the continued presence of an overturning cell involving the northward transport of surface waters, but a short residence time could also be accomplished by rapid flushing of deep or intermediate waters either from the north or the south. Lynch-Stieglitz et al. [1999b] find a reduced vertical shear in the geostrophic velocity in the Florida Straits, consistent with a weaker surface branch of the overturning circulation. This does not contradict the idea of continued production of deep and intermediate water masses in the North Atlantic during the LGM, but does suggest that either the overturning cell was weaker, or that the upper and lower deep water masses were formed by mechanisms which did not draw large quantities of surface water northward through the Florida Straits.
In this study, we attempt to directly measure the strength of the meridional overturning cell at 30°S in the LGM Atlantic by reconstructing the cross-basin density gradient associated with the geostrophic shear in the overturning circulation.
2. Methods and Materials
The large-scale surface-to-deep meridional overturning circulation is reflected in a density contrast across the upper North Atlantic (denser waters on the western margin for a given water depth), and a similar density contrast across the upper South Atlantic (denser waters on the eastern margin for a given water depth) (Figures 2 and 3). Specifically, the vertical shear in the geostrophic transport integrated across the basin is proportional to the east-west density difference at any given depth, [(fρ0)/g]∂z(Lx) = ρe − ρw, where f is the Coriolis parameter, ρ is the density of seawater, Lx is the width of the basin and is the average velocity across the basin [Marotzke et al., 1999]. The geostrophic shear associated with the subsurface branch of the shallower wind-driven overturning cells [McCreary and Lu, 1994] is also reflected in the density contrast across the ocean basins [Lynch-Stieglitz, 2001; Veronis, 1981]. The shear associated with the shallow wind-driven cells disappears where the average wind stress over the ocean basin is zero (generally around 30°N and 30°S where the zonal wind direction reverses). For this reason, if we are primarily interested in the shear associated with the large scale surface to deep overturning, it is best to monitor the density contrast near 30°N or 30°S. However, as one moves from north to south within the Atlantic basin the changes in east-west density contrast, which reflect the changing geostrophic shear and Coriolis parameter, are reflected almost exclusively by changes in density along the western margin [Lynch-Stieglitz, 2001]. The flows near the eastern boundary of the Atlantic are weak, and are in geostrophic balance. The requirement that water not flow into or out of the ocean margin then requires that the density not change in a north-south direction along the eastern boundary [Veronis, 1973].
We can use the δ18O from the calcite tests of benthic foraminifera preserved in ocean sediments to estimate density on the ocean margins because both the δ18O of calcite and density increase as a result of increasing salinity or decreasing temperature [Lynch-Stieglitz et al., 1999a]. For times in the geologic past, our ability to reconstruct density from the δ18O of foraminiferal calcite is most limited by our knowledge of the relationships between the δ18O of seawater and salinity, as well as the relationship between temperature and salinity.
Here we use oxygen isotope measurements on benthic foraminifera to examine the density contrast across the South Atlantic at 27°S for the Holocene and last glacial maximum (LGM). We focus on sediment cores above 2 km water depth because the subtle Holocene gradient below 2 km cannot be resolved using the foraminifera data. In addition, the relationship between δ18O of seawater and salinity is different in the deep sea than in the main thermocline due to the impact of sea ice formation on the δ18O and salinity of these coldest deep waters. It would be difficult to assess how this relationship may have changed in the past.
A depth transect of sediment cores on the western margin between 26–28°S off Brazil were collected and oxygen isotope measurements on benthic foraminifera were reported by Curry and Oppo . The Curry and Oppo  isotope measurements were made on Cibicidoides and Planulina species which have been shown to accurately record seawater δ18O and calcification temperature [Duplessy et al., 2002; Lynch-Stieglitz et al., 1999a], including C. wuellerstorfi, C. kullenbergi, C. pachyderma, C. floridanus, and P. ariminensis. These data were supplemented by oxygen isotope data on C. lobatulus from an additional piston core (RC12-279) at 35.35°S for western margin vertical transect over a water depth of 441 to >2000 meters. We use the published Holocene and LGM time slices which were chosen on the basis of benthic and planktonic oxygen isotope stratigraphy as well as radiocarbon dates on selected cores [Curry and Oppo, 2005]. Because changes in the cross basin density contrast must be accommodated along the western margin of the ocean, a reduced shear in the overturning circulation will, all else being equal, be associated with denser waters (higher δ18O in foraminifera) along the western margin. Because the LGM values in the Curry and Oppo  transect were chosen to be the δ18O maximum, it is possible that these values reflect periods of low overturning within the LGM rather than average LGM values. However, most of the oxygen isotope maxima appear to cover a broad period of time, suggesting a relatively constant profile during the LGM.
Because the density does not change much in the north-south direction along the eastern margin, we can use cores from a wider latitude range to reconstruct the vertical density structure at the corresponding latitude (27°S) along the eastern (African) margin. We have investigated all of the piston cores from the Lamont core collection in the depth range 200–1000 m between the equator and the southern tip of Africa. While many of these cores are unsuitable (no Holocene to LGM section within the core, insufficient foraminifera, etc.), there were a number of cores which were identified as covering the transition between LGM and present on the basis of planktonic and/or benthic foraminifera oxygen isotope stratigraphy (Table 1, Figure 4). Most measurements on benthic foraminifera were, like the western margin cores, from specimens of the genera Planulina or Cibicidoides. For two of the cores (V19-248, V19-249) that had very low abundances of Cibicidoides and Planulina, we also analyzed Bolivina, applying an offset of 0.83‰ for comparison with the other benthic foraminifera measurements [Herguera et al., 1992; McCorkle et al., 1997].
Table 1. South Atlantic Sediment Cores Contributing to the Profiles Shown in Figure 5
All oxygen isotope measurements from the eastern margin cores in this study except for ODP 1079 were made at Lamont-Doherty Earth Observatory on a Micromass Optima with Multiprep individual acid bath carbonate preparation device. Data were calibrated to PDB using NBS-19, NBS-18 and an in-house standard. Sample intercalibration yields no systematic differences between measurements made on this system and those for the western margin which were made primarily at Woods Hole Oceanographic Institution [Curry and Oppo, 2005]. Measurements on ODP 1079 were made at Scripps Institution of Oceanography on a Finnigan MAT252 equipped with a common acid bath carbonate preparation device were calibrated to PDB using NBS-19 and an in-house coralline carbonate standard. Replicate measurements at selected depths in ODP 1079 were made at Woods Hole Oceanographic Institution, and were found to be 0.18‰ heavier than those made at Scripps. A correction of 0.18‰ was applied to the data from ODP 1079 for better consistency with the data measured at WHOI and LDEO. The new eastern margin data were supplemented with published benthic δ18O records from the literature to form the eastern margin profile (Table 1).
Average profiles for the Holocene and LGM benthic δ18O for the eastern margin were constructed. The time intervals were chosen on the basis of the planktonic and benthic δ18O records (Figure 4). By choosing the maximum δ18O values for the eastern margin cores, we will reconstruct the maximum possible LGM density gradient across the basin. If, in fact, during the LGM the benthic δ18O on the eastern margin were lower, our reconstruction would overestimate the magnitude of the shear in the overturning circulation. Similarly, if there was a divergence in values between the species, we chose to use the species which gave the most positive δ18O values which also will produce a maximum reconstruction of the density gradient and shear in the overturning circulation.
3. Results and Discussion
On the African margin sedimentation rates were low and benthic foraminiferal abundances showed a dramatic shift from a Holocene dominance of Planulina and Cibicidoides to a LGM dominance of Bolivina and other benthic foraminifera common in areas of high overlying productivity and low oxygen concentration. The combination of abundance changes, low sedimentation rates, bioturbation, and the possibility of down slope transport on the continental margins leads to down core records which show significant differences depending on the species analyzed, and large amounts of scatter when species become rare (Figure 4).
However, the Holocene profiles do show consistently higher δ18O values on the eastern margin than on the western margin in the upper 1 km, consistent with the values predicted from modern hydrographic data (Figure 5). For the LGM, despite our efforts to choose the most positive δ18O values for the eastern margin profile, the eastern margin profile shows δ18O that is either the same as, or even lower than, the δ18O in the western margin profile.
While the relationship between δ18O, T and S in the South Atlantic today associates higher δ18O in the foraminifera with waters of higher density, was this necessarily true in the past? Presuming a stable water column, the fact that there is a general increase in δ18O of the benthic foraminifera with depth for both profiles suggests that this was indeed the case. Could a change in relationship between δ18O in foraminifera and density account for a collapse in the cross-basin δ18O gradient with no reduction in cross basin density gradient? Because the density structure in the main thermocline is dominated by temperature, a complete flattening of the δ18O/Salinity relationship (no change in δ18O for a large change in salinity) due to a very isotopically heavy fresh end-member (from sea ice formation for example) would yield a reduction in the cross basin δ18O of the benthic foraminifera at 750 m water depth from 0.3‰ to 0.25‰. There is no reason to suspect that the main thermocline would not be temperature dominated during the LGM, especially given the continued presence of strong sea surface temperature gradients in the subpolar North and South Atlantic where the main thermocline ventilation occurs. This assumption could be checked using an independent proxy for temperature in these profiles such as the Mg/Ca or Sr/Ca ratios in benthic foraminifera [Lear et al., 2002; Rosenthal et al., 2006].
Today the T-S-δ18O relationship in the main thermocline is relatively constant across the narrow South Atlantic basin due to the presence of mixing along isopycnals and the lack a proximal water mass source. However, if there were a strong source of intermediate water production on one side of the South Atlantic, this could decouple the T-S-δ18O properties from one side to the next. Depending on the source of the water masses on either side, it is possible in this scenario that differences in the δ18O of seawater from one side to the next could lead to a reduced cross basin gradient in the δ18O of foraminifera in the presence of a cross basin density gradient similar to today. G. A. Gebbie and P. Huybers (Meridional circulation during the Last Glacial Maximum explored through a combination of South Atlantic δ18O observations and a geostrophic inverse model, submitted to Geochemistry, Geophysics, Geosystems, 2006; hereinafter referred to as Gebbie and Huybers, submitted manuscript, 2006) demonstrate that some very different temperature and salinity combinations could produce glacial δ18O values along the eastern margin that are lower despite being from waters of higher density. However, the required horizontal gradients in temperature and salinity are much larger than today's, and could be checked with an independent paleo-temperature proxy. Clearly further constraints on the T-S-δ18O relationship in the LGM ocean from pore waters and paleo-temperature proxies will be quite useful in better interpreting the foraminifera data.
In the discussion that follows we will presume that the LGM δ18O data represents a collapse or reversal in the cross-basin density gradient. A northward shift of the wind systems could reduce the northward geostrophic transport in the upper 500 meters or so, which would reduce the magnitude of the shear in the meridional flow. However, all else being equal, this reduction would not be enough to change the sense of the shear associated with the meridional overturning circulation and could not account for the entire reduction in density contrast especially at deeper levels [Lynch-Stieglitz, 2001]. The elimination or reversal of the cross basin density gradient would seem to require a different sort of circulation altogether, one in which warm surface waters do not enter the Atlantic basin to compensate the export of cold deep waters (NADW) or intermediate waters (GNAIW).
Simply shoaling the NADW overturning cell during the glacial (Glacial North Atlantic Intermediate Water) would tend to increase the shear and the density gradient between the eastern and western margins (Figure 6b), in contrast with our observations. Even if the flow were somewhat weaker, the compression to shallower depths would tend to increase the shear and thus the density gradient across the basin. If the shear associated with the upper limb of NADW were compressed from 0–3 km depth to 0–2 km depth (a 50% increase) the density gradient, which is directly proportional to the shear should also increase by 50%. Some general circulation and intermediate complexity models show a strong, shallow overturning which is closed within the Atlantic basin for the LGM, implying that the return circulation is not derived from waters outside of the Atlantic. If this were the case, we might not see a density gradient associated with the maximum in the overturning this far south. Many of these low-resolution models also show many of the streamlines of the modern NADW overturning cell rising from the deep ocean into the warm surface waters. This behavior in the relatively low resolution general circulation and intermediate complexity models is due to fact that they are more diffusive than the real ocean, which shows relatively little mixing across isopycnals and the associated upwelling of deep waters at low latitudes (Figure 1). If the closed deep overturning cells observed in these low-resolution models do not accurately reflect the real ocean circulation, we do not necessarily expect that the closed shallow overturning cells simulated for glacial conditions are realistic either.
On the other hand, it is difficult to envision that GNAIW could gain their high δ13C, low nutrient signal without a significant contribution from warm surface waters. Perhaps while today the large-scale MOC in the Atlantic is not closed at 30° (surface and intermediate waters are imported into the Atlantic and deep waters are exported), a shallower GNAIW cell was closed to the north of 30°S. This could happen if the wind systems and oceanographic regimes in the Southern Ocean shifted northward. However, subtropical fauna existed at these latitudes during the LGM as they do today, and the reconstructed sea surface temperature patterns do not suggest a migration of the Subtropical Front northward of 30°S [Niebler et al., 2003]. Even if there were a GNAIW circulation which was closed within the Atlantic Basin, the weaker flow through the Florida Straits during the LGM suggests that it would have probably been considerably weaker than the overturning associated with NADW today [Lynch-Stieglitz et al., 1999b]. In short, we find the ocean margin δ18O data presented here very difficult to reconcile with the inferences based on other proxies of a shallower version of today's Atlantic overturning circulation during the LGM.
While a collapse in the density gradient in the upper 2 km of the South Atlantic is consistent with a scenario in which there is no surface to deep overturning cell extending to the South Atlantic (Figure 6c), a reversal of the density gradient would require a circulation completely unlike the modern Atlantic. Some models with a completely collapsed overturning circulation show a vigorous northward flow of Antarctic Intermediate Water balanced by a southward flow of surface waters in the South Atlantic [e.g., Weaver et al., 2003]. This would imply a shear in the meridional circulation that is consistent with the ocean margin δ18O data that suggest a density reversal above 1 km (Figure 6d). A vigorous import of AAIW would, however, seem to require a shear in the opposite sense below the core of AAIW (1 km). This is not seen in the deeper data, but if the shear were relatively weak, it could be hidden within the errors and large scatter of the data.
Can these scenarios without a surface to deep overturning extending to the South Atlantic be consistent with the Pa/Th data which support a weakened but significant flushing of the Atlantic deep waters during the LGM [McManus et al., 2004; Yu et al., 1996]? Are they consistent with the argument that the sharp nutrient boundary in the deep North Atlantic and the large bathymetric gradients in δ13C require vigorous renewal of both water masses [Curry and Oppo, 2005]? We believe that it is possible to have a significant ventilation of deep waters in the North Atlantic during glacial times accomplished without an overturning cell involving the northward transport of surface waters. Our data allows the possibility that the high nutrient deep waters below 2 km in the glacial Atlantic could be well ventilated from the south (increased production of AABW). The shear associated with a vigorous AABW cell would be below 2 km and not be reflected in a density difference between the profiles above this depth. A more vigorous AABW cell would be consistent with the bathymetric changes in δ13C in both the North and South Atlantic, which appear to decrease by 0.7 to 1‰ per km below 2 km [Curry and Oppo, 2005]. Bathymetric gradients like this are difficult to maintain without active advection because vertical mixing is so strong near the mid-Atlantic Ridge [Ledwell et al., 2000; Mauritzen et al., 2002; Polzin et al., 1997]. As discussed in the introduction, deep waters formed in the North Atlantic would have contributed to this deeper water mass as well.
It is also difficult, but not impossible, to envision vigorous export of GNAIW without a strong northward transport of surface waters. Perhaps this water mass is ventilated in the well developed cyclonic subpolar gyre north of the polar front (which lay significantly farther to the south during the LGM), where upwelled cold water is made more dense by cooling and/or sea-ice formation at the sea surface. This mechanism could also be associated with little shear in the upper 2 km at lower latitudes. Or, as discussed above, a GNAIW overturning cell was confined in the Atlantic basin, north of 30°S.
Clearly we still have no scenario for the circulation of the glacial Atlantic that is completely consistent with our understanding of all of the existing data. However, since it is the temperature contrast between the northward moving warm surface waters and cold deep waters which is associated with the large amounts northward heat transport in the Atlantic Ocean, a severe reduction of shear in the upper 2 km of the glacial South Atlantic would imply a large reduction in northward ocean heat transport in the South Atlantic regardless of the specific scenario.
The quality of the records do not give us a high degree of confidence in the LGM density profile along the eastern margin, and it is possible that better records on the eastern margin could yield higher δ18O on the eastern margin during the LGM, consistent with a vigorous surface to deep overturning. It is also possible that local formation of intermediate water masses in the South Atlantic could lead to different relationships between δ18O of foraminifera and density on the eastern and western margins, confounding the use of δ18O to infer cross-basin density gradients (e.g., Gebbie and Huybers, submitted manuscript, 2006). It is also likely that there were large changes in the strength of the overturning on millennial timescales during glacial times and on the deglaciation, which we could never resolve with these data. For example our “LGM” profiles may include data for H1 and H2, times when deepwater circulation may have been dramatically different than during the LGM proper.
We examined oxygen isotope data from benthic foraminifera on both sides of the South Atlantic, and find that the cross-basin gradient seen in the Holocene is absent or even reversed during the last glacial maximum. If the reduction in the cross basin gradient in the δ18O of benthic foraminifera reflects a reduction in cross basin density, these data imply a reduction in the shear associated with the Atlantic meridional overturning circulation. We feel that the quality of the materials we analyzed here do not provide a definitive assessment of the cross basin density gradient during glacial times, but rather show the potential of this approach. Hopefully further work will serve to confirm/refute and quantify the scenario we outline here with well dated, high-resolution isotope records from the margins of the Atlantic Ocean, coupled with independent paleo-temperature estimates. However, it is very hard to reconcile the ocean margin δ18O data we present in this paper with a strong overturning cell involving the northward transport of surface waters in the glacial Atlantic.
We acknowledge the technical assistance of Pat Malone, Linda Baker, Heather Griffiths, Christa Farmer, and Jennifer Mayer. This work was supported by NSF award OCE-9984989/OCE-0428803 to J.L.-S., NSF award OCE-9986748 to D.W.O. and W.B.C., NSF OCE-0222111 to C.D.C., and SEGRF fellowship at LLNL to J.M. Sample material used in this project provided by the Lamont-Doherty Earth Observatory Deep-Sea Sample Repository. Support for the curation of the core collection is provided by the National Science Foundation through OCE-0350504. Comments from Claire Waelbroeck and an anonymous reviewer improved this manuscript.