Deep Atlantic Circulation During the Last Glacial Maximum and Deglaciation

Early seafaring nations recognized the practical and economic benefits of mapping surface currents and winds in great detail. However, knowledge of the deep oceans, their properties, and their climatic significance has been acquired relatively recently. The first field program to systematically measure physical and chemical properties of all the world's deep oceans took place from 1973–1978. Subsequent measurements revealed that properties of deep water in key regions vary from decade to decade, and that these changes are linked to oscillations in surface climate (Dickson et al. 1996, Zhang 2007). Unfortunately the observations are too limited to provide insight into how the deep oceans and climate interact on the longer time scales of ocean circulation and also how this interaction might change in response to rising greenhouse gases. Instead, scientists use computer climate models to predict how the Earth's climate will change. Reconstructions of past ocean circulation using the geochemistry of microfossils preserved in marine sediments provide critical information to test these models.

The Atlantic Ocean is the only ocean basin that features the transformation of surface-to-deepwater near both poles. Warm salty tropical surface waters flowing northward in the western Atlantic cool in transit to and within the high-latitude North Atlantic, releasing heat to the overlying atmosphere and increasing seawater density. Once dense enough, these waters sink and flow southward between ~ 1000 and 4000m. This North Atlantic Deep Water (NADW), as it is called, flows from the Atlantic to the Southern Ocean where much of it upwells — or rises to the surface — around Antarctica, and some of it circulates Antarctica before entering the rest of the world's deep oceans. Antarctic Bottom Water (AABW), which is formed close to Antarctica, is denser than NADW, and flows northward in the Atlantic below NADW. AABW is confined to water depths below 4000 meters in the tropical and North Atlantic. Antarctic Intermediate Water (AAIW) flows northward above NADW. The presence of these three water masses in the Atlantic Ocean is evident in cross-sections of many water properties, including salinity, phosphate concentration and carbon isotope ratios (Figure 2). The residence time of deepwater in the western Atlantic is approximately 100 years (Broecker 1979), meaning that the average water parcel spends about a century in the deep Atlantic.

Warren (1983) first noted that the difference in salinity between the North Pacific and the North Atlantic (Figure 1) was the principal reason deep water formation occurs today only in the North Atlantic. Salty water, when cooled, achieves a higher density and is thus able to sink to greater depth in the water column. Wintertime cooling occurs in both the North Atlantic and North Pacific, but since the surface waters of the North Atlantic are much closer in salinity to the mean of the ocean's deep water, they achieve a density high enough to sink to great water depths. Warren (1983) noted that the salinity of the North Pacific was low because of relatively low evaporation, little exchange with salty tropical waters, and an influx of fresh water from precipitation and river runoff. Emile-Geay et al. (2003) reevaluated the Warren (1983) results and fundamentally confirmed his thesis, noting that atmospheric moisture transport from the Asian monsoon was also an important source of fresh water to the North Pacific not originally considered by Warren. Interestingly, Warren also noted that the North Atlantic had much greater river runoff than the North Pacific, so its higher surface salinities must be the result of greater evaporation in the Atlantic basin.

Broecker et al. (1990a) noted that higher Atlantic salinities are the result of a net transfer of water vapor from the Atlantic to the Pacific over the Isthmus of Panama, equivalent to approximately 0.35 Sverdrup (10⁶ m³ per second). In the absence of other processes, this would raise the salinity of the Atlantic by about 1 salinity unit each 1000 years. If the Atlantic salinity is in balance, then it must be exporting the excess salt (enough to compensate for the lost fresh water) through ocean circulation processes. Today this is occurring through the production and export of North Atlantic Deep Water.

At times in the past, rapid melting of ice sheets surrounding the North Atlantic was great enough to alter surface salinities, likely reducing the density of deep water formed, and slowing the export of deep water from the North Atlantic. Broecker et al. (1990b) hypothesized that natural oscillations in the rate of water vapor exchange between the Atlantic and the Pacific during the last glacial period were responsible for the rapid, short term fluctuation ocean circulation linked to the abrupt millennial-scale Dansgaard-Oeschger Events seen in the Greenland ice cores (Figure 9).

There are three main pathways for water to return to the North Atlantic and renew NADW, a warm-water route and two cold water routes (Figure 3). The "warm-water route" begins with the flow of surface and thermocline water from the Pacific to the Indian Ocean through the Indonesian Seas. Both colder return flows involve Antarctic Intermediate Water (AAIW), described above. AAIW enters the southern South Atlantic through the Drake Passage between Antarctica and South America, with some flowing into the Atlantic and some flowing into the Indian Ocean. AAIW also enters the Indian Ocean from south of Tasmania and flows westward towards Africa, where it joins the warm-water flow and the other branch of AAIW before rounding southern Africa, entering the South Atlantic, and flowing northward (Gordon 1985, Speich et al. 2002). Along its transit to the North Atlantic, AAIW from the Drake Passage, flowing above Tasman AAIW, mixes with overlying water and contributes to the "warm-water route" (Gordon 1986). These return flows provide a significant source of heat to high northern latitudes. Together, southward flow of water in the deep Atlantic and its shallower return flows are a large component of what is known as the global Meridional Overturning Circulation (MOC).

Reconstructions of past ocean circulation have relied heavily on the chemistry of foraminifera, single-celled organisms that secrete a shell made of calcium carbonate (CaCO₃). The shells of foraminifera that live on the sea floor, or "benthic" foraminifera, record many chemical properties of the overlying seawater. Critical for paleoceanographic reconstructions, benthic foraminifera incorporate carbon to make their shells in approximately the same carbon-13 to carbon-12 isotope ratio as the overlying seawater (Curry et al. 1988). Similarly, benthic foraminifera incorporate cadmium (Cd) into their shells in a known proportion to seawater Cd, and so cadmium/calcium (Cd/Ca) measured in benthic foraminifera enable estimates of seawater Cd (CdW) (Boyle 1988).

CdW and the ratio of carbon-13 to carbon-12 (¹³C/¹²C, reported as δ¹³C) in deep water covary with nutrient content. In the deep Atlantic, phosphate content is correlated to salinity, which helps define the different water masses (Figure 2), and so measuring δ¹³C or Cd/Ca in fossil foraminifera with known ages can tell us whether the properties and boundaries of the important deep water masses — NADW, AABW, and AAIW — have changed in the past.

Foraminifera also record the abundance of carbon-14 in seawater. Carbon-14 is a radioactive carbon isotope produced in the atmosphere, and enters the surface ocean through air-sea gas exchange. The difference between its abundance in deep water and in surface water provides a measure for how much time has passed since the deep water was last at the surface and in contact with the atmosphere. This "ventilation age" can be estimated in the past by measuring the difference in the abundance of ¹⁴C between coexisting benthic foraminifera and foraminifera that lived at the ocean surface, or "planktic" foraminifera (Benthic-Planktic radiocarbon age). Radiocarbon measurements therefore provide a way to evaluate whether the renewal rate of deepwater by surface water was slower or faster in the past (Broecker & Peng 1982). Ventilation age estimates are complicated by variations through time in the production of radiocarbon in the atmosphere (Adkins & Boyle 1997) and by variations in surface ocean radiocarbon due to changes in oceanography (e.g., mixing with deeper, older water) (Bard et al. 1994). These complications can be circumvented to some extent if there is an independent timescale for a sediment record that does not rely on the radiocarbon chronology (e.g., Thornalley et al. 2011).

To reconstruct deep ocean circulation using the geochemistry of foraminifera, scientists use sediment cores, taken from a range of water depths along the continental margins, mid-ocean ridges, and other bathymetric highs in the ocean basins. This strategy allows them to sample sediments that intersect the main water masses: AAIW, NADW, and AABW (Figure 4). Once they have acquired the sediment cores, they use a variety of methods, including radiocarbon dating of foraminifera, to identify sediments that were deposited at times in the past, like the last Ice Age, when climate was very different from today.

There is strong evidence that the circulation of the deep Atlantic during the peak of the last Ice Age, or the Last Glacial Maximum (LGM; ~22,000 to 19,000 years ago) was different from the modern circulation (Boyle & Keigwin 1987, Duplessy et al. 1988, Marchal & Curry 2008). Compilations of deepwater δ¹³C and CdW for the LGM (Figure 5) show several features that contrast with their modern distributions. Whereas much of the modern deep western Atlantic has similar δ¹³C values because it is filled with NADW, during the LGM, the range of δ¹³C values was larger than today, with higher values in NADW and lower values in AABW. The main core of high-δ¹³C, low-CdW NADW was at least 1000 meters shallower than today, probably because the density difference between surface waters and deep water was reduced — surface salinity may have decreased as a result of less evaporation due to colder glacial temperatures, and as a result of input of freshwater from glaciers surrounding the North Atlantic (Boyle & Keigwin 1987). In the western Atlantic, depths below ~2 km were filled with AABW. Radiocarbon data suggest that deepwater was older (Keigwin & Schlegel 2002), consistent with less NADW and more AABW as indicated by the δ¹³C and CdW of benthic foraminifera. Glacial δ¹³C data from the eastern Atlantic suggest that the boundary between glacial AABW and glacial NADW may have been shallower than in the western Atlantic (Sarnthein et al. 1994), although the difference may be the result of local effects caused by increased glacial productivity and higher rates of remineralization of low-δ¹³C organic carbon in the eastern basin. Inferences using other kinds of proxy data of deep Atlantic circulation are consistent with the changes inferred from δ¹³C, Cd/Ca and ¹⁴C of benthic foraminifera (Lynch-Steiglitz et al. 2007).

Carbon isotope and Cd data from the western South Atlantic indicate that like today, AAIW was centered at ~1000 meters water depth (Curry & Oppo 2005, Makou et al. 2010), but data defining how far north into the Atlantic it flowed are still lacking. Defining the northward extent of AAIW is an area of active investigation, as doing so will help scientists understand whether AAIW was still an important return flow for NADW during the LGM. If not, it implies even larger differences between glacial and modern circulation than currently appreciated.

The melting of the vast continental ice sheets, which began ~20,000 years ago due to gradual changes in the seasonal and spatial distribution of the Sun's energy (Broecker & Von Donk 1970), was interrupted by several abrupt cold climate events. The two largest deglacial events in the North Atlantic — known as Heinrich Stadial 1 and the Younger Dryas — occurred approximately 17,500–14,600 and 13,000–11,500 years ago respectively (Figure 6) (Heinrich 1988, Bond et al. 1992, Grootes et al. 1993).

Evidence from marine sediment cores suggests large changes in deep ocean circulation associated with these events. For example, relatively high CdW values in the deep North Atlantic require a reduced relative contribution of NADW to the deep Atlantic during both the Heinrich Stadial and the Younger Dryas (Figure 6a). Relatively low CdW values at ~ 750m in the Florida Strait, where nutrient-rich AAIW reaches today, imply that AAIW did not penetrate as far north during these events (Figure 6a). These simultaneous decreases in both NADW and AAIW link the northward flow of AAIW to variations in the southward flow of NADW during the deglaciation, consistent with the role of AAIW as a supplier of NADW (Came et al. 2008).

Deglacial deepwater evolution based on δ¹³C exhibits some similarities to the CdW records, but also some important differences (Figure 6b). Like CdW, the δ¹³C records suggest that both the contribution of NADW to the deep Atlantic and northward AAIW penetration decreased during the Heinrich Stadial. Although δ¹³C records suggest that the contribution of NADW to the deep Atlantic was reduced during the Younger Dryas, δ¹³C records do not provide clear evidence for an associated reduced northward penetration of AAIW. The AMOC recovery following the Heinrich Stadial weakening is recorded ~ 16,000 years ago in the intermediate-depth δ¹³C record and both CdW records, but ~1,000 years later in the deepwater δ¹³C record.

Other proxies, including Benthic-Planktic ¹⁴C records also suggest that the contribution of NADW to the deep Atlantic decreased during these events (e.g., Boyle & Keigiwn 1987, Thornalley et al. 2011, Lynch-Stieglitz et al. 2007, Robinson et al. 2005) whereas the response of AAIW during these events is more controversial (e.g., Pahnke et al. 2008). Resolving why differences occur between proxy records is necessary in order to fully understand the link between deep ocean circulation and climate.

The prevailing view of the Heinrich Stadial is that instabilities in the Northern Hemisphere ice sheets resulted in catastrophic iceberg discharges into the North Atlantic Ocean (Bond et al. 1992). These freshened and reduced the density of North Atlantic surface waters, significantly curtailing surface-to-deepwater transformation and reducing northward transport of heat to the region. Expanded sea ice may have amplified cooling in the North Atlantic region. Once warming began at the end of the events, quick northward displacement of sea ice may have triggered an abrupt end of the cold events (Dansgaard et al. 1989, Li et al. 2005).

It is generally believed that freshening of the surface North Atlantic also initiated the Younger Dryas cooling. Fresh water from ice sheets melting into the North Atlantic, perhaps by way of the Arctic Ocean (e.g., Murton et al. 2010), reduced surface-to-deepwater transformation, and the associated northward heat transport. Expanded sea ice and meltwater from iceberg discharge may also have sustained the event.

Although there are fewer data than for the LGM, compilations of δ¹³C data from the western Atlantic during the Heinrich Stadial highlight significant differences in deepwater mass geometry relative to both the modern and the LGM transects (Figure 7). While the modern and glacial transects clearly show the core of nutrient-poor, high-δ¹³C NADW at 3000m and 2000m respectively, the high-δ¹³C North Atlantic waters were restricted to depths shallower than 1000 m during the Heinrich Stadial. In addition, modern and glacial NADW can be traced to the South Atlantic by their high δ¹³C values, but during the Heinrich Stadial high-δ¹³C northern source water may not have reached the equatorial Atlantic. These data strongly suggest that vigorous surface-to-deepwater transformation akin to modern or glacial NADW did not occur during this portion of the Heinrich Stadial. As in the glacial ocean, lowest δ¹³C values, suggesting the presence of AABW, were found below 4000m. Decreasing δ¹³C values below 1000m suggest a progressive increase in the proportion of nutrient-rich southern ocean waters relative to nutrient-poor upper ocean waters. Given the water mass distributions and properties in the subpolar North Atlantic at the time, the higher nutrient waters observed at shallow depths during the Heinrich Stadial must have originated in the southern ocean. Shoaling of southern ocean waters already present in the glacial deep Atlantic provides a simple way to explain these low δ¹³C values, but without additional data, we cannot rule out the alternative possibility of enhanced advection of low-δ¹³C, southern intermediate waters to the high-latitude North Atlantic (Rickaby & Elderfield 2005, Thornalley et al. 2011).

One of the most exciting and important discoveries relating to abrupt climate changes is the finding that when the North Atlantic region cooled during abrupt events, the southern hemisphere warmed (Figure 8) (Blunier et al. 1998). The reason for the opposite temperature response is related to changes in heat transport associated with deepwater variability. The decrease in surface-to-deepwater transformation in the North Atlantic during Heinrich Stadial 1 and the Younger Dryas caused a decrease in the northward flow of warm tropical surface water, cooling the North Atlantic region. Less heat was transported from the tropics to the North Atlantic in the upper ocean to renew NADW, so the heat accumulated in the southern hemisphere and tropical Atlantic. The simultaneous cooling in one hemisphere and warming in the other, due to deepwater variability and associated changes in upper ocean heat transport, has been named "The Bipolar Seesaw" (Broecker 1998). Ice core evidence suggests that the bipolar seesaw also operated during earlier stadial events during the glacial period (Figure 9). Simulations with numerical models of the ocean-atmosphere system show that a reduction in the AMOC would cool the high latitudes of the North Atlantic while warming the South Atlantic, consistent with the bipolar seesaw mechanism (e.g., Manabe & Stouffer 1988).

The discovery of rapid and abrupt climate oscillations in Greenland ice cores in the 1980s changed the way researchers thought about climate changes on longer time scales. At the time, the prevailing view was that slow changes in the earth's orbit around the sun (the Milankovitch hypothesis) altered the seasonal and latitudinal distribution of the sun's energy, which forced the slow growth and somewhat more rapid decay of the ice sheets during the ice ages (Broecker & van Donk 1970). While this orbital theory is still generally accepted to explain climate variations on 10,000 to 100,000-yr time scales, the records of climate found in the ice cores showed that the climate system, at least in the North Atlantic region, exhibited much more rapid and frequent changes than could be accounted for by external orbital forcing (Figure 9). These short-term, rapid changes in climate occurred during the glacial phase of the ice ages. They reflected millennial-scale oscillations in air temperature, with some of the warmings occurring in as few as ten years. The oscillations became known as Dansgaard-Oeschger (D-O) Events, in honor of the two distinguished ice core researchers who were most responsible for discovering them (Dansgaard et al. 1982, 1984, Oeschger et al. 1984).

At nearly the same time, studies of marine sediments in the subpolar North Atlantic revealed evidence of several events of massive discharge of the land-based ice sheets into the North Atlantic Ocean. At times during the last glacial period, large numbers of icebergs calved from the surrounding glacier systems, leaving in the underlying sediments a telltale signature of ice-rafted minerals with a North American origin. The ice-rafted sediments were the undeniable record of large-scale calving of the ice sheets and as a result, the rapid, large-scale addition of fresh water into the entire subpolar North Atlantic. These ice-rafting events were less common than the D-O events, and appeared to occur after a series of increasingly colder D-O events (Bond et al. 1993). Referred to as Heinrich Events (after their discoverer Hartmut Heinrich), their occurrence was often associated with evidence for major reductions in the production of deep water in the North Atlantic. The two most recent events, Heinrich Event (Stadial) 1 and the Younger Dryas (although technically not a Heinrich event, it is sometimes referred to as "Heinrich Event 0") occurred during the last deglaciation and were each responsible for a major change in North Atlantic circulation and climate (Figure 6).

We know from modern observations that rainfall migrates north and south with the seasons, towards the warmer hemisphere (Waliser & Gautier 1993). There is evidence to suggest that this was also true on longer time scales, for example during the Heinrich Stadial, when the warming of the South Atlantic relative to the North Atlantic caused a southward shift in South American rainfall (Figure 10). Evidence from many sources suggests that changes in the hydrologic cycle occurred throughout the tropics, including large drying in the Asian monsoon region. The spatial distribution of the increased aridity and moisture, however, suggests that, in addition to the southward migration of the Intertropical Convergence Zone and monsoon systems, another mechanism, possibly a generally weaker hydrologic cycle due to cooler sea surface temperatures, is needed to explain anomalies over Asia and Africa (Stager et al. 2011). Climate simulations with computer models suggest ways that a reduction in North Atlantic overturning and the related sea surface temperature changes could have influenced global tropical climate (Zhang & Delworth 2005). However, there are still many uncertainties, and an important role for the tropics in abrupt climate change is also possible (Seager & Battisti 2007).

Although Although beyond the scope of this article, geologic data suggest that deep ocean circulation changes we describe above played an important role in raising atmospheric CO₂ from glacial levels of ~ 200 parts per million to pre-industrial interglacial values of ~ 280 parts per million (e.g., Sigman et al. 2007).

The instrumental record is short, and only spans a very limited range of climate states. Deep ocean circulation operates on much longer time scales. The ability to understand ocean and climate connections over a larger range of climate states is especially important in the context of predicting future climate. Complex computer climate models are based on the laws of physics, but there are many processes that are not specifically modeled because, for example, they occur on spatial scales smaller than a model grid box (e.g., cloud formation processes, mixing in the ocean). Because modelers use relationships based on modern observations to parameterize these processes, there is some risk that the models are tuned to remain close to the modern climate. On the other hand, if a computer model can simulate ocean and climate conditions for the LGM or the Heinrich Stadial, for example, then scientists have more confidence that its predictions for a greenhouse world are reliable.