Well-dated benthic foraminifer oxygen isotopic records (δ18O) from different water depths and locations within the Atlantic Ocean exhibit distinct patterns and significant differences in timing over the last deglaciation. This has two implications: on the one hand, it confirms that benthic δ18O cannot be used as a global correlation tool with millennial-scale precision, but on the other hand, the combination of benthic isotopic records with independent dating provides a wealth of information on past circulation changes. Comparing new South Atlantic benthic isotopic data with published benthic isotopic records, we show that (1) circulation changes first affected benthic δ18O in the 1000–2200 m range, with marked decreases in benthic δ18O taking place at ∼17.5 cal. kyr B.P. (ka) due to the southward propagation of brine waters generated in the Nordic Seas during Heinrich Stadial 1 (HS1) cold period; (2) the arrival of δ18O-depleted deglacial meltwater took place later at deeper North Atlantic sites; (3) hydrographic changes recorded in North Atlantic cores below 3000 m during HS1 do not correspond to simple alternations between northern- and southern-sourced water but likely reflect instead the incursion of brine-generated deep water of northern as well as southern origin; and (4) South Atlantic waters at ∼44°S and ∼3800 m depth remained isolated from better-ventilated northern-sourced water masses until after the resumption of North Atlantic Deep Water (NADW) formation at the onset of the Bølling-Allerod, which led to the propagation of NADW into the South Atlantic.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 In order to decipher causal links between changes in climate and ocean circulation, deep-sea marine records have to be placed in the same time frame as polar ice and continental records. Ideally, all these climatic records should be dated in calendar years B.P., so that phasing between the climatic response and insolation forcing can be assessed. The last deglaciation period is especially interesting in that respect, because it belongs to the time span covered by 14C dating technique. The 14C dates can be converted into calendar ages using calibration curves describing the variations in the atmospheric 14C/12C ratio over the past 30 to 50 kyr [Reimer et al., 2009]. The 14C dating of marine records is however problematic in certain regions where the surface reservoir age has not remained constant over time [Siani et al., 2000; Sikes et al., 2000; Waelbroeck et al., 2001]. This can be corrected in oceanic regions surrounding polar ice sheets by synchronizing sea surface temperature (SST) recorded in marine cores with air temperature recorded in polar ice cores, based on the assumption of thermal equilibrium between surface water and overlying air [Govin, 2008; Govin et al., 2009; Skinner and Shackleton, 2004; Skinner et al., 2010; Waelbroeck et al., 2001].
 Oxygen stable isotopic composition (18O/16O) of foraminifer tests has been used as a chronostratigraphical tool for the last 40 years [Imbrie et al., 1984; Lisiecki and Raymo, 2005, 2009; Shackleton and Opdyke, 1973]. On orbital time scales, and averaged over the whole ocean, planktonic and benthic 18O/16O (δ18O, expressed in ‰ versus VPDB) reflect changes in global ice volume. However, because δ18O of foraminifer calcite is a function of the temperature and isotopic composition of the water in which calcification takes place, it also records changes in local temperature and water δ18O. As deep water temperature and δ18O vary less than surface water temperature and δ18O, benthic foraminifer δ18O has been preferentially used as a global correlation tool [Lisiecki and Raymo, 2005, 2009]. However, a significant fraction of the benthic δ18O signal does not derive from changes in global ice volume and hence is not a global signal [Labeyrie et al., 2005; Skinner and Shackleton, 2005; Waelbroeck et al., 2002, 2008]. Downcore pore fluid δ18O measurements have shown that only 50% to 70% of the total benthic δ18O glacial-interglacial amplitude can be attributed to the mean ocean δ18O signal resulting from changes in global ice volume, while the remainder of the benthic δ18O glacial-interglacial amplitude is explained by changes in local bottom water temperature and δ18O [Adkins et al., 2002; Schrag et al., 1996]. Furthermore, the timing of deglacial changes in these two components need not be in phase at a given site, or synchronous globally.
 Here we present a new well-dated benthic isotopic record from a deep South Atlantic site and compare it with published Atlantic benthic isotopic records versus calendar age from different water depths and latitudes. We demonstrate that not only the amplitude, but also the timing of the benthic δ18O glacial-interglacial transition greatly varies from one depth range to another within the Atlantic Ocean basin. Using additional information from benthic foraminifer carbon isotopic composition, we interpret the differences in benthic δ18O glacial-interglacial amplitude and timing in terms of changes in the distribution of different water masses within the Atlantic basin.
2. Material and Methods
2.1. Study Area
 We present new data from core MD07-3076Q (44°09.19′S, 14°13.70′W, 3770 m water depth), retrieved on the eastern flank of the Mid-Atlantic Ridge, in the Atlantic sector of the Southern Ocean. Site MD07-3076 is currently bathed in a predominantly southward flowing mixture of modified North Atlantic Deep Water (NADW) and Lower Circumpolar Deep Water (LCDW), which feeds into the core of eastward flowing Circumpolar Deep Water (CDW) [Talley, 2008] (Figure 1). We compare core MD07-3076 benthic isotopic records with well-dated published benthic isotopic records from various depths between 1000 and 5000 m in the Atlantic Ocean and Nordic Seas (Figure 1 and Table 1). We selected Cibicides isotopic records with sufficiently high temporal resolution and low dating uncertainty over the 10–21 ka time interval to allow assessment of the phasing between the different records over that time interval. Site ENAM93-21 is located in the Nordic Seas, north of the Faeroe-Shetland Channel, close to a major area of outflow of Norwegian Sea Deep Water into the North Atlantic [Rasmussen et al., 1998]. Site NA87-22 is situated in the northern North Atlantic, on the southern side of the Iceland-Scotland Ridge and is currently bathed by a mixture of Upper NADW and Norwegian Sea Deep Water [Vidal et al., 1997]. Site MD95-2037 is located further south on the eastern slope of the Mid-Atlantic Ridge. At present, site MD95-2037 is bathed at depth by the southward flow of the Upper NADW [Gherardi et al., 2009]. Site MD99-2334K, at 3146 m depth on the Iberian Margin, is currently bathed in northward flowing Northeast Atlantic Deep Water, a mixture of about 50% NADW and 50% Lower Deep Water which is derived from Antarctic Bottom Water (AABW) [van Aken, 2000]. Site CH69-K09 is located at 4100 m depth and ∼42°N in the Western Atlantic, and is currently bathed by the Lower NADW contour current [Labeyrie et al., 1999]. Site KNR31GPC-5 located at 4583 m depth on the Bermuda Rise is currently also bathed in Lower NADW [Talley, 2008]. Site M35003 is located on the Atlantic side of the Caribbean sill and is currently in the transition zone between Antarctic Intermediate Water (AAIW) and Upper NADW [Rühlemann et al., 2004]. Core GeoB1711 was taken from the Cape Basin, but above the Walvis Ridge water depth, so that the site is currently under the influence of the southward flow of Upper NADW [Vidal et al., 1999].
 Core MD07-3076 age model is based on 49 monospecific planktonic radiocarbon dates that have been corrected for variable surface reservoir age effects using ice core and marine chronostratigraphic tie points [Skinner et al., 2010]. Age models of the other Atlantic cores used in this study were published in previous studies (see references in Table 1). They all rely on 14C dates, with corrections in case of variable reservoir ages for cores located north of 40°N [Waelbroeck et al., 2001] and south of 40°S [Skinner et al., 2010]. In the latter cases, the average uncertainty on the final calendar age over the 10–21 ka time interval is larger, ranging from 450 to 950 when a correction for variable reservoir age has to be applied, whereas it is comprised between 270 and 410 years for cores that do not necessitate variable reservoir age corrections (Table S1 in the auxiliary material).
2.3. Isotopic Analyses
 Benthic foraminifera Cibicides kullenbergi were picked in the >150 μm fraction. C. kullenbergi18O/16O and 13C/12C ratios (δ18O and δ13C respectively, expressed in ‰ versus Vienna Pee-Dee Belemnite (VPDB)) were measured at LSCE (Gif-sur-Yvette) on Finnigan Δ+ and Elementar Isoprime mass spectrometers. VPDB is defined with respect to NBS-19 calcite standard (δ18O = −2.20‰ and δ13C = +1.95‰) [Coplen, 1988]. The mean external reproducibility (1σ) of carbonate standards is ±0.05‰ for δ18O and ±0.03‰ for δ13C; measured NBS-18 δ18O is −23.2 ± 0.2‰ VPDB and δ13C is −5.0 ± 0.1‰ VPDB.
 The δ13C in the epifaunal benthic foraminifer Cibicides has been shown to record the bottom water dissolved inorganic carbon (DIC) δ13C [Duplessy et al., 1984]. Deep water δ13C values reflect the physical mixing of different water masses, their initial 13C content, controlled by biological processes in the surface ocean and air-sea CO2 exchanges [Lynch-Stieglitz et al., 1995], and the remineralization at depth of 13C-depleted organic matter raining from the surface [Kroopnick, 1985]. A decrease in Cibicides δ13C thus generally corresponds to an increase in bottom water nutrient content. As nutrient content largely follows water mass structure and circulation in the modern ocean, Cibicides δ13C can be used to trace water masses [Duplessy et al., 1988]. A decrease in Cibicides δ13C may thus be interpreted as a decrease in bottom water ventilation, and conversely. [Mackensen et al., 1993] have proposed that increased seasonally variable productivity driven rain rates of organic matter to the ocean floor produce low Cibicides δ13C values in the South Atlantic. However, the fact that enhanced glacial-interglacial benthic δ13C amplitudes are observed across the whole South Atlantic, independently of sedimentary environments, primary productivity regimes, or chosen benthic foraminifera species, has been used to argue that extremely low glacial benthic δ13C values in the Southern Ocean reflect changes in ocean circulation [Hodell et al., 2003; Mackensen et al., 2001; Michel et al., 1995; Ninnemann and Charles, 2002].
3. Asynchrony Between North and South Atlantic Benthic δ18O Signals Below 3000 m Depth
3.1. Observed Phase Shifts
 In order to define the timing of the beginning of the last glacial-interglacial transition in deep benthic δ18O signals, we computed the average benthic δ18O over the period 20–17 calendar kyr B.P. (ka) in cores located below 3000 m and determined at what time benthic δ18O became significantly lighter than its 20–17 ka value (i.e., lighter than the mean 20–17 ka value minus 1 standard deviation). Applying this approach to the two high-resolution continuous benthic δ18O records of Figure 2 yields an onset of the last glacial-interglacial transition at 17.0 ka (±0.4 kyr) in North Atlantic core MD99-2334K, and at 16.0 ka (±1.4 kyr) in South Atlantic core MD07-3076. Although these two dates are not significantly different due to relatively large dating uncertainties induced by the uncertainty on South Atlantic surface reservoir age correction [Skinner et al., 2010], the two records differ strongly in the subsequent evolution of the benthic δ18O signal: benthic δ18O in Iberian Margin core MD99-2334K decreases by ∼0.6‰ from 17 to 15 ka, whereas the decrease in South Atlantic core MD07-3076 benthic δ18O is of only ∼0.2‰ over the same time interval, before markedly accelerating at about 14.9 ka (Figure 2 and 3). A similar marked decrease in benthic δ18O has been dated at about 15.8 ka in core RC11-83 (41°36′S, 9°48′E, 4718 m) [Charles et al., 1996]. The younger age of our record results from the increase in surface water reservoir age over that period, as interpolated by Skinner et al. . However, significant uncertainties in the magnitude and variability of surface reservoir ages both in MD07-3076 and RC11-83 remain, in particular over the 17–13 ka time interval. It is thus not possible at present to determine if phase shifts exist between the different deep South Atlantic benthic δ18O records, or if the discrepancies instead represent poorly constrained surface reservoir age variability across the South Atlantic. These issues will only be resolved in future through concerted efforts to further constrain surface reservoir age variability in the region.
 Core MD99-2334K 0.6‰ benthic δ18O decrease is much larger than the change in mean ocean δ18O resulting from sea level rise: sea level had risen by only one third of its total glacial-Holocene rise by 15 ka [e.g., Lambeck and Chappell, 2001], which corresponds to a decrease in mean ocean δ18O of about 0.3‰ [Adkins et al., 2002; Schrag et al., 1996] (Figure 3). This early decrease in benthic δ18O at ∼3150 m depth on the Iberian Margin therefore reflects changes in this site bottom water temperature and/or δ18O. In particular, nonuniform mixing of meltwater could induce large local water δ18O signals.
 In contrast to the situation prevailing in Iberian Margin core MD99-2334K, the benthic δ18O record of South Atlantic core MD07-3076Q does not indicate any large change in bottom water properties at that site until the HS1/B-A transition (Figures 2 and 3). Changes in bottom water δ18O arising from changes in sea level might have accounted for a portion of core MD07-3076Q Cibicides δ18O signal. The heaviest Cibicides δ18O values likely reflect very cold and salty waters, in agreement with pore fluid measurements in ODP core 1093 located at ∼3630 m water depth in the South Atlantic, a few degrees South and East from site MD07-3076Q [Adkins et al., 2002]. The slight decrease in core MD07-3076 Cibicides δ18O from ∼17 to ∼15 ka could result from a slight bottom water warming and/or water δ18O decrease that could be transferred from Antarctic surface waters to depth by deep water formation around the Antarctic continent in response to the deglacial warming in the high southern latitudes, as Antarctic ice and sub-Antarctic SST records indicate that surface warming began at ∼18 ka in high southern latitudes and was already midway to its Holocene values by ∼15 ka [Barker et al., 2009; Calvo et al., 2007; Lamy et al., 2004; Lemieux-Dudon et al., 2010; Skinner et al., 2010].
 Although we lack well-dated continuous high-resolution records from North Atlantic cores located below 3200 m, the available records suggest that the onset of the glacial-interglacial decrease in benthic δ18O occurred first at ∼3000 m and later below (Figure 2), confirming the results of Labeyrie et al. . A later onset of the benthic δ18O deglacial transition in deeper cores than in shallower cores is consistent with a reduction in advection below ∼3000 m depth of the surface light δ18O meltwater signal. Such a reduction is in turn consistent with the reduction in the rate of export of northern sourced deep waters deduced from North Atlantic Pa/Th records during Henrich Stadial 1 (HS1) [Gherardi et al., 2009].
3.2. Deglacial Circulation Changes in the Atlantic Ocean Below 3000 m
 In this section, we examine the additional information contained in the dated Cibicides δ13C records from Atlantic cores located below 3000 m depth. The occurrence of marked decreases in North Atlantic deep cores Cibicides δ13C records indicates the presence of a poorly ventilated deep water mass in the North Atlantic during the HS1 and Younger Dryas (YD) cold intervals (Figure 3). These decreases in Cibicides δ13C have been interpreted as the signature of the northward expansion of southern-sourced deep water (glacial version of AABW) in response to decreases in northern-sourced deep water (glacial version of NADW) formation [Elliot et al., 2002; Oppo and Lehman, 1995; Rickaby and Elderfield, 2005; Skinner and Shackleton, 2006; Vidal et al., 1997]. This interpretation is consistent with sediment Pa/Th measurements [Gherardi et al., 2009; McManus et al., 2004] and radiocarbon data [Robinson et al., 2005; Skinner and Shackleton, 2004] showing that the rate of export of northern-sourced deep waters out of the North Atlantic basin decreased during HS1 and the YD.
 Benthic foraminifer Mg/Ca measurements conducted on Iberian Margin core MD99-2334K indicate that changes in deep water temperature paralleled changes in surface water temperature in the North Atlantic region, with colder water reaching site MD99-2334K during the HS1 and YD cold intervals than during the Bølling-Allerod (B-A) or Holocene periods [Skinner et al., 2003; Skinner and Shackleton, 2006]. Skinner et al.  and Skinner and Shackleton  demonstrated that the water mass bathing site MD99-2334K during HS1 and the YD was characterized by low temperatures, δ18O, DIC δ13C and Δ14C values, a signature they interpreted as reflecting the local dominance of southern-sourced deep water during these periods. They also showed that the 0.6‰ decrease in core MD99-2334K benthic foraminifer δ18O from 17 to 15 ka is entirely explained by a decrease in bottom water δ18O [Skinner et al., 2003].
 A simple increase in the influence of southern-sourced deep water in the North Atlantic during HS1 is difficult to reconcile with our new South Atlantic record: over the time interval 17–15 ka, benthic δ18O decreased by only ∼0.2‰ in South Atlantic core MD07-3076, whereas it decreased by ∼0.6‰ in core MD99-2334K. One might suggest that the weak decrease in South Atlantic benthic δ18O could result from the counteracting impact of a ∼0.6‰ decrease in deep water δ18O and of a 1.5 to 2°C deep water cooling over the 17–15 ka interval. However, the temperature of Southern Ocean deep water was close to the freezing point at the LGM (18–24 ka) [Adkins et al., 2002], so that it cannot have further decreased by 1.5 to 2°C over the 17–15 ka time interval. Therefore, the 0.2‰ decrease recorded in core MD07-3076 benthic δ18O cannot be explained by a 0.6‰ decrease in bottom water δ18O that would have been compensated by a large local cooling over that interval. We therefore suggest that cold, low-δ18O, and low-δ13C deep waters of northern origin have contributed to deep water composition at ∼3100 m depth on the Iberian Margin over that time interval.
 South Atlantic core MD07-3076 Cibicides δ13C data exhibit extremely low and relatively stable values centered on −1.06 ± 0.12‰ (1 σ, n = 64) over the 30–15 ka time interval which indicate that bottom water at this site was extremely poorly ventilated during the glacial and HS1 (Figure 3). Similar glacial Cibicides δ13C values have been recorded in other South Atlantic cores [Hodell et al., 2003; Mackensen et al., 2001; Ninnemann and Charles, 2002], indicating that the water mass that bathed site MD07-3076 during the last glacial and HS1 occupied a large portion of the South Atlantic, extending vertically from the seafloor to ∼2500 m depth south of 43°S, and equatorward to ∼41°N below 4000 m depth [Martínez-Méndez et al., 2009]. The presence of very poorly ventilated deep water in the South Atlantic at the end of the last glacial has recently been further supported by radiocarbon measurements on core MD07-3076 benthic foraminifera [Skinner et al., 2010].
 The large decrease in Cibicides δ18O recorded in core MD07-3076 site starting at ∼14.9 ka is accompanied by a large increase in Cibicides δ13C marking an abrupt increase in bottom water ventilation at site MD07-3076 (Figure 3). The onset of this Cibicides δ13C increase is synchronous, within dating uncertainties, with the onset of the Cibicides δ13C increase marking the HS1/B-A transition in North Atlantic cores (Figure 3). These observations suggest that the increase in deep water formation in the North Atlantic that took place at the HS1/B-A transition led to an increase in the influence of well ventilated northern-sourced water at site MD07-3076 over a ∼900 year interval, from ∼14.7 to 13.8 ka. This transient increase in bottom water δ13C at site MD07-3076 is coeval with a minimum in the benthic-atmosphere age offset [Skinner et al., 2010], which is consistent with the arrival of younger waters to the South Atlantic site.
 From 13.6 ka on, bottom water δ13C at site MD07-3076 progressively increased, reaching a second transient maximum between ∼11.9 and ∼9.9 ka. During the 13–10 ka interval, the difference between Cibicides δ13C of northern and southern Atlantic cores was noticeably reduced, indicating that site MD07-3076 bottom waters were less isolated from 13C rich water than prior to the HS1/B-A transition (Figure 3). In addition, radiocarbon ages measured on core MD07-3076 benthic foraminifera indicate that at that time, the very poorly ventilated glacial deep Southern Ocean water had substantially retreated from the Atlantic sector of the Southern Ocean at ∼3800 m depth [Skinner et al., 2010]. Therefore, the state of the ocean circulation that prevailed prior to the YD/Holocene transition was very different from the glacial state that prevailed prior to the HS1/B-A transition.
 The episode of high bottom water δ13C at site MD07-3076Q over the YD and YD/Holocene transition could have multiple causes: an increase in southern-sourced deep water δ13C in response to either, increased surface waters δ13C due to sea ice retreat, increased deep water formation around Antarctica, a change in the location of deep water formation as suggested by [Mackensen et al., 2001], or a combination of these three factors. However, part of the δ13C signal could also result from an increased contribution of northern deep water masses to MD07-3076 bottom water composition, especially after the YD/Holocene transition (at ∼11.7 ka) in response to increased deep water export out of the North Atlantic as evidence in Pa/Th [Gherardi et al., 2009; McManus et al., 2004] and other North Atlantic records.
 After 10 ka, core MD07-3076 Cibicides δ13C decreased to reach Holocene values of ∼−0.1‰, i.e., about 1‰ lower than North Atlantic deep cores Cibicides δ13C values, indicating that the Atlantic Meridional Ocean Circulation (AMOC) adjusted to another circulation regime, characterized by lower ventilation in the South Atlantic at ∼3800 m than in the North Atlantic at ∼4500 m. This new circulation regime seems to have prevailed until ∼5 ka (Figure 3). After 5 ka, the Bermuda Rise Cibicides δ13C record indicates that ventilation progressively decreased at ∼4500 m in the North Atlantic, suggesting an increasing contribution of AABW to bottom water composition at Bermuda Rise deep site.
4. Asynchrony Between Shallower and Deeper Atlantic Benthic δ18O Signals
 Atlantic benthic δ18O signals from the 1300–2200 m depth range lead Atlantic benthic δ18O signals from cores located below 3000 m over the last glacial-interglacial transition (Figures 4 and 5). More specifically, a relatively large and rapid decrease in benthic δ18O is observed at the beginning of HS1 (∼17.5 ka), followed by a marked slow down or even a reversal in the trend of the benthic δ18O signal for the shallowest cores (Figure 5). A second decrease in δ18O, analogous to the one occurring at the beginning of HS1, takes place at the beginning of the YD (∼13 ka) (Figure 5).
 The large light benthic δ18O peaks seen in Norwegian Sea core ENAM93-21 during HS1 and the YD have been shown to result from brine formation in a context of iceberg discharges and meltwater input to the surface ocean [Dokken and Jansen, 1999]. Sea-ice formation causes an increase in surface water density by salt rejection with almost no change in water δ18O. Therefore, during HS1 and the YD, brine formation transmitted the light δ18O values resulting from surface water freshening to depth. In addition, because freshening of the surface ocean reduced or even stopped open-ocean convection during stadials, deep waters underneath the fresh water lid were characterized by low 13C and 14C concentrations [Dokken and Jansen, 1999; Elliot et al., 2002], as shown by the large decreases in benthic δ13C (Figure 5). The density of these brine-generated deep waters was high enough to create overflows across the sills [Dokken and Jansen, 1999] so that brine-generated deep waters entered the North Atlantic basin and formed part of North Atlantic intermediate and deep water masses during HS1 and the YD [Meland et al., 2008; Thornalley et al., 2010; Waelbroeck et al., 2006].
 All North Atlantic benthic isotopic records of Figure 5 show the imprint of these brine-generated waters. These records reflect various levels of mixing of brine-generated waters with other water masses, depending on the water depth of the core and its proximal or distal location with respect to the Nordic Seas (Figure S1 in the auxiliary material).
 Benthic isotopic records of North Atlantic cores NA87-22 and MD95-2037, located downstream from the Norwegian deep water overflow, clearly exhibit the brine-generated water low δ18O and low δ13C signature during HS1 and the YD [Labeyrie et al., 2005; Waelbroeck et al., 2006]. Core M35003 Cibicides δ18O decreases over HS1 and the YD have been interpreted as reflecting intermediate water warming [Rühlemann et al., 2004]. According to modeling experiments, a strong reduction of the AMOC in response to freshwater input in the North Atlantic generates a temperature increase in middepth tropical Atlantic caused by a reduced inflow of cold intermediate and deep waters in conjunction with downward mixing of heat from the thermocline [Rühlemann et al., 2004]. However, models used in this type of studies do not account for brine formation and cannot therefore simulate the impact of brine-generated dense waters on the salinity and temperature fields. Moreover, in the absence of additional data, it is not possible to establish whether the observed Cibicides δ18O decreases result from warming, a decrease in water δ18O, or a combination of both. The decreases observed in Cibicides δ13C records of cores NA87-22, MD95-2037 and M35003 at 17.5 ka (Figure 5) are too large to be explained by changes in surface biological productivity alone. One could argue that the Cibicides δ13C signal reflects a decrease in site M35003 bottom water ventilation resulting from decreased overturning circulation at ∼1300 m depth during HS1. However, North Atlantic Pa/Th data suggest that there was on the contrary an increase in water export out of the North Atlantic basin above 1700 m depth during HS1 with respect to the Holocene [Gherardi et al., 2009]. Therefore, available evidence supports the hypothesis that a mixture of brine-generated and overlying North Atlantic waters reached site M35003 over HS1.
 Because South Atlantic site GeoB1711 is located at ∼1970 m depth, above the Walvis Ridge water depth, it is under the influence of the southward flowing water masses originating in the North Atlantic. Core GeoB1711 Cibicides δ18O deglacial decrease starts around 17 ka, exhibits a plateau during the B-A period and resumes over the YD, similarly to what is observed in core MD95-2037 (Figure 5). However, core GeoB1711 Cibicides δ13C signal is different from North Atlantic signals: comparing Cibicides δ13C records from North Atlantic, South Atlantic and Southern Ocean cores from about 2000 m depth reveals that ventilation was higher in the North Atlantic than in the South Atlantic sites during the last glacial and Holocene. The relatively low Cibicides δ13C values recorded in core GeoB1711 over the last glacial indicate mixing at site GeoB1711 of northern-sourced bottom waters with southern-sourced, less ventilated, CDW waters [Waelbroeck et al., 2006]. Importantly, although core GeoB1711 Cibicides isotopic record is at lower resolution than the other records from the 1000–2200 m depth range examined in the present study, the timing of core GeoB1711 Cibicides δ18O is nevertheless clearly similar to the latter records and distinct from that of the deeper benthic δ18O records.
 North Atlantic isotopic records from the 1300–2200 m depth range exhibit the imprint of low δ18O and low δ13C brine-generated waters during HS1, and during the YD to a lesser extent. The fact that there is no apparent brine-generated water signal at site M35003 during the YD could reflect the much smaller amplitude of the meltwater event and associated disruption of the AMOC during the YD than during HS1. Also, because sea level had already risen by 50% with respect to its LGM level at the beginning of the YD (whereas it had only risen by 15% at the beginning of HS1), shelf areas with water depths of less than 200 m and high potential for formation of large volumes of brine waters [Meland et al., 2008] were reduced during the YD with respect to HS1.
 During HS1, the signature of depleted δ18O water has been observed on the continental slope just south of the Iceland-Scotland Ridge, down to 3500 to 4050 m depth [Barker et al., 2004]. Although the amount of brine-generated water dense enough to reach those water depths, was probably limited, the ∼0.6‰ decrease in bottom water δ18O observed over the course of HS1 at site MD99-2334K, located at ∼3100 m on the Iberian Margin, argues for a contribution of this very dense brine-generated water to the composition of North Atlantic deep waters at ∼3100 m, in addition to southern sourced deep waters during HS1.
 In section 3 and 4, we have described changes in water mass distribution that took place in the Atlantic over the last deglaciation, we will now briefly discuss what may have caused these changes. Available climate records versus calendar age over the last 30 kyr comprise U-Th dated sea level data, 14C-dated SST records, Greenland ice core records dated by annual-layer counting [Rasmussen et al., 2006; Svensson et al., 2008] and Antarctic ice core records in the new glaciological chronology of Lemieux-Dudon et al. , which is consistent for both Greenland and Antarctic ice cores. U-Th dated sea level data indicate that the onset of sea level rise over the last deglaciation took place at 19.01 ka ± 0.14 kyr [Lambeck and Chappell, 2001], which is earlier than any of the circulation changes described above (Figure 3). A recent study showed that Northern Hemisphere ice sheets and mountain glaciers started to retreat at 19–20 ka [Clark et al., 2009], which is consistent with sea level data. Examining three of the most widely proposed mechanisms that induce changes in ice volume and feedback with the climate system (i.e., high northern latitude insolation, atmospheric CO2, and tropical Pacific SSTs), these authors further showed that northern summer insolation was the primary mechanism for triggering the onset of Northern Hemisphere deglaciation. The earliest onset of deglacial surface warming observed in sub-Antarctic [Barker et al., 2009; Calvo et al., 2007; Lamy et al., 2004] and some tropical [e.g., Levi et al., 2007] SST records, took place at ∼18 ka, in phase, within dating uncertainty, with the onset of the increase in air temperature over Antarctica and in atmospheric CO2 (respectively at 17.9 ka ± 0.3 kyr and 17.6 ka ± 0.3 kyr [Lemieux-Dudon et al., 2010]) and significantly later than the onset of sea level rise. The present study indicates that the first detectable changes in Atlantic Ocean circulation took place at the onset of the HS1 cold interval (∼17.5 ka), also in phase, within dating uncertainty, with the onset of the increase in air temperature over Antarctica and atmospheric CO2. These initial changes in Atlantic Ocean circulation occurred in the 1000–2200 m depth range, with marked decreases in North Atlantic benthic δ18O, indicating the southward propagation of brine waters generated in the Nordic Seas in a context of iceberg discharges and meltwater input to the surface ocean. We suggest that these changes in Atlantic circulation occurred in response to northern ice sheet melting.
 In summary, benthic isotopic records from the present study indicate that the first detectable changes in Atlantic Ocean circulation took place at the onset of the HS1 cold interval and that the subsequent evolution of deep water circulation in the Atlantic Ocean was characterized by transient changes over the B-A, YD and early Holocene that lead to the Holocene ocean circulation regime starting at ∼10 ka. Ocean circulation changes appear thus to have interacted with ice sheet melting, sea level rise, and changes in the global carbon cycle, to bring the climate system from a glacial to an interglacial state, in response to the initial triggering effect of the increase in northern summer insolation.
 We have shown that benthic isotopic records from different water depth ranges and locations within the Atlantic Ocean exhibit distinct patterns and significant differences in timing over the last deglaciation. This confirms that benthic δ18O cannot be used as a global correlation tool with millennial-scale precision. However, the corollary of this finding is that independently dated benthic isotopic records provide a wealth of information on past circulation changes, where these are expressed in terms of temperature and water-δ18O changes.
 Over the last deglaciation, such independent time scales can be established based on 14C dating and correlation to absolutely dated ice core records. Using this approach, we have shown the following.
 1. Changes in benthic δ18O that are attributable to circulation changes first occurred in the 1000–2200 m range, with marked decreases in North Atlantic benthic δ18O taking place at ∼17.5 ka due to the southward propagation of brine waters generated in the Nordic Seas during HS1 (∼17.5–14.5 ka).
 2. The arrival of δ18O-depleted deglacial meltwater took place later at North Atlantic sites located below 3000 m depth than at sites located above 2200 m depth, as a result of the reduction in advection of northern-sourced deep waters during HS1.
 3. The prevailing view that the low benthic and/or water δ18O and DIC δ13C values recorded in North Atlantic cores below 3000 m during HS1 is a signature of southern-sourced deep water alone is not compatible with our new South Atlantic deep isotopic record. In particular, northern sourced brine-generated deep water contributed to the lowering of North Atlantic water δ18O at ∼3100 m depth on the Iberian Margin over HS1.
 4. South Atlantic waters at ∼44°S and ∼3800 m depth remained isolated from better ventilated northern-sourced water masses until after the HS1/B-A transition. At that time, the increase in deep water formation in the North Atlantic led to an increase in the influence of well ventilated northern-sourced water at site MD07-3076. This situation however only prevailed for ∼900 years. After this episode of increased bottom water δ13C, site MD07-3076 fell back under the influence of low δ13C Southern Ocean deep waters before its bottom water δ13C progressively increased to reach a second, longer transient maximum over the YD and YD/Holocene transition. The timing of this second increase in site MD07-3076 bottom water δ13C suggests that it was likely of southern origin, with a possible subsequent northern contribution attributable to the increase in NADW export that took place after the end of the YD.
 We acknowledge G. Isguder and J. Tessier for processing the samples. We are grateful to A. Mackensen and R. Gersonde for sharing information on South Atlantic coring sites and to A. Mackensen for fruitful comments on an earlier version of this article. We also would like to acknowledge the crew of the research vessel Marion-Dufresne and the French Paul Emile Victor Institute (IPEV) for collecting the core studied here. This work was supported by the EPICA-MIS (STREP-003868) project funded by the European Union, by the PICC project funded by the ANR, and by the CNRS, CEA and IPEV, as well as the IMAGES program. We are grateful to the anonymous reviewers and the Editor, C. Charles, for their constructive comments on an earlier version of the manuscript. This is contribution 4274 of the LSCE.