Impact of abrupt climate change in the tropical southeast Atlantic during Marine Isotope Stage (MIS) 3

Authors


Abstract

[1] High resolution planktonic foraminifera Mg/Ca paleotemperatures and oxygen isotopes of seawater of Ocean Drilling Program (ODP) Site 1078 (off Angola) have been reconstructed and reveal insights into the seasonal thermal evolution of the Angola Current (AC), the Angola-Benguela Front (ABF), and the Benguela Current (BC) during the last glacial (50–23.5 ka BP). Special emphasis is put on time intervals possibly associated with the North Atlantic Heinrich Stadials (HS), which are thought to lead to an accumulation of heat in the South Atlantic due to a reduction of the Atlantic Meridional Overturning Circulation (AMOC). Within dating uncertainties, Globigerinoides ruber (pink) Mg/Ca-based sea surface temperature (SST) estimates that represent southern hemisphere summer surface conditions show several warming episodes that coincide with North Atlantic HS, thus supporting the concept of the bipolar thermal seesaw. In contrast, the Mg/Ca-based temperatures of Globigerina bulloides, representing the SST of the ABF/BC system during southern hemisphere winter, show no obvious response to the North Atlantic HS in the study area. We suggest that surface water cooling during the winter season is due to enhanced upwelling or upwelling of colder water masses which has most likely mitigated a warming of the ABF/BC system during HS. We further speculate that the seasonal asymmetry in our SST record results from seasonal differences in the dominance of atmospheric and oceanic teleconnections during periods of northern high latitude cooling.

1. Introduction

[2] The tropical southeast Atlantic surface circulation is dominated by a wind-driven current system that is crucial in the cross-equatorial heat transport [Gordon, 1986; Peterson and Stramma, 1991]. The BC represents a crucial component of the upper level warm water route of the thermohaline circulation, carrying heat northward across the equator. As one of the World's four major eastern boundary currents the BC is dominated by a coastal upwelling system, the Benguela upwelling system [Nelson and Hutchings, 1983; Shannon, 1985]. The northern boundary of the Benguela system is defined by the occurrence of a sharp thermal front, the ABF, situated at present-day and depending on the season between 13° and 17°S off southwest Africa [Shannon et al., 1987; Kostianoy and Lutjeharms, 1999; Lass et al., 2000] (see Figure 1).

Figure 1.

Modern southeast Atlantic seasonal SST pattern and surface water oceanography. AC, Angola Current; ABF, Angola-Benguela Front; BCC, Benguela Coastal Current; BOC, Benguela Ocean Current; BC, Benguela Current. Star indicates location of ODP Site 1078.

[3] During the last glacial intensified trade winds led to a strengthening of the BC upwelling which is accompanied by a cooling of the surface waters [Diester-Haass, 1985; Summerhayes et al., 1995; Schneider et al., 1996; Mix and Morey, 1996; Little et al., 1997; Kirst et al., 1999; Kim et al., 2003; Niebler et al., 2003; Farmer et al., 2005; Kucera et al., 2005]. North of the ABF the intensified trade winds result in a stronger intrusion of AC waters leading to a warming of the surface waters [Kim et al., 2003]. Regarding the glacial position and intensity of the ABF contradicting results exist. Kim et al. [2003] proposed an enhanced temperature gradient across the ABF as a result of intensified trade winds but a steady position of the front. In contrast, Jansen et al. [1996] suggested a strong northward shift of the ABF during the last glacial. The authors conclude that the entire southeast Atlantic Current system was latitudinally displaced due to a shift of the Subtropical Front and possibly the Subpolar Front.

[4] Enhanced upwelling, cooling of the surface waters, intensified trade winds and possible shifts in the annual mean position of the ABF represents only a fraction of effect the last glacial had on the tropical southeast Atlantic. Of particular interest and importance are the effects of millennial-scale climate fluctuation associated with North Atlantic freshwater perturbations, the so-called Heinrich Stadials (HS), which have been found in records from various places around the globe including the tropics [e.g., Leuschner and Sirocko, 2000; Dupont et al., 2008; González et al., 2008; Tjallingii et al., 2008; Zhou et al., 2008]. Based on the concept of a thermal bipolar seesaw a shut-down or reduction in the strength of the Atlantic Meridional Overturning Circulation (AMOC) has been suggested for periods associated with HS, leading to a cooling of the North Atlantic realm and an accumulation of heat in the South Atlantic [Broecker, 1998; EPICA Members, 2006; Barker et al., 2009]. In contrast to the cooling in the northern hemisphere, SST records from the southeast South Atlantic reveal a warming due to the reduction of the AMOC and/or a potential intensification of the Agulhas Leakage during HS [e.g., Barker et al., 2009]. However, very little is known about seasonal changes associated with HS in the tropical southeast Atlantic. Although there are several records that indicate changes in the surface water conditions of the southeast Atlantic in relation to HS [Little et al., 1997, Kim et al., 2003, Farmer et al., 2005] none of these records provide seasonal estimates. This, however, is of particular importance since a number of multiproxy studies suggest a proxy-dependent discrepancy in the temporal pattern of surface water temperatures that are most likely attributed to differences in the seasonality of the proxy carriers used in these studies (e.g., planktonic foraminifera, alkenone producing algae) [e.g., Mix, 2006; Steinke et al., 2008; Leduc et al., 2010b]. Yet, none of these studies deal with HS in the tropical southeast Atlantic and its complex oceanography.

[5] Here, we present high-resolution Mg/Ca-based SST records of G. bulloides and pink-pigmented G. ruber in order to resolve the winter and summer SST development during the last glacial (50–23.5 ka BP) and particularly HS in the tropical southeast Atlantic. Stable oxygen isotopes have been obtained from G. bulloides which provide, when removing the temperature and ice volume effect on the stable isotopes, insight into possible salinity variations during the last glacial and HS. Analyzed sediments originate from Ocean Drilling Program Site 1078 located in the Bight of Angola.

2. Modern Oceanographic Setting

[6] The modern oceanography in the southeast Atlantic is dominated by a seasonal varying trade wind system. As one of the four major eastern boundary currents, the BC, is the main oceanographic feature controlling the regional surface water circulation in the southeast Atlantic (Figure 1) [Nelson and Hutchings, 1983; Shannon, 1985; Stramma and Peterson, 1989; Peterson and Stramma, 1991]. Its water masses are formed at the subtropical convergence zone between 30° and 40°S at a depth of 100 to 300 m. Subantarctic surface water which is added at the Subtropical Front, and Antarctic Intermediate Waters (AAIW) that enter the South Atlantic through the Drake Passage contribute significantly to the BC [Nelson and Hutchings, 1983; Stramma and Peterson, 1989; Garzoli and Gordon, 1996]. The contribution of Agulhas waters appears to be small but continuous [Nelson and Hutchings, 1983; de Ruijter et al., 1999].

[7] The Benguela Current system flows north-northwest along the coast of southwest Africa, dividing around 25°S into the coastal oriented Benguela Coastal Current and the northwestward flowing Benguela Ocean Current (Figure 1). Seasonal components of the BC extend up to 12°–13°S defining its northernmost boundary (Figure 1) [Moroshkin et al., 1970]. Atmospheric conditions over the BC region are mainly influenced by the semi-permanent South Atlantic high pressure cell and a low pressure system developing over southern Africa during the southern hemisphere summer. Prevailing winds from the south and southeast induce coastal upwelling of cold, nutrient rich waters [Nelson and Hutchings, 1983; Shannon, 1985; Stramma and Peterson, 1989]. The annual position and intensity of the BC upwelling system varies as a consequence of seasonal shifts in the position and intensity of the South Atlantic High and changes in the atmospheric pressure over the African subcontinent.

[8] The warm AC which is fed by the southeastern branch of the South Atlantic Counter Current is flowing southward along the shelf of Angola [Moroshkin et al., 1970]. Between 13° and 17°S, the waters of the AC encounter the upwelled waters of the BC which results in the development of a sharp thermal front (temperature drop from 27° to 20.5°C [Locarnini et al., 2006]), the ABF (Figure 1). The meridional migration of the ABF is induced by the seasonal varying wind system. Coupled to the equatorward extension of the coastal upwelling zone the ABF reaches its northernmost position of 12° to 13°S during the southern hemisphere winter season (Figure 1). During this time of the year, the ABF is less well defined and warm waters of the AC intrude regularly as far south as 22°S. During southern hemisphere summer when the ABF is located in its southernmost position at about 17°S the convergence between the AC and BC is highest [Shannon et al., 1987].

3. Strategy and Proxy Variables

[9] Two different planktonic species of foraminifera, G. bulloides and G. ruber (pink), were selected to reconstruct variations in surface water conditions in the tropical southeast Atlantic during the last glacial. Temperature variations of near surface waters estimated using Mg/Ca in G. bulloides often relate to upwelling environments [Regenberg et al., 2009]. The Mg/Ca of G. ruber (pink) is used to reconstruct SST variations in a warm water environment [Hemleben et al., 1989]. Relative variations in the seawater salinity in the upwelling regime during the southern hemisphere winter season were estimated using stable oxygen isotope measurements of G. bulloides. Although more abundant than G. ruber (white), there were not enough specimens of G. ruber (pink) to perform additional stable oxygen isotope measurements.

[10] The spatial distribution of planktonic foraminifera depends largely on parameters such as water temperature, thermal stratification and nutrients [Bé and Tolderlund, 1971; Hemleben et al., 1989]. The occurrence of G. bulloides is strongly related to poorly stratified eutrophic waters in subpolar and upwelling regions [Hemleben et al., 1989; Ufkes et al., 1998]. Accordingly, its abundance in the southeast Atlantic is highest in the ABF and BC [Ufkes et al., 1998; Jansen et al., 1996]. In contrast, G. ruber (pink), typical for oligotrophic warm tropical waters, mainly occurs in the warm waters of the AC [Jansen et al., 1996; Ufkes et al., 1998].

[11] The seasonal shift of the ABF provides a dominant control on the temporal distribution pattern and composition of planktonic foraminifera assemblages in our research area. Sediment trap studies on G. bulloides and G. ruber (pink) from different regions of the world oceans confirm that both species become important components of the foraminiferal population during different seasons with different environmental conditions [e.g., Žarić et al., 2005; Wilke et al., 2009]. It has been shown that during periods of upwelling, high productivity and cool surface waters, fluxes of G. bulloides are highest [e.g., Žarić et al., 2005; Wilke et al., 2009]. During southern hemisphere winter when the ABF is its northernmost position the ambient waters of ODP Site 1078 are mainly influenced by waters from the ABF and BC. In contrast, during southern hemisphere summer when surface water temperatures are highest, the upper water column is strongly stratified, G. ruber (pink) becomes an important part of the warm water assemblages [Thompson et al., 1979; Žarić et al., 2005; Wilke et al., 2009]. The ABF is at its southernmost position during southern hemisphere summer allowing for the AC to penetrate further south into the southeast Atlantic (Figure 1). G. ruber (pink) is most common in the upper 25 m but occurs occasionally down to 50 m water depth as shown by plankton tow studies in the Bight of Angola [Oberhänsli et al., 1992]. In the Benguela upwelling system, G. bulloides occurs mainly in the upper 50 m of the water column but was also found in lesser abundance at a water depth of 100–150 m [Oberhänsli et al., 1992].

[12] The different ecological and seasonal preferences of the planktonic foraminifera species used in this study enable us to derive seasonal SST estimates for the tropical southeast Atlantic. Thus, G. ruber (pink) Mg/Ca based temperatures indicate changes of the AC during southern hemisphere summer. G. bulloides Mg/Ca based SST estimates allow to monitor the ABF/BC system during the southern hemisphere winter season.

4. Material and Methods

4.1. Site Description and Age Model

[13] The planktonic foraminifera tests used in this study were obtained from sediments of ODP Site 1078 Hole C located inside the Bight of Angola (11°55′S, 13°24′E) in a water depth of 426 m (Figure 1). The sediment is predominately composed of moderately bioturbated olive-gray silty clay with varying amounts of nannofossils and planktonic and benthic foraminifera [Wefer et al., 1998].

[14] The chronostratigraphy of the studied time interval (22−>45 ka BP) developed by Dupont et al. [2008] is based on 6 AMS 14C dates. All 14C ages were calibrated using the Fairbanks0107 calibration data set [Fairbanks et al., 2005].

[15] All radiocarbon ages were corrected using an average oceanic reservoir age of 400 years [Hughen et al., 2004]. Although simulations for the glacial ocean indicate an increase in the surface water oceanic reservoir age of up to 600 years [Butzin et al., 2005], an increase of the marine reservoir age of 200 years in the glacial dates of ODP Site 1078 would not significantly change the age model [Dupont et al., 2008] and the interpretation of our SST records. Additionally, model simulations show that temporal fluctuations in the marine reservoir age are less significant in the tropical Atlantic compared to the northern high-latitudes [Hughen et al., 2004; Franke et al., 2008]. In the following all ages refer to calibrated ka before present (cal ka BP).

[16] The Greenland (NGRIP) and Antarctica (EDML) stable oxygen isotope records shown in Figure 2 act as reference records and indicate the different course of temperature fluctuations in both hemispheres. However, we want to emphasize that no tuning of the age model of ODP Site 1078 to either the NGRIP and/or EDML record has been conducted.

Figure 2.

Proxy records of ODP Site 1078 for the time interval 50–23.5 ka BP compared to the Greenland (NGRIP) and Antarctic (EDML) oxygen isotope records for the same period. (a) NGRIP δ18O is displayed on the GICC05 time scale [Andersen et al., 2006; Svensson et al., 2006] and given in per mil (‰). The NGRIP record acts as reference showing the abrupt nature of millennial-scale climate change in the northern high latitudes. (b) Mg/Ca SST of G. ruber (pink) in °C. Standard deviation of Mg/Ca SST (°C) is indicated with error bar (labeled SD) on graphs. Error of temperature calibration of ∼1°C [Anand et al., 2003] is indicated next to the temperature scale. Arrow on scale indicates average modern day southern hemisphere summer SST (26.8°C) [Locarnini et al., 2006]. (c) Mg/Ca of G. ruber (pink) in mmol/mol. (d) Mg/Ca SST (°C) of G. bulloides calibrated after Mashiotta et al. [1999]. Standard deviation of Mg/Ca SST (°C) is indicated with error bar (labeled SD) on graphs. Error of temperature calibration of ∼1°C [Mashiotta et al., 1999] is indicated next to the temperature scale. Arrow on scale indicates average modern day southern hemisphere winter SST (∼21.9°C) [Locarnini et al., 2006]. (e) Mg/Ca of G. bulloides in mmol/mol. (f) Seawater δ18O of G. bulloides given in per mil (‰). The temperature driven component of δ18O was removed using the empirical equation of Shackleton [1974]. The effect of ice volume was removed using data of Waelbroeck et al. [2002]. All original data are available at www.pangaea.de. Data of G. ruber (pink) and G. bulloides have been smoothed with a 5-point running average. The running standard deviation for both G. ruber (pink) and G. bulloides is mostly between 0.5 and 1, hence actual temperature changes exceed the data scatter in most parts of the record. (g) EDML δ18O is displayed on the GICC05 time scale [Andersen et al., 2006; Svensson et al., 2006] and is given in per mil (‰). It acts as a reference record showing the more gradual signal of southern hemisphere climate change. The age model is illustrated as the one sigma age range calibrated according to Fairbanks et al. [2005] including an oceanic reservoir age correction of 400 years. Heinrich Stadials (HS5 = 49.2–46.8, HS4 = 40.1–38.2, HS3 = 32.5–28.9) are after Wolff et al. [2010].

4.2. Analytical Methods

[17] Samples (n = 155) for both species were taken for Mg/Ca analysis with a sampling interval of 2 cm (average sample resolution ∼200 years) and a volume of 10 cm3. Approximately 40 specimens of G. bulloides (250–315 μm) and ∼15 to 40 specimens of G. ruber (pink) (250–400 μm) were picked, gently crushed, and cleaned following the cleaning protocol of Barker et al. [2003]. Clays were removed by rinsing the samples five times with distilled deionized water and two times with methanol (Ultrapure) while after each rinse samples were briefly ultrasonicated. Samples were then oxidized with an alkali buffered 1% H2O2 solution and heated for 10 min to remove organic matter. Possible gaseous build-ups were released by tapping the micro-tubes every 2.5 min, while after 5 min micro-tubes were briefly placed in an ultrasonic bath to maintain the contact between reagent and sample. This step was repeated after renewing the oxidation solution to ensure the removal of all organic matter. After removing the oxidation solution, rinsing with distilled deionized water (SERALPUR), and transferring the samples into new clean micro-tubes a weak acid leach with 250 μl 0.001 M QD HNO3 was applied including a 30 s of ultrasonic treatment. After two rinses with distilled deionised water the samples were dissolved in 0.075 M QD HNO3 and placed in the ultrasonic bath to support the dissolution. After dissolution, any potentially remaining solids were removed by centrifuging the samples for 10 min (600 rpm) and transferring the supernatant into clean micro-tubes for dilution and analysis. Samples were measured with an ICP-OES Perkin Elmer Optima 3300 RL with autosampler and ultrasonic nebulizer (U-5000 AT Cetac Technologies Inc., Department of Geosciences, University of Bremen). Instrumental precision of the ICP-OES was ensured by analysis of an in-house standard solution (Mg/Ca = 2.92 mmol/mol) every fifth sample. In order to allow for interlaboratory comparison an international limestone standard (ECRM752–1) with a specified Mg/Ca of 3.75 mmol/mol [Greaves et al., 2008] was routinely analyzed twice before each batch of 50 samples. Analytical precision was monitored calculating the relative standard deviation which is on average ±0.195%. Duplicate measurements for G. bulloides resulted in a mean standard deviation (n = 5) of ±0.24 for Mg/Ca (mmol/mol), the mean standard deviation for G. ruber (pink) (n = 2) was ±0.14 for Mg/Ca (mmol/mol). To monitor the efficiency of the cleaning and/or possible silicate contaminations Al/Ca and Fe/Ca have been also determined [Barker et al., 2003]. There has been no correlation between Al/Ca, Fe/Ca and Mg/Ca detected, indicating that terrigenous materials were fully removed by the cleaning process. Cleaning protocols including a reductive cleaning step [e.g., Boyle and Keigwin, 1985] are known to remove possible Mn-Fe coatings on the tests but also systematically lower the Mg/Ca ratio due to partial dissolution [Barker et al., 2003]. The cleaning protocol of Barker et al. [2003] without reductive cleaning step was applied to avoid such bias of Mg/Ca.

[18] Mg/Ca were converted to SST by means of the species-specific calibration for G. bulloides and G. ruber (pink) developed by Mashiotta et al. [1999] and Anand et al. [2003], respectively.

[19] For the oxygen isotope measurements G. bulloides tests were picked from the same samples as for Mg/Ca analysis. Depending on the availability of material, between 6 and 12 specimens were picked from the size fraction 250–315 μm. Stable oxygen isotope analyses were performed using a Finnigan MAT 251 mass spectrometer with an automated carbonate preparation device at the MARUM, University of Bremen. The external standard error of the stable oxygen isotope analyses is < 0.06‰. Values are reported relative to the Vienna Pee Dee Belemnite (VPDB), calibrated by using the National Bureau of Standards (NBS) 18, 19, and 20 standards. To calculate the δ18O of seawater (δ18Osw), the temperature driven component of changes in the G. bulloides δ18O record was removed using the empirical equation of Shackleton [1974]. The effect of continental ice volume on the stable oxygen isotopes was removed using data of Waelbroeck et al. [2002].

[20] Dissolution is known to affect the Mg/Ca of planktonic foraminifera in a systematic way, with increasing dissolution causing a progressive decrease in the Mg/Ca [Regenberg et al., 2006]. Purely based on its shallow location at 426 m water depth a possible effect of carbonate dissolution is negligible [Regenberg et al., 2006]. Due to its location in an upwelling sensitive area it can, however, not be excluded that some degree of supra-lysoclinal dissolution has taken place. But, we consider the impact minor due to the very good preservation of the tests of G. bulloides and G. ruber (pink), which were classified by Berger [1968] as very susceptible for dissolution.

5. Results

[21] Throughout the covered time period the Mg/Ca based surface water temperatures indicated by G. ruber (pink) range between 25° and 27°C (Figure 2). Temperature peaks during three intervals, 46–44 ka BP (25°–26.5°C), 40–37 ka BP (26°–27.5°C) and 30–28 ka BP (25°–26°C). Highest temperatures (up to 27.5°C) are recorded between 40 and 37 ka BP. In general Mg/Ca based temperatures of G. ruber (pink) increase between 50 and 40 ka (25°–26.5°C), followed by a decrease (27°–25°C) between 37 and 23.5 ka BP (Figure 2).

[22] The Mg/Ca of G. ruber (pink) vary between 3.5 and 4.0 mmol/mol. The Mg/Ca exceed 4.0 mmol/mol between 36.5 and 38.5 ka BP. Maximum values of up to 4.5 mmol/mol are recorded around 38 ka BP (Figure 2).

[23] G. bulloides Mg/Ca based temperatures range between 16.5° and 20°C over the investigated time period (Figure 2). SSTs are on average cooler (16.5–19.5°C) during the period 50–38 ka BP and changes are of higher amplitude than during the period 38–23.5 ka BP (Figure 2). Between 50 and 38 ka BP G. bulloides temperatures range between 16.5° and 19.5°C whereas coldest temperatures are recorded at 49 ka BP (16.5°C) and around 42–44 ka BP (17°C). However, Mg/Ca based SST of G. bulloides generally increase between 50 and 38 ka (16.5°–19°C). Higher temperatures and only minor fluctuations (19°–20°C) are observed for the period 40–23.5 ka BP.

[24] The Mg/Ca of G. bulloides range between 2.8 and 4.2 mmol/mol (Figure 2). Between 50 and ∼40 ka BP values are usually lower (2.8–3.7 mmol/mol) whereas the period between ∼40 and 23.5 ka BP is marked by higher ratios fluctuating between 3.4 and 4.2 mmol/mol.

[25] The seawater δ18O-values vary between 1.26 and 0.11‰ (Figure 2). Similar to the Mg/Ca temperature record of G. bulloides, the δ18Osw record reveals two distinct phases: Pronounced variations occur between 50 and 38 ka BP with lowest values between 44 and 42 ka BP (0.1–0.5‰). Highest values (∼1.25‰) occur between 39 and 37.5 ka BP. The period between 38 and 23 ka BP is marked by less distinct variations and values are ranging between 0.5 and 0.1‰. However, a distinct decrease in δ18Osw (0.5–1‰) occurs at 30.5–29 ka BP.

6. Discussion

[26] In the following the Mg/Ca record of G. ruber (pink), representing SST estimates during southern hemisphere summer and the dominance of the AC on the core location is discussed at first. Subsequently, the G. bulloides derived SST estimates are discussed to give insight into the glacial development of the ABF/BC system during southern hemisphere winter.

6.1. Southern Hemisphere Summer

[27] Several paleotemperature reconstructions of the glacial southeast Atlantic surface waters indicate an overall decrease of SST between 3.5°C and 5°C in comparison to modern times (26.8°C [Locarnini et al., 2006]) [Schneider et al., 1995; Kirst et al., 1999; Chen et al., 2002; Kim et al., 2003; Farmer et al., 2005].

[28] Our G. ruber (pink) Mg/Ca temperature record represents the SST during the southern hemisphere summer, which is the season when the influence of the warm AC on the core site is dominant. The reconstructed glacial temperatures are up to 3°C lower compared to today, which is consistent with other glacial temperature reconstructions [e.g., Schneider et al., 1995; Farmer et al., 2005]. Most significantly, our G. ruber (pink) Mg/Ca based SST record reveals considerably increased temperatures of 1–2°C around 46 ka, 38 ka and 30 ka. Within dating uncertainties, these warming events are possibly associated with the North Atlantic HS. Most pronounced is the warming between 40 and 37 coinciding with HS4 (Figure 2). This increase in SST can be attributed to mode shifts in the Atlantic heat transport regime associated with changes in the intensity of the AMOC [EPICA Members, 2006; Barker et al., 2009]. According to the concept of the thermal bipolar seesaw, a slow down or reduction in the intensity of the AMOC during HS causes a warming in the South Atlantic while the temperatures in the North Atlantic decrease [Stocker, 1998; Broecker, 1998; Seidov and Maslin, 2001; Barker et al., 2009].

[29] In contrast to the abrupt nature of millennial-scale climate change recorded in the northern high-latitudes, the course of temperature fluctuations in Antarctica is significantly wider [NGRIP, 2004; EPICA Members, 2006; Barker et al., 2009]. Similar to the Antarctic (EDML) surface air temperature record the warming events in the tropical southeast Atlantic recorded in our G. ruber (pink) Mg/Ca SST record are more gradual than the abrupt (millennial-scale) changes in Greenland during these time intervals (Figure 2). Therefore, the SST fluctuations observed in the G. ruber (pink) record resemble the temperature variations recorded over Antarctica (EDML) record and hence seem to follow the southern hemisphere signal during MIS 3.

[30] The warming of the South Atlantic due to thermal anti-phase behavior seems to be strongest in the southeast Atlantic off the coast of west and/or southwest Africa where modeling experiments indicate the formation of a warm water pool [Rühlemann et al., 2003; Prange et al., 2004; Merkel et al., 2010]. Simulated SSTs reveal an increase of about 1–2°C for meltwater perturbation experiments [Prange et al., 2004; Merkel et al., 2010]. This is consistent with our results that show a 1–2°C increase in SST during periods, within dating uncertainties, associated with HS (Figure 2). A rapid and intense warming of intermediate waters during HS1 at ODP Site 1078 is indicated by benthic foraminiferal oxygen isotopes [Rühlemann et al., 2003; Rühlemann et al., 2004]. The authors suggest that the warming occurs due to a reduced ventilation of cold intermediate and deep waters in conjunction with a downward mixing of heat from the thermocline. Similarly, warming of the surface waters in conjunction with HS4 and HS5 recorded in a planktonic δ18O record south of Walvis Ridge (GeoB1711) has been attributed to a reduced production of North Atlantic Deep Waters in the northern high latitudes [Vidal et al., 1999].

6.2. Southern Hemisphere Winter

[31] Comparable to the summer season SST the G. bulloides based SST, which are indicative for the dominance of water masses of the ABF and BC at the core location during southern hemisphere winter, show significantly cooler average temperatures than today (21.9°C; Locarnini et al., 2006). The Mg/Ca temperatures of G. bulloides vary between 16.5° and 20°C and are therefore 2°–5°C below modern temperatures (Figure 2). There are several mechanisms potentially responsible for the strong cooling of the BC and ABF water masses during glacial southern hemisphere winters in comparison to the summer.

[32] First, the temperature of the upwelled waters of the Benguela coastal current were generally cooler during the last glacial than today, perhaps due to a northward expansion of Antarctic cold waters [Keeling and Stephens, 2001; Gersonde et al., 2005]. Compared to present-day conditions, Antarctic winter sea ice is reconstructed to have expanded about 10° northward reaching up to 47°S during the last glacial [Gersonde et al., 2005]. Consequently, the Southern Ocean frontal system would have been shifted northward, giving way to a northward expansion of the Antarctic cold water realm [Keeling and Stephens, 2001; Gersonde et al., 2005]. A northward shift of the Subtropical Front could have also cut of the warm water route via the Agulhas Leakage leading to a reduced supply of warm waters to the Benguela Current [Gersonde et al., 2003, 2005]. However, contradicting results are given by Matsumoto et al. [2001] arguing against a glacial migration of the Southern Ocean Polar Front. This study not only indicates a largely unchanged position of the Southern Ocean oceanic fronts but generally little change in the overall Southern Ocean hydrography. On a more local scale a northward shift of the South Atlantic Fronts would results in a more northward position of the ABF [Jansen et al., 1996]. Hence, the ABF would have moved closer to our core location or even passed over it toward a more northerly position resulting in lower SST due to an enhanced influence of the BC. However, contrasting results are given by Kim et al. [2003] and Colberg and Reason [2006] arguing against a shift of the ABF but rather a stable glacial position and a pronounced temperature gradient across the ABF due to enhanced trade winds.

[33] The intensification of the Benguela upwelling as a result of enhanced trade winds would have also led to decreased SSTs during the last glacial [Diester-Haass, 1985; Mix et al., 1986; Summerhayes et al., 1995; Schneider et al., 1996; Mix and Morey, 1996; Little et al., 1997; Kirst et al., 1999; Shi et al., 2001; Kim et al., 2003; Farmer et al., 2005; Kucera et al., 2005]. Due to enhanced upwelling rates cold water would intrude further to the north influencing the ambient waters of ODP Site 1078 more directly and/or extend the duration of the cold water influence from the south. Kirst et al. [1999] and Schneider et al. [1996] indicate that the upwelling intensity was at its maximum between 50 and 35 ka and Shi et al. [2001] reconstructed enhanced trade wind and upwelling rates between 56 and 42 ka BP. During these periods coolest temperatures have been recorded [Schneider et al., 1996; Kirst et al., 1999; Shi et al., 2001] which is in good agreement with our G. bulloides Mg/Ca-based SST record (Figure 2).

[34] We suggest that the decreased austral winter SSTs in the study area during the last glacial reflect the combined effect of an intensification of the Benguela coastal upwelling, and enhanced northward intrusion of cold BC waters and a decreased inflow of warm waters via the Agulhas Leakage due to changes in glacial boundary conditions.

6.3. Millennial-Scale Climate Change During Southern Hemisphere Winter?

[35] An impact of millennial-scale climate change on the ABF and BC during southern hemisphere winters potentially reflected in the Mg/Ca temperature of G. bulloides is less obvious than in the G. ruber (pink) record. However, it seems that the surface waters of the BC and ABF tend to warm during periods of HS4 and HS5 which would agree with the bipolar thermal seesaw hypothesis [Broecker, 1998; EPICA Members, 2006; Barker et al., 2009]. Yet the duration of the warming considerably exceeds the time frame of HS which makes a connection to northern high-latitude climate change questionable. Therefore, we assume that the long-term warming trend recorded in the winter SST between 50 and 38 ka BP (Figure 2) is not related to HS but likely attributed to (1) a general weakening of the trade winds associated with a decrease in upwelling intensity [Shi et al., 2001] or (2) a change in the southeast Atlantic current configuration due to a more northward position of the ABF and/or Subtropical Front [Berger and Wefer, 1996; Jansen et al., 1996]. However, the surface water temperatures in the BC region remain well below its modern level indicating the persistence of enhanced upwelling due to more windy conditions [Shi et al., 2001]. These inferences, however, should be explored further with a broader geographical coverage of surface water temperature records that span the last glacial.

[36] While the general difference in the temperature development recorded by both planktonic foraminifera species is most likely attributed to ecological responses of G. bulloides and G. ruber (pink) to changes in winter and summer surface water conditions, respectively, the differences in the response to millennial-scale climate change are less obvious. In the following we provide several possible mechanisms shaping our southern hemisphere winter record:

[37] 1. It could simply be that there are no millennial-scale changes during southern hemisphere winter in the Benguela upwelling system. While the G. bulloides Mg/Ca-based SST record from the southern southeast Atlantic (TN057–21 [Barker et al., 2009]) does show a clear warm event during HS1, our Mg/Ca-based SST record of G. bulloides as well as a G. bulloides δ18O record [Romero et al., 2003], both located in the Benguela upwelling system, give no indication for an impact of abrupt climate change. However, we are aware that the characteristics of different HS do not necessarily match.

[38] 2. Although the impact of HS on the southern hemisphere winter SST appears to be doubtful the oxygen isotope composition of seawater reconstructed using G. bulloides tests indicates a freshening of the surface waters possibly associated with HS3 (Figure 2). The opposite pattern, a slight increase in the seawater δ18O-values, is detectable during the period likely associated with HS4. However, no significant variations in the seawater δ18O is apparent in relation with HS5.

[39] Collins et al. [2011] recently showed that the distribution of rainfall in tropical Africa varied along with climate changes in the northern high latitudes and AMOC. They further argue against a southward shift of the ITCZ during HS as suggested by several paleoclimate studies [e.g., Mulitza et al., 2008], but rather a symmetrically contraction of the African rainbelt leading to an increase in precipitation at 12°S compared to the LGM [Collins et al., 2011], thus a salinity decrease in the southeast Atlantic. The prevailing fresh surface water conditions in the southeast Atlantic between 44 and 42 ka BP can, however, not be attributed to increased rainfall rates due to the contraction of the African rainbelt during HS.

[40] If we consider the slight increase in the seawater δ18O-values at about 37.5 to 39 ka BP as a real response to HS4 a contraction of the African rainbelt [Collins et al., 2011] or southward shift of the ITCZ [Mulitza et al., 2008; Lee et al., 2011], both leading to an increase in precipitation over southwest Africa, does not apply for HS4. The salinity increase likely associated with HS4 could be the result of higher evaporation rates due to warmer surface waters, as it is suggested by our G. bulloides record.

[41] 3. The effect of increasing SST due to HS could potentially be masked or subdued by mechanisms that cancel out each other. For instance, periods of high abundances of left coiling planktonic foraminifera Neogloboquadrina pachyderma in the northern Benguela upwelling system, the so-called PS (pachyderma sinistral)-events, correlate with North Atlantic HS and indicate increased intensity and zonality of SE trade winds [Little et al., 1997]. These PS events or periods of intensified upwelling are further interpreted to be the primary control on the SST patterns in the northern Benguela region during HS and would be capable to mitigate the expected HS warming.

[42] 4. Recent modeling studies suggest that atmospheric processes are highly important in terms of inter-hemispheric climate coupling during HS [Lee et al., 2011]. They postulate that freshwater induced cooling in the North Atlantic led to the reorganization of the atmospheric circulation resulting in a southward shift of the ITCZ, a weakening of the southern branch of the Hadley cell, and an alternation of subtropical jet which finally allows the midlatitude westerlies to increase the wind stress particularly during southern hemisphere winter at the latitude of the Drake Passage. Since parts of the water masses of the BC originate from AAIW [e.g., Reid, 1989; Garzoli and Gordon, 1996] of the Drake Passage loop [Stramma and England, 1999], the increase in wind stress at this latitude could increase the volume transport of cold AAIW into the BC lowering its temperature during HS and possibly damping the effect of warming due to bipolar seesaw. This assumption is supported by the study of Anderson et al. [2009] interpreting the deglacial minimum of δ13C in planktonic foraminifera, which is occurring around HS1 and concurrent with their deglacial maximum in opal burial in equatorial upwelling regions, as period of enhanced entrainment of deep waters during the formation of Subantarctic Mode Waters which eventually surface in tropical upwelling regimes.

[43] 5. Leduc et al. [2010a] found a correlation between BC SST minima and warm Antarctic air temperatures as well as warm periods in the Moroccan upwelling system when analyzing recent climate anomalies like the Little Ice Age which has comparable characteristics of HS, however, with smaller amplitude. The inter-hemispheric thermal anti-phase behavior occurring on millennial and multicentennial timescales is suggested to modulate the land-ocean atmospheric pressure gradient and the intensity of winds blowing along the coastline. Therefore, the inter-hemispheric bipolar seesaw together with local land-sea interactions would lead to an increase in the coastal upwelling when regional climate warms [Leduc et al., 2010a]. Assuming that such scenario also applies for HS during MIS3, the regionally enhanced winds would increase the BC upwelling, thus mitigate the warming in the South Atlantic expected from the bipolar seesaw [Broecker, 1998].

[44] In summary we speculate that enhanced upwelling rates and/or the upwelling of colder water masses during southern hemisphere winter which are associated with an increased equatorward advection of cold waters from the subtropical convergence during the HS, most likely masked or subdued increased SST in the study area as expected from the bipolar seesaw concept. However, though this assumption seems to be plausible the exact mechanisms involved need to be further investigated.

[45] We further speculate that the seasonal asymmetry in our southeast Atlantic SST record is due to a seasonal difference in the dominance of the oceanic and atmospheric transmitted HS signal. While during the southern hemisphere winter season the atmospheric mechanisms proposed by Lee et al. [2011] dominate over the oceanic bipolar seesaw effect and mitigate the expected warming signal in the BC, oceanic teleconnections as the bipolar seesaw concept of Broecker [1998] seem capable of shaping our temperature record during southern hemisphere summer.

7. Conclusions

[46] The G. ruber (pink) and G. bulloides derived Mg/Ca SST records of ODP Site 1078 off Angola indicates the temperature development and their differences during the southern hemisphere summer and winter between 23.5 and 50 ka BP. Due to the seasonal and ecological preferences of pink-pigmented G. ruber, its Mg/Ca SST reflects the thermal behavior of the AC which dominates the ambient waters of the core location during southern hemisphere summer. The G. ruber Mg/Ca based SST record shows generally cooler temperatures compared to today and a significant but gradual warming during periods likely associated with HS. These results are on the one side in good agreement with other southeast Atlantic temperature reconstructions showing a cooling during the last glacial [e.g., Mix and Morey, 1996; Niebler et al., 2003; Kucera et al., 2005]. On the other side the more gradual course of warming recorded in the Mg/Ca based SSTs of G. ruber (pink) is comparable to the Antarctic (EDML) air surface temperature record indicating a likely association with HS. In addition, the recorded warming supports the concept of the bipolar thermal seesaw [Broecker, 1998; EPICA members, 2006; Barker et al., 2009].

[47] The glacial evolution of southern hemisphere winter temperatures coinciding with the prevalence of the ABF/BC system on the core location is reflected in the G. bulloides record. Similar to the summer season, southern hemisphere winter temperatures between 50 and 23.5 ka BP are lower compared to modern SST, hence agreeing with other paleotemperature reconstructions [e.g., Mix and Morey, 1996; Niebler et al., 2003; Kucera et al., 2005]. The generally cooler SSTs during the investigated time interval in the ABF/BC system might result from an intensification of the coastal upwelling due to enhanced trade winds combined with an increase in the northward intrusion of cold BC waters. A decrease of the inflow of warm waters via the Agulhas Leakage due to a more northern position of the Subtropical Front [e.g., Berger and Wefer, 1996; Gersonde et al., 2005] may have also played a role. In contrast to the G. ruber (pink) record, the winter SST estimates based on G. bulloides do not show an obvious response to changes associated with HS. Within dating uncertainties, periods of increased SST are recorded during HS4 and HS5 but not during HS3. However, since these warming periods exceed the time frame of both HS4 and HS5, we are careful about their association to HS but also do not want to eliminate the possibility. Assuming that the warming periods are not associated with HS their absence in the southern winter temperatures may be explained by several mechanisms. However, we certainly cannot identify the exact factors involved in shaping our southern hemisphere winter SST record, but we can raise some speculations.

[48] First, we speculate that enhanced upwelling rates alone or combined with an increased contribution of cold water masses to the BC due to increased wind stress in the latitude of the Drake Passage masked or subdued the expected increasing SST in the BC region during the winter season of HS.

[49] Second, we speculate that the seasonal asymmetry in our Mg/Ca based SST records is based on seasonal difference in the dominance of ocean-atmosphere teleconnections during HS. Atmospheric mechanisms proposed by Lee et al. [2011] seem to dominantly shape the southern hemisphere winter SST. In contrast, the comparability of our southern hemisphere summer SST record with the Antarctic EDML δ18O record suggest that oceanic processes related to the bipolar seesaw concept of Broecker [1998] transmitted the HS signal from the northern high latitude to the tropical southeast Atlantic.

[50] That the HS signal is not entirely absent during the winter season is shown by the freshening of surface waters during HS3 which is conform with the hypothesis of a southward shifting ITCZ [Mulitza et al., 2008; Lee et al., 2011] or a contraction of the African rainbelt [Collins et al., 2011] both, however, lead to an increase of precipitation in the southern hemisphere African tropics.

Acknowledgments

[51] We thank the Ocean Drilling Program for providing samples. This study was funded by the Deutsche Forschungsgemeinschaft (DFG) as part of the German contribution to the Integrated Ocean Drilling Program (SPP 527) and by the Deutsche Forschungsgemeinschaft as part of the DFG-Research Center - Cluster of Excellence “The Ocean in the Earth System.” We thank Stefan Mulitza and Mahyar Mohtadi for the discussions and suggestions which clearly improved the manuscript. Furthermore, we want to thank Rainer Zahn as well as the three anonymous reviewers for their detailed and helpful comments which significantly improved the manuscript. Isotope measurements have been conducted by Monika Segl at the MARUM, University of Bremen.