An ensemble of new, high-resolution records of surface ocean hydrography from the Indian-Atlantic oceanic gateway, south of Africa, demonstrates recurrent and high-amplitude salinity oscillations in the Agulhas Leakage area during the penultimate glacial-interglacial cycle. A series of millennial-scale salinification events, indicating strengthened salt leakage into the South Atlantic, appear to correlate with abrupt changes in the North Atlantic climate and Atlantic Meridional Overturning Circulation (AMOC). This interhemispheric coupling, which plausibly involved changes in the Hadley Cell and midlatitude westerlies that impacted the interocean transport at the tip of Africa, suggests that the Agulhas Leakage acted as a source of negative buoyancy for the perturbed AMOC, possibly aiding its return to full strength. Our finding points to the Indian-to-Atlantic salt transport as a potentially important modulator of the AMOC during the abrupt climate changes of the Late Pleistocene.
Rapid changes are a prime expression of the inherently unstable Earth's climate [Broecker, 1997]. One notable example of this instability is the millennial-scale and abrupt ocean and climate variability of the Late Pleistocene [e.g., McManus et al., 1999; Pahnke et al., 2003; Jouzel et al., 2007; Martrat et al., 2007; Hodell et al., 2008; Charles et al., 2010; Barker et al., 2011] that is commonly ascribed to changes in the strength of the Atlantic Meridional Overturning Circulation (AMOC) [Broecker et al., 1990] and of its associated interhemispheric transport of oceanic heat [Trenberth and Caron, 2001]. Synchronization of ice core records from both polar regions to a common time scale has shown that these oscillations did not occur synchronously across the hemispheres; that is, the abrupt and high-amplitude temperature shifts in Greenland were out of phase with respect to the gradual changes in Antarctica [Blunier and Brook, 2001]. This “bipolar seesaw” [Broecker, 1998] is consistent with the contrasting response simulated for each hemisphere to reorganizations of the cross-equatorial heat transport dictated by AMOC strength variations [e.g., Knutti et al., 2004] and with the notion that the ocean played a pivotal role in millennial-scale climate fluctuations. Transient AMOC shifts and attendant bipolar seesaw behavior were also recurrent features of the Late Pleistocene glacial-interglacial transitions, with implications for the deglacial exchange of carbon between the deep ocean and the atmosphere [Cheng et al., 2009; Barker et al., 2011].
Millennial-scale reorganizations of the coupled ocean-atmosphere system are therefore a central component of the Late Pleistocene climate dynamics, but their underlying controls remain elusive, notably for the abrupt switches between cold (stadial) and warm (interstadial) conditions that punctuated the North Atlantic climate during the last several glacial-interglacial cycles [Martrat et al., 2007; Barker et al., 2011]. Broecker et al.  hypothesized that millennial-scale AMOC changes were modulated by salinity oscillations altering ocean water density (buoyancy) in the proximity of the North Atlantic Deep Water (NADW) formation sites [Renold et al., 2010]. Freshwater discharge from collapsing circum–North Atlantic ice sheets [Clark et al., 2002] and the Atlantic-to-Pacific moisture transport [Broecker, 1997; Leduc et al., 2007] are generally regarded as main drivers of the North Atlantic salt budget. However, mounting evidence also points to the salt transport through the Indian-Atlantic oceanic gateway (I-AOG), via the Agulhas Leakage, as a likewise instrumental controlling factor [Lutjeharms, 2006; Beal et al., 2011]. The I-AOG currently “leaks” ~5–15 sverdrup (Sv, 106 m3 s−1) of warm and saline Subtropical Indian Ocean water into the South Atlantic and is considered the dominant source of the upper branch of the AMOC, connecting the southern tip of Africa to the North Atlantic [Gordon, 1986; Lutjeharms, 2006; Beal et al., 2011].
Paleoceanographic reconstructions of the I-AOG circulation are currently limited to orbital time scales [Peeters et al., 2004; Martínez-Méndez et al., 2010; Caley et al., 2012] and lack the temporal resolution and/or the paleoceanographic detail needed to decipher the potential connection between salt oscillations in the south and millennial-scale and abrupt AMOC and climate shifts. We present new and high-resolution paleoceanographic records from the western sector of the I-AOG (Figure 1), resolving a series of abrupt and high-amplitude surface ocean hydrographic (temperature and salinity) changes that recurred on a millennial time scale in the Agulhas Leakage area. These time series, which span the penultimate glacial-interglacial cycle (265–77 ka, Marine Isotope Stages, MIS, 8–5a) at multicentennial resolution, are discussed along with previously published records to explore the role of the Agulhas Leakage in the abrupt ocean and climate changes of the Late Pleistocene.
2 Materials and Methods
Sediment cores MD96-2080 and MD02-2588 were retrieved with the RV Marion Dufresne from the Agulhas Bank and Agulhas Plateau, respectively (Figure 1). Core MD96-2080 (36°19.2′S, 19°28.2′E, 2488 m water depth) is ideally positioned to capture past changes in the Agulhas Leakage [Lutjeharms, 2006], and by focusing on penultimate climate cycle, we avoid a sedimentation gap in more recent sections of its sediment sequence [Martínez-Méndez et al., 2010]. Core MD02-2588 (41°19.9′S, 25°49.7′E, 2907 m water depth) is located close to the subtropical front (STF), which separates the warm subtropical waters of the eastward flowing Agulhas Return Current from the colder subantarctic waters to the south [Lutjeharms, 2006].
We present a paired record (N = 412) of stable oxygen isotopes (δ18O) and magnesium-to-calcium ratios (Mg/Ca) for the planktic foraminifer Globigerinoides ruber (sensu stricto, white) along core MD96-2080. These profiles are complemented with Mg/Ca data (N = 512) for the planktic foraminifer Globigerina bulloides and ice-rafted detritus (IRD) counts (N = 229) in core MD02-2588.
The marine and ice core records used in this study were synchronized to a common time scale (see supporting information), using the Antarctic ice core chronologies [Parrenin et al., 2007; Ruth et al., 2007], which are tied to the new time scale constructed for the Greenland ice cores North Greenland Ice Core Project, North Greenland Eemian Ice Drilling [Capron et al., 2010; Dahl-Jensen et al., 2013] and are consistent with the time scale of the synthetic δ18Oice record of Barker et al. .
2.2 Trace Element and Stable Oxygen Isotope Analyses in Planktic Foraminifera
Sixty-five tests per sample for G. ruber (300–355 µm size fraction) and 55 for G. bulloides (250–315 µm size fraction) were picked, crushed between methanol-cleaned glass microscope slides, and subsequently mixed to homogenize them before splitting into two aliquots used for δ18O and trace element analyses. G. ruber is closely associated with the subtropical waters conveyed by the Agulhas Current to the tip of Africa [Peeters et al., 2004], and its data reflect the influence of the Agulhas Leakage on the hydrography of the Agulhas Bank more reliably than those derived from other foraminiferal species [e.g., Martínez-Méndez et al., 2010]. G. bulloides is ubiquitous in the South Atlantic [Niebler and Gersonde, 1998] and is adapted to the strong hydrographic gradients that are associated with the regional oceanic frontal systems, such as the STF in the vicinity of the Agulhas Plateau [Lutjeharms, 2006].
Trace element analyses were carried out following standard cleaning procedures [Barker et al., 2003], complemented with the reductive step in core MD96-2080. Element ratios were determined by inductively coupled plasma-optical emission spectroscopy [de Villiers et al., 2002] and inductively coupled plasma-mass spectrometry [Yu et al., 2005] at the University of Cambridge and at the Universitat Autònoma de Barcelona. Long-term instrumental precision of element ratio data, determined by replicate analyses of standard solutions containing Mg/Ca = 1.3 mmol mol−1 (Cambridge) and Mg/Ca = 2.6 mmol mol−1 (Barcelona), was ±0.46% and ±1.30%, respectively. The accuracy of Mg/Ca determinations was confirmed by an interlaboratory study [Greaves et al., 2008]. The Mg/Ca data from I-AOG cores were tested for reproducibility by replicating ~20% of the records (see supporting information). Previously published calibrations [Mashiotta et al., 1999; Anand et al., 2003] were used to convert G. ruber and G. bulloides Mg/Ca values (Mg/CaG. ruber, Mg/CaG. bulloides) to calcification temperatures, which, due to the near-surface habitat of these species [Hemleben et al., 1989], approximate sea surface temperatures (SST) at the core sites.
G. ruber δ18O (δ18OG. ruber) was determined at Cardiff University using a ThermoFinnigan MAT 252 mass spectrometer linked online to a single acid bath CarboKiel-II carbonate preparation device. External precision was better than ±0.08‰ (1σ). Isotope values are referred to the Vienna Peedee Belemnite scale through calibration to the NBS-19 carbonate standard.
2.3 Conversion of Paired Mg/Ca-δ18O Data to Local Seawater δ18O
The seawater δ18O (δ18OSW) record from core MD96-2080 is used as a proxy of past surface ocean salinity variations at the Agulhas Bank, building on the linear relationship between δ18OSW and salinity that is observed at the regional scale in the modern ocean [LeGrande and Schmidt, 2006]. We derive paleo-δ18OSW time series from paired Mg/CaG. ruber-δ18OG. ruber data (Figure S4, see also the supporting information for details on the analysis of uncertainties). The record of calcification temperature based on Mg/CaG. ruber is used to correct δ18OG. ruber for the temperature-dependent (0.21 ± 0.01‰ °C−1) water to carbonate δ18O fractionation [Bemis et al., 1998]. Next, we subtract from the residual δ18OSW the mean ocean δ18OSW, which reflects variations in global ice volume and obtain the local δ18OSW. For this, sea level records [e.g., Rohling et al., 2009] are converted into mean ocean δ18OSW by assuming a 18O enrichment of the global ocean by 0.008‰ per meter sea level lowering [Schrag et al., 2002].
2.4 Analyses of Ice-Rafted Detritus in Core MD02-2588
Lithic fragments (IRD) were counted under a binocular microscope from the >150 µm size fraction, in order to limit interference by fine-grained terrigenous material stemming from non-ice rafting transport mechanisms (e.g., bottom currents) [Teitler et al., 2010]. Relatively large splits were counted to increase the likelihood that IRD occurring in low numbers appears in the results. The mineralogy of the IRD grains is mainly quartz, with some grains being feldspar or of mafic origin, and scanning electron microscope analysis reveals typical glacial microfeatures on the surface of the grains, such as grooves, deep troughs, linear steps, and chatter marks (Figure S5).
The δ18OG. ruber profile from the Agulhas Bank core MD96-2080 displays a clear glacial-interglacial modulation of the signal (Figure 2a), with glacial values that are heavier than interglacial ones by up to 1.4‰. Oscillations of δ18OG. ruber on a millennial time scale appear far less pronounced, generally in the order of 0.2‰ (see also Figure S6). By contrast, Mg/CaG. ruber based SST and local δ18OSW profiles from the same site exhibit a prominent millennial-scale variability with abrupt increases by 2–5°C and 0.5–1.5‰, respectively, standing out as a series of distinct peaks (Figures 2b and 2c).
Millennial-scale SST variability is only weakly expressed at the Agulhas Plateau core MD02-2588 (Figure 2d). This site is located near the southern border of the I-AOG (Figure 1) and is marginally influenced by the retroflected Agulhas waters which return to the Indian Ocean paralleling the STF [Lutjeharms, 2006; Diz et al., 2007]. Hence, the gradual and low amplitude warming shifts (slightly exceeding 1°C) that we observe are indicative of transient southward displacements of the STF. SST decreases by up to 2°C coincide with enhanced IRD deposition (Figure 2e), which testifies to episodes of pronounced northward advection of polar waters during circum-Antarctic cold phases [e.g., Pahnke et al., 2003], with the northward shift of the regional oceanic fronts supporting iceberg survivability as far north as ~41°S.
4.1 Millennial-Scale Hydrographic Variability at the Tip of Africa
Our results reveal that between MIS8 and MIS5 the surface ocean hydrography in the western sector of the I-AOG underwent a series of millennial-scale oscillations, which were particularly pronounced at the Agulhas Bank core site (Figures 2b–2e). The latitudinal migrations of the STF, inferred from the Agulhas Plateau SST and IRD records (Figures 2d and 2e), allude to changes in wind forcing associated with shifts of the Southern Hemisphere westerlies [De Deckker et al., 2012], with potential consequences for the surface circulation within the I-AOG and, in turn, for the leakage of warm and saline waters into the South Atlantic [Lutjeharms, 2006; Beal et al., 2011]. Indeed, the high-amplitude SST and δ18OSW increases observed at the Agulhas Bank (Figures 2b and 2c) testify to positive heat and salt anomalies during millennial-scale episodes of southward migration of the STF. The Agulhas Bank is located directly in the path of the rings, eddies, and filaments constituting the Agulhas Leakage [Lutjeharms, 2006]. Accordingly, the δ18OSW and SST maxima at this site can be seen as hydrographic fingerprints of the enhanced transport to the southern tip of Africa of warm and saline waters, which are sourced in the subtropical Indian Ocean.
The Agulhas Leakage feeds with heat and salt the upper branch of the AMOC that flows northward, compensating the formation and export of NADW to the Indian and Pacific Oceans [Gordon, 1986]. However, sustained heat loss to the atmosphere as the Agulhas waters exit the I-AOG causes the heat anomaly coupled with the leakage to dissipate [Lutjeharms, 2006], thereby amplifying the impact of salinity on the buoyancy of the waters advected to the South Atlantic and beyond [Beal et al., 2011]. In addition, both theoretical and numerical models hint at seawater salinity as the critical parameter that modulates millennial-scale buoyancy changes at the northern end of the AMOC's upper branch [Broecker et al., 1990; Renold et al., 2010]. Hence, in the following, we focus on the characteristics and large-scale implications of the Agulhas Bank δ18OSW oscillations. We identify 18 transient δ18OSW increases, dubbed Agulhas salt-leakage maxima (ASM), which are numerically coded on the basis of their correlation with the isotopic maxima in the Antarctic deuterium (δDice) record (Figure 2e). The most evident and long-lasting δ18OSW shifts display positive amplitudes of up to ~1.5‰ and mark the end of full glacial conditions of MIS8 and MIS6, corroborating the notion of a systematic deglacial strengthening of the Agulhas Leakage [Peeters et al., 2004; Caley et al., 2012]. However, δ18OSW fluctuations at the Agulhas Bank are not confined to glacial-interglacial transitions, large positive amplitude, but shorter-lasting events punctuate the glacial and interglacial sections alike (Figures 2 and 3). The significance of some of these δ18OSW anomalies (e.g., 1ASM23-24, 2ASM2-3) is equivocal as they remain within or barely exceed the bounds of the propagated (2σ) uncertainties, but in particular, during early MIS6 and MIS5a–b, their amplitudes are remarkably large, reaching up to two thirds of the major deglacial shifts.
4.2 Ocean-Atmosphere Interhemispheric Linkages
The evidence of transient and high-amplitude oscillations in the Indian-to-Atlantic saltwater transport raises important questions concerning their relationship with the interhemispheric climate and AMOC dynamics on a millennial time scale. To explore such links, we place our I-AOG records into a framework of abrupt ocean-atmosphere changes from the wider North Atlantic region (Figures 3a–3f). Although the interhemispheric phasing discussed below is consistent with that identified in the nearby Cape Basin by Charles et al. , the uncertainties (1σ, 0.4–1.3 kyr between 185 and 77 ka; Figure 3g) associated with our chronological framework hamper the clear identification of the timing of the ASM relative to their North Atlantic counterparts, at least for the short-lasting events. We consequently focus on the multimillennial ASM to build our interpretation and advance conceptual ideas concerning the I-AOG dynamics and its significance for interhemispheric teleconnections. Based on this data comparison, we note that the salinity buildup and peak maxima at the Agulhas Bank appear to correlate with millennial-scale cold phases and subsequent abrupt stadial-to-interstadial transitions documented in the Greenland air temperatures and in the Iberian Margin tetraunsaturated (C37:4) alkenone profile from the Iberian Margin (Figures 3a and 3b). The latter traces the advection of cold surface Arctic waters to this region [Martrat et al., 2007]. These changes reflect the response of the circum–North Atlantic to AMOC variations [Broecker et al., 1990] coupled with ice sheet instabilities and attendant freshwater perturbation [McManus et al., 1999; Oppo et al., 2006].
The prominent δ18OSW increase (1ASM21) at 87–84 ka is associated with the millennial-scale IRD pulse in the subpolar North Atlantic [Oppo et al., 2006] and with the incursion of surface Arctic waters to the Iberian Margin [Martrat et al., 2007]. This is a clear expression of the climate system's bipolar seesaw behavior, in which cooling (and freshening) in the North Atlantic coincides with warming over Antarctica [Blunier and Brook, 2001]. Other examples of ASM that correlated with abrupt swings in the north are as follows: (i) the 2ASM1 salt buildup at ~147 ka (Figure 3c); (ii) the double-peaked salinity increase between 138 and 130 ka, coinciding with Heinrich event 11 [Oppo et al., 2006]; (iii) the series of five consecutive ASM between ~158 and 181 ka (events 2ASM2–6); and (iv) the 1ASM22–25 salt events during MIS5. Events 2ASM2–6 also match the structure of the stadial-interstadial fluctuations in the Greenland δ18Oice composite, perhaps suggesting an interhemispheric coupling similar to that established above for the longer-lasting events during the penultimate deglaciation and 1ASM21.
The curtailment of NADW (or its glacial equivalent) formation in response to positive surface buoyancy forcing during episodes of freshwater perturbation in the North Atlantic [e.g., McManus et al., 1999; Clark et al., 2002] has been shown to promote a basin-scale reconfiguration of the deep branch of the AMOC [Renold et al., 2010]. These changes are recorded as negative shifts in the stable carbon isotopes (δ13C) of bottom-dwelling foraminifera (Figure 3f) at Integrated Ocean Drilling Program Site U1308 [Hodell et al., 2008]. Site U1308 is well positioned (Figure 1b) to capture the northward incursions of nutrient-laden (low δ13C) southern component waters (equivalent to modern Antarctic bottom water) into the deep North Atlantic, which is otherwise ventilated by nutrient-depleted (high-δ13C) NADW [Curry and Oppo, 2005]. Numerical simulations show that the retention of oceanic heat in the south coupled with an AMOC slowdown [e.g., Knutti et al., 2004] is compensated by the intensification of the northward heat transport via the atmosphere [Lee et al., 2011], which ultimately weakens the Hadley Cell in the southern tropics and strengthens the midlatitude westerlies that shift toward Antarctica [Toggweiler and Lea, 2010; Lee et al., 2011]. This reconfiguration of the atmospheric circulation in the Southern Hemisphere potentially “drags” the STF to a southerly position [De Deckker et al., 2012], as exemplified by the surface ocean warming at the Agulhas Plateau (Figure 3d), and widens the I-AOG [Beal et al., 2011]. The latter would have then maximized the Agulhas Leakage into the South Atlantic, salinifying the surface ocean at the Agulhas Bank (Figure 3c). The resumption of NADW formation (Figure 3f) and of its associated heat piracy [Knutti et al., 2004] at the onset of interstadials (Figure 3a) would cause this sequence of interhemispheric developments to reverse [Toggweiler and Lea, 2010; Lee et al., 2011]. Specifically, an equatorward migration of the STF over the Agulhas Plateau (Figures 3d and 3e) is expected to narrow the I-AOG and to hinder the Agulhas salt leakage to the South Atlantic, in line with the rapidly dissipating salinity anomalies at the Agulhas Bank (Figure 3c).
Our analysis suggests that a dynamic connection between hydrographic changes in the I-AOG and in the North Atlantic climate is consistent with recently proposed adjustment processes of the coupled ocean-atmosphere system to a freshwater-bound weakening of the AMOC [Knutti et al., 2004; Toggweiler and Lea, 2010; Lee et al., 2011]. The salinity buildup in the Agulhas Leakage area during North Atlantic stadials, possibly superimposed upon an underlying salinification of the wider South Atlantic linked to AMOC weakening [Lohmann, 2003], may have had important implications for the buoyancy budget of the Atlantic Ocean and, in turn, for the AMOC [Beal et al., 2011]. It has been proposed that the salt sourced in the I-AOG and advected to the NADW formation sites has the potential to kick-start the AMOC [Knorr and Lohmann, 2003]. This raises the possibility that the Agulhas salt-leakage events that are documented in our records do not simply reflect a passive response of the I-AOG to the interhemispheric bipolar seesaw but either actively triggered the AMOC resumption or (at least) modulated its stability characteristics upon resumption [Knorr and Lohmann, 2007].
5 Concluding Remarks
Results presented in this study indicate that transient and high-amplitude oscillations of the Agulhas salt leakage were a persistent feature of the I-AOG circulation in the past and were possibly dynamically linked with abrupt reorganizations of AMOC and interhemispheric climate. It emerges that the salt-leakage events at the southern tip of Africa occurred when the North Atlantic was cold and fresh (stadials) and the AMOC was weak and that the salt leakage maximized during the abrupt (stadial/interstadial) resumptions of the AMOC. This relationship involves the impact of changing AMOC strength and its coupled interhemispheric climate adjustment of the Southern Hemisphere westerlies, which plausibly altered the I-AOG circulation and impacted the Agulhas Leakage. Coupled climate models are needed to determine whether the salt leakage actively triggered the recovery of the AMOC from a perturbed state or it played a more passive role by contributing to the strengthening of the AMOC that was triggered elsewhere. The correlation of salt oscillations in the I-AOG and abrupt climate changes in the North Atlantic, however, poses a challenge to the canonical view ascribing abrupt climate swings exclusively to the processes taking place in the northern North Atlantic and/or in the Northern Tropics.
L. Rodríguez-Sanz, M. Greaves, G. Mortyn, and C. Lear provided analytical advice. I. Villarroya, C. Cantero, and J. Becker assisted during geochemical analyses. G. Martínez-Méndez, F. Peeters, J.R. Toggweiler, and C. Charles (Editor) are acknowledged for their constructive comments that helped improve the manuscript. We thank L. Rodríguez-Sanz for her advice on the statistical analysis of the data. Financial support was provided by the Universitat Autònoma de Barcelona (grant PS-688-01/08) to G.M., by the European Commission Seventh Framework Programme (EC FP7) collaborative project “Past4Future” (G.M., P.Z., R.Z.), by the EC FP7 Initial Training Network “GATEWAYS” (C.P.P., I.H., G.K., G.M., M.Z., P.Z., R.Z.), by the Natural Environment Research Council, UK (I.H., H.E.), by the European Research Council (grant 2010-NEWLOG ADG-267931, H.E.), and by the Helmholtz Climate Initiative “Regional Climate Change” (REKLIM, G.K.).