High-resolution records of alkenone-derived sea surface temperatures and elemental Ti/Ca ratios from a sediment core retrieved off northeastern Brazil (4°S) reveal short-term climate variability throughout the past 63,000 a. Large pulses of terrigenous sediment discharge, caused by increased precipitation in the Brazilian hinterland, coincide with Heinrich events and the Younger Dryas period. Terrigenous input maxima related to Heinrich events H6–H2 are characterized by rapid cooling of surface water ranging between 0.5° and 2°C. This signature is consistent with a climate model experiment where a reduction of the Atlantic meridional overturning circulation (AMOC) and related North Atlantic cooling causes intensification of NE trade winds and a southward movement of the Intertropical Convergence Zone, resulting in enhanced precipitation off northeastern Brazil. During deglaciation the surface temperature evolution at the core site predominantly followed the Antarctic warming trend, including a cooling, prior to the Younger Dryas period. An abrupt temperature rise preceding the onset of the Bølling/Allerød transition agrees with model experiments suggesting a Southern Hemisphere origin for the abrupt resumption of the AMOC during deglaciation caused by Southern Ocean warming and associated with northward flow anomalies of the South Atlantic western boundary current.
 Asynchronous climate variability between northern and southern high latitudes of the Atlantic is documented in various paleoclimatic records, especially for the last deglaciation [Kanfoush et al., 2000; Shemesh et al., 2002; Sachs and Anderson, 2005]. Freshwater perturbation experiments with climate models suggest that decreased North Atlantic Deep Water formation and associated reduction in cross-equatorial northward Atlantic heat transport are the main drivers of this interhemispheric climate coupling [Crowley, 1992; Stocker, 1998; Ganopolski and Rahmstorf, 2001; Prange et al., 2004; Knutti et al., 2004]. An asymmetric temporal behavior between the hemispheres is also found for the onset of the large-scale ocean circulation during deglaciation. One hypothesis is that a gradual warming in the Southern Ocean induces an abrupt resumption of the thermohaline circulation with high temperatures in the North Atlantic, in agreement with marine and ice core records [Knorr and Lohmann, 2003]. For the tropical Atlantic Ocean, there is equivocality about the evolution of surface temperatures associated with rapid and large millennial-scale reorganizations of the thermohaline circulation during the last glacial period. Sea surface temperature (SST) records from the north Brazilian continental margin [Arz et al., 1999; Weldeab et al., 2006], the Tobago Basin [Rühlemann et al., 1999; Hüls and Zahn, 2000], the western Caribbean Sea [Schmidt et al., 2004], and the Gulf of Mexico [Flower et al., 2004] suggest warming of the western tropical Atlantic and the Caribbean Sea during Heinrich event H1 and the Younger Dryas although significant differences exist with respect to timing and magnitude of the SST changes. In contrast, SST reconstructions from the Cariaco Basin indicate deglacial temperature variability in synchrony with the northern high latitudes [Lea et al., 2003]. These various tropical paleotemperature records are all limited to the last glacial-interglacial transition, when the large Northern Hemisphere ice sheets vanished and climate boundary conditions rapidly and drastically changed. For MIS 3, with its relatively stable background climate, information about millennial-scale ocean temperature variability in the tropical Atlantic is much rarer but ambiguous as well. On one hand, for a Ceará Rise sediment core, Curry and Oppo  inferred cooling of the western tropical Atlantic during periods of reduced overturning circulation from analysis of stable oxygen and carbon isotopes, determined on co-occurring planktic and benthic foraminifera. On the other hand, Hüls and Zahn  derived SSTs from planktic foraminiferal transfer functions and suggested warm anomalies during several Heinrich events for a high-resolution sediment core from the Tobago Basin.
 To further assess the coupling of millennial-scale climate variability between the northern North Atlantic and the western tropical Atlantic during the last glacial, we studied a sediment core spanning the past 63,000 a, which was retrieved from the upper continental slope off northeastern Brazil (4°S). The core site is directly influenced by the North Brazil Current (NBC), which transports warm and salty upper layer waters from the warm-water reservoir of the western tropical Atlantic northward. We reconstructed sea surface temperatures from alkenone unsaturation ratios and used Ti/Ca ratios to infer past changes in precipitation and continental runoff. Experiments with a coupled atmosphere-ocean model were performed in order to study the change in tropical Atlantic SSTs and atmospheric circulation associated with weakened Atlantic overturning circulation. Sediment data and modeling results consistently indicate abrupt cooling and large southward shifts of the Intertropical Convergence Zone (ITCZ) off northeast Brazil during Heinrich events, when the meridional overturning circulation was strongly reduced.
2. Study Area
 Sediment core GeoB 3910-2 was recovered from the continental slope about 100 km off northeastern Brazil (4°14.7′S, 36°20.7′W) from a water depth of 2362 m (Figure 1). Modern surface circulation in the western tropical Atlantic is dominated by the northwestward flowing NBC, which originates from the northern branch of the South Equatorial Current. Volume transport and velocity of the NBC depend on the seasonal variations in trade wind intensity with maximum transport during the austral winter [Hastenrath and Merle, 1987]. Sea surface temperatures at the core site range between 28.5°C in February and 26.5°C in August (annual mean is 27.3°C) [Levitus et al., 1994]. At present, there is little riverine input because of the relatively dry climate of the Brazilian hinterland. The rainy season is related to the southward migration of the ITCZ during northern winter, when NE trade wind intensity increases. Anomalously dry periods coincide with negative SST anomalies south of and positive anomalies north of the equator [Ratisbona, 1976; Hastenrath and Heller, 1977]. This meridional SST anomaly pattern strengthens SE trade winds, displacing the ITCZ to the north.
3. Material and Methods
 The stratigraphy of sediment core GeoB 3910-2 is based on 20 accelerator mass spectrometry (AMS) 14C age control points of mainly monospecific samples of Globigerinoides sacculifer (Table 1 and Figure 2). Radiocarbon ages were uniformly corrected for a reservoir age of 400 a [Bard, 1988] and calibrated to calendar years with the program CalPal using the CalPal2004_Jan calibration curve that is mainly based on various radiocarbon data sets related to the Greenland Ice Sheet Project 2 (GISP2) (http://www.calpal.de/calpal/files/CalCurveComparisons.pdf). Maxima in Ti/Ca ratios almost perfectly correlate with the GISP2 ice δ18O minima in the radiocarbon-dated section of GeoB 3910-2 and were thus used to adjust the older part (47–63 calendar years) of the sediment core to the GISP2 timescale. Further confirmation for our age model comes from the U/Th-dated growth intervals of Brazilian speleothems and travertines [Wang et al., 2004]. These growth intervals, which coincide with the North Atlantic Heinrich events, indicate periods of high rainfall over northeast Brazil and correspond to the phases of increased terrigenous supply (high Ti/Ca ratios) to core site GeoB 3910 (Figure 3). The best estimate for the age uncertainty of the older interval of core GeoB 3910-2 is on the order of 1–2 ka. The youngest age determined for the end of the wet phase related to Heinrich event H6, for example, is 59.6 (±0.5) ka in the U/Th-dated speleothem record TBV63 of Wang et al. , while the assigned age for the associated decline in Ti/Ca in core GeoB 3910-2 is 58.0 ka (Table 1). Sedimentation rates of core GeoB 3910-2 range between 4 and 43 cm/ka, with an average of 8 cm/ka (Figure 2).
Table 1. AMS 14C Ages and Correlation Points to the GISP2 Oxygen Isotope Record Used as Age Control Points for the GeoB 3910 Stratigraphya
Core Depth, cm
14C AMS Age, a B.P.
Calibrated Age, a B.P.
AMS, accelerator mass spectrometry. Reservoir age of 400 a removed from original radiocarbon ages before calibration.
 Samples of 5 g freeze-dried and homogenized sediment were mixed with an internal standard and ultrasonically extracted for 3 min (UP200H ultrasonication disrupter probe; S3 micropoint, amplitude 0.5, pulse 0.5), using successively less polar mixtures of methanol and methylene chloride (CH3OH, CH3OH/CH2Cl2 1:1, and CH2Cl2). After centrifuging, the supernatants were combined, desalted with deionized water, dried with Na2SO4, and rotary evaporated to complete dryness. The residues were dissolved in CH2Cl2 and were additionally purified using a silica cartridge (Varian Bond Elut; 1 cm3/100 mg). To eliminate interference with wax esters, the clean extracts were hydrolyzed with 0.1 N KOH in methanol (90/10 CH3OH/H2O) at 80°C for 2 h, and the neutral fraction containing the alkenones was obtained by partitioning into hexane. Finally, the extracts were concentrated under N2 and taken up in 25 μL of the 1:1 CH3OH/CH2Cl2 mixture. Gas chromatography was performed using a HP5890 series II gas chromatograph equipped with a split/splitless injector, a 60 m × 0.32 mm × 0.1 μm nonpolar fused silica capillary column DB-5MS and flame ionization detector. An aliquot of 3 μL was injected in split mode (1/10), with helium as the carrier gas. The oven temperature was programmed from 50° to 250°C at 25°/min, from 250° to 290°C at 1°/min, and a final heating from 290° to 310°C at 30°/min. Compounds were quantified using octacosane acid methyl ester as an internal standard and the relative response of the C38n-alkane. The ketone unsaturation index U37K′ was converted to temperature according to the calibration of Prahl et al. :
The analytical precision (±1σ), which was based on duplicates and multiple extractions of a sediment sample used as a laboratory internal reference sample, was better than 0.003 U37K′ units or 0.1°C.
3.3. X-Ray Fluorescence Spectrometry
 Bulk sediment chemistry was determined at intervals of 0.4 cm (average time resolution of 45 a) employing nondestructive, profiling X-ray fluorescence (XRF) spectrometry. The measurements were made on the CORTEX scanner at the Bremen Integrated Ocean Drilling Program (IODP) core repository [Röhl and Abrams, 2000]. This automated scanning method allows for a rapid qualitative determination of the geochemical composition of the sediment in very high resolution [Jansen et al., 1992]. Ratios of Ti and Ca XRF intensities were used to quantify the terrigenous sediment components in core GeoB 3910-2.
3.4. Model Setup and Experimental Design
 To examine the thermal response of the tropical Atlantic and changes in the evaporation-precipitation balance during reduced ocean circulation, we carried out a meltwater perturbation experiment with a coupled atmosphere-ocean model consisting of the atmospheric circulation model ECHAM3 [Roeckner et al., 1992], the ocean primitive large-scale geostrophic ocean circulation model [Maier-Reimer et al., 1993], and the runoff model of Sausen et al. . In the meltwater experiment we introduced a large freshwater anomaly in the northern North Atlantic. The model was integrated over 250 a, with a linearly increasing amount of freshwater ranging from 0 to 0.625 Sv (1 Sv = 106 m3 s−1). After 250 a, freshwater forcing linearly decreased to zero until year 500. The meltwater was released in equal parts to two grid points on the Canadian coast of the Labrador Sea (for a detailed experimental description, see Schiller et al.  and Lohmann ). For all types of analyses the model output of years 201–250 are taken for the control integration and the freshwater experiment.
 Six outstanding maxima in Ti/Ca ratios of tropical Atlantic sediment core GeoB 3910-2 coincide with Heinrich events H1 to H6 as seen in many high-resolution North Atlantic sediment cores (summarized by Hemming ) and are consistent with the pattern of climate fluctuations documented from the GISP2 ice core (Figure 3). Two minor peaks are related to Heinrich event H5a [Rashid et al., 2003] and the Younger Dryas period. Several other small peaks in between the large pulses of titanium during MIS 3 appear to be related to the cold stadials between Dansgaard-Oeschger events although the radiocarbon-based stratigraphy for ages >25 ka is not sufficiently precise for an unequivocal attribution of these small-scale variations.
 The sea surface temperature record of core GeoB 3910-2 is characterized by several rapid temperature shifts ranging between 25° and 27°C during the last glacial (Figure 3). The largest cooling events (1°–2°C) coincide with Ti/Ca maxima co-occurring with Heinrich events H6, H4, and H2 whereas H5, H5a, and H3 are associated with minor temperature decreases between 0.5° and 1°C. Heinrich event H1 shows internal short-term variability but no clear trend of cooling or warming. Two exceptional cooling events in core GeoB 3910-2 at 26.5 ka (1.2°C) and 19.5 ka (0.6°C), prior to H2 and H1, respectively, are not related to increased Ti/Ca values and do not have equivalents in the GISP2 or Byrd oxygen isotope record. Additional small-scale temperature shifts, most distinct between 45 and 26 ka, co-occur with small Ti/Ca peaks and may be related to the North Atlantic Dansgaard-Oeschger events. The abrupt rise in temperature of 1.2°C at 15.5 ka B.P. precedes the onset of the Bølling/Allerød (BA) warm period in the Northern Hemisphere by about 1 ka. After the initial warming, ocean surface waters off NE Brazil slightly cooled during the BA until the onset of the Younger Dryas, when temperatures increased by about 0.5°C. The early Holocene is characterized by a steady temperature increase of about 1°C followed by a short SST drop of 0.7°C around 6 ka and a slight cooling trend after the SST maximum at 5 ka. Late Holocene SSTs of about 27.2°C almost perfectly match present-day annual mean values of 27.3°C off northeastern Brazil [Levitus et al., 1994]. The temperature difference of 2°C between late Holocene and Last Glacial Maximum (LGM) (ΔSSTHOL-LGM) values is 1°C lower than determined by a recent Mg/Ca-SST reconstruction on the planktic foraminifer Globigerinoides ruber (white) from a nearby sediment core [Weldeab et al., 2006]. This slight discrepancy may be related to the shallower living depth of the alkenone producers as compared to G. ruber. In general, however, our study confirms earlier results obtained from alkenone and Mg/Ca temperature records from the western tropical Atlantic and the Caribbean Sea [Wolff et al., 1998; Rühlemann et al., 1999; Lea et al., 2003; Schmidt et al., 2004] which place the range of ΔSSTHOL-LGM for the western Atlantic warm pool between 2° and 4°C.
 The Ti/Ca record exhibits high variability over an order of magnitude during the last glacial. Calcium mainly reflects the biogenic carbonate content whereas titanium, which is related to siliciclastic sediment components, varies directly with the terrigenous fraction of the sediment. Changes in glacial carbonate production and dissolution between 2000 and 3000 m water depth in the western tropical Atlantic were relatively minor [Rühlemann et al., 1996; Gerhardt et al., 2000]. Therefore maxima in Ti/Ca ratios characterize periods when precipitation in the Brazilian hinterland abruptly increased and erosion and fluvial transport of terrigenous matter to the shelf and continental slope were strongly enhanced [Arz et al., 1998]. Clay mineralogic and palynologic studies from nearby marine sediment core sites as well as speleothem and travertine deposits confirm the occurrence of more humid phases in NE Brazil during Heinrich events [Behling et al., 2000; Jennerjahn et al., 2004; Wang et al., 2004].
 These humid periods are evidently related to changes in sea surface temperatures off NE Brazil. A comparison of the alkenone-derived SSTs with the GISP2 and Byrd ice core δ18O records reveals that surface temperatures off NE Brazil mirror many of the North Atlantic climate events during the last glacial period, especially the Heinrich events, whereas SSTs varied in concert with the Byrd record during deglaciation after the rapid warming at 15 ka and during the early Holocene (Figure 3). Hence the North Brazil Current displays close climatic teleconnection to both high northern and southern latitudes, probably dependent on the general climatic background conditions (e.g., global temperature distribution and ice sheet size).
 Modeling experiments provide further insight into the physical conditions under which the climate of the western tropical Atlantic off NE Brazil is linked to either of the two hemispheres and can help elucidate the processes relating increased precipitation over NE Brazil to cool climate conditions in the North Atlantic. With the model experiment we investigated the climatic response of the western tropical Atlantic to a sustained freshwater input into the North Atlantic. In the model ocean the freshwater input into the Labrador Sea forces a transient shutdown of the Atlantic meridional overturning circulation (AMOC) [Lohmann, 2003]. The ceased NADW formation causes a strong reduction of the northward heat transport and associated decrease in northern North Atlantic temperatures, in accordance with the results from freshwater perturbation experiments with other climate models [e.g., Stocker et al., 1992; Manabe and Stouffer, 1997; Vellinga and Wood, 2002]. Furthermore, we see a net precipitation increase south of the equator at the core site of GeoB 3910-2 and a decrease north of the equator off Venezuela (Figure 4). Lohmann  found that along with the cooled Northern Hemisphere the Hadley cells move southward and the associated southward migration of the Intertropical Convergence Zone causes this tropical freshwater dipole. The predicted antiphase behavior in precipitation is corroborated by the comparison of Ti/Ca ratios in core GeoB 3910-2 with a high-resolution iron record from OPD core 1002 recovered in the Cariaco Basin at 12°N (Figure 1). Peterson et al.  showed that the Fe fluctuations in Cariaco Basin sediments are predominantly controlled by changes in precipitation over the Orinoco catchment area. Ti/Ca maxima off NE Brazil coincide with the most pronounced Fe minima off northern South America (Figure 5), indicating antiphased precipitation changes due to recurrent southward shifts of the tropical rain belt during Heinrich events. The agreement of proxy data and model results concerning the migration of the Hadley cell during thermohaline circulation slowdown thus provides additional evidence that the Ti/Ca maxima in core GeoB 3910-2 correspond to periods of strengthened precipitation over NE Brazil and increased freshwater discharge into the North Atlantic.
Figure 6 shows the anomalous horizontal surface velocities and sea surface temperatures during the meltwater experiment, indicating that the hydrography off NE Brazil has changed synchronously with North Atlantic temperatures during the last glacial. The temperature signal in the eastern North Atlantic is propagated southwestward to the tropical Atlantic toward the region of site GeoB 3910. The initial freshening applied to the North Atlantic creates an instantaneous change in density to which the ocean circulation has to adjust. This adjustment occurs through coastal Kelvin and Rossby wave propagation on timescales from several months to decades [Kawase, 1987; Johnson and Marshall, 2002], in conjunction with anomalous advection. The velocity anomalies in the North Atlantic (Figure 6) follow the anomalous high pressure in the North Atlantic [Lohmann, 2003, Figure 8]. A different situation occurs during the deglaciation. According to the hypothesis of Knorr and Lohmann , a gradual warming and accompanying sea ice retreat in the Southern Ocean initiates an abrupt resumption of the large-scale ocean circulation. The warming generates a negative density anomaly in the Southern Ocean, producing anomalous westward velocities between 30° and 50°S. Off the Brazilian coast the anomalous flow is along with pressure gradients to the north (nearly opposite to the anomalous flow of the South Atlantic western boundary current in Figure 6) and induces a Southern Hemisphere influence on the western tropical Atlantic Ocean. The abrupt temperature increase prior to the onset of the Bølling/Allerød in the northern North Atlantic, as seen in the alkenone record of GeoB 3910-2, is consistent with the early warming of the Southern Hemisphere and the northward transfer of the temperature signal during the restart of the AMOC (see supplementary material of Knorr and Lohmann , http://www.palmod.uni-bremen.de/∼gerrit/filme/movie2.mov).
 For Heinrich event H1 we do not observe distinct changes in alkenone temperature at site GeoB 3910 whereas farther to the north in the Caribbean Sea, H1 is characterized by clear warming in most of the SST reconstructions [Rühlemann et al., 1999; Schmidt et al., 2004; Flower et al., 2004]. If the western tropical Atlantic indeed acted as a heat reservoir during weakened overturning circulation, then a temperature response similar to that of the Caribbean Sea could also be expected for the site off NE Brazil. This apparent discrepancy may have been caused by an increase in local upwelling off NE Brazil, which would have suppressed the warming signal during H1. Wind anomalies during the freshwater experiment are to the southeast along the North Brazilian coast [Lohmann, 2003]. Such a change in wind direction may have triggered coastal upwelling. On the other hand, a Mg/Ca record measured on planktic foraminifera from a nearby sediment core shows a distinct warming of 2.5°C during H1 [Weldeab et al., 2006]. This disagreement between alkenone and Mg/Ca temperatures off NE Brazil points to postdepositional resuspension and transport of the alkenones, with bottom currents as an alternative explanation for the unexpected alkenone temperatures during H1.
 The two cooling events at 26.5 and 19.5 ka off NE Brazil (Figure 3) can be related to the “precursor events” that preceded the major sediment pulses of Heinrich events H2 and H1 in the North Atlantic [Grousset et al., 2000; Scourse et al., 2000]. While Heinrich layers H5–H1 were dominated by ice-rafted debris from the Laurentide ice sheet, the precursory iceberg discharges were derived from non-Laurentide (Fennoscandian and Icelandic) sediment sources. Grousset et al.  showed that these precursor events led the Laurentide iceberg discharges by about 1.5 ka, equal to the time span by which the two cooling events at 26.5 and 19.5 ka in core GeoB 3910-2 precede the Ti/Ca pulses and SST reductions associated with Heinrich events H2 and H1. It seems that the North Atlantic precursor events occurred only prior to Heinrich events H2 and H1 [Jullien et al., 2006], which would explain why similarly pronounced cooling events off NE Brazil are lacking prior to Heinrich events H5–H3. The amount of freshwater discharged into the North Atlantic during these precursor events is unknown. Considering the large SST reductions off NE Brazil, however, it probably had a significant effect on the intensity of the AMOC. Indeed, excess 231Pa/230Th ratios measured on two sediment cores from the subtropical North Atlantic [McManus et al., 2004] and the northeastern North Atlantic [Hall et al., 2006] indicate a distinct reduction of the AMOC between 19 and 19.5 ka that predates the major Heinrich meltwater event H1 by about 1.5 ka.
 Variations in SST and XRF intensities of Ti and Ca, as observed in a sediment core from the upper continental slope off northeastern Brazil over the past 63,000 a, imply a link between marine and continental climate signals. Periods of more humid continental conditions, as indicated by high Ti/Ca ratios, coincide with low SST off NE Brazil during the last glacial and generally correspond with North Atlantic Heinrich events. Both data and modeling results suggest a southward migration of the ITCZ and intensification of NE trade winds during the Heinrich events. Anomalously cold temperatures off NE Brazil throughout these periods of strongly reduced meridional overturning circulation were caused by southward propagation of the North Atlantic temperature signal. The comparison of the SST record of GeoB 3910-2 with ice core δ18O records from Greenland and Antarctica revealed that under full glacial and late glacial climate conditions, SSTs off NE Brazil were primarily responding to Northern Hemisphere climate dynamics. In contrast, SSTs were climatically linked to the southern high latitudes during the last deglaciation. Our data indicate that the temperature response in the western tropical Atlantic is more heterogeneous than previously thought, spatially but also temporally at a single site. Consistent with model simulations, the hydrography is affected by both Northern and Southern Hemisphere processes of millennial-scale changes in the large-scale circulation.
 We thank Stefan Mulitza for discussion and Dietmar Grotheer and Ralph Kreutz for assistance with the gas chromatography. Ulla Röhl and her team are acknowledged for help with the X-ray fluorescence spectrometry. Two anonymous reviewers gave insightful comments on the manuscript. Radiocarbon analyses were performed at the Kiel Leibniz Laboratory for Radiometric Dating and Stable Isotope Research. This work was supported through funding from the Deutsche Forschungsgemeinschaft (DFG) and the Gary Comer Science and Education Foundation. The data are available from the PANGAEA database (http://www.pangaea.de/home/cruehlemann).