Parallelisms between sea surface temperature changes in the western tropical Atlantic (Guiana basin) and high latitude climate signals over the last 140 000 years

Sea surface temperatures (SST) in the Guiana basin over the last 140 ka were obtained by measuring the C37 alkenone unsaturation index (7 N, 53 W). The resulting dataset is unique for this period in the western tropical Atlantic region. SSTs range from 25.1 to 28.9 ◦ C, i.e. glacial-to-interglacial amplitude of 5 3.8 ◦ C, which is common in tropical areas. During the the studied trace rapid transmission of the climate variability from arctic-to-tropical vice-versa. During these showed a remarkable parallelism with temperature changes observed in Greenland 10 and SST records of North Atlantic cores. The last deglaciation in Guiana is particularly revealing. MIS2 stands out as the coldest period of the interval analysed, with SSTs reaching as low as 25.1 ◦ C. It contains reminders of northern latitude events such as the Bølling-Allerød warming and the Younger Dryas cooling which ensued. These oscillations were previously documented 15 in the δ 18 O of the Sajama tropical ice core and are present in Guiana with rates of ca. 6 ◦ C ka − 1 and changes of over 2 ◦ C. During

3.8 • C, which is common in tropical areas.
During the last two interglacials (MIS1 and MIS5e) and warm long interstadials (MIS5d-a), the sediments studied trace rapid transmission of the climate variability from arctic-to-tropical latitudes and vice-versa. During these periods, MD03-2616 SSTs showed a remarkable parallelism with temperature changes observed in Greenland 10 and SST records of North Atlantic cores.
The last deglaciation in Guiana is particularly revealing. MIS2 stands out as the coldest period of the interval analysed, with SSTs reaching as low as 25.1 • C. It contains reminders of northern latitude events such as the Bølling-Allerød warming and the Younger Dryas cooling which ensued. These oscillations were previously documented 15 in the δ 18 O of the Sajama tropical ice core and are present in Guiana with rates of ca. 6 • C ka −1 and changes of over 2 • C. During the glacial interval, significant abrupt variability is observed; e.g. oscillations of 0. 5-1.2 • C during MIS3, i.e. about 30 % of the maximum glacial-interglacial SST change. Nevertheless, in the MD03-2616 record it is hard to identify unambiguously ei- 20 ther the Dansgaard-Oeschger type of oscillations described in northern latitudes or the SST drops associated with the Heinrich events characterising North Atlantic records. Although these specific events form the background of the climate variability observed, what truly shapes SSTs in Guiana is a long-term tropical response to precessional changes, which is modulated in the opposite way to polar variability. This lack of syn-

Introduction
Abrupt climate changes have been recorded in a variety of environmental sensors and archives. Examples of these are: (i) isotopic composition of foraminifera (Bond et al., 1993;Lisiecki et al., 2005;McManus et al., 1994;Peterson et al., 2000;Shackleton et al., 2000) and sea surface temperatures (SST) derived from alkenones (Herbert and 5 Schuffert, 2000;Martrat et al., 2004Martrat et al., , 2014 or Mg/Ca measured in marine sediments (Cacho et al., 2006;Marino et al., 2013;Martinez-Mendez et al., 2010), (ii) isotopic composition of speleothems (Cheng et al., 2009;Wang et al., 2001), (iii) isotopes and greenhouse gases trapped in continental polar and tropical ice (EPICA, 2004;North Greenland Ice Core Project members, 2004;Jouzel et al., 2007;Loulerge et al., 2008;10 Wolff et al., 2010;Thompson et al., 1998). Diverse locations across both hemispheres, from Greenland and the North Atlantic through South America and Antarctica, among others, record these abrupt climate changes. Hence, there is currently little doubt that the Atlantic has experienced changes from warm to cold conditions and vice-versa in sub-millennial time-scale events (Barker et al., 2011) which punctuated the orbital- 15 driven glacial-to-interglacial evolution (Berger, 1978;Jouzel et al., 2007). Variability in the Atlantic Meridional Overturning Circulation (AMOC) is detected as one of the primary causes of these abrupt climate variations (Ganopolski and Rahmstorf, 2001;Gherardi et al., 2009;Hendy et al., 2002). In the past, AMOC reductions brought about a decrease in the transfer of heat to northern latitudes, with climate changes occurring at polar latitudes, whether northern or southern. Most of the evidence to explain Atlantic climate processes has been obtained from sediment cores located at mid-to-high latitudes (Allen et al., 1999;Bond et al., 1993;Martrat et al., 2007;McManus et al., 2004;Shackleton et al., 2000). However, several aspects of these changes are as yet to be explored, among them, the occurrence of abrupt cli-10 mate transitions in tropical latitudes (Seager and Battisti, 2007). Climate changes in tropical Atlantic regions have received less attention, particularly those encompassing variability beyond the last deglaciation, e.g. Dubois et al. (2014), Herbert and Schuffert (2000), Jaeschke et al. (2007), Keigwin and Boyle (1999), Schmidt et al. (2004). Relevant topics for consideration are (i) identification of climate processes leading to rapid 15 changes, (ii) detection of feedback mechanisms involved in the polar-to-tropical transmission of high latitude abrupt climate signals without loss of intensity, (iii) evaluation of the potential role of atmospheric/oceanic reorganisations in sustaining this variability. These questions require investigation into whether the tropics, as the main global store of heat and salt, may have also played an active role in abrupt climate change 20 development or at least contributed to several stages of the process. This justifies wider exploration of the less studied tropical areas, going into greater detail as to the origin and geographic impact of rapid climate variability (Broecker, 2006).
The Guiana basin belongs to a tropical region confined between Arctic and Antarctic oceanographic influence. Its hydrography is modulated by the water discharges of high-latitude surface waters and the North Atlantic climate because of their influence on thermohaline circulation (Schmidt et al., 2004;Ritz et al., 2013).
Alkenones synthesized by haptophyte algae have been very successful for SST monitoring, particularly in the Atlantic Ocean (Jaeschke et al., 2007;Martrat et al., 2007;Müller et al., 1998). The alkenone unsaturation index U k 37 is used here to estimate SSTs 10 (Brassell et al., 1986;Müller et al., 1998) during the past 140 ka in the western tropical Atlantic. These SST variations trace abrupt climate events and may help to identify connections with northern or southern Atlantic processes and evaluate the sensitivity of tropical areas to the changes occurring at high latitudes. 15 Core MD03-2616 was recovered in Guiana basin (7.4875 • N, 53.0080 • W by about 650 km off the coast, at 1233 m below sea level) during the PICASSO cruise on board the R/V Marion Dufresne (Fig. 1). The core has a total length of 39 m. Most of the sediment was formed by olive green clay, rich in foraminifera and organic matter with little bioturbation (Shipboard Scientific Party, 2003).

Atmospheric circulation
The Guiana region is situated north of South America (Fig. 1)  summer -early autumn and early spring). This spatial and seasonal variability in the ascending branch of the Hadley cell has an impact on the vegetation and hydrology of the area, involving maximum runoff when the ITCZ is over the basins of the Amazon, Orinoco, Maroni and Oyapock rivers (Masson and Delecluse, 2001;Muller-Karger et al., 1989). Trade winds predominate in the region and change their direction de-5 pending on the ITCZ position ( Fig. 1). South-east trade winds prevail when the ITCZ is in its northern position (drier continental climate; short rainfalls in Guiana). There is a predominant opposite flow of north-east trade winds when the ITCZ is in its southern position (wetter oceanic climate; long rainfalls in Guiana). 10 According to the Levitus database, the present average annual at the MD03-2616 location is 27.6 • C (Reynolds et al., 2002). The GC washes the coastline from south-east to north-west ( Fig. 1) and pushes the Amazon river plume towards the Caribbean Sea (Masson and Delecluse, 2001;Muller-Karger et al., 1988, 1995. This main current extends from the North Brazil Current (NBC), which branches off from the South Equato- 15 rial Current (SEC). The NBC provides salty, warm waters to the western tropical Atlantic north of the Equator (Stramma and Schott, 1999). The NBC is also influenced by the ITCZ. When the convergence zone is in its northern position it undergoes a retroflection, generating the North Equatorial Counter Current (NECC) and decreasing the GC flow. The formation and strengthening of the NBC diverts part of the Amazon plume 20 sediment from the Caribbean Sea towards the Central Atlantic (Rühlemann et al., 2001;Zabel et al., 2003), thereby decreasing the sediment supply to the area in which core MD03-2616 is located, i.e. divergence area between the NECC and GC (Fig. 1)

River's run off
The Amazon is the main river in South America. Its annual mean flow of 200 000 m 3 s −1 contributes with 6 × 10 12 m 3 yr −1 of fresh water to the tropical Atlantic (Muller-Karger et al., 1988). Guiana's rivers (Maroni and Oyapock) have much lower runoff, 1600 and 800 m 3 s −1 , respectively, and lower influence in the area (Masson and Delecluse, . The Amazon River plume is rich in nutrients and suspended sediments and forms coastal mud banks. These mud banks are associated with salinity variations and have an effect on the development of coastal ecosystems such as mangroves (Lambs et al., 2007). The material accumulated in the continental shelf is transported to the Guiana basin by the GC in a continuous band of 100-150 km. The GC carries much of 10 the Amazon river plume northward to the Caribbean Sea (Muller-Karger et al., 1995). These river waters are rich in sediments and organic compounds generated in the Amazon forests (Saliot et al., 2001). The river is also a major contributor of nutrients to the marine system which provide appropriate habitats for algal growth, including haptophyte algae (López-Otálvaro et al., 2009).

Lipids and SSTs
Sediment samples (2.5 g) from MD03-2616 were taken every 3 cm. The procedure for analysis of organic compounds, including C37 alkenones, has been previously described (Villanueva and Grimalt, 1997). Briefly, samples were freeze-dried and n- 20 nonadecan-1-ol, n-hexatriacontane and n-dotetracontane were added as internal standards. The sediments were then extracted with dichloromethane in an ultrasonic bath. The extracts were saponified with 10 % potassium hydroxide in methanol to eliminate interfering compounds such as fatty acids, ester waxes, aminoacids and proteins. The neutral lipid phase was recovered from this alkaline digestion with hexane, which was Introduction

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | evaporated to near dryness under a gentle nitrogen stream. The lipid mixture was redissolved with toluene, derivatized with bis(trimethylsilyl) trifluoroacetamide and analyzed by gas chromatography and a flame ionisation detector (GC-FID). Instrumental analyses were performed with a Varian 3400 equipped with a CPSIL-5 CB column coated with 100 % dimethylsiloxane (film thickness of 0.12 µm). Hydro-5 gen was the carrier gas (50 cm s −1 ). The oven was programmed from 90 to 170 • C at 20 • C min −1 , then to 280 • C at 6 • C min −1 (holding time 35 min), to 300 • C at 10 • C min −1 (holding time 7 min) and finally to 320 • C at 10 • C min −1 (holding time 3 min). The injector was programmed from 90 • C (holding time 0.3 s) to 320 detector was maintained at 320 • C. 10 Selected samples were analysed by gas chromatography coupled with mass spectrometry (GC-MS; Thermo DSQ II Instruments). The instrument was equipped with a CPSIL-5 CB column and helium was used as carrier gas. Injection conditions were as described above for GC-FID. Mass spectra were acquired in the electron impact mode (70 eV) scanning from 50 to 700 mass units in cycles of 1 s. 15 The U k 37 index was obtained from the concentrations of C 37 alkenones, C 37:2 /(C 37:2 + C 37:3 ) (C 37:2 refers to heptatriaconta-8E ,15E ,22E -trie2-one and C 37:3 to heptatriaconta-15E ,22E -die2-one), and used to calculate the SST (U k 37 = 0.033×SST+ 0.044; Müller et al., 1998). 20 Greenland climatic variability (Fig. 2a) and precessional oscillations (Fig. 2c) are shown as a reference for presenting the chronology applied to the MD03-2616 SSTs (Fig. 2b). From 5.9 to 34.5 ka, the MD03-2616 age-model is based on 6 AMS-14 C-dates measured in shelves of planktonic foraminifera Globigerinoides sacculifer, calibrated using the Marine13 curve (Reimer et al., 2013;  MD03-2616 SSTs display a well-defined orbital modulation of glacial and interglacial reference marine isotope stages (MIS): the last interglacial complex MIS5e-a (from 127.3 to 71.6 ka BP), glacial stages from MIS4 to MIS2 (from 71.6 ka to 11.5 BP) and the present interglacial or MIS1 (from 11.5 to 0 ka BP). Sedimentation rates over time in Guiana (Fig. 2f) are supposed to be influenced by the sediment output of the Amazon 10 river. The chronology used suggests that during MIS4 and MIS3 much of the sediment discharged remained in the Amazon fan and the sedimentary particle flow arriving in the Guiana basin was relatively small, average of 8 cm ka −1 . Apparently, sedimentation rates during the last interglacial complex (MIS5d-a) and deglacial events (late MIS2) showed higher values ranging from 4 to 30 cm 2 ka −1 .

SST glacial/interglacial patterns
Alkenone-derived SSTs range from a minimum of 25.1 • C during MIS2 to a maximum of 28.9 • C in MIS5e (Fig. 2b). SST glacial-to-interglacial amplitude may appear subtle (3.8 Trends between perihelion passage in the NH summer and winter solstices are shown as follows ( Fig. 2a and b): oscillations at warming in red (in Guiana, MIS3 and late MIS2; in Greenland, MIS5b and late MIS2), changes at cooling in blue (in Guiana, MIS 5e-a, MIS4 and early MIS2; in Greenland, MIS5d-c, and from MIS5a to MIS3) and, finally, trends of less than 1 % of the maximum change in yellow (in Greenland, early 10 MIS2). During the last two interglacials (MIS1 and MIS5e) and warm long interstadials (MIS5d-a), both Greenland and Guiana experienced parallel cooling trends ( Fig. 2a and b). SSTs around MIS5b (82.7 ka) present inverse long-term trends between Greenland and Guiana: polar latitudes showed a warming while the tropical site experienced a cooling. This is interesting, given that MIS5b comprises one of the most extreme 15 events recorded in southern European pollen sequences (Tzedakis et al., 2003), in line with the maximum extension of the Barents-Kara and Scandinavian ice sheets (Ehlers et al., 2011) which blocked drainage of north-east European rivers owing to large proglacial lakes (Krinner et al., 2004). Additionally, note the opposite trends between Guiana and Greenland during MIS3 and early MIS2 (Table 2 includes rates of 20 Change and number of samples used).  (Bard et al., 2000;Martrat et al., 2007) or −6 • C in the Alboran Sea (Martrat et al., 2004;Cacho et al., 1999). Lesser amplitudes in MD03-2616 are consistent with 5 the narrower SST range found in tropical regions. Previous studies identified abrupt changes based on the fastest rate of change associated with the last deglaciation (Martrat et al., 2004;Rahmstorf, 2003). In Guiana, this interval presents a rate of change of +2 • C ka −1 (3.1 • C in 1550 years in the MD03-2616 record; Fig. 3b; Table 3). Thus, in this study, events with a warming/cooling speed higher than ±0.5 and ±2 • C ka −1 were 10 considered abrupt. Most events were found in the glacial period when instability was higher (Fig. 3b). Some relevant SST oscillations are detected at transitional phases such as MIS5d (+2.0  Table 3). The intra-MIS5e variability previously reported in the North Atlantic (Oppo et al., 2001(Oppo et al., , 2006 is also observed in Guiana (Fig. 4b). From MIS5c to MIS5a (GS and GI from 25 to 19), SSTs followed a pace of events similar to those of Greenland. Generally, SST oscillations did not exceed 0.5 • C, though some remarkable exceptions are 20 observed around GS-24 (cold event C23 in McManus et al., 1994, GS-22 (cold event C21 in McManus et al., 1994 and GS-25 (cold event C24 in McManus et al., 1994. Transitions from MIS5a to MIS4 and from MIS3 to MIS2 were abrupt (e.g. cooling of −1.5 • C in 0.4 ka; Table 3) and presented high instability, i.e. warming and cooling events occurred rapidly (in less than 1.5 ka). The MIS3 transition started 25 with a rapid warming at 57 ka (+1.4 • C in 0.6 ka) and exhibited high Variability (Fig. 3b). Late MIS2 presents a warming trend ( Fig. 2b; respond to the stadials associated with Heinrich event 1 (H1) and the YD respectively, as described at higher latitudes (Fig. 3b).

Rapid tropical-pole connections during warm, stable periods
The fact that SSTs in MIS5e (28.9 • C) are higher than in the MIS1 (28.3 • C) further  Table 2) has been attributed to a strengthening in thermohaline circulation (McManus et al., 2002). High latitude climate changes in the Northern Hemisphere were transmitted towards the southern Atlantic Ocean even at latitudes of 7 • N during periods when precessional changes increased in amplitude. When the transport of salt from the Caribbean Sea to North Atlantic latitudes was strong and the AMOC was 15 active, close connection between high latitudes and tropical regions was enabled.
During the latest stages of the last deglaciation, MD03-2616 SSTs are a reminder of oscillations observed in Greenland (North Greenland Ice Core Project members, 2004), with structures similar to the B-A and the stadial associated with H1 respectively ( Fig. 3a and b). Once again, these SST changes were of lesser intensity in the Introduction Possible links between the MD03-2616 SST record and those from the cores in the Agulhas area should be considered, given that the Guiana core is located in the area of influence of the NBC originating from the SEC and providing salty warm waters to the western tropical Atlantic north of the equator ( Fig. 1; Stramma and Schott, 1999). The SEC is ultimately fed by leakage from the Agulhas current (Bard and Rickaby, Conversely, the coupling between SST change in MD03-2616, the Greenland temperatures and the SST of northern Atlantic latitudes in the interglacials is consistent with the model describing an AMOC dependence on global mean air temperature 15 anomalies and North Atlantic SSTs (Ritz et al., 2013). Analogous SST evolution between tropical areas and Greenland suggests that ocean processes in Guiana are directly related to the AMOC strength during the last two interglacials (MIS5e and MIS1) and warm long interstadials (MIS5d-a). This parallel behaviour is in line with the amplification of thermohaline circulation resulting from the movement of salty trop-20 ical waters into the North Atlantic, as observed in cores from the Caribbean Sea (12 • N, 78 • W; Schmidt et al., 2004). The coupling of the west tropical Atlantic waters with these processes was probably necessary for the supply of salty waters to the Caribbean Sea prior to concentration and advection towards the North Atlantic. The coupling is observed irrespectively of the higher amounts of sediment from the Amazon CPD 11,2015 Parallelisms between sea surface temperature changes in the western tropical Atlantic

Tropical abrupt SST changes during transitional intervals
While, in the North Atlantic, abrupt changes occurred throughout MIS3 (Martrat et al., 2004), in MD03-2616 they are mostly to be found at the end of this stage. Hence, most abrupt changes occur during deglaciation periods ( Fig. 3b; Table 3). This pattern is somewhat consistent with the events described above. When the AMOC is active, the 5 climate also undergoes abrupt variability. In this respect, MD03-2616 exhibits abrupt oscillations around the B-A. This feature has also been observed in the δ 18 O ice record of continental ice accumulated in Sajama (Bolivia; Thompson et al., 1998), which reinforces the evidence of links between the climate changes in the North Atlantic and in central and south America during the end of the last deglaciation ( Fig. 3a and b). 10 A strong SST variability in the YD has been identified in core MD03-2616. Bearing in mind that the YD most likely resulted from the massive discharge of cold freshwater into the North Atlantic, causing a decrease in the AMOC (Broecker and Hemming, 2001;Teller et al., 2002), it is feasible that such huge freshwater inputs could modify oceanic circulation in the tropical Atlantic. The influence of these northern waters may have had 15 an effect on latitudinal displacements of the ITCZ which may have also resulted in SST variations in Guiana. The onset of this cold period was very abrupt at the Guiana site, with SST decreases of ca. −6 • C ka −1 and changes over 2 • C (Table 3).
During glacial periods, the SST record of MD03-2616 shows significant variability, with oscillations of 0.5-1. ability of MD03-2616 in MIS3 is therefore lower than in the cores retrieved further north in the North Atlantic. Changes at high latitudes are stronger than in the tropics due to sea-ice albedo feedbacks (Menviel et al., 2014). In this respect, it is hard to identify un-CPD 11,2015 Parallelisms between sea surface temperature changes in the western tropical Atlantic ambiguously either the Dansgaard-Oeschger type of oscillations described in northern latitudes or the SST drops associated with the Heinrich events characterising North Atlantic records.

Glacial see-saw between the tropics and Greenland
Previously published datasets are available to assess the significance of trends and 5 events observed in Greenland and in Guiana during the glacial ( Fig. 5a and b). Long term trends in the ODP 1002C reference core from the Cariaco basin (ca. 72 radiocarbon dates; 10 Guiana supports that these tropical cores present reminders of Greenland rapid oscillations but also a robust response to precessional forcing (Fig. 5d). Terrestrial records of Central America from Lake Peten Itza (Guatemala, 17 • N, 89 • W) also follow the MIS3 15 abrupt variability recorded in Greenland ice (Hodell et al., 2008). Cold conditions over the North Atlantic and strong trades induce a southward shift of the ITCZ over the Atlantic region with hydrological perturbations simulated over the northern part of South America (Menviel et al., 2014). The same is the case for terrestrial climate signals involving contributions from pollen, fern spores and lithogenic deposition (e.g. Ti/Ca 20 or Fe/Ca ratios, or continental organic matter inputs) in Brazilian cores, which follow sedimentation pulses paralleling those recorded during Heinrich events (Jennerjahn et al., 2004;Nace et al., 2014). Hence, the influence of abrupt climate variations in the North Atlantic and Greenland encompassed a large extension of tropical regions. This evidence suggests that the marine and continental climate of northern South America connected with polar variability during glacial periods, though reacting in a muted way and mainly dominated by precessional forcing.
CPD 11,2015 Parallelisms between sea surface temperature changes in the western tropical Atlantic Lack of synchrony between trends in tropical SST records and Greenland temperatures are also observed in high resolution profiles from sites located in the Agulhas current (not shown), such as in cores MD96-2080(Simon et al., 2013 and MD02-2594(Dyez et al., 2014 or tropical cores located in the eastern Atlantic (Gulf of Guinea, 2 • N, 9 • E; Weldeab et al., 2007). SST dynamics of these sites have been attributed eas. SSTs during the MIS1 (28.3 • C) were lower than in MIS5e, which is consistent with observations from previous studies at North Atlantic latitudes. MIS5b and MIS5d decreases are much smaller than those observed in the Atlantic Ocean at higher latitudes, though proportional to the subdued glacial-to-interglacial SST range of tropical regions.

20
From MIS5e to MIS5a, SSTs in Guiana show a remarkable parallelism with the temperature changes observed in Greenland, suggesting a close connection between tropical and arctic Atlantic latitudes in these periods. A possible mechanism to explain this connection is the transport of salt from the Caribbean Sea to North Atlantic latitudes when the whole AMOC was active and strong, thereby facilitating the thermohaline 25 transmission role resulting from the transfer of salty tropical waters into the North Atlantic. 11,2015 Parallelisms between sea surface temperature changes in the western tropical Atlantic Abrupt transitions have been identified in core MD03-2616. Some of these changes are observed in MIS5d and MISb but are much more commonly found during transitional periods from MIS4 to MIS2. The influence of northern waters during deglaciation periods may have had an effect on the latitudinal displacements of the ITCZ, which could also have increased SST variability in Guiana. MD03-2616 SSTs exhibit a strong 5 abrupt warming and cooling changes coincident with the B-A. This variability has also been observed in the δ 18 O ice profile of Sajama, a Bolivian ice core. Both sites show a very abrupt end of the YD (rates of 4 • C ka −1 and more than a 2. During MIS3 and early MIS2, the SST record in Guiana appears to balance changes in the characteristic long-term trend observed at higher latitudes. When Greenland experienced a cooling trend, Guiana showed a warming; or vice versa, Greenland remained 15 stable when Guiana experienced a cooling trend. This lack of synchrony is consistent with SST records in northern and southern locations of the Atlantic Ocean (Cariaco and Brazil, respectively) and evidence the decoupling between these areas when the AMOC weakens.