Salinity changes in the western tropical South Atlantic during the last 30 kyr

e 57 (2007) 383–395www.elsevier.com/locate/gloplacha

Global and Planetary Chang

Salinity changes in the western tropical South Atlanticduring the last 30 kyr

Felipe A.L. Toledo a,⁎, Karen B. Costa a, María A.G. Pivel b

a Laboratório de Paleoceanografia do Atlântico Sul, Departamento de Oceanografia Física, Química e Geológica, Instituto Oceanográfico,Praça do Oceanográfico 191, Universidade de São Paulo, SP, CEP 05508-900, Brazil

b Programa de Pós Graduação em Oceanografia Química e Geológica, Instituto Oceanográfico, Praça do Oceanográfico 191,Universidade de São Paulo, SP, CEP 05508-900, Brazil

Received 26 July 2006; received in revised form 14 December 2006; accepted 2 January 2007Available online 17 January 2007

Abstract

The hydrographic changes in the western tropical South Atlantic during the last 30 kyr were reconstructed based in the faunaland isotopic analyses of planktonic foraminifera of three cores taken along the Brazilian Continental Margin between 14°S and25°S. The application of the SIMMAX–MAT method on faunal counts data provided the sea surface temperature estimates. Seasurface salinity estimates were based on the oxygen isotope composition of Globigerinoides ruber (white). Additionally, theabundance record of the planktonic foraminifera Globorotalia truncatulinoides (right) was used as a proxy for vertical mixing ofsurface waters. Sea surface temperature estimates suggest a relative stability of the area during the last 30 kyr. However, significantchanges in the isotopic composition of G. ruber (white) suggest that the isotopic signal is dominated by the influence of sea surfacesalinity changes. The observed salinity changes are related to both the local hydrological balance and global circulation. Orbitalforcing and sea surface salinity changes were responsible for considerable changes in the stability of the upper water column andconsequently in the depth of the mixed layer, as indicated by the abundance record of G. truncatulinoides (right).© 2007 Elsevier B.V. All rights reserved.

Keywords: paleotemperature; paleosalinity; paleoceanography; South Atlantic

1. Introduction

Oceanic circulation is driven by the wind field at thesea surface and the density of seawater. Density dif-ferences that drive the global thermohaline circulationare ultimately generated by heat and freshwater fluxes at

⁎ Corresponding author. Tel.: +55 11 3091 6675; fax: +55 11 30916610.

E-mail addresses: [email protected] (F.A.L. Toledo),[email protected] (K.B. Costa), [email protected](M.A.G. Pivel).

0921-8181/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.gloplacha.2007.01.001

the ocean surface (Rahmstorf, 1996). Although coolingoccurs at all places in high latitudes, seawater is notalways dense enough to sink to the bottom of the oceaneven when cooled to the freezing point. Therefore, be-sides sea surface temperature, salinity plays a significantrole in determining where deep water can form (Boyle,1990).

The distribution of sea surface salinity depends on thefreshwater fluxes (evaporation, precipitation and conti-nental runoff) and on ocean dynamics. Maximumsalinities occur in the sub-tropics associated to the highpressure belts where evaporation exceeds precipitation.In the area of the Intertropical Convergence Zone (ITCZ)

384 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

high precipitation determines a negative evaporation–precipitation (E–P) budget, and thus lower salinities.Salinity also decreases towards high latitudes and locallynear coasts due to continental runoff. The large riversthat discharge into the Atlantic Ocean contribute toalleviate a positive E–P balance caused by the net exportof moisture from the Atlantic to the Pacific, but theycannot compensate for the low level of rainfall over thesea surface (Tomczak and Godfrey, 2003). As a result,the Atlantic Ocean is significantly saltier than the otherocean basins. The salinity contrast between the Atlanticand Pacific ocean basins is considered one of the mostcrucial factors governing the present global thermoha-line circulation (Seidov and Haupt, 2003). Changes inthe longitudinal distribution of E–P and thus, salinity,may be responsible for the dramatic changes occurred inthe global circulation patterns during the last 20 kyr(Bigg, 2003).

Given the important role of salinity in global oceaniccirculation and therefore in global climate, a completepaleoceanographic reconstruction must include esti-mates of past temperature and salinity changes. Unfortu-nately, unlike paleotemperatures which can be estimatedthrough a series of geochemical and faunal proxies, thereis not a direct proxy to estimate paleosalinities. The firstattempts to provide estimates of paleosalinities werebased on the analysis of relative abundances of plank-tonic foraminifera (e.g. Imbrie and Kipp, 1971; Cullen,1981), comparable to the methods usually applied forpaleotemperature estimates. Indeed, foraminiferal as-semblages are determined by several environmentalfactors, including temperature, salinity, seasonality andproductivity. However, temperature is the dominantfactor, and since salinity and temperature are correlated,an estimate based in salinity would not be independentfrom the temperature effect (Wolff et al., 1999). Still,foraminiferal assemblages can provide useful estimateswherever temperature changes are relatively low andsalinity is the major environmental variable influencingthe faunal assemblages and where salinity is not highlycorrelated to other environmental variables such as tem-perature, nutrient levels and productivity as in the north-ern Indian Ocean (Cullen, 1981).

Another method that has been used to estimatepaleosalinities is based on residual variations of theoxygen isotopic composition of foraminifera. Thisapproach has been applied to provide estimates of seasurface salinity (SSS) changes between present day andLast Glacial Maximum (LGM) conditions (e.g. Duplessyet al., 1991) and also to provide continuous down-coreSSS estimates (e.g. Maslin et al., 1995; Wang et al.,1995; Kroon et al., 1997; Chapman et al., 2000). The

carbonate isotopic content primarily reflects changes intemperature and in the isotopic composition of seawater.Whenever an independent sea-surface temperature(SST) proxy is available, it is possible to exclude thetemperature effect from the isotopic signal and thusobtain the isotopic composition of seawater (δ18Ow).Subsequently, δ18Ow estimates can be transformed intoSSS estimates based on the existing relationship betweenδ18Ow and salinity. Salinity and δ18Ow show a quasi-linear relationship, with the slope depending on thegeneral evaporation and precipitation characteristics ofthe region investigated (Craig and Gordon, 1965). Theslope of the δ18Ow–salinity relationship ranges from 0.1for tropical surface waters to 1 for high-latitude surfacewaters (Paul et al., 1999).

The residual method has several limitations. Proba-bly the most controversial is the fact that it relies on theassumption that the relationship between δ18Ow andsalinity remained constant both in space and time. Yet,this relationship must have changed at least, in responseto global changes of salinity and δ18Ow as discussed inSection 3. The confidence of paleosalinity estimates isalso limited by the precision of the paleotemperatureestimates.

In this paper we apply the residual method to provideestimates of SSS for the western tropical South Atlanticfor the last 30 kyr. Once the limitations of the methodare acknowledged, we analyze the general trend of sa-linity changes based on the qualitatively record of SSSchanges. We also investigate the surface layer stabilityusing the relative abundance of G. truncatulinoides(right coiling) as a proxy for mixed layer depth.

2. Methods

This study is based on the analysis of three pistoncores taken from intermediate depths in the westernSouth Atlantic Ocean between 14° and 25°S (Fig. 1,Table 1). Modern hydrographic conditions were ob-tained from the Geochemical Ocean Section Study(GEOSECS) database (Bainbridge, 1981).

Sediment cores were sampled at intervals of 5 to10 cm. Our study covers the last 30 kyr, correspondingto 36 samples at SAN-76, 17 at CMU-14 and 16 at ESP-08. Raw sediment samples were shaken in distilledwater and sieved over a 63 μm mesh.

2.1. Chronostratigraphy

The chronology of the analyzed cores is based on thecorrelation of the isotopic record of benthic foraminiferawith the SPECMAP chronology of Martinson et al.

Table 2Radiocarbon age control points obtained by accelerator massspectrometry (AMS)

14

Fig. 1. Location of the three cores analyzed in this study and Core GeoB3129/3911 (Weldeab et al., 2006).

385F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

(1987) and inAcceleratorMass Spectrometer radiocarbondatings on eight samples of monospecific planktonicforaminifera Globigerinoides ruber (N250 μm). The 14CAMS datings were performed at NOSAMS–WHOI labo-ratory facility, and all the 14C ages were corrected for areservoir effect of 400 yr (Bard, 1988) and transformedinto calendar years (Table 2).

Sedimentation rates of core ESP-08 were constantduring the last 30 kyr with values of ∼3 cm/kyr. Forcore CMU-14, the sedimentation rate was ∼6.5 cm/kyrduring the LGM and deglacial and ∼3 cm/kyr duringthe Holocene. For core SAN-76, the sedimentation ratesare significantly higher; during the LGM the values are23 cm/kyr, deglacial 11 cm/kyr and Holocene 4 cm/kyr.The sedimentation rates for the LGM are in general, atleast twice higher than the Holocene. This difference in

Table 1Location and water depth of the piston cores analyzed

Piston cores Latitude Longitude Water depth (m)

CMU-14 14°24′00″S 38°49′12″W 965SAN-76 24°25′48″S 42°16′48″W 1682ESP-08 20°57′00″S 39°31′48″W 1995

the sedimentation rates between LGM and Holocene aremainly attributed to the sea level changes (−130 m)(Kowsmann and Costa, 1979) that exposed the conti-nental shelf, favouring the riverine terrigenous inputdirectly to the continental slope off Brazil (Damuth,1977).

2.2. Oxygen isotope analyses

All isotope analyses were performed at the WoodsHole Oceanographic Institution (WHOI) laboratory

Core Depth in core (cm) C age (yr) Calendar age (yr)

SAN-76 13 3010 2756148 13,450 15,516320 19,300 22,374416 29,400 33,583

ESP-08 07 3370 321989 24,000 27,738

CMU-14 12 4010 3987121 17,850 20,756

386 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

facilities, using a Finnigan MAT252 with the automatedKiel device. The standard deviation of the isotope valuesof the National Bureau of Standards (NBS) carbonatestandard NBS-19 was ±0.08‰ NBS-19 isotope valueswere used to calibrate to PeeDee Belemnite (δ18O=−2.2Vienna Pee Dee Belemnite (VPDB)).

Data on the oxygen isotopic composition of plank-tonic foraminifera is based in the analysis of ten spe-cimens of Globigerinoides ruber (white) larger than212 μm in order to avoid size-dependent effects on theoxygen isotopes values. For the analysis of benthicforaminifera, 1–2 Cibicidoides specimens larger than250 μm with individual weight ranging from 50 to150 μg were picked.

2.3. Faunal analyses and estimates of paleo-SST

SST estimates based in relative species abundancesconstitute the only non-geochemical, and thus trulyindependent paleothermometer (Malmgren et al., 2001).Our paleo-SST estimates were obtained based in theSIMMAX–MAT method (Pflaumann et al., 1996). Thismethod differs from the original Modern AnalogTechnique (MAT) in the way best analogs are foundand treated. The census counts were performed on theN150 μm fraction which is the most used in studies ofplanktonic foraminifera assemblages for paleoclimaticand paleoceanographic reconstructions. Specimenssmaller than 150 μm are excluded from the faunalanalyses because they cannot be accurately identified tospecies level. The N150 μm fraction was then sub-sampled with a microsplitter until approximately 300specimens were left for binocular microscopic investi-gation. Criteria used for species identification comefrom Bé (1977) and Hemleben et al. (1989).

2.4. Estimates of paleo-SSS

Several paleotemperature equations have been pro-posed to express the relationship between the isotopiccomposition of calcium carbonate deposited by marineorganisms and the temperature at the time of deposition(e.g. Epstein et al., 1953; O'Neil et al., 1969; Shack-leton, 1974). Among all the equations available, wechose the one based in inorganically precipitated calciteby Kim and O'Neil (1997) (Eq. (1)) which is the mostappropriate when working with Globigerinoides ruber(white) (Schmidt and Mulitza, 2002).

Tiso ¼ 16:1−4:64ðd18Oc−d18O

wÞ

þ0:09ðd18Oc−d18O

wÞ2

ð1Þ

where Tiso is the isotopic temperature, δ18Ow is theisotopic composition of seawater and δ18Oc is the iso-topic composition of calcite.

The above equation is usually used to estimate pasttemperatures, yet, if we have an independent paleotem-perature estimate such as one based on faunal assem-blages of planktonic foraminiferawe are able to rearrangeEq. (1) to solve for δ18Ow. Before doing so, it is necessaryto calibrate the data set of δ18O of planktonic foraminifercalcite. The calibration is made by comparing the δ18Ovalues of the chosen species from surface sediments, inthis case G. ruber, and modern sea surface temperatures(SST) (Levitus, 1982). This calibration has alreadybeen made by Wang et al. (1995) (Eq. (2)), who foundthat in the low-latitude Atlantic, δ18O values of G. ruber(δ18Oruber) mainly reflect the summer temperature andsalinity at the 0–50 m water depth interval.

Tiso ¼ 3:147þ 0:963⁎Tm ð2Þwhere Tm is the measured (real) temperature.

Eq. (2) shows the relationship between the isotopictemperature derived from δ18Oruber and the measuredSST. Rewriting Eq. (1) to solve for δ18Ow and using thecalibration for G. ruber (Eq. (2)) we have:

d18Ow ¼ d18Oruber

− 25:78

þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið16:87þ 0:347 4 TmÞ

p= 0:18

ð3Þ

In order to transform the results obtained with Eq. (3)into SSS estimates we first need to analyze the relation-ship between δ18Ow and salinity. We applied three dif-ferent δ18Ow–salinity relationships, all of them based indata from the upper (b250 m) Atlantic. The latitudinalranges considered in each of the three equations are:(a) 40°N–40°S, as in Wang et al. (1995), (b) 25°N–25°S, and (c) 0°–45°S. Data include samples fromGeochemical Ocean Sections Study (GEOSECS,Östlund et al., 1987), World Ocean Circulation Experi-ment (WOCE, Meredith et al., 1999), Craig and Gordon(1965) and Pierre et al. (1991). All data are available athttp://www.giss.nasa.gov/data/o18data/. Fig. 2 showsthe regression lines of SSS versus δ18Ow (converted tothe PDB scale).

ðaÞ SSS ¼ 2:334⁎d18Owþ 34:66 ð4Þ

ðbÞ SSS ¼ 2:458⁎d18Owþ 34:61 ð5Þ

ðcÞ SSS ¼ 1:863⁎d18Owþ 34:95 ð6Þ

Eqs. (4)–(6) represent the modern relationship be-tween SSS and δ18Ow for different geographic areas. In

Fig. 3. Comparison of SSS estimates for core SAN-76 using threedifferent δ18Ow–salinity relationships.

Fig. 2. Modern δ18Ow–salinity relationships tested in this study(Eqs. (4)–(6)).

387F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

order to calculate past SSS values it is necessary toassume that this relationship has remained constantthrough time. Sea surface salinities are then calculatedby inserting Eq. (3) into one of the equations describingthe local relationship between δ18Ow and salinity. Forinstance, inserting Eq. (3) into Eq. (6) yields the follow-ing equation:

SSS¼ 34:95þ 1:863⁎½d18Oruber

− 25:78

þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið16:87þ 0:347 4 TmÞ

p= 0:18�

ð7Þ

Finally, to calculate past local SSS changes it isnecessary to extract the ice volume-effect from theδ18Ow value (Eq. (8)). The global δ18Ow ice effect usedhere is based on the estimates of Labeyrie et al. (1987).

SSS¼ 34:95þ 1:863⁎½d18Oruber

−d18Oice− 25:78

þffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffið16:87þ 0:347⁎TmÞ

p=0:18�

ð8Þ

Since the SST estimates used here are for the summerseason and since the δ18O values of G. ruber mainlyreflect the summer SST (Wang et al., 1995) the SSSestimates provided here are also for the summer season.

3. The residual method

Fig. 3 shows the SSS estimates for core SAN-76,using the three different relationships between δ18Ow

and salinity as explained in Section 2.4. Although interms of absolute values the different regressions ap-plied here yield different results, the pattern of changesis the same. The differences in the three curves de-monstrate the sensitivity of the residual method to theuse of different relationships between SSS and δ18Ow.Eqs. (4) (Wang et al., 1995) and (5) have similar slopes

and intercepts so they both provide very similar absolutevalues. Estimates for modern conditions using bothequations are higher than real salinity values. Eq. (6)yields a similar trend but smaller absolute values andmodern salinities closer to real values. Since we areinterested in the trends and not in absolute values, all thediscussion of the downcore estimates will be made usingonly the curves yielded by Eq. (6) which seems the mostrealistic.

The salinity–δ18Ow relationship is the result of themixing line between an oceanic and a freshwater end-member. The slope of the relationship used in paleo-salinity reconstructions is actually a spatial relationshipapplicable to a certain region. Ideally, the temporalvariations of these two variables should be considered(Delaygue et al., 2001) (i.e. the slope relating the timevariations of δ18Ow and salinity which are verydifficult to estimate). Past changes in the global mixingline can be estimated assuming that the high-latitudefreshwater end-member of precipitation was differentand that the linear regression still passed through theocean mid point which it does at present (Maslin et al.,1995; Schmidt, 1999). However, for low latitudes, it ismore appropriate to consider the local mixing lineinstead of the global one, and although the uncertaintyregarding the freshwater end-member is smaller in lowlatitudes (since it cannot be too different from theaverage tropical precipitation), the reduced slope of theδ18Ow–salinity relationship translates into larger errors(Schmidt, 1999).

Both the oxygen isotopic composition of seawaterand salinity are determined by the freshwater budgetincluding the processes of evaporation, precipitation andrunoff, but unlike salinity, the oxygen isotopic compo-sition of seawater also depends on the isotopic com-position of the freshwater being added or withdrawn,thus, the relationship between both variables is unlikelyto have remained constant through time. Taking as an

388 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

example the extreme changes of the LGM, we know thatthe global salinity increased by around 1.2‰ due to thesea level drop (Fairbanks, 1989) and that the oxygenisotopic composition of seawater increased 1.05±0.20‰ (Duplessy et al., 2002) due to the preferentialstorage of the lighter isotope in the ice sheets. If no otherchanges were present, these global variations wouldhave simply shifted the modern δ18Ow–salinity rela-tionship to a glacial one with the same slope (Schäfer-Neth, 1998) (Fig. 4)). Nevertheless further regionalvariations did most likely exist probably leading tochanges in the slope of the δ18Ow–salinity relationship.Localized temporal changes in the δ18Ow–salinityrelationship ultimately affect the entire ocean, includingmore “stable” areas far from the effects of sea-icefreezing and melting as discussed by Rohling and Bigg(1998). Also, when using a local relationship betweenδ18Ow and salinity it is assumed that it reflects the localhydrological budget although for many regions, theδ18Ow–salinity relationship is actually determined byadvection and mixing (Rohling and Bigg, 1998; Wadleyet al., 2002).

The final error associated to the paleosalinity esti-mates presented here is around 1.6‰, considering theerrors associated with the isotopic composition offoraminifera, the temperature proxy, the paleotempera-ture equation, and the isotopic temperature versus realtemperature. This error does not include uncertainties inthe ice volume effect and in the freshwater end-member(see Schmidt, 1999 for a detailed discussion on the errorpropagation through the residual method for paleo-SSSestimates). However, given the importance of salinity,even a semiquantitative salinity proxy is valuable (Mixet al., 2000), so, having considered the limitations of themethod we proceed to analyze the results.

Fig. 4. The δ18Ow–salinity relationship for modern conditions andcorrected for the global salinity and isotopic changes of the LastGlacial Maximum.

4. Results and discussion

4.1. SST estimates

The application of the SIMMAX–MAT methodyields low temperature changes for the western tropicalAtlantic during the last 30 kyr (Figs. 5a–7a). Estimatesfor cores CMU-14 (14.4°S) and ESP-08 (20.95°S) sug-gest that the region experienced only minor (∼1 °C)temperature changes. The southernmost core SAN-76(24.43°S) is the one that shows the larger (but stillmoderate) changes (∼2 °C) with the highest tempera-tures in the Holocene, between 4 and 6 kyr BP and inpresent conditions, and lowest temperatures in the LGM,around 21 kyr BP. The paleo-SST estimates obtained forthe LGM are consistent with other recent temperatureestimates (e.g. Crowley, 2000, Niebler et al., 2003)which suggest that tropical SST changes were relativelymodest (around 2°C) but not as small as previouslysuggested by CLIMAP. Relatively low tropical coolingduring glacial times reflects the stability of this regionduring the last 30 kyr.

4.2. SSS estimates

The paleotemperature reconstructions suggest thatmost of the isotopic signal is related to changes in theisotopic composition of seawater, and thus, salinity.Salinity variations may reflect changes in the hydrolog-ical balance (evaporation and precipitation) and/orchanges in ocean dynamics (advection and mixing).Since the three cores are away from the influence oflarge rivers, the effect of continental runoff is assumedto be negligible. Figs. 5c–d–7c–d show the salinityestimates for each core, including the “total” sea surfacesalinity and the “local” salinity changes after excludingthe global salinity signal resulting from the ice volumeeffect.

The three cores analyzed show similar trends in pastchanges of SSS (Figs. 5c–7c). They all present higherSSS values during MIS 2 especially between 22 and28 kyr BP and a gradual decrease during the course ofthe Holocene. A freshening of the tropical Atlanticduring the Holocene is consistent with the intensificationof the thermohaline circulation which would reestablishthe heat and salt export to the northern Atlantic.

The salinity estimates for all cores are consistent,with similarities between them reflecting their geo-graphical location. The salinity estimates for core ESP-08 show an intermediate trend between those of SAN-76and those of CMU-14. Despite the low resolution ofESP-08 and CMU-14 the record between 8 and 13 kyr

Fig. 5. Downcore record and estimates for SAN-76: (a) sea surface temperature (°C) estimates based on the SIMMAX–MAT method, (b) isotopiccomposition of the planktonic foraminifer G. ruber (white) (‰ PDB), (c) sea surface salinity estimates based on the residual method, (d) local seasurface salinity estimates excluding the ice volume effect, and (e) relative abundance of the planktonic foraminifer G. truncatulinoides (right) (%).

389F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

BP of both cores is remarkably similar. However, duringthe Holocene, CMU-14 seems to have been more stableand the trend of ESP-08 resembles more with SAN-76.

The last 20 kyr are analyzed in more detail in Fig. 8along with the data obtained byWeldeab et al. (2006) forcore GeoB3129/3911, which is located farther north inthe western tropical Atlantic (Fig. 1). Although we

cannot compare absolute values – at least with theestimates from core GeoB3129/3911 which were ob-tained through different calculations – Fig. 8 indicatesthat the latitudinal gradient with maximum salinitiesrecorded at the southernmost core, SAN-76 was main-tained throughout the whole period investigated. Thismeridional gradient results from the evaporation–

Fig. 6. Downcore record and estimates for ESP-08: (a) sea surface temperature (°C) estimates based on the SIMMAX–MAT method, (b) isotopiccomposition of the planktonic foraminifer G. ruber (white) (‰ PDB), (c) sea surface salinity estimates based on the residual method, (d) local seasurface salinity estimates excluding the ice volume effect, and (e) relative abundance of the planktonic foraminifer G. truncatulinoides (right) (%).

390 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

precipitation (E–P) budget, which in turn is related tothe atmospheric circulation (Hadley cell). Two SSS lowsaround 19, 14 kyr BP were observed in the three coresunder study and in core GeoB3129/3911 (Weldeabet al., 2006). A sharp decrease in salinity around 10 kyrBP is observed for cores ESP-08 and CMU-14. LowSSSs could be related either to a decrease in the E–P

balance related to shifts in the ITCZ or to a strengthen-ing of the thermohaline circulation and export of surfacewaters to the northern hemisphere.

Despite the different resolution of cores CMU-14 andGeoB3129/3911 (the former being a low-resolution re-cord and the latter a high resolution record), the simi-larity between the general trend of SSS estimates of both

Fig. 7. Downcore record and estimates for CMU-14: (a) sea surface temperature (°C) estimates based on the SIMMAX–MAT method, (b) isotopiccomposition of the planktonic foraminifer G. ruber (white) (‰ PDB), (c) sea surface salinity estimates based on the residual method, (d) local seasurface salinity estimates excluding the ice volume effect, and (e) relative abundance of the planktonic foraminifer G. truncatulinoides (right) (%).

391F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

cores is noteworthy (Fig. 8). Both records show twopronounced SSS increases around 17.5 and 12.5 kyr BP.A third peak of pronounced increase in salinity recordedby GeoB3129/3911 around 14.5 kyr BP and absent inthe record of CMU-14 could have been easily misseddue to the much lower resolution of the latter core.Weldeab et al. (2006) demonstrate that the salinity peaks

around 17.5 and 12.5 kyr BP are related to the northernhemisphere Younger Dryas and Heinrich 1 events. Theauthors also show that these high salinity events aresynchronous with southern high latitude warmings andthat they lead approximately 500 yr the cooling eventsfrom the northern hemisphere. These proxy records areconsistent with modeling studies (Manabe and Stouffer,

Fig. 8. Comparison of the SSS estimates obtained in this study andthose obtained by Weldeab et al. (2006) for the last 20 kyr in thewestern tropical Atlantic.

392 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

1997; Rühlemann et al., 1999; Dahl et al., 2005) ac-cording to which the deglacial meltwater events wouldhave weakened the thermohaline overturn, reducing thenorthward advection of warm, saline surface waters and,thus, causing a warming of the upper layers in thetropical ocean. However, modeling results indicate thatthe maximum warming anomaly probably occurred inintermediate water depths rather than in the most super-ficial waters (Dahl et al., 2005). This would explain therelatively stable sea surface temperatures observed inour records.

Although CMU-14 is closer to ESP-08 than to coreGeoB3129/3911 (Fig. 1), the similarities between therecords of cores CMU-14 and GeoB3129/3911 suggestthat both sites were under similar environmental con-ditions, which were slightly different from those of thecores located southwards. Today, core GeoB3129/3911is fully under the influence of the North Brazil Currentand CMU-14 is located close to the area where theSouth Equatorial Current bifurcates into a northwestbranch (North Brazil Current) and a southwest branch(Brazil Current). The situation might have changed inthe course of the last deglaciation due to changes in theITCZ location and in trade wind strength, andconsequently in the wind-driven circulation. However,even if both sites were influenced by different currents(i.e. the North Brazil and Brazil currents), they wouldhave still been bathed by the same Tropical Water, sothe fact that the salinity trends are the same at both sitesindicates that the hydrological conditions were alsosimilar.

4.3. Mixed layer depth

The extent of vertical mixing depends on the strengthof the winds blowing on the sea surface, and on thedensity (and thus, temperature and salinity) stratifica-tion. The abundance record of the planktonic foramin-ifer Globorotalia truncatulinoides (right) can be used asa proxy for vertical mixing of surface waters (Mulitzaet al., 1997; Dürkoop et al., 1997). In low latitudesG. truncatulinoides dwells at considerable depths,except during early austral spring when it ascends intothe upper water column to reproduce (Hemleben et al.,1989). Subsequently, during summer and fall it de-scends into deeper layers. In principle, the comparisonbetween the surface environmental parameters SST andSSS should not necessarily be related with the abun-dance of a species that dwells in deeper waters. How-ever, the upper layer density stratification directlyaffects this deep-living species, since it depends to asignificant degree on the mixing of the upper ocean inorder to return juveniles produced at depth to surfacewaters (Lohmann and Schweitzer, 1990). Therefore,rather than temperature, its distribution is dependent onthe scale of vertical mixing (Lohmann, 1992).

G. truncatulinoides corresponds in fact to a complexof four genetic species adapted to particular hydrograph-ic conditions (de Vargas et al., 2001). There are two cold-water species (sp.3 and sp.4) living respectively in thesubantarctic convergence and in the subantarctic zoneand two warm-water species (sp.1 and sp.2) with lessclear ecological preferences. Both warm-water speciesare found along the area covered by the analyzed cores.Warm-water species 2 is the only one including bothright and left coiled specimens. The other three speciescomprise left-coiled specimens only. Apparently, leftcoiled sp.1 inhabits nutrient-depleted centers of thesubtropical gyres, whereas right-coiled sp.2 prefers moreproductive environments such as the margins of thesubtropical gyres (Renaud and Schmidt, 2003). Today,maximum abundances of G. truncatulinoides (right) insediments from the western South Atlantic are foundaround the 30°S latitude (Mulitza et al., 1997; Kemle-von Mücke and Hemleben, 1999).

In general terms, the three cores present similartrends of changes in the relative abundances ofG. truncatulinoides (right) (Figs. 5e–7e). Again, as inthe comparison of SSS estimates, ESP-08 presents anintermediate trend between CMU-14 and SAN-76.Maximum abundances occurred around 9 to 11 kyrBP in the three cores and also around 21 kyr BP at thesite of CMU-14. The peak around 10 kyr BP coincidesat SAN-76 with high SSSs. In fact, all the downcore

393F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

record of the relative abundance of G. truncatulinoides(right) at SAN-76 is congruent with the SSS estimatesfor this site suggesting that at least at the region of coreSAN-76, SSS changes play an important role indetermining the extent of vertical mixing.

On the other hand, the peak around 10 kyr BPobserved in the records of ESP-08 and CMU-14, coin-cides with low SSSs. This means that in this case, windstrength must have played a more significant role thanSSS in determining the mixed layer depth. The increasein wind strength during this period can be explained byorbital forcing. In terms of the precessional cycle, theEarth's orbital parameters today are similar to those at21 kyr BP and opposite to the ones at 11 kyr BP. Today,the southern hemisphere is farther from the Sun in June(austral winter) and closer in December (austral sum-mer) resulting in large seasonal temperature contrasts.Conversely, 11 kyr ago the Earth was closer to the Sunin June and farther in December, therefore reducing theseasonal contrast in the Southern Hemisphere (Bergerand Loutre, 1991). As a result, the ITCZ was locatedfarther north, probably intensifying the penetration ofpolar advections (Martin et al., 1997), which wouldcontribute to the instability and mixing of the surfacelayers. Subsequently, over the course of the Holocene,the insolation changes related to the precessional cycleseem to have caused a general southward migration ofthe ITCZ (Haug et al., 2001). The general decrease inthe G. truncatulinoides (right) abundance during theHolocene is explained by the combination of morestable conditions along with higher SSTs and lowerSSSs which reduce sea surface density and thus promoteupper layer stratification inhibiting the growth of thisdeep-dweller foraminifer. This explains the absence ofthis species from the records in the late Holocene.

The peak in the abundance of G. truncatulinoides(right) at 21 kyr BP at CMU-14 cannot be explained byorbital forcing. This maximum is more likely related tothe combination of high SSSs, a local reduction in SST(Fig. 7) and stronger trade winds which have beensuggested by Kim and Schneider (2003) for this timeperiod.

5. Conclusions

The analyses of three cores recovered from the Bra-zilian Continental Margin allowed to reconstruct thesurface hydrography of the area during the last 30 kyr. Interms of SST, the western tropical South Atlantic seemsto have experienced only minor changes, of around 1 °Cin the location of cores CMU-14 and ESP-08 and ∼2 °Cat the southernmost core, SAN-76. The low SST

changes suggest that the oxygen isotopic signal ofG. ruber (white) is dominated by changes in the oxygenisotopic composition of surface waters, and thus, SSS.Superimposed on the general trend of reduction of SSSfrom the last glacial to modern conditions due to theglobal reduction of water stored as ice, the local SSSchanges also indicate a freshening of the westerntropical Atlantic during the Holocene which is consis-tent with the reestablishment of the heat and salt exportto the northern Atlantic. Besides the role of globalcirculation and salinity changes, the SSS also variedlocally as a function of changes in the hydrologicalbalance. The abundance record of G. truncatulinoides(right) showed that both the changes in SSS and inorbital forcing were responsible for considerablechanges in the stability of the upper water column andconsequently in the depth of the mixed layer.

Despite the large errors associated to the residualmethod, the paleosalinity estimates presented here aresupported by the general agreement between the trendsobserved in the estimates for the three cores, theconsistency of the latitudinal gradients and the remark-able similarity between our estimates for core CMU-14and those presented by Weldeab et al. (2006) for coreGeoB3129/3911.

Acknowledgments

We thank the Brazilian National Council for Re-search and Scientific Development (CNPq) for a grantto M.A.G. Pivel and to PETROBRAS for providing thesamples. Financial support was provided by FAPESP(Fundação de Amparo à Pesquisa do Estado de SãoPaulo) process 04/02819-8. We thank two anonymousreviewers for their contribution to the improvement ofthis manuscript. This is Laboratório de Paleoceanogra-fia do Atlântico Sul (LaPAS), IO-USP, contribution 07.

References

Bainbridge, A.E., 1981. GEOSECSAtlantic Expedition, Hydrographicdata, 1972–1973. National Science Foundation. U.S. GovernmentPrinting Office, Washington, D.C.

Bard, E., 1988. Correction of accelerator mass spectrometry 14C agesmeasured in planktonic foraminifera: paleoceanographic implica-tions. Paleoceanography 3, 635–645.

Bé, A.W.H, 1977. An ecological, zoogeographic and taxonomicreview of Recent planktonic Foraminifera. In: Ramsay, A.T.S.(Ed.), Oceanic Micropaleontology, vol. 1. Academic Press, Lon-don, pp. 1–100.

Berger, A., Loutre, M.F., 1991. Insolation values for the climate of thelast 10 million years. Quat. Sci. Rev. 10, 297–317.

Bigg, G.R., 2003. The Oceans and Climate, 2nd Ed. CambridgeUniversity Press, Cambridge, UK.

394 F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

Boyle, E.A., 1990. Quaternary deepwater paleoceanography. Science249, 863–870.

Chapman, M.R., Shackleton, N., Duplessy, J.C., 2000. Sea surfacetemperature variability during the last glacial–interglacial cycle:assessing the magnitude and pattern of climate change in the NorthAtlantic. Palaeogeogr. Palaeoclimatol. Palaeoecol. 157, 1–25.

Craig, H., Gordon, L.I, 1965. Deuterium and oxygen 18 variations inthe ocean and the marine atmosphere. In: Tongiorgi, E. (Ed.),Stable isotopes in oceanographic studies and paleotemperatures.Consiglio Nazional delle Richerche, Spoleto, Italy, pp. 9–130.

Crowley, T.J., 2000. CLIMAP SSTs re-revisited. Clim. Dyn. 16,241–255.

Cullen, J.L., 1981. Microfossil evidence for changing salinity patternsin the Bay of Bengal over the last 20.000 years. Palaeogeogr.Palaeoclimatol. Palaeoecol. 35, 315–356.

Dahl, K.A., Broccoli, A.J., Stouffer, R.J., 2005. Assessing the role ofNorth Atlantic freshwater forcing in millennial scale climate vari-ability: a tropical Atlantic perspective. Clim. Dyn. 24, 325–346.

Damuth, J.E., 1977. Late Quaternary sedimentation in the westernequatorial Atlantic. Geol. Soc. Amer. Bull. 88, 695–710.

Delaygue, G., Bard, E., Rollion, C., Jouzel, J., Stiévenard,M.,Duplessy,J.C., Ganssen, G., 2001. Oxygen isotope/salinity relationship in thenorthern Indian Ocean. J. Geophys. Res. 106 (C3), 4565–4574.

de Vargas, C., Renaud, S., Hilbrecht, H., Pawlowski, J., 2001.Pleistocene adaptative radiation in Globorotalia truncatulinoides:genetic, morphologic, and environmental evidence. Paleobiology27, 104–125.

Duplessy, J.C., Labeyrie, L., Juillet-Leclerc, A., Maitre, F., Duprat, J.,Sarnthein, M., 1991. Surface salinity reconstruction of the NorthAtlantic Ocean during the last glacial maximum. Oceanol. Acta 14,311–324.

Duplessy, J.C., Labeyrie, L., Waelbroeck, C., 2002. Constraints on theocean isotopic enrichment between the Last Glacial Maximum andthe Holocene: paleoceanographic implications. Quat. Sci. Rev. 21,315–330.

Dürkoop, A., Hale, W., Mulitza, S., Pätzold, J., Wefer, G., 1997. LateQuaternary variations of sea surface salinity and temperature in thewestern tropical Atlantic: evidence from δ18O of Globigerinoidessacculifer. Paleoceanography 12 (6), 764–772.

Epstein, S.R., Buchsbaum, R., Lowenstam, H.A., Urey, H.C., 1953.Revised carbonate-water isotopic temperature scale. Geol. Soc.Amer. Bull. 64, 1315–1325.

Fairbanks, R.G., 1989. Glacio-eustatic sea level record 0–17,000 yearsbefore present: influence of glacial melting rates on Younger Dryasevent and deep ocean circulation. Nature 342 (6250), 637–642.

Haug, G.H., Hughen, K.A., Sigman, D.M., Peterson, L.C., Röhl, U.,2001. Southward migration of the Intertropical Convergence Zonethrough the Holocene. Science 293, 1304–1308.

Hemleben, C., Spindler, M., Anderson, O.R., 1989. Modern Plank-tonic Foraminifera. Springer-Verlag, Berlin. Heidelberg NewYork.

Imbrie, J., Kipp, N.G., 1971. A new micropaleontological method forpaleoclimatology: application to a late Pleistocene Caribbean core.In: Turekian, K.K. (Ed.), The late Cenozoic glacial ages. YaleUniversity Press, New Haven, pp. 71–181.

Kemle-von Mücke, S., Hemleben, C., 1999. Foraminifera. In: Boltov-skoy, D. (Ed.), South Atlantic zooplankton. Backhuys Publisher,Leiden, The Netherlands, pp. 43–73.

Kim, S.-T., O'Neil, J.R., 1997. Equilibrium and nonequilibriumoxygen isotope effects in synthetic carbonates. Geochim. Cosmo-chim. Acta 61, 3461–3475.

Kim, J.-H., Schneider, R.R., 2003. Low-latitude control of interhemi-spheric sea-surface temperature contrast in the tropical Atlantic

over the past 21 kyears: the possible role of SE trade winds. Clim.Dyn. 21, 337–347.

Kowsmann, R.O., Costa, M.P.A., 1979. Sedimentação quaternária daMargem Continental Brasileira e das áreas oceânicas adjacentes.CENPES/PETROBRÁS, Rio de Janeiro.

Kroon, D., Austin, W.E.N., Chapman, M.R., Ganssen, G.M., 1997.Deglacial surface circulation changes in the northeastern Atlantic:temperature and salinity records off NW Scotland on a centuryscale. Paleoceanography 12 (6), 755–763.

Labeyrie, L.D., Duplessy, J.C., Blanc, P.L., 1987. Variations in modeof formation and temperature of oceanic deep waters over the past125000 years. Nature 327, 477–482.

Levitus, S., 1982. Climatological atlas of the world ocean. NOAAProfessional Paper, vol. 13. US Government Printing Office,Washington, D.C.

Lohmann, G.P., 1992. Increasing seasonal upwelling in the subtropicalSouth Atlantic over the past 700000 yrs: evidence from deep-livingplanktonic foraminifera. Mar. Micropaleontol. 19, 1–12.

Lohmann, G.P., Schweitzer, P.N., 1990. Growth and chemistry asprobes of the past thermocline, 1. Shell size. Paleoceanography 5,55–75.

Malmgren, B.A., Kucera, M., Nyberg, J., Waelbroeck, C., 2001.Comparison of statistical and artificial neural network techni-ques for estimating past sea surface temperatures fromplanktonic foraminifer census data. Paleoceanography 16 (5),520–530.

Manabe, S., Stouffer, R.J., 1997. Coupled ocean–atmosphere modelresponse to freshwater input: comparison to the Younger Dryasevent. Paleoceanography 12, 321–336.

Martin, L., Bertaux, J., Corrège, T., Ledru, M.P., Mourguiart, P.,Sifeddine, A., Soubiès, F., Wirrmann, D., Suguio, K., Turcq, B.,1997. Astronomical forcing of contrasting rainfall changes in tro-pical South America between 12.400 and 8.800 cal yr B.P. Quat.Res. 47, 117–122.

Martinson, D.G., Pisias, N.G., Hays, J.D., Imbrie, J.T.C., Moore, J.,Shackleton, N.J., 1987. Age dating and the orbital theory of the IceAges: development of a high-resolution 0 to 300,000-year chrono-stratigraphy. Quat. Res. 27, 1–29.

Maslin, M.A., Shackleton, N.J., Pflaumann, U., 1995. Surfacewater temperature, salinity and density changes in the northeastAtlantic during the last 45.000 years: Heinrich events, deepwater formation and climatic rebounds. Paleoceanography 10,527–544.

Meredith, M.P., Grose, K.E., McDonagh, E.L., Heywood, K.J., Frew,R.D., Dennis, P.F., 1999. Distribution of oxygen isotopes in thewater masses of Drake Passage and the South Atlantic. J. Geophys.Res. 104 (C9), 20949–20962.

Mix, A.C., Bard, E., Eglinton, G., Keigwin, L.D., Ravelo, A.C., Rosenthal,Y., 2000. Alkenones and multiproxy strategies in paleoceanographicstudies. Geochem. Geophys. Geosyst. 1 (2000GC000056).

Mulitza, S., Niebler, H.S., Dürkoop, A., Hale, W., Wefer, G., 1997.Planktonic foraminifera as recorders of past surface-waterstratification. Geology 25, 335–338.

Niebler, H.-S., Arz, H.W., Donner, S., Mulitza, S., Pätzold, J., Wefer,G., 2003. Sea surface temperatures in the equatorial and SouthAtlantic Ocean during the Last Glacial Maximum (23–19 ka).Paleoceanography 18 (3), 1069.

O'Neil, J.R., Clayton, R.N., Mayeda, T.K., 1969. Oxygen isotopefractionation in divalent metal carbonates. J. Chem. Phys. 51,5547–5558.

Östlund, H.G., Craig, H., Broecker, W.S., Spencer, D., 1987.GEOSECS Atlantic, Pacific and Indian Ocean expeditions: shore

395F.A.L. Toledo et al. / Global and Planetary Change 57 (2007) 383–395

based data and graphics. Tech. rep., vol. 7. Nat. Sci. Found,Washington, D.C.

Paul, A., Mulitza, S., Pätzold, J., Wolff, T., 1999. Simulation of oxygenisotopes in a global ocean model. In: Fischer, G., Wefer, G. (Eds.),Use of proxies in paleoceanography: Examples from the SouthAtlantic. Springer-Verlag, Berlin, pp. 655–686.

Pflaumann, U., Duprat, J., Pujol, C., Labeyrie, L.D., 1996. SIMMAX:a modern analog technique to deduce Atlantic sea surface tem-peratures from planktonic foraminifer in deep-sea sediments.Paleoceanography 11, 15–36.

Pierre, C., Vergnaud-Grazzini, C., Faugeres, J.-C., 1991. Oxygen andcarbon stable isotope tracers of the water masses in the CentralBrazil Basin. Deep-Sea Res. 38, 597–606.

Rahmstorf, S., 1996. On the freshwater forcing and transport of theAtlantic thermohaline circulation. Clim. Dyn. 12, 799–811.

Renaud, S., Schmidt, D.N., 2003. Habitat tracking as a response of theplanktonic foraminifer Globorotalia truncatulinoides to environ-mental fluctuations during the last 140 kyr. Mar. Micropaleontol.49, 97–122.

Rohling, E.J., Bigg, G.R., 1998. Paleosalinity and δ18O: a criticalassessment. J. Geophys. Res 103 (C1), 1307–1318.

Rühlemann, C., Multiza, S., Müller, P.J., Wefer, G., Zahn, R., 1999.Warming of the tropical Atlantic Ocean and slowdown ofthermohaline circulation during the last deglaciation. Nature 402,511–514.

Schäfer-Neth, C., 1998. Changes in the seawater salinity–oxygenisotope relation between Last Glacial and Present: sediment coredata and OGCM modelling. Paleoclimates 2 (2–3), 101–131.

Schmidt, G.A., 1999. Error analysis of paleosalinity reconstructions.Paleoceanography 14 (3), 422–429.

Schmidt, G.A., Mulitza, S., 2002. Global calibration of ecologicalmodels for planktic foraminifera from coretop carbonate oxygen-18. Mar. Micropaleontol. 44, 125–140.

Seidov, D., Haupt, B.J., 2003. Freshwater teleconnections and oceanthermohaline circulation. Geophys. Res. Lett. 30 (6), 62-1–62-4.

Shackleton, N.J., 1974. Attainment of isotopic equilibrium betweenocean water and benthonic foraminifera genus Uvigerina: isotopicchanges in the ocean during the last glacial. Colloques Internationauxdu Centre National de la Recherche Scientifique, vol. 219. CNRS,Paris, pp. 203–209.

Tomczak, M., Godfrey, J.S., 2003. Regional Oceanography: AnIntroduction, 2nd Ed. Daya Publishing House, Delhi.

Wadley, M.R., Bigg, G.R., Rohling, E.J., Payne, A.J., 2002. Onmodelling present-day and last glacial maximum oceanic δ18Odistributions. Glob. Planet. Change 32, 89–109.

Wang, L., Pflaumann, U., Sarnthein, M., 1995. Paleo sea surfacesalinities in low latitude Atlantic: the δ18O record ofGlobigerinoidesruber (white). Paleoceanography 10 (4), 749–761.

Weldeab, S., Schneider, R.R., Kölling, M., 2006. Deglacial sea surfacetemperature and salinity increase in the western tropical Atlantic insynchrony with high latitude climate instabilities. Earth Planet. Sci.Lett. 241, 699–706.

Wolff, T., Grieger, B., Hale, W., Dürkoop, A., Mulitza, S., Pätzold, J.,Wefer, G., 1999. On the reconstruction of paleosalinities. In:Fischer, G., Wefer, G. (Eds.), Use of proxies in paleoceano-graphy: Examples from the South Atlantic. Springer-Verlag,Berlin, pp. 207–228.