Sea Surface Temperature Calculations in the Equatorial East Atlantic From

PALEOCEANOGRAPHY, VOL. 15, NO. 1,PAGES 124-134, FEBRUARY 2000

Paleo-seasurface temperature calculationsin the equatorial east Atlantic from Mg/Ca ratios in planktic foraminifera- A comparisonto seasurface temperature estimates from U37 ', oxygen isotopes,and foraminiferal transfer function

D. Niirnberg and A. Miiller GEOMAR ResearchCenter, Kiel, Germany

R. R. Schneider FachbereichGeowissenschaften, University of Bremen, Bremen, Germany

Abstract. We presenttwo -270 kyr paleo-seasurface temperature (SST) recordsfrom the EquatorialDivergence and the SouthEquatorial Current derived from Mg/Ca ratiosin the plankticforaminifer Globigerinoides sacculifer. The presentstudy suggests that the magnesiumsignature of G. sacculiferprovides a seasonalSST estimatefrom the upper -50 m of the watercolumn generated during upwelling in australlow-latitude fall/winter. Common to both down-core recordsis a glacial-interglacialamplitude of-3ø-3.5øC for the last climaticchanges and lower Holoceneand glacial oxygenisotope stage 2 temperaturescompared with interglacial stage 5.5 andglacial stage 6 temperatures,respectively. The comparisonto publishedSST estimatesfrom alkenones,oxygen isotopes, and foraminiferal transfer function from the samecore material pinpoints discrepancies and conformities between methods.

1. Introduction 1.2. Foraminiferal Mg/Ca Ratios as Paleotemperature Indicators 1.1. Paleo-Sea Surface Temperature Estimations Althoughvarious SSTu•7' approaches were developed to Becauseof the incompatibilityof methodsand their reconstructsea surface SSTu•' temperatures (SST), the glacial inconsistentresults, the developmentof additionalSST recordis especiallydebated. Ice corerecordsfrom equatorial proxiesseems useful since climate models largely rely on mountains[Thompson et al., 1995] and glacial snowline accurateestimations of the oceanpaleotemperature. The depression[Broecker and Denton, 1989] suggestterrestrial interestin magnesiumin biotic calcite for reconstructing temperaturesto be as muchas -5øC coolerthan during the paleowater temperatures is continuously growing, although Holocene.While Stute et al. [1995]inferred a 5øC temperature the viewson the processescontrolling its uptakeand drop for the Last GlacialMaximum from noble gas distribution,its susceptibilityto dissolution,and its measurementsin groundwater, the revisionof the groundwater applicabilityfor paleoreconstructions remain controversial. dataonly leadto a 1.9ø-2.5øCtemperature decline [Ballentine Many studiescould not revealany relationshipbetween andHall, 1998]. A large temperaturedecline of -4ø-5øCis magnesiumand temperature [e.g., Krinsley, 1960; Savinand supportedby corrallineSt/Ca records[Beck et al., 1992; Douglas, 1973]. Delaney et al. [1985] concludedthat Guildersonet al., 1994] and, recently,by temperatureadditional environmental parameters may affect the calcium reconstrutionsfrom planktic tropical •5•80 records [Curry and substitutionby magnesium. In contrast, other investigators Oppo,1997]. Theseresults contradict the reducedglacial suggesteda relationshipbetween water temperatureand temperaturedrop of only 2øC deducedfrom studieson marine magnesiumfrom the investigation of core top biogenic calcite faunalassemblages [Climate: Long-Range Investigation, [e.g.,Puechmaille, 1985; Izuka, 1988]. The application for Mapping,and Planning (CLIMAP), 1981] and stable oxygen SSTreconstructions is recently growing, partly stimulated by isotopes in foraminifera[Broecker, 1986]. Unsaturated cultureexperiments with planktic foraminifera showing a alkenonedata provide evidence for a 2ø-4øCtemperature temperature-related Mg/Ca shell ratio [Narnberg et al., 1996a, decline[Rostek et al., 1993;Sikes and Keigwin, 1994; Rosell- b]. Meld, 1998]. Cronbladand Malmgren [1981] were the firstto report down-coremagnesium variations in plankticforaminiferal Copyright2000 by the AmericanGeophysical Union. tests,correlating with late Quaternaryclimatic oscillations. Recently,Hastings et al. [1998]presented Mg/Ca-based SST Papernumber 1999PA000370. recordsfrom the equatorialAtlantic and the Caribbean 0883-8305/00/1999PA000370512.00 pointingout that the reliabilityof Mg/Ca recordsis not

124 125 NUERNBERGET AL.: Mg/Ca- DERIVED PALEO-SST perturbedby dissolution. Mashiotta et al. [1999] used a dissolution, which should be critically consideredwhen Mg/Ca-basedSST recordto reconstructchanges in global ice performing magnesiumanalyses. volume. Recently, culture experiments with living G. sacculifer In fact, calcite dissolution may significantly alter the revealeda significantmagnesium increase in the foraminiferal magnesiumconcentrations within foraminiferaltests and thus testsfor a 10øCincrease in watertemperature [Niirnberg et al., may prevent the applicability of magnesiumas a tracer for 1996a,b]. It could also be shownthat gametogeniccalcite has water mass properties [Rosenthaland Boyle, 1993]. Brown a significantlyhigher magnesiumconcentration than primary and Elderfield [1996] recently estimated the effect of calcite, althoughsecreted at the sametemperature. A sac-like magnesium content on the solubility of low-magnesium chamber, which may be built during the foraminifer's final biogenic calcite. They concludedthat the depth of the stageof development[Bd et al., 1983; Hemlebenet al., 1987], saturationhorizon for magnesiancalcites is a function of their however, does not show enhancedmagnesium concentrations magnesiumcontent. Lohmann [1995] showedin addition that comparedto a normal final chamber[Niirnberg et al., 1996a]. G. sacculifermay start losing shell masseven -2 km abovethe lysocline. Besides partial dissolution, other biases on the 2.2. Core Samples magnesiumsignal introducedby salinity and pH variations, Mg/Ca studieswere performedon two sedimentcores from amount of gametogenic calcite, and contaminant phases still the tropical Atlantic, for which major stratigraphical, needto be evaluatedbefore foraminiferalMg/Ca ratios may be sedimentological,and geochemicalinformation was available. acceptedas a paleothermometer. Sedimentcores GeoB 1105-3/4 from the Equatorial Divergence The objective of this study is to pursue further the and GeoB 1112-3/4 located -400 km south in the low investigation on the potential of magnesium for productiveSouth Equatorial Current (SEC) were recoveredfrom paleothermometry. In continuation to Niirnberg et al. the Guinea Basin from 3225 (1ø39.9'S, 12ø25.7'W) and [1996a, b], who showed a well-constrained Mg/SST 3122m water depth (5ø46.2'S, 10ø44.7'W), respectively calibrationcurve for primary calciteof cultivatedG. sacculifer, (Figure 1). Stable isotope geochemistry, foraminiferal this study comparesmagnesium-derived SST recordsover the assemblageanalyses and related SST reconstructions,fluxes of last -270 kyr to different paleo-SST proxies (foraminiferal sedimentarybiogenic components, grain size analyses, and transfer function, oxygen isotopes, and alkenone organiccarbon fluxes were previously publishedby Meinecke concentrations) on the same core material in order to resolve [ 1992], Bickert and Wefer [1996], Schneideret al. [ 1996], and discrepanciesand conformitiesbetween methods. Wefer et al. [1996]. A record of alkenone SST estimates is availableonly for core GeoB 1105-3/4 [Schneideret al., 1996] 2. Study Area, Material, and Methods because the content of alkenones was insufficient for the propercalculation of U3• in coreGeoB 1112. 2.1. Foraminiferal Species Selected for Analyses For both cores, stratigraphicinformation is available from the PANGAEA Paleoclimate Data Center (Alfred-Wegener- G. sacculifer inhabits shallow euphotic waters in the Institute,Bremerhaven). The age modelsare basedon graphic tropical and subtropical oceans [Shackleton and Vincent, correlationof Cibicidoideswuellerstorfi /5180 records to the 1978; Hernleben and Spindiet, 1983] because of the /5180standard record of Martinsonet al. [1987].We choosethe photosyntheticrequirements of their dinoflagellatesymbionts benthic record for stratigraphic correlation to minimize sea [Leeand Anderson, 1991]. Inferred from both abundancesand surfacetemperature effects that may occur betweenupwelling isotopic temperatures,the depth habitat lies at-25-75 m (Equatorial Divergence) and non upwelling regions (South [Fairbankset al. , 1980; Erez and Honjo, 1981; Fairbanks et Equatorial Current) and may include a time lag between al., 1982; Hemleben et al., 1989] with prevailing habitat planktic isotope signals. For detailed stratigraphic temperaturesof 20ø-30øC.For the equatorialAtlantic a depth information,see Bickert and Wefer [1996] and Schneideret al. habitatof-0-50 m is proposedby Erniliani [1954] and Hecht [1996]. The depth-age relationships for both cores [1971]. From plankton tow studies close to our area of demonstratenearly 2 times higher overall sedimentationrates investigation, Ravelo et al. [1990] concludethat highest (4.16 cm/kyr) in core GeoB 1105 comparedwith core GeoB abundancesof G. sacculiferoccur in -0-40 m. Salinities from 1112 (2.54 cm/kyr) due to the enhanced biogenic 24 to 47 are tolerated[Hemleben et al., 1989]. Reproductionis sedimentationunder the equatorialupwelling areaat site GeoB linked to the lunar cycle [Bijma and Hernleben,1994]. In order 1105 [Bickert and Wefer, 1996]. to reproduce,G. sacculifermoves far below the euphoticzone. In contrast to the western South Atlantic basins and the During gametogenesis a secondary layer of calcite, called Cape Basin, the deep-seaareas that are mainly filled by Lower gametogenic calcite, is secretedin 80-100 m [Bijma and CircumpolarDeep Water (LCDW) tindersaturatedwith respectto Hemleben,1994]. Accordingto modelresults, nearly a third of carbonateconcentration, the deepestparts of the GuineaBasin the populationeven adds gametogeniccalcite within the main are dominatedby slightly supersaturatedNorth Atlantic Deep thermoclineat 300-800 m water depth[Duplessy et al., 1981]. Water (NADW) [Broecker and Peng, 1982]. This causes the Growth rate and frequencyof gametogenesisare a function of lysoclineof calcite to stay much deepercompared to the other feeding and illumination [Bd, 1980; Spindler et al., 1984]. South Atlantic basins [Biscaye et al., 1976; Thunnel, 1982]. G. sacculifer commonly feeds on copepods[Spindler et al.. Delta CO32-values from the nearestGeochemical Ocean 1984] and prefers a low-nutrient oligotrophic environment SectionsStudy (GEOSECS) site 109 [Bainbridgeet al., 1981 ], [Bijma and Hernleben, 1994]. Accordingto the ranking of indicating the difference between the in situ carbonate ion Berger [1970], G. sacculifer is fairly susceptible to concentrationand the pressure-correctedsaturation carbonate NUERNBERGET AL.' Mg/Ca- DERIVED PALEO-SST 126

.40 ø .20 ø 0 o 10 ø 10 ø

o o

0 ø 0 ø

.5 o

-10 ø -10 ø

-15 ø -15 ø

10 ø 10 ø GD

o o

0 ø 0 ø

GeoBl105 .5 o .5 o GeoB1112 AG

-10 ø .10 ø AC

-15 ø -15 ø .40 ø .20 ø 0o Figure1. Areaof investigationschematically SSTu•,' indicating the horizontal distribution pattern of currentsfor thetropical surfacelayer at -0-100 m water depthfor (a) northernspring and (b) northernfall accordingto Strammaand Schott[1999]. Core locationsare indicatedby large dots.NEC, North Equatorial Current; NECC, North Equatorial Countercurrent;GD, Guinea Dome; GC, Guinea Current; SEC, SouthEquatorial Current with the northern (NSEC), equatorial (ESEC), central (CSEC), and southernbranches (SSEC); EUC, Equatorial Undercurrent;SEUC, SouthEquatorial Undercurrent; SECC, South EquatorialCountercurrent; NBC, North Brazil Current;GCUC, Gabbon-CongoUndercurrent; AG, Angola Gyre; AD, Angola Dome; AC, Angola Current.The hatchedarea indicates potential upwelling in the studyarea.

ion concentration(with respectto calcite),point towardwater cleaning step and, again, distilled deionized water rinses depthsof •-4800 m for the calcite saturationhorizon. Even followed,alternated with ultrasonicalcleaning. Subsequently, duringthe entirelate Pleistoceneinterval, the lysoclineis not sampleswere treated with a hot (80øC)oxidizing NaOH/H202 expectedto haverisen above •-3800 m waterdepth [Bickert and solution(30 ml 0.1 N NaOH (analyticalgrade); 10 pL 30% Wefer,1996], suggestingthat the coresare well aboveany H20: (supraput))for 30 min. Every 10 min, ultrasonical reasonablecalcite saturation horizon. The magnesiancalcite cleaningsteps (2 min) wereapplied. Afterward, three steps of horizoncalculated for Mg/Ca = 10 mmol/mol (the upper distilled deionizedwater rinsing andultrasonical treatment as limit of whatis appropriatefor planktic foraminiferalcalcite) above followed. The clean foraminiferal fragmentswere could have been as much as 150 m above the calcite saturation brought into new vials previouslycleaned with 1 N HCI horizon. (subboileddestilled) and subsequentlydissolved by 0.1 N HNO3 (subboileddestilled) during ultrasonic treatment. The 2.3. Preparation and Analytical Approaches samplesolution was then diluted with distilled deionizedwater Specimens of G. sacculifer were selected from the up to 3 mL. 250-500pm size fraction. The large size ensures that In order to assessthe efficiency and necessity of our specimenscan be easily identified and that a relatively small cleaning procedure we subjected sieved foraminifera number of specimens is sufficient for the geochemical (Globigerinabulloides and Orbulinauniversa, 250-500 pm) investigation. Specimens visibly contaminated by from core top samples off Portugal (39ø04'N, 10ø40'W; ferromanganeseoxides were discarded.Magnesium and calcium 1605 m waterdepth) to sequentiallymore rigorous cleaning. analyses were carried out by inductively coupled plasma Variousforaminiferal cleaning protocols described by Boyle optical emissionspectrometry (ICP-OES). [1981], Boyle and Keigwin [1985], Brown [1996] and Approximately 0.5-1.2 mg sample material usually Hastingset al. [1996]were adopted and/or modified. Cleaning consisting of •-30 specimens of G. sacculifer were gently step 1 comprisesthree rinses with destilled deionizedwater crushedbetween glass plates in orderto openthe chambersand alternatedwith sonication(2 min). Cleaningstep 2 subjects were subsequently placed into vials. In order to remove the foraminiferato three rinses with destilleddeionized water, contaminantphases the material was rinsed three times with one rinse with methanol,and two subsequentrinses with distilled deionized water with ultrasonical cleaning (2 min) destilleddeionized water, each step alternatedby sonication steps after each rinse. One methanol (subboiled destilled) (2 min). This step significantly reducedthe magnesium 127 NUERNBERGET AL.: Mg/Ca- DERIVED PALEO-SST contentsof G. bulloidesand O. universain comparisonto step glacial SST• and wouldreduce the glacial-interglacial 1. Further cleaning (step 3 as describedabove) with hot SST•otop•amplitude by maximumof 1øC. alkalineperoxide for 30 min with alternatingsonication after each10 min did not changethe magnesiumcontent within the 3. Results and Discussion measurementerror nor did step 4, which applies a solution of NHnC1instead of hot alkaline peroxide.Step 5 is similar to 3.1. Mg/Ca Signal in Planktic Foraminifera step3, but hasno methanolrinse. Finally, the most extensive From the Equatorial Divergence and cleaning,step 6, comprises,besides sonication and rinsing the South Equatorial Current with distilled dionized water and methanol, the treatment with NHnCLsolution and hot alkalineperoxide. All cleaningsteps The down-coreMg/Ca recordsof both Guinea Basin cores after step 2 did not changethe Mg/Ca ratios significantly. GeoB 1105 and 1112 cover the last --270,000 years(oxygen However,since we expectstronger contamination in the down- isotopestages 1-8), thus comprisingthree glacial-interglacial coresamples compared to coretops,we systematicallyapplied changes.In both cores, Mg/Ca ratios range between2.6 and cleaning step 3. 4.4 mmol/molat most(Tables A1 andTable A2•). The long- Analyseswere run on an ICP-OES(ISA Jobin Yion-Spex term variability in Mg/Ca ratios by far exceedsthe standard InstrumentsS.A. GmbH) with polychromatorapplying yttrium error of the measurements and thus has to be considered as real. asan internalstandard. We selectedelement lines for analyses Although the short-term scatter of the Mg/Ca ratios is also which appeared most intensive and undisturbed significantfrom the analyticalpoint of view, its paleothermal (Ca:317.93nm; Mg: 279.55nm; and Y: 371.03 nm). relevance is questioned.It is likely that these small-scale Elementdetection was performedwith photomultipliers,the variations are dueto the intraspecificinhomogeneity of the hightension of which was adapted to each element magnesiumincorporation. Narnberg [1995] already statedthat concentration range. The relative standarddeviation is < 1% the large rangeof magnesiumconcentrations within a single for magnesiumand calcite. The Mg/Ca reproducibilityof 15 foraminiferaltest, which may exceeda factorof 3 along replicatesamples of G. sacculiferis good(? = 0.98, mean chamberwalls, may lead to quantitativedifferences between standarddeviation is 0.04 mmol/mol, and maximum standard differenttests of one sample.Replicate ICP-OES analyses deviation is 0.11 mmol/mol). The error in terms of SST is varying on averageby --0.05 mmol/mol contribute to the maximum of + 0.4øC. scatterof the magnesiumdata down-core.In order to exclude The accuracyof ourcalcium and magnesium analyses can be the short-term scatter of the data from further discussion we evaluatedwith literaturedata. Hastings et al. [ 1998] published applied a simple seven-point smoothing procedure(least inductivelycoupled plasma mass spectrometry (ICP-MS) and squaressmoothing filter) [Paillardet al., 1996]on the original atomicabsorption spectrometry (AAS) derivedMg/Ca ratiosof Mg/Ca records(Figures 2 and 3). G. sacculifer selected from an equatorial Atlantic sediment The comparisonof Mg/Ca ratios in both coresreveal that core, the glacial-interglacialrange of which lies between --2.9 the Mg/Ca ratiosin the EquatorialDivergence (GeoB 1105) are and 3.9 mmol/mol, thus being consistentwith our ICP-OES- lower comparedto the SouthEquatorial Current (core GeoB based data (--2.9-4.0 retool/tool). 1112). This observation coincides with the observation of generallyheavier stable oxygen isotopevalues in coreGeoB 2.4. SST From Oxygen Isotopes 1105 [Meinecke, 1992] (Figures2 and 3). The down-core variationsof both Mg/Ca ratiosand oxygenisotopes match For core GEOB 1112 we calculated SST from the available fairly well, thereby clearly reflecting glacial-interglacial •5•80record of G. sacculiferby applyingthe Erezand Luz changes.Highest Mg/Ca ratiosoccur during interglacials, and [1983] temperatureequation: lowest values occur during glacials with pronounced T = 17.0- 4.52(•80•- •80,,) + 0.03(•80•- •O,,) 2, (1) concentrationgradients at stageboundaries 8/7, 6/5, and 2/1. Spectral analysis [Paillard et al., 1996] reveals that the where•180•is the oxygen isotopic composition of thecalcite Mg/Ca variability in core GeoB 1112 is dominatedby (permillePeedee belemnite (PDB) and •5•80•v is the oxygen Milankovitch frequencieswith periodsclose to 100, 41, and isotopic composition of seawater (permille PDB; the 23 kyr. The lengthof time seriesof--270,000 years, in fact, is conversionfrom •5•80•v SMOW to •5•80•vPDB is performed too short to definitely resolve for variability in the accordingto Hut [1987]. The •5•80•vis assessedfrom the eccentricityband (100-kyr cycle). The cross-spectralanalysis salinityversus •5•80•v relationship established by Wanget al. [Blackmanand Tukey,1958] of theMg/Ca and the •5•SO records [1995] for the low-latitude Atlantic. Assuming that derivedfrom G. sacculiferfrom the samesamples indicates G. sacculiferinhabits the upper-100 m of the watercolumn, that the two time seriescorrelate (R =-0.54) with Mg/Ca we consideredsalinities of 35.5-36.1 at 100-0 m waterdepth leadingthe •5 lSO signal. The comparison of phase angles of [Levitusand Boyer, 1994] leadingto a rangeof •5180•vvalues. Mg/Caand •5 lsO reveals that changes in Mg/Ca lead changes in Glacial-interglacialvariations of seawater•5180•v, which are •80 by 12.0+2.1kyrat the 100 kyr orbitalperiod, by drivenby ice volumechanges are taken from the mean•5•80•v 5.5+1.0kyr at the41 kyr orbitalperiod, and by 2.6+0.7 kyr recordof Vogelsang[1990]. Vogelsang[1990] proposesa meanocean •5180•v increase by 1.1%oduring the LastGlacial Maximum,which is in accordancewith pore water •5•O •SupportingTables AI andA2 are available electronically atWorld Data measurementsin sedimentcores from the equatorialAtlantic Center-Afor Paleoclimatology, NOAA/NGDC, 325 Broadway, Boulder, [Schraget al., 1996]. Assuminga largermean ocean •5•sO•v Colorado (e-mail: paleo•mail.ngdc.noaa.gov; URL: http:// increaseof 1.2-1.3%o[Fairbanks, 1989] would result in warmer www.ngdc.noaa.gov/paleo). NUERNBERGET AL.: Mg/Ca- DERIVEDPALEO-SST 128

GeoB 1112 South5o46.:-sEquatorial •0o4•.7-w, Current smm(SEC) water depth Mg/Ca SST (øC) (mrnol/mol) SST (øC) 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 20 22 24 26 28 16 18 20 22 24 26 28

...

30'

ß 60'

90' Isotopic 120- temp. 150.

180

210-

240- ;ST std. dev. ß dev. 270 1 0 -1 SSTMg/Ca SSTMg/Ca --*--- fi180 (%0PDB) C.sacculifer

Figure2. (left) Down-coreMg/Ca record of coreGeoB 1112-3/4 for thelast -270,000 years in comparisonto the stable oxygenisotope curve (G. sacculifer)[Wefer et al., 1996].Replicate Mg/Ca analyses are indicatedby squares.The smoothed recordis includedas a thickline. (middle) SST temperature curve from oxygen isotopes in comparisonto the smoothedSSTMedC a record.The broad band of SSTIsotopegives reference todifferent salinities applied for calculating 15•8Ow. (right) Calculated $ST of coreGeoB 1112 versus age in comparisonto $$T derivedfrom foraminiferal transfer functions [Wefer et al., 1996]. Mg/Ca

GeoB 1105 Equatorial1ø39'9'S12ø25'7'W'3325 Divergence mwater depth Mg]Ca (mmol/mol) SST (øC) SST (øC) 2.6 2.8 3.0 3.2 3.4 3.6 23 24 25 26 27 28 15 17 19 21 23 25 27

30'

90 •,..,•120 < •50.

180'

Mg/Ca 210 $td.dev.

240 o

270 2.0 1.5 1.0 0.5 0.0 '0.5 SSTMv/C:• ---*-- •80 (%0PDB)a. ruber

Figure 3. (left) Down-core Mg/Ca record of core GeoB 1105-3/4 for the last -270,000 years in comparisonto the stable oxygenisotope curve (G. ruber) [Weferet al., 1996].Replicate Mg/Ca analysesare indicatedby squares.The smoothedrecord is includedas a thickline. (middle) SST temperature curve from alkenones in comparison to the smoothed SSTMg/C a record. (right)Calculated $$T of coreGeoB 1105versus age in comparisonto $$T derivedfrom foraminiferaltransfer functions [Weferet al., 1996]. MgtCa 129 NUERNBERGET AL.: Mg/Ca- DERIVED PALEO-SST at the 23 kyr orbital period. The coherencyof the two time indicatedas SSTMg/Ca),which Mg/SST relationship is applied. seriesis 0.87 at 1/!00 kyr, 0.82 at 1/41 kyr, and0.78 at 1/23 Up to now, only a few species-specificMg/SST calibration kyr. All thesecoherencies exceed the 95% confidenceinterval curvesfrom cultivatingexperiments exist. For primary calcite for coherency(= 0.62). The correlationcoefficient of the two of G. sacculifer,Niirnberg et al. [1996a, b] suggesteda strict time seriesimproves to R =-0.71, whenassuming a time lag relationship between the foraminiferal Mg/Ca ratios and of 5500 years(at the 41 kyr orbital period)between both the temperature.They chosean exponentialmodel to fit their data, equallysampled and smoothed Mg/Ca and • 180records. althoughstudies on the partition coefficient of magnesiumin In coreGeoB 1112 the stage8/7 boundary(Termination III) magnesiancalcite overgrowthsare still ambivalent, showing exhibits a pronouncedMg/Ca increase from -3.2 to both linear and exponentialrelationships between magnesium -3.9 mmol/mol. At stage 6/5 boundary (Termination II), and temperature [Chilingar, 1962; Kinsman and Holland, Mg/Ca ratiosincrease from 3.1 to up to 3.9 mmol/mol, andat 1969; Katz, 1973; Fiichtbauer and Hardie, 1976; Mucci, 1987; stage1/2 boundary(Termination I)they increasefrom 3.0 to Oomori et al., 1987]. Since the foraminiferal Mg/Ca ratios are 3.8 mmol/mol. The overall glacial-interglacial amplitudeof orders of magnitude lower compared to inorganically the smoothedMg/Ca recordsis -0.7-0.8 mmol/mol. precipitated calcite, it was suggestedthat the foraminifera are For core GeoB 1105 we observea Mg/Ca ratio increasefrom capable of physiologically controlling the magnesium -2.7 to 3.2 mmol/mol at stage boundary8/7, from 2.8 to concentrationin their shells [Niirnberg et al., 1996a]. Such 3.5 mmol/mol at stage boundary 6/5, and from 2.6 to physiological control is apparently the dominant driving 3.4 mmol/mol at stageboundary 2/1, thus showinga glacial- force, althoughtemperature seems to affect theseprocesses the interglacial amplitude of 0.5-0.8 mmol/mol, which is way that a direct magnesium-temperaturerelationship is comparableto the valuesin coreGeoB 1112. pretended. The slight deviation of the relationship applied here from 3.2. SST Information From Foraminiferal Mg/Ca the one originally proposedby Niirnberget al. [1996b] stems Ratios in Comparison to Recent Conditions from the correction made to adjust the microprobe It is of considerable importance, for the paleo-SST measurementsof Niirnberg et al. [1996b] to the ICP-OES data reconstruction from down-core Mg/Ca records (hereafter of this study:

Table 1. CoreTop SST Estimatesand Glacial-Interglacial SST AmplitudesFrom Various Proxies

SST SST Method GeoB1105, øC GeoBl112, øC Period

Levitus [1994] 26.2-20.4 26.5-23.2 May-June Levitus [1994] 24.8-19.9 25.1-22.8 May-Aug. Levitus [1994] 26.2-22.9 25.7-23.3 Sept.-April Levitus [1994] 26.2-22.3 25.5-23.3 annual

Mg/Ca•moo• 24.1 26.1 U•'37 25,7 n.d. TF,o,d 23.5 25.4 TF•m 26.1 27.3 (5•0•. n.d. 24.7 8•Om•. n.d. 25.5

SST Amplitude,GeoB 1105, øC SST Amplitude,GeoB 1112, øC

Method MIS 2 / MIS 1 MIS 6 / MIS 5 MIS 8 / MIS 7 MIS 2 / MIS 1 MIS 6 / MIS 5 MIS 8 / MIS 7

Mg/Ca•moo• 3,4 2.9 2.9 2.9 2.9 1.9 U•'37 3.5 3.1 3.8 n.d. n.d. n.d. TF,•d 7.2 6.1 6.5 8.0 7.9 6.3 TF,,•m 5.2 3.9 3.7 4.9 5,1 2.7 •5•gO•. 5.1 7,7 6,5 •5•gOm•. 5,1 7,7 6,5

(top) SST estimatesare for coresGeoB 1105and GeoB 1112in comparisonto modemLevitus and Boyer (1994) watertemperatures. The Levitus andBoyer (1994) temperaturesof theupper 50 m of thewater column are given for thebeginning equatorial upwelling (May to June),the upwelling period(May to August),the remainingyear (September to April), andthe entireyear. (bottom) Glacial-interglacial SST amplitudesof variousSST proxiesare comparedfor TerminationI, (MIS 2/1), II, (MIS 6/5), and III, (MIS 8/7), differentiatedfor both cores.Here n.d. is no data. NUERNBERGET AL.: Mg/Ca- DERIVEDPALEO-SST 130

T = (logMg/Ca - log0.491)/0.033 R2 = 0.92. (2) [Oberhiinsliet al., 1992]. Ufkes et al. [1998] report high abundancesduring austral low-latitude spring (October to The uppermostHolocene samplesin cores GeoB 1112 and November), when a deep mixed layer has establishedin the GeoB 1105 reveal a-2øC offset in SSTM•cawith lower eastern South Atlantic. temperaturesat core GeoB 1105 (Table 1), which generally Taken together, the-2øC differencebetween the two core persistsover the entire time period investigated.Today, only sitesand the presumedabundance maximum during austral low- during equatorial upwelling from May to August/beginning latitudefall/winter suggest that at site GeoB1105 the SSTMg/Ca September,SST in the EquatorialDivergence are considerably signalreflects the thermalsituation within the upper -50 m of cooler than in the South Equatorial Current at 0-50 m water the water columnduring upwelling (May to August),while site depth [Levitusand Boyer, 1994] (Figure 4). GeoB 1112 is not affected by upwelled waters. The latest Unfortunately, information about abundancemaxima of the holoceneSSTMg•c aof -26 ø +0.4øC and24 ø +0.4øC at core oligotrophic G. sacculifer during that time is sparse. sites GeoB 1112 and GeoB 1105, respectively, fairly well B. Donner (personalcommunication, 1998) gathereda long- reflectthe upwellingsituation [Levitus and Boyer, 1994]. The term abundancerecord for planktic foraminifera from the errorin SST•ecacalculations to lowervalues, which may raise equatorial East Atlantic revealing an abundance peak of from the presenceof varying portions of gametogeniccalcite G. sacculiferduring the beginningupwelling from australlow- enrichedin magnesium[Niirnberg et al., 1996b], may provide latitude late fall/early winter. During summer, instead, no enough amount of scope to even explain the rapid SST drop, G. sacculifer were observed in the eastern South Atlantic which takes place during the ongoing upwelling in July and lastsuntil August [Levitusand Boyer, 1994] (Figure 4). From September on when upwelling ceases, SST slowly July/ May/ increasesagain in both areas by -1-3øC, and surfacemixed August June T (øC) layer temperaturesin the Equatorial Divergence never fall 13 15 17 19 below the ones observedwithin the South Equatorial Current 0 i . i . i _ i (Figure 4). We thereforeconclude for the area of investigation that the SSTuvcasignal reflects the austral low-latitude 30 fall/winter upwelling situationwithin the uppermost-50 m of the water column (the assumedhabitat of pregametogenic G. sacculifer). 60 . Whether salinity changesperturb the magnesiumsignal is I still a matter of debate. From live culturing of OrbMina 90 I universa, Lea et aL [1999] infer an increase of 4 _+3% in Mg/Ca per salinityunit at 22øC. For G. sacculifer,Narnberg et •120 UpwellingI Mayto August I al. [1996a] pointed out that pronounced salinity changes (greaterthan -10) apparently dominate over the temperature II(a) 150 effect with respect to the magnesiumuptake, possibly caused i SouthEquatorial Current (SEC) by an enhanced metabolic activity at high salinity levels...... EquatorialDivergence Small-scale salinity differences below 3, instead, are not 13 15 17 19 21 23 25 27 resolvedby the Mg/Ca ratios [Niirnberget al., 1996a]. Since, 0 I ß I ß I ß I . I ß I ß I ...... •..,:.•:•,....-• ...... •:?:...... • ...... • ...... ,...... •:..,• ...... •..• at the study sites, annual salinity variations of-35.1-36.1 [Levitus and Boyer, 1994] within the upper 100 m water ao column (which is the assumeddepth habitat of G. sacculifer) give rise to Mg/Ca ratio variations of-0.25 mmol/mol at '• 60 most, which is close to the analytical error, we do not expect salinity to have any pronouncedeffect on the foraminifera/ Mg/Ca ratio. • 90 oo-øø I 3.3. Paleo-SST Information From Foraminiferal •120 SeptembertoApril Mg/Ca Ratios The paleo-SSTu•carecords of both coresresemble each 150•'/ SeptembertoApril i(b) other, although the SSTu•caestimates of coreGeoB 1105 are Figure 4. Watertemperature data for the EquatorialDivergence (thin generally lower in comparisonto core GeoB 1112. It is dashed lines) and the South Equatorial Current (thick solid lines) averagedfor (a) the upwellingseason (May to August)and (b) the apparentthat both SSTMg•c arecords are -1ø-2øC higher than the remainingyear [Levitusand Boyer,1994]. For the upwellingseason we correspondingpaleosummer SST• recordsof Wefer et al. differentiatedbetween the beginning(May to June) and the remaining [1996] derivedfrom planktic foraminifera/ transfer functions upwelling(July to August).The shadedarea marks the presumedhabitat (Figures2 and3). The overallrange of thedown-core SSTu,•c ais of pre-gametogenicG. sacculifer(upper 50 m of the water column). smaller (3.6 ø and 4.0øC in cores GeoB 1112 and GeoB 1 105, Time and depth at which the EquatorialDivergence is significantly coolerthan the SouthEquatorial Current and at which the Mg/Ca signal respectively)than the maximumrange observed from transfer is presumablybeing generated are markedby the hatchedarea. The functions, which even exceeds 8.0øC. In fact, the glacial- arrow pointsto considerablesurface water coolingduring ongoing interglacialSSTMg/C a amplitudes at stage8/7, 6/5, and 2/1 upwellingin australlow-latitude fall/winter. boundaries are-3øC for each of the boundaries (Table 1), 131 NUERNBERGET AL.: Mg/Ca- DERIVEDPALEO-SST

28 I 8 ' Entiredata set 27

• 26

[• 25

a) 22

27 GeoB 1112 SSTMg/Ca

r• 26 o

• 25

t$3 24

23 GeoB 1105 SSTMg/ca b) 22 0 30 60 90 120 150 180 210 240 270 Age (kyr) Figure5. (a) Comparisonof SSTMyca and SSTu•; from core GeoB 1105 and (b) comparisonof SSTMg/C a for bothcores investigated.The hatched area marks the timeperiod for whicha goodcorrelation between SSTMg/c a and SSTu•'exists (R = 0.78). Marine oxygenisotope stages are indicated.

whereasthe SSTvFamplitudes are larger and often exceed5øC. SSTMg/Ca and warm SSTvF. In pre-Holocenesections, SSTisotop e The magnitudeof the glacial-interglacialSSTMg/c a gradient reconstructionssignificantly deviate in amplitude from the correspondsto that of Hastings et al. [1998], who describeda warmSSTTr data, although the SSTi.•otop• amplitude better fits t o temperaturedrop of-2.5øC during glacial stage 2 in the warmSSTTr than to SSTMyca.The temporalamplitude of the equatorialAtlantic based on SSTMg/ca.Studies of VanCarnpo et SSTM•ais considerablysmaller than for SSTisotope (Table 1). It al. [1990], who pointedout that African sea level temperatures needsto be pointedout that in contrastto SSTM•athe accuracy were 3ø-4øClower during the Last Glacial Maximum, further of SST•.•otop•suffers considerably from assumptionsregarding supportthe magnesium-derivedSST reconstructions. theice effect,the recent, and the paleo-5•SO Wvalues. Since the Both the smaller amplitude and the generally higher magnesiumsignal in G. sacculiferdefinitely leads the oxygen temperatureestimates may result from the fact that the SSTMr./caisotope signal and a considerableportion of the 5•sOsignal signal reflects seasonal temperatureconditions from the reflectsglobal ice volumechanges, we concludethat changes uppermostwater column, whereas the SST,r integratessummer in $ST precedechanges in globalice volumeat this site. temperatures from the surface waters down to the $$T estimates obtained from alkenone concentrations subthermocline when considering the entire planktic (SSTu•,') establishedfor core GeoB l105 [Schneideret al., foraminiferal faunal spectrum [Meinecke, 1992]. Also, 1996] exceedthe SSTM•a by -lø-2øC, while for mostof the foraminiferal abundancesin this specific area might be related time spancovered, the temporalpattern of both temperature to vertical water column hydrographyand nutrient distribution reconstructionsmatch well (Figure 5). The overall correlation rather than to sea surface temperature alone [Ravelo and betweenboth the smoothedSSTM•/c a time seriesand the Fairbanks, 1992; Sikes and Keigwin, 1994]. smoothedSSTu•; time seriesis R = 0.49. The correlationfor Holocene SST estimates derived from oxygen isotope data samplesyounger than 90,000 years, for which a good visual of G. sacculifer(SSTisotope) from coreGEOB 1112 rangefrom correlationexists, improves to R = 0.78. The overall glacial- -23ø-25.5øC (Figure 2) and thus are considerably lower than interglacialamplitudes at TerminationsI andII rangefrom 2.9ø NUE•ERG ET AL.: Mg/Ca- DERIVED PALEO-SST 132 to 3.5øCin both the SSTMg/ca and SSTu}'records, while at functionsare furtherexpressed by the fact that both chemically TerminationIII the amplitudeof the SSTMgrais only -2.9øC derived temperature reconstructions reveal (1) higher comparedwith the 3.8øC amplitude of theSSTug/c arecord (Table interglacial stage 5e temperaturesin comparisonwith the 1). Major discrepanciescan be observedin oxygen isotope Holocene and (2) colder glacial stage 2 conditions in stage 7 and basal stage 6. While the SSTu•i record clearly comparison with glacial stage 6. Similar conditions are indicateswarming of surfacewaters, the SSTuvcastays low. reportedfrom SSTu[' reconstructionsof the northernNorth Also, in substage5.3 the SSTug'record indicates a significant Atlantic [Villanueva et al., 1998] and the Indian Ocean [Rostek warming, which is not reflected by the magnesiumsignal. et al., 1993]. The temperaturerecords derived from transfer For coreGeoB 1112 the SST•r•ca estimates fall into the functions, instead, reveal warmer Holocene and stage 2 range of the SSTu}' published for core GeoB 1105, being temperaturescompared to stages5e and 6 conditions. slightly higher (maximum of 0.5øC) duringpeak interglacials (Figure5). The glacial-interglacialSST•r•c aamplitude of-3øC 4. Conclusions is comparableto the SSTu•( amplitudeof core GeoB 1 105. Similar to core GeoB 1105, discrepanciescan be observed Since the disparity between terrestrial and marine SST duringlate stage7 andbasal stage6, wherethe SSTu,•;record estimatesis most obvious in tropical regions, this study definitelyshows sea surface warming, while the SSTMgrastay attemptsto contributenew aspectsby applying Mg/Ca ratios cool. Also, substage5.1, which exhibits warm SSTu,•(,is not in the planktic foraminifer G. sacculifer as a tool for reflectedin SSTMg/ca. estimating SST in the past to two sedimentcores from the Both the absolutetemperature offset and the discrepancies easternequatorial Atlantic covering the last -270 kyr. The betweenthe down-corerecords of SSTu•dand SST•, in the down-coreMg/Ca variationswere calibratedusing the species- EquatorialDivergence are most presumablyexplained by the specific, exponential Mg/SST relationship establishedfor time of signal generationand the different depth habitats of primary calcite of G. sacculifer[Niirnberg et al., 1996a]. SST the U 37•' signalproducing coccolithophorids on the one hand estimationsreveal an accuracyof approximately+0.4øC. and foraminifera on the other hand. Conte et al. [1993], Sikes Holocene SST estimates amount to -26øC for the South and Keigwin [1994], Herbert et al. [1998], and Miiller et al. EquatorialCurrent and---24øC for the EquatorialDivergence

[1998] consistentlyconcluded that the sedimentaryU •'37 ratio suggestingthat the chemicalsignature must have generated essentially reflects annual mean mixed layer temperatures duringupwelling in australlow-latitude fall/winter in -0-50 m

(-0-10 m waterdepth). In fact, the U •'37 temperaturecalculated water depth. Within the South Equatorial Current, pre- for the uppermost sample in core GeoB l105 of 25.7øC Holocene SST vary between ---24øC and 27.5øC at most, matchesthe annualmean temperature of the uppermost--- 10 m whereasthe coolerEquatorial Divergence exhibits SST of only of the water column of 25.6øC [Levitus and Boyer, 1994]. 22ø-25.5øC for the last 270,000 years. Common to both Seasonal variations of 24ø-27.5øC observed in this area are recordsis a glacial-interglacialamplitude of -3-3.5øC for the apparentlyaveraged out. The SST•g/c, of-24øC at thesame site, last climatic changes, thus being consistent t o instead, reflects the austral low-latitude fall/winter signal paleotemperature estimates derived from unsaturated beneaththe mixedlayer downto ---50m waterdepth [Levitus alkenones.Furthermore, both recordsimply lower Holocene and Boyer, 1994]. andglacial oxygenisotope stage2 temperaturescompared to The prevailing temperatureoffset of-lø-2øC through time interglacial stage 5.5 and glacial stage 6 temperatures, betweenEquatorial Divergence and South EquatorialCurrent respectively. (Figure 5) points to continuous upwelling within the Thecomparison of SSTMg/C ato otherSST proxies underlines EquatorialDivergence, comparable to the recent situation. The the applicability of foraminiferal Mg/Ca ratios for increasingtemperature offset observed in stage7, stage6, and paleothermalreconstructions and thus may reinforce the substage5.3 maybe interpretedas the thermaldecoupling of a discussionon SSTreconstructions. Our study emphasizes the warm thin surfacelayer from cooler water massesbeneath and need for caution when applying single paleotemperature thusa strongeror longerperiod of stratification.In contrast,a techniquesto reconstructclimate dynamicsand stressesthe decreasingtemperature offset mainly observedduring the cold importanceof a multiproxyapproach. eventsof glacial and interglacial stageswould propose less strong seasonaldifferences in the upper -50 m of the water Acknowledgements.For intensivediscussion and improvementof column. the manuscript we thank Jelle Bijma, Ralf Tiedemann, and ChristineC. NiJmberg.For critical reviews we aregrateful to E. Rohling Thecorrespondance of the SSTMycaand SSTu,•' records and and two anonymousreviewers. This study was fundedby the German their disparity to temperaturerecords derived from transfer ScienceFoundation (DFG) grantNu60/4-1/2.

References

Bainbridge,A. E., W. S. Broecker,D. Spencer,and re-examined,Mineral. Mag., 62A, 100-102, sacculifer(Brady), Micropaleontology, 29(3), H. Harmon, GEOSECS Atlantic Ocean 1998. 310-325, 1983. Expedition,vol. 1, Hydrographic Data, 1972- B6, A.W.H., Gametogenic calcification in a Beck,J. W., R.L. Edwards,E. Ito, F. W. Taylor, 1973, 121 pp., Natl. Sci. Found., Washington, spinoseplanktonic foraminifer, Globigerinoides J. R6cy, F. Rougerie,P. Joannot,and C. Henin, D.C., 1981. sacculifer(Brady), Mar. Micropaleontol., 5, Sea-surfacetemperature from coral skeletal Ballentine, C.J., and C.M. Hall, A non-linear 283-310, 1980. strontium/calciumratios, Science, 257, 644-647, inverse technique for calculating B6, A.W.H., O.R. Anderson, W.W. Faber Jr., 1992. palaeotemperatureand other variables using and D. A. Caron, Sequenceof morphological Berger, W.H., Planktonic foraminifera: Selective noble gas concentrationsin groundwater:The andcytoplasmic changes during gametogenesis dissolutionand the lysocline, Mar. Geol., 8, LastGlacial Maximum in the continentaltropics in the planktonicforaminifer Globigerinoides 111-138, 1970. 133 NUERNBERGET AL.: Mg/Ca- DERIVED PALEO-SST

Bickert, T., and G. Wefer, Late Quaternarydeep Erez, J., and S. Honjo, Comparison of isotopic foraminifera,Micropaleontology, 63(3), 297- water circulation in the South Atlantic: composition of planktonic foraminifera in 300, 1960. Reconstruction from carbonate dissolution and plankton tows, sediment traps and sediments, Lea, D.W., T.A. Mashiotta, and H.J. Spero, benthic stable isotopes, in The South Atlantic: Palaeogeogr. PalaeoclimatoL PalaeoecoL, 33, Controlson magnesiumand strontiumuptake in Present and Past Circulation, edited by G. 129-156, 1981. planktonic foraminifera determined by live Wefer et al., pp. 599-620, Springer-Verlag,New Erez, J., and B. Luz, Experimentalpaleo-temperature culturing, Geochim. Cosmochim. Acta, 63, York, 1996. equation for planktonic foraminifera, Geochim. 2369-2379, 1999. Bijma, J., and C. Hemleben, Population dynamics Cosmochim. Acta, 47, 1025-1031, 1983. Lee, J.J., and O.R. Anderson, Biology of of the planktic foraminifer Globigerinoides Fairbanks,R. G., A 17,000-year glacio-eustaticsea foraminifera,Academic, San Diego, Calif., 1991. sacculifer (Brady) from the central Red Sea, level record:Influence of glacialmelting rates o n Levitus, S., and T. Boyer, World Ocean Atlas 1994 Deep Sea Res., Part L 41, 485-510, 1994. the Younger Dryas event and deep-ocean vol. 4, Temperature,NOAA Atlas NESDIS 4, Biscaye, P.E., V. Kolla, and K.K. Turekian, circulation, Nature, 342, 637-642, 1989. Natl. Ocean.and Atmos. Admin., Silver Spring, Distribution of calcium carbonate in surface Fairbanks, R.G., P. H. Wiebe, and A.W.H. B6, Md., 1994. sedimentsof the Atlantic Ocean, J. Geophys. Vertical distribution and isotopic composition Lohmann, G.P., A model for variation in the Res., 81, 2595-2603, 1976. of living planktonic foraminiferain the western chemistry of planktonic foraminifera due to Blackman, R.B., and J.W. Tukey, The North Atlantic, Science, 207, 61-63, 1980. secondary calcification and selective Measurementof Power Spectra From the Point Fairbanks, R.G., M. Sverdlove, R. Free, dissolution, Paleoceanography, 10(3), 445- of View of CommunicationEngineering, 190 P. H. Wiebe, and A.W.H. B6, Vertical 457, 1995. pp., Dover Publications,Mineola, N.Y., 1958. distribution and isotopic fractionationof living Martinson,D. G., N. G. Pisias,J. D. Hays, J. Imbrie, Boyle, E. A., Cadmium,zink, copperand bariumin planktonic foraminiferafrom the PanamaBasin, T. C. Moore, and N.J. Shackleton, Age dating foraminifera tests, Earth Planet. Sci. Lett., 53, Nature, 298, 841-844, 1982. and the orbital theory of the ice ages: 11-35, 1981. Fiichtbauer,H., and L.A. Hardie, Experimentally Developmentof a high-resolution0 to 300,000 Boyle, E.A., and L.D. Keigwin, Comparison of determined homogeneous distribution years chronostratigraphy,Quat. Res., 27, 1-27, Atlanticand Pacific paleochemicalrecords for the coefficientsfor precipitatedmagnesium calcites, 1987. last 215,000 years: Changes in deep ocean Abstr. Annu. Prog. Meet. Geol. Soc. Amer., 8, Mashiotta, T.A., D.W. Lea, and H.J. $pero, circulation and chemical inventories, Earth 877, 1976. Glacial-interglacialchanges in Subantarcticsea Planet. Sci. Lett., 76, 135-150, 1985. Guilderson, T.P., R.G. Fairbanks, and surface temperatureand •80-water using Broecker,W.S., Oxygen isotope constraintson J. L. Rubenstone, Tropical temperature foraminiferalMg, Earth Planet. Sci. Lett., 170, surfaceocean temperatures, Quat. Res., 26, 121- variationssince 20,000 years ago: Modulating 417-432, 1999. 134, 1986. interhemisphericclimate change, Science,263, Meinecke, G., Spiitquartiire Oberfliichenwasser- Broecker, W.S., and G. H. Denton, The role of 663-665, 1994. temperaturenim 0stlichen iiquatorialenAtlantik, ocean-atmospherereorganizations in glacial Hastings, D.W., S. Emerson, J. Erez, and Bet. Fachber. Geowiss. 29, 181 pp., Univ. cycles, Geochim. Cosmochim. Acta, 53, 2465- B. K. Nelson, Vanadium in foraminiferal calcite: Bremen,Bremen, Germany, 1992. 2501, 1989. Evaluation of a method to determine paleo Mucci, A., Chemistry of low-temperatureabiotic Broecker,W. S., and T. H. Peng, Tracersin the Sea, seawater vanadium concentrations, Geochim. calcites:Experimental studies on coprecipitation, 690 pp., Lahmont-Doherty Earth Obs., Cosmochim. Acta, 60, 3701-3715, 1996. stability,and fractionation,Rev. Aquat. Sci., 3, Palisades, N.Y., 1982. Hastings,D. W., A. D. Russell, and S. R. Emerson, 217-254, 1987. Brown, S. J., Controls on the tracemetal chemistry Foraminiferalmagnesium in G. sacculifer as a Miiller, P. J., G. Kirst, G. Ruhland, I.V. Storch, and of foraminiferalcalcite and aragonite, Ph.D. paleotemperatureproxy in the equatorialAtlantic A. Rosell-Me16, Calibration of the alkenone thesis, 231 pp., Univ. of Cambridge, and Carribean surface oceans, paleotemperatureindex U•; basedon core-tops Cambridge,England, 1996. Paleoceanography,13(2), 161-169, 1998. from the easternSouth Atlantic and the global Brown, S. J., and H. Elderfield,Variations in Mg/Ca Hecht, A.D., Morphological variation, diversity ocean (60øN-60øS), Geochim. Cosmochim.Acta, and Sr/Ca ratios of planktonic foraminifera and stable isotope geochemistry of Recent 62, 1757-1772, 1998. caused by postdepositional dissolution: planktonicforaminifera from the North Atlantic, N'tirnberg, D., Magnesium in tests of Evidenceof shallowMg-dependant dissolution, Ph.D. thesis, 222 pp., Case Western Reserve Neogloboquadrina pachyderma sinistral from Paleoceanography,11(5), 543-551, 1996. Univ., Cleveland, Ohio, 1971. high northern and southern latitudes, J. Chilingar, G.V., Dependanceon temperatureof Hemleben,C., and M. Spindler,Recent advancesin Foraminiferal Res., 25, 350-368, 1995. Ca/Mg ratio of skeletal structuresof organisms research on living planktonic foraminifera, Niirnberg, D., J. Bijma, and C. Hemleben, and direct chemicalprecipitates out of seawater, Micropaleontology, 21, 334-341, 1983. Assessing the reliability of magnesium in Bull. South. California Acad. Sci., 61, 45-60, 1962. Hemleben,C., M. Spindler,I. Breitinger,and R. Ott, foraminiferalcalcite as a proxy for water mass Morphological and physiological responsesof temperatures,Geochim. Cosmochim. Acta, 60, Climate:Long-Range Investigation, Mapping, and 803-814, 1996a. Prediction (CLIMAP), Seasonal reconstructions Globigerinoides sacculifer (Brady) under of Earth's surface at the Last Glacial Maximum, varying laboratory conditions, Mar. Niirnberg, D., J. Bijma, and C. Hemleben, Erratum Micropaleontology, 12, 305-324, 1987. GSAMap and Chart Ser. MC-36, Geol. Soc. of to "Assessingthe reliability of magnesiumin Am., Boulder, Colo., 1981. Hemleben, C., M. Spindler, and O.R. Anderson, foraminiferalcalcite as a proxy for water mass Conte, M. H., G. Eglington, and L. A. S. Madureira, Modern Planktonic Foraminifera, 363 pp., temperatures",Geochim. Cosmochim.Acta , 60, 2483-2484, 1996b. Long-chain alkenones and alkyl alkenoatesas Springer-Verlag,New York, 1989. palaeotemperatureindicators: Their production, Herbert,T. D., J. D. Schuffert,D. Thomas,C. Lange, Oberh'ansli,H., C. B6nier, G. Meinecke, H. Schmidt, flux, and early sedimentarydiagenesis in the A. Weinheimer, A. Peleo-Alampay, and R. R. Schneider, and G. Wefer, Planktic easternNorth Atlantic, Org. Geochem.,19, 287- J. C. Herguera, Depth and seasonality of forarninifers as tracers of ocean currents in the 298, 1993. alkenone production along the California easternSouth Atlantic, Paleoceanography, 7(5), Cronblad, H.G., and B. A. Malmgren, Climatically margin inferred from a core top transect, 607-632, 1992. controlledvariation of $r and Mg in Quaternary Paleoceanography,13(3), 263-271, 1998. Oomori, T., H. Kaneshima, Y. Maezato, and planktic foraminifera, Nature, 291, 61-64, Hut, G., Stable isotope referencesamples for Y. Kitano,Distribution coefficient of Mg2+ ions 1981. geochemicaland hydrologicalinvestigations, between calcite and solution at 10-50øC, Mar. Curry, W.B., and D.W. Oppo, Synchronous, reportto the DirectorGeneral, 42 pp., Int. At. Chem., 20, 327-336, 1987. high-frequency oscillations in tropical sea- Energy Agency, Vienna, 1987. Paillard, D., L. Labeyrie, and P. Yiou, Macintosh surfacetemperatures and North Atlantic deep Izuka, S. K., Relationshipof magnesiumand other program performs time-series analysis, Eos waterproduction during the last glacial cycle, minor elements in tests of Cassidulina Trans. AGU, 77(39), 379, 1996. Paleoceanography,12(1), 1-14, 1997. subglobosaand corriangulata to physical Puechmaille,C., Teneursen strontiumet magn6sium Delaney, M. L., A. W. H. B6, and E. A, Boyle, Li, oceanicproperties, J. Foraminiferal Res., 18, dansles tests de foraminiferesplanctoniques-- Sr, Mg, Na in foraminiferal calcite shells from 151-157, 1988. Premiersindices de l'alt6ration post-mortem, laboratoryculture, sediment traps, and sediment Katz, A., The interactionof magnesiumwith calcite Bull Inst. Geol. BassindAquitaine, 38, 81-94, cores, Geochim. Cosmochim. Acta, 49, 1327- during crystal growth at 25-90øC and one 1985. 1341, 1985. atmosphere,Geochim. Cosmochim. Acta, 37, Ravelo, A.C., and R.G. Fairbanks, Oxygen Duplessy, J.C., P.L. Blanc, and A.W.H. B6, 1563-1586, 1973. isotopic compositionof multiple species of Oxygen-18 enrichmentof planktic foraminifera Kinsman, D.J.J., and H.D. Holland, The co- planktonicforaminifera: Recorders of the modern due to gametogenic calcification below the precipitationof cationswith CaC03, IV, The co- photic zone temperature gradient, euphoticzone, Science,213, 1247-1249, 1981. precipitationof Sr 2* with aragonitebetween 16 ø Paleoceanography,7(6), 815-831, 1992. Emiliani, C., Depth habitatsof some speciesof and 96øC, Geochim. Cosmochim.Acta, 33, 1- Ravelo, A.C., R.G. Fairbanks, and pelagic foraminiferaas indicated by oxygen 17, 1969. S. G. H. Philander, Reconstructing tropical isotoperatios, Am. J. Sci., 252, 149-158, 1954. Krinsley,D., Traceelements in the testsof planktic Atlantic hydrography using planktonic NUERNBERG ET AL.: Mg/Ca- DERIVED PALEO-SST 134

foraminifera and an ocean model, sea surfacetemperature for the last 30 kyr: A and L. Labeyrie, Assessment of sea surface Paleoceanography,5, 409-431, 1990. comparisonof U,•' , 15•80,and foraminiferal temperature variations in the central North Rosell-Me16,A., Projecttakes a new look at past sea assemblage temperature estimates, Atlantic using the alkenone unsaturationindex surfacetemperatures, Eos Trans. AGU., 79(33), Paleoceanography,9(1), 31-45, 1994. ( U•' ), Geochim.Cosmochim. Acta, 62, 2421- 393-394, 1998. Spindler, M., C. Hemleben, J.B. Salomons, and 2427, 1998. Rosenthal,Y., and E. Boyle, Factorscontrolling the L. P. Smit, Feedingbehavior of someplanktonic Vogelsang, E., Pal•io-Ozeanographie des fluoride content of planktic foraminifera:An foraminifers in laboratory cultures, J. Europ/iischen Nordmeeres anhand stabiler evaluation of its paleoceanographic Foraminifaral Res., 14, 237-249, 1984. Kohlenstoff- und Sauerstoffisotope, Ph.D. applicability, Geochim. Cosrnochim.Acta, 57, Stramma,L., and F. Schott, The mean flow field of thesis, Rep. SFB 313, 23, 136 pp., Univ. of 335-346, 1993. the tropicalAtlantic Ocean, Deep-Sea Res., Part Kiel, Kiel, Germany, 1990. Rostek, F.G., G. Ruhland, F.C. Bassinot, II, 46, 279-303, 1999. Wang, L., M. Sarnthein, J.-C. Duplessy, P. J. Miiller, L.D. Labeyrie, Y. Lancelot, and Stute, M., M. Forster, H. Frischkorn, A. Serejo, H. Erlenkeuser, S. Jung, and U. Pflaumann, E. Bard, Reconstructingsea surfacetemperature J. F. Clark, P. Schlosser, W.S. Broecker, and Palco sea surface salinities in the low-latitude and salinityusing 15•80and alkenonerecords, G. Bonani, Cooling of tropical Brazil (5øC) Atlantic:The 15'•Orecord of Globigerinoides Nature, 364, 319-321, 1993. duringthe Last Glacial Maximum, Science,269, ruber (white). Paleoceanography, 10(4), 749- Savin, S.M., and R. G. Douglas, Stable isotope 379-383, 1995. 761, 1995. and magnesiumgeochemistry of Recentplanktic Thompson, L.G., E. Mosley-Thompson, Wefer, Get al., Late Quaternarysurface circulation of foraminiferafrom the South Pacific, Geol. Soc. M. E. Davis, P.N. Lin, K.A. Henderson, the South Atlantic: The stable isotope record Am. Bull., 84, 2327-2342, 1973. J. Cole-Dai, K. B. Bolzam, and K. B. Lin, Late and implications for heat transport and Schneider, R.R., P.J. Miiller, G. Ruhland, glacial stage and Holocene tropical ice-core productivity, in The South Atlantic: Present G. Meinecke, H. Schmidt, and G. Wefer, Late recordsfrom Huascaran,Peru, Science, 269, 46- and Past Circulation, edited by G. Wefer et al., Quaternarysurface temperatures and productivity 5O, 1995. pp. 461-502, Springer-Verlag,New York, 1996. in the East-EquatorialSouth Atlantic:Response Thunnel,R. C., Carbonatedissolution and abyssal to changesin Trade/Monsoonwind forcing and hydrographyin the Atlantic Ocean, Mar. Geol., surface water advection, in The South Atlantic: 47, 165-180, 1982. A. Mailer and D. Narnberg,GEOMAR Present and Past Circulation, edited by Ufkes, E., J. H. F. Jansen, and G.J.A. Brummer, ResearchCenter, Wischhofstrasse 1-3, D-24148 G. Wefer et al., pp. 527-551, Springer-Verlag, Living planktonic foraminifera in the eastern Kiel, Germany. ([email protected]; New York, 1996. SouthAtlantic during spring:Indicators of water [email protected]) Schrag,D.P., G. Hampt, and D. W. Murray, Pore masses,upwelling and the Congo (Zaire) River R. R. Schneider,Fachbereich Geowissen fluid constraintson the temperatureand oxygen plume, Mar. Micropaleontol., 33, 27-53, 1998. isotopic composition of the glacial ocean, Van Campo, E., J.C. Duplessy, W.L. Prell, schafien,University of Bremen,Klagenfurther Science, 272, 1930-1932, 1996. N. Barratt, and R. Sabatier, Comparison of Strasse, D-28334 Bremen, Germany. Shackleton,N.J., and E. Vincent, Oxygen and terrestrialand marine temperature estimates for the (rschneider•uni-bremen.de) carbon isotope studies in Recent foraminifera past 135 ky off southeastAfrica: A testfor GCM from the southwest Indian Ocean, Mar. simulationsof palcoclimate,Nature, 348, 209- (ReceivedJanuary 28, 1999; Micropaleontol., 3, 1-13, 1978. 212, 1990. revised October 12, 1999; Sikes,E. L., and L. D. Keigwin, EquatorialAtlantic Villanueva, J., J. O. Grimalt, E. Cortijo, L. Vidal, acceptedOctober 19, 1999.)