the closing of a seaway: ocean water masses and global climate change

The closing of a seaway:ocean water masses and global climate change

Caroline H. Lear a;�, Yair Rosenthal a, James D. Wright b

a Institute of Marine and Coastal Sciences and Department of Geology, Rutgers University, 71 Dudley Road, New Brunswick,NJ 08901, USA

b Department of Geology, Wright Laboratories, Rutgers University, 610 Taylor Road, Piscataway, NJ 08854, USA

Received 5 July 2002; received in revised form 27 January 2003; accepted 20 March 2003

Abstract

The Late Neogene witnessed various major paleoceanographic changes that culminated in intense NorthernHemisphere Glaciation (NHG). The cause and effects of these changes are still debated. We use a multiproxyapproach to determine the relative timing of the closure of the Panama gateway, changes in Atlantic circulation,global cooling and ice sheet growth. Benthic foraminiferal Mg/Ca records from a Pacific and an Atlantic Site havebeen produced and are interpreted in terms of bottom water temperatures. These Mg-temperature records arecombined with published benthic N

13C, N18O and erosion records to reconstruct the flow of proto-North AtlanticDeep Water (proto-NADW) over the past 12 Ma. The results suggest that between 12.5 and 10.5 Ma, and againbetween about 8.5 and 6 Ma, a nutrient-depleted water mass that was colder (by 1^2‡C) and fresher than theintervening deep water mass filled the Atlantic basin. This proto-NADW became warmer (by V1‡C) and saltierbetween 6 and 5 Ma, coincident with the restriction of surface water flow through the Central American Seaway. TheMg-temperature records define a subsequent global cooling trend of V3.5‡C between 5 Ma and today. Early NHG inthe late Miocene was perhaps related to the formation of the relatively cold, fresh proto-NADW. The formation ofthe warmer and saltier proto-NADW in the early Pliocene may have initially limited Northern Hemisphere icegrowth. However, the increased moisture released at high northern latitudes associated with formation of ‘warm’proto-NADW, coupled with the global temperature decrease of deep (and hence polar surface) waters, likely helpedinitiate the intense NHG of the Plio^Pleistocene.@ 2003 Elsevier Science B.V. All rights reserved.

Keywords: North Atlantic Deep Water; benthic foraminifera; Mg/Ca; ocean gateway; Ocean Drilling Program

1. Introduction

Ocean circulation patterns a¡ect global climate

via the distribution of heat, moisture and CO2.Today, the evaporative cooling associated withthe formation of North Atlantic Deep Water(NADW) supplies heat and moisture to highnorthern latitudes, thereby a¡ecting the stabilityof Northern Hemisphere ice sheets [1]. Tradition-ally, reconstructions of past deep water circula-tion are based on the carbon isotopic composition

0012-821X / 03 / $ ^ see front matter @ 2003 Elsevier Science B.V. All rights reserved.doi:10.1016/S0012-821X(03)00164-X

* Corresponding author. Tel. : +1-732-932-6555, ext. 255;Fax: +1-732-932-8578.E-mail address: [email protected] (C.H. Lear).

EPSL 6629 5-5-03

Earth and Planetary Science Letters 210 (2003) 425^436

www.elsevier.com/locate/epsl

mailto:[email protected]

http://www.elsevier.com/locate/epsl

(N13C) of benthic foraminifera, which record inter-basin di¡erences in N

13C. Site-to-site N13C gra-

dients are related to ocean circulation patternsvia nutrient distribution by water mass agingand/or mixing of di¡erent water masses. Thismethod can be complicated by the fact that thecarbon isotopic composition of deep waters is alsoa¡ected by changes in the global carbon cycle, thetotal deep ocean nutrient inventory, and the over-all rate of oceanic overturn. Using integrated bio-,magneto-, and isotope stratigraphies, these com-plications were minimized, showing that deepwaters sourced from the North Atlantic probablyformed during much of the late Miocene andprobably during the latter part of the middle Mio-cene [2,3].

The key to understanding how deep watersformed and hence their e¡ect on climate lies inreconstructing their relative densities. While car-bon isotopes may point to source regions andcirculation paths, it has not been possible to de-termine the physical properties (i.e. temperatureand salinity) of proto-NADW using traditionalpaleoceanographic proxies such as the oxygen iso-topic composition (N18O) of benthic foraminiferalcalcite. Although a robust and valuable proxy, theN18O of fossil calcite re£ects both the temperatureand N

18O of seawater, so for example modernbenthic foraminifera bathed in NADW have sim-ilar N

18O to those bathed in Antarctic BottomWater (AABW), despite the V2‡C temperaturedi¡erence between the two water masses. Mg/Ca-paleothermometry in benthic foraminifera isa relatively new tool that has been shown tohave potential in providing salinity-independentestimates of past bottom water temperatures(BWTs) [4^7]. Here we use this tool to compare

the physical properties of the water masses bath-ing an Atlantic and a Paci¢c site over the last12 Myr. Our working hypothesis is that parallelvariations in Mg/Ca re£ect global temperaturechange whereas di¡erences between the two sitesre£ect changes in ocean circulation. We linkchanges in the physical properties of NADW tothe tectonic closure of the Panama Gateway andglobal climatic change.

2. Methods

2.1. Site selection and age models

Ocean Drilling Program (ODP) Site 806(2500 m) on the Ontong Java Plateau (OJP) inthe western equatorial Paci¢c was chosen to rep-resent average global ocean temperatures. Con-versely, ODP Site 926 (3500 m) on the CearaRise (CR) in the eastern equatorial Atlantic waschosen as a site that is today within the mixingzone between NADW and AABW, and thereforewould be sensitive to changes in the £ow of proto-NADW (Table 1). Samples were obtained fromthese two sites covering 0^12 Ma at a temporalresolution of roughly one sample per 0.2 Myr. Anage model suitable for this low resolution wasobtained for Site 806 using linear regressions be-tween the depths of biostratigraphic events [8] andages for these events that were compiled from theliterature by the ODP Leg 199 Shipboard Scien-ti¢c Party [9] (Fig. 1). An orbitally tuned time-scale [10] was used for dating the 926 samplesfor the 4^11 Ma interval. The same method asused for Site 806 was used to date the 0^4 Mainterval of Site 926.

Table 1Locations of sites (new and published data)

Site Ocean region Latitude Longitude Present water depth(m)

DSDP 289 Paci¢c 0‡29.9PS 158‡30.7PE 2206DSDP 563 North Atlantic 33‡38.5PN 43‡46.0PW 3786DSDP 608 North Atlantic 42‡50.2PN 23‡5.3PW 3526ODP 806 Paci¢c 0‡19.1PN 159‡21.7PE 2521ODP 926 Equatorial Atlantic 3‡43.1PN 42‡54.5PW 3598

EPSL 6629 5-5-03

C.H. Lear et al. / Earth and Planetary Science Letters 210 (2003) 425^436426

2.2. Sample preparation and analysis

Sediments were disaggregated by shaking for48 h in de-ionized water and wet-sieved through60 Wm mesh. Samples were dry-sieved to obtainaround 10^15 benthic foraminifera from the250^355 Wm size fraction. Occasionally, andwhere necessary, foraminifera were picked fromthe s 355 Wm size fraction to increase samplesize. The same species (Cibicidoides wuellerstor¢,Cibicidoides mundulus ( =C. kullenbergi), and Or-idorsalis umbonatus) were picked from each site.Foraminiferal tests were cleaned using a protocolto remove clays, organic matter, metal oxides [11],although about half of the CR samples werecleaned without the reducing step that removesmetal oxides, and with only one weak acid leachas opposed to four separate leaches (Fig. 2). Aftercleaning, the foraminifera were gradually dis-solved in trace metal clean 0.065 N HNO3 (Seast-ar, Vancouver, BC, Canada) and 100 Wl of thissolution was diluted with 300 Wl trace metal clean0.5 N HNO3 to obtain a Ca concentration of 3R 1mmol l31. Measured Mg/Ca varies slightly withCa concentration of the solution and this matrixe¡ect was corrected for as described in [12]. Thesesamples were analyzed for Mg/Ca, Sr/Ca andMn/Ca by Finnigan MAT Element Sector Fieldinductively coupled plasma mass spectrometer(ICP^MS) operated in low resolution (m/vm= 300) following the method outlined in [13] and

modi¢ed by [12]. Benthic foraminiferal Sr/Ca ra-tios are presented here to assess diagenetic alter-ation and are discussed in more detail in a sepa-rate publication [14]. In both sites, benthicforaminiferal Mn/Ca increases between the sea-£oor and about 10 m burial depth and subse-quently decreases down-core. The shorter cleaningtechnique results in higher and more variable Mn/Ca, presumably re£ecting the presence of Mn ox-ides. The Mn/Ca pro¢les bear no resemblance tothose of Mg/Ca and Sr/Ca and in this case are notregarded as a useful indicator of compromisedMg/Ca data.

The long-term (several months) precision of aconsistency standard with Mg/Ca of 1.10 mmolmol31 was R 3.7% (r.s.d.), the precisions of con-sistency standards with Mg/Ca of 2.40 mmolmol31 and 6.10 mmol mol31 were R 1.5% and

Fig. 1. Age model for ODP Site 806. Biostratigraphic eventsrecorded by [8] are tied to assigned ages from literature com-piled by the ODP Leg 199 Shipboard Scienti¢c Party [9](mbsf =meters below sea£oor).

Fig. 2. Mg/Ca of benthic foraminifera (top panel, C. wueller-stor¢ ; bottom panel, O. umbonatus) from ODP Site 926(CR). Mg/Ca values appear una¡ected by choice of cleaningtechnique or method of analysis. Square symbols representsamples cleaned without the reducing step at CambridgeUniversity. Of these samples, the open symbols were ana-lyzed by ICP^AES at Cambridge University using Cambridgestandards, and the closed symbols were analyzed by ICP^MSat Rutgers University using Rutgers standards. The crossesrepresent samples cleaned with the reducing step and ana-lyzed at Rutgers University.

EPSL 6629 5-5-03

C.H. Lear et al. / Earth and Planetary Science Letters 210 (2003) 425^436 427

R1.6%, respectively. A portion of the CR recordwas prepared without the reducing step in thecleaning process, and was analyzed either byICP-MS at Rutgers University or by simultaneousICP-AES (inductively coupled plasma atomicemission spectroscopy) at Cambridge University.There is no systematic di¡erence in benthic fora-miniferal Mg/Ca resulting from either the di¡er-ent cleaning or analytical methods (Fig. 2).

2.3. Stable isotope analysis

Samples for isotope analysis were loaded into aMulti-prep peripheral device, reacted in 100% or-thophosphoric acid at 90‡C, and analyzed on anOptima mass spectrometer at Rutgers University.Seventy samples were analyzed in the GodwinLaboratory (University of Cambridge, UK).These samples were soaked in a 3% solution ofhydrogen peroxide, ultrasonicated and dried over-night prior to reaction with 100% orthophosphor-ic acid at 90‡C and analysis by a VG PRISMmass spectrometer. Values are reported versusV-PDB through the analysis of NBS19 with val-ues of 1.95x and 32.20x for N

13C and N18O,

respectively, as reported by [15]. The precision ofthe isotope analyses is better than R 0.06x forN13C and better than R 0.08x for N

18O. We usethe species adjustment factors of [16] to convertN18O values to ‘equilibrium’ values.

2.4. Determination of Mg-paleotemperatures

We use the revised calibration equations forC. wuellerstor¢ and O. umbonatus [12] to estimateBWT. Multispecies records clearly show a largeo¡set between C. mundulus and C. wuellerstor¢Mg/Ca; for this reason C. mundulus Mg/Ca arenot used in core-top Cibicidoides calibrations. Forthis study therefore, we have assumed thatC. mundulus has the same temperature sensitivityas C. wuellerstor¢ but a di¡erent pre-exponentialconstant. This is a reasonable assumption becausethere is a growing body of evidence that the in-corporation of Mg into the calcite tests of mostepibenthic and also planktonic foraminifera ap-pears to have a temperature sensitivity of 10R1% per ‡C (e.g. [4,7,12,17^19]). We estimated the

pre-exponential constant using the ratio of themean C. mundulus Mg/Ca and mean C. wueller-stor¢Mg/Ca for the interval between 4 and 14 Maat ODP Site 806 to obtain a temperature calibra-tion for C. mundulus :

Mg=Ca ¼ 0:9expð0:11UBWTÞ ð1Þ

Assuming little or no change in seawater Mg/Ca, the temperature estimates calculated hereshould be accurate to within R 1‡C (2 S.E.) [12].

3. Results

ODP Sites 926 (CR) and 806 (OJP) both dis-play a general trend of increasing benthic forami-niferal Mg/Ca with sediment depth (Figs. 3 and4). The Atlantic record displays a decrease inbenthic foraminiferal Mg/Ca between 150 and200 meters composite depth (mcd), followed byan interval of elevated Mg/Ca between 200 and250 mcd. These features are not observed in thePaci¢c record at either the same burial depth orage. In the Atlantic core, the Mg/Ca of C. mun-dulus is more similar to O. umbonatus thanC. wuellerstor¢, whereas the reverse is true forC. mundulus in the Paci¢c core. This is possibly

Fig. 3. Multi-species benthic foraminiferal trace metal data,Mg/Ca (panel a) and Sr/Ca (panel b), from ODP Site 926(CR) versus depth in sediment (mcd=meters compositedepth).

EPSL 6629 5-5-03


a result of genetic di¡erences that may occur be-tween the ‘same’ species across both time andspace. These di¡erences are responsible for someof the scatter in the Mg-temperature records.

Unlike Mg/Ca, benthic foraminiferal Sr/Cafrom each site generally decreases with increasingsediment depth, reaching a broad minimum ataround 200 m burial depth, (V6^8 Ma), althoughthe minimum is more pronounced in the Atlanticsite than in the Paci¢c site (Figs. 3 and 4). Sub-sequently, Sr/Ca slowly increases. Sr/Ca of O. um-bonatus is low relative to the Cibicidoides species.As with Mg/Ca, C. mundulus appears to have aslightly di¡erent inter-species o¡set for Sr/Ca inthe Atlantic compared to the Paci¢c (Figs. 3 and4).

Benthic foraminifera were analyzed for N18O tocomplement published [16,20] stable isotope re-cords (Fig. 5). The Pleistocene is characterizedby large amplitude oscillations in N

18O on the or-der of 1.0x at each site. The CR record alsodisplays large oscillations in N

18O prior to thePleistocene, but the di¡erent resolution of the re-cords prevents comparison of the magnitude andfrequency of these oscillations. However, bothsites display a broad maximum on the order of

0.5x between 7 and 5 Ma and a general increaseof 0.8^1.0x through the Pliocene.

4. Discussion

4.1. Geochemical preservation and changes inseawater Mg/Ca

We will address the issue of post-depositionalalteration of calcite Mg/Ca and the e¡ect of sec-ular variations in seawater Mg/Ca. Descriptionsof foraminifera from Sites 926 and 806 in additionto the quality of published isotope records showthat the foraminifera at these sites have notundergone wholesale recrystallization (e.g. [8,16,21]). It has been argued that calcite saturationstate may potentially a¡ect the trace metal con-tent of benthic foraminifera, either through disso-lution or through a precipitation e¡ect (e.g. [22]).The e¡ect of post-depositional dissolution onthe trace metal content of benthic foraminiferais still a matter of debate (e.g. [4,23]). However,the contrast between the trends in the benthic fo-raminiferal Mg/Ca and Sr/Ca records presentedhere suggests that the major features have notbeen caused by calcite saturation-related processes(Figs. 3 and 4). Measured foraminiferal calcitechemistry may be a¡ected by adhering inorganiccalcite cements. Such cements would result in anincrease in Mg/Ca and a decrease in measuredSr/Ca cements [24]. No such cements were visiblethrough inspection by binocular microscope.Nevertheless, using a mass balance approach as-suming cement Mg/Ca of V8 mmol/mol (Lear,unpublished data), inorganic cements comprisingup to 1% of the total calcite could introduce anuncertainty in the accuracy of the Mg tempera-tures of around 1‡C, and Sr/Ca values could de-crease by V1%.

Seawater Mg/Ca is assumed to be constant be-tween 0 and 12 Ma. However, it should be notedthat if the benthic foraminiferal Sr/Ca records re-£ect changes in seawater Sr/Ca, it is likely thatMg/Ca has changed over the same interval, asCa has the shortest residence time in the oceanof all three elements. The largest change inbenthic foraminiferal Sr/Ca is a 10^15% increase

Fig. 4. Multi-species benthic foraminiferal trace metal data,Mg/Ca (panel a) and Sr/Ca (panel b), from ODP Site 806(OJP) versus depth in sediment (mbsf =meters below sea-£oor).

EPSL 6629 5-5-03


between 5 and 0 Ma. If this change were causedby only a change in seawater Ca concentration, itwould result in an underestimation of the Mg-temperature decrease of about 1‡C over the5 Myr interval.

4.2. Temperatures

The benthic foraminiferal Mg/Ca data are in-terpreted solely in terms of variations in BWT(Fig. 5). Using the species-speci¢c temperaturecalibrations [12] results in more consistent temper-ature estimates (generally R 1‡C) than the ‘con-

stant o¡sets’ method [5]. Much like N18O paleo-

thermometry, the ‘constant o¡sets’ method nor-malizes all Mg/Ca data to a single species be-fore using a generic temperature calibration.However, the inter-speci¢c o¡sets in Mg/Ca ap-pear to be sensitive to changes in temperature.Di¡erences between species-speci¢c temperaturesestimated from a single sample could be the resultof natural Mg/Ca variability, sampling resolution(one sample represents V5 cm of core), imperfectcalibrations, evolutionary changes within a spe-cies, or changes in seawater Mg/Ca. We combinethe temperatures derived from all three species to

Fig. 5. Mg-temperatures from three benthic foraminiferal species and composite benthic foraminiferal N18O for (a) ODP Site 926cored on the CR, Atlantic and (b) ODP Site 806 cored on the OJP, Paci¢c. Published stable isotope data [16,20] are supple-mented with new data (this study). Benthic foraminiferal oxygen isotope data have been adjusted to ‘equilibrium values’.

EPSL 6629 5-5-03


avoid any species bias and produce a robust rec-ord of the temperature variations at each site. Weuse the Kaleidagraph0 weighted ¢t function(10%) which uses the locally least-squared errormethod to ¢t the best smooth curve through thecompiled data (Fig. 6). The accuracy of these tem-perature estimates is R 1‡C (2 S.E.) given minimalchanges in seawater Mg/Ca [12].

The Atlantic and Paci¢c Mg-temperatures areindistinguishable within the limits of the methodfor most of the past 12 Ma (Fig. 6). However,there are two intervals (12^11 Ma and 8.5^6Ma) over which the Atlantic Mg-temperatureswere appreciably lower (by 1^2‡C) than the Pacif-ic Mg-temperatures. BWT estimates from Paci¢cSite 806 were relatively stable between around 11and 5 Ma, at about 5.5‡C. In contrast, the deepAtlantic Site 926 displays £uctuations in benthicforaminiferal Mg/Ca, which cannot be interpretedin terms of global temperature or seawater Mg/Cachange. BWTs calculated from benthic foraminif-eral Mg/Ca at ODP Site 926 increased about 3‡Cacross the Serravallian^Tortonian boundary, from

V3‡C at 11.5 Ma to V6‡C at 10 Ma. The Mg-temperatures remained at around 6‡C until 8.5Ma, except for a V1‡C temperature decrease cen-tered at 9.5 Ma. Subsequently, Site 926 BWTdecreased towards the Tortonian^Messinianboundary, reaching around 3.5‡C at 7.1 Ma.Temperatures gradually increased by around1.5‡C throughout the Messinian, reaching V5‡Cat 5.3 Ma. We interpret these Atlantic tempera-ture changes in terms of changing ocean circula-tion, in particular changes in the £ow of proto-NADW. Both the Atlantic and the Paci¢c sitesdisplay a concurrent decrease in benthic forami-niferal Mg/Ca through the Plio^Pleistocene. Weinterpret this trend as re£ecting a global decreasein BWT of around 3.5‡C (Fig. 6).

4.3. Nature and cause of circulation change

A seismic erosion re£ector ‘Merlin’ has beenidenti¢ed along the western margin of the NorthAtlantic, and its age constrained to be between13 and 10.5 Ma [25]. This re£ector has been cor-related to an increase in the £ux of proto-NADWduring the late middle Miocene [3]. Carbon iso-tope data indicate that this phase of proto-NADW production began about 12.5 Ma [3].This erosion event was also coincident with theinitiation of drift accumulation on the Hattonand Snorri drifts [26] and relatively low (V4‡C)Mg-temperatures at CR Site 926 (Fig. 6). Site-to-site N13C reconstructions have shown that between10 and 8.5 Ma there was little di¡erence betweenthe North Atlantic, Paci¢c and Southern oceans,indicating little to no £ow of proto-NADW. Dur-ing this interval, CR Mg-temperatures rose byV2‡C to temperatures similar to those at the Pa-ci¢c site. Between 8.5 and 7 Ma, the inter-oceanN13C gradients increased, recording the £ow ofproto-NADW to the south [27,28], and CRODP Site 926 Mg-temperatures decreased byV2‡C (Fig. 7). Several seismic re£ectors of lateMiocene to early Pliocene age have been identi¢edin the northern North Atlantic by various authors[29,30].

The limited spatial and temporal resolution ofthe paleoceanographic records coupled with theuncertainties inherent in each of the various prox-

Fig. 6. 10% weighted ¢t through the compiled Mg-tempera-tures (Fig. 5) for ODP Sites 806 and 926. Accuracy of tem-perature estimates is about R 1‡C (2 S.E.) given minimalchange in seawater Mg/Ca over this interval (see Section 4).The snow£akes and water droplets indicate increased and re-duced occurrences of IRD, respectively, on the Iceland andVVring plateaus [42].

EPSL 6629 5-5-03


ies complicate the paleoceanographic interpreta-tion. However, there is one paleoceanographicscenario that is consistent with the erosion, car-bon isotopic and Mg/Ca records described above,and which at present we believe is most likely torepresent the evolution of proto-NADW over thepast 12 Myr. This scenario invokes the £ow ofproto-NADW between 12.5 and 10 Ma, followedby an interval of little to no £ow between 10 and8.5 Ma, and the resumption of proto-NADW£ow at V8.5 Ma. The Mg-temperature records

presented here suggest that, unlike today, proto-NADW was colder (by about 2‡C) and fresher(by an amount equivalent to N

18Ow of 0.5x)than the competing southern-sourced deep waterin the late Miocene.

From around 9 Ma, the Paci¢c benthic N13C

record decreases, while the North Atlantic N13C

records remain relatively constant (Fig. 7). Rela-tive (site-to-site) changes are more important thanabsolute changes in N

13C in reconstructing circu-lation paths, as changes in the average isotopiccomposition of the global carbon reservoir aree¡ectively canceled out. For example, the decreasein the Paci¢c deep ocean N

13C record at V9 Malikely re£ects a global change in the ocean carbonreservoir that is masking a more localized increasein N

13C at CR coincident with the resumptionof proto-NADW. Interestingly, the increase inbenthic N

13C of the North Atlantic relative tothe Paci¢c between V9 and 7 Ma (Fig. 7C) iscoincident with an increase in the intensity of car-bonate dissolution at CR Site 926 [31,32]. Thisobservation implies that the increased dissolutionintensity at CR Site 926 was the result of a globalincrease in the oceanic carbon inventory ratherthan the e¡ect of water mass switching (as impliedby [10]). This hypothesis is supported by the co-incident decrease in the carbonate content andsediment accumulation rate of ODP Site 804, sit-uated below the lysocline on the OJP, Paci¢cOcean [33].

There are perhaps two major hypotheses for thecause(s) of the initiation of NADW. One is thatthe closure of the Panama Gateway ¢nally al-lowed the North Atlantic to become salineenough for the formation of NADW [34]. Theother hypothesis is that changes in the £ux ofNADW are directly linked to changes in thedepth of the Greenland^Iceland^Scotland Ridge,which varies with the intensity of the mantleplume below Iceland [28]. These hypotheses arenot incompatible. It has been shown thatNADW formation may have been related to theclosure of the Panama Gateway, but that its out-£ow was controlled by the topography of theGreenland^Scotland Ridge [35]. The same authorsshowed in their model that while NADW mayform with a partially restricted Panama Gateway,

Fig. 7. (a) Mg-temperature records from CR Site 926(dashed line) and OJP Site 806 (solid line). (b) Publishedbenthic foraminiferal N13C records from ODP Sites 563 and608 (dashed line) in the North Atlantic and DSDP Site 289(solid line) on the OJP, Paci¢c (Table 1) [27]. (c) Di¡erencebetween the Atlantic and Paci¢c N

13C records in panel b.

EPSL 6629 5-5-03


convection in the Labrador Sea would be signi¢-cantly reduced, or close to zero. There has been agreat deal of e¡ort aimed at determining the tim-ing of both the initiation of convection in theLabrador Sea, and the initiation of NADW, inparticular through the use of lead and neodymi-um isotopes (e.g. [36]). However, it has recentlybeen shown that the isotopic composition ofsource sediments is greatly a¡ected by changesin the weathering regimes that occurred atroughly the same time period under consideration[37].

Taken at face value, the CR Mg/Ca record sug-gests a cooling between 8.5 and 7 Ma on CRassociated with the major divergence of the Paci¢cand Atlantic carbon isotope records (Fig. 7). This¢nding suggests that proto-NADW was coolerand fresher (by an amount equivalent to a de-crease in N

18Ow of about 0.5x) than the compet-ing southern-sourced deep water. Today, watertemperatures in the Greenland^Iceland^Norwe-gian (GIN) seas are colder than averageNADW, which is also largely in£uenced by deepwater sourced from the warmer and saltier Lab-rador Sea [38]. Therefore, the cooling observed atSite 926 between 8.5 and 7 Ma is perhaps com-patible with the resumption of proto-NADW pri-or to convection in the Labrador Sea, with themajority of this deep water sourced from theGIN seas. An alternative explanation for a rela-tively cold and fresh proto-NADW in the lateMiocene is that the open Panama Gateway mayhave reduced the residence time of surface watersin this low-latitude region. This would prevent thesurface water salinity from increasing as substan-tially as today prior to northward transport viathe Gulf Stream. Consequently, the surface watersdense enough to sink at high northern latitudes toform proto-NADW would have been colder andfresher than today. Between 6 and 5 Ma, CR Site926 BWTs gradually became warmer and saltier,likely related to further restriction of the PanamaGateway. Severe restriction of the Atlantic^Paci¢cconnection as indicated by the sea surface salinitycontrast either side of the isthmus occurred atV4.7 Ma [39], coincident with the warmestBWT at Site 926 since 8 Ma, and the initiationof sediment drifts in the Labrador Sea [40].

4.4. Caveats regarding the paleoceanographicinterpretation

There are certain complications inherent ineach of the proxy records mentioned here.Changes in seawater Mg/Ca will impact the abso-lute magnitude of the Mg-temperature changes,although the site-to-site temperature relationshipsare robust. We note that calcite saturation-relatede¡ects have been postulated to a¡ect benthic fo-raminiferal Mg/Ca [7,22]. At present there is node¢nite evidence for, or quanti¢cation of, suche¡ects. Therefore, we interpret the Mg/Ca recordspresented here in the most straightforward way(i.e. as representing a dominant temperature con-trol). We note that the temperature changes ob-served at CR may have resulted from a verticalmovement of the water masses. Rather than a‘switching on’ or ‘switching o¡’ scenario for pro-to-NADW, the temperature changes at CR couldbe viewed more realistically as an intensi¢cationor reduction in the £ow of proto-NADW. Lastly,it is not clear to what extent ODP Site 806, situ-ated at 2500 m on the OJP, represents the averagePaci¢c Ocean. In particular, it is possible that inthe late Miocene, an open Panama Gateway mayhave allowed proto-NADW to pass into the Pa-ci¢c, potentially reaching this site [41].

4.5. Changes in global ice volume

Between V7 and 6 Ma, the Paci¢c and Atlanticbenthic foraminiferal N

18O records increase byabout 0.5x. This increase does not coincidewith an excursion in the Mg-temperature record(Fig. 5). The simplest way to interpret this featureis therefore by a 0.5x increase in Nw and henceglobal ice volume during this period, which issupported by an observed increase in NorthernHemisphere IRD (ice-rafted debris) at this time[42] (Fig. 6). Between V6 and 5 Ma, there is anoverall decrease of about 0.5x in the stable iso-tope record. The Mg-temperature records perhapssuggest 9 1‡C warming associated with the de-crease. Thus, the remaining portion of the N

18Orecord (v 0.3x) must be interpreted in terms ofa decrease in ice volume, which is supported by adecrease in IRD content in the Iceland^Norwe-

EPSL 6629 5-5-03


gian Sea between about 5.5 and 5.0 Ma [42] (Fig.6). The benthic foraminiferal N18O records fromSites 806 and 926 increase through the Pliocene,and it appears that this increase is a result of botha decrease in BWTs (V3.5‡C assuming constantseawater Mg/Ca) and an increase in global icevolume.

4.6. E¡ect of circulation change on global climate

It is likely that either decreased polar temper-atures and/or increased moisture £ux to the Arcticregion induced the initiation of intense NorthernHemisphere Glaciation (NHG). It has been sug-gested that closure of the Isthmus of Panama en-hanced moisture transport to the Eurasian conti-nent, leading to increased freshwater delivery tothe Arctic and hence the NHG [43]. Our datasuggest that although complete closure of thePanama Gateway occurred after NADW beganto form, the closure may be responsible for thechanges in the physical characteristics of NADWtowards higher temperatures and salinity. EarlyNHG between 7 and 6 Ma evidenced by an in-crease in benthic N

18O and Northern HemisphereIRD records was perhaps associated with the ini-tiation of a relatively cold proto-NADW. Thesubsequent warming of proto-NADW associatedwith the closure of the Panama Gateway mayhave caused the temporary decay of these icesheets around 5 Ma (Fig. 6). Our Mg/Ca datadocument a signi¢cant cooling throughout thePliocene in both the Atlantic and Paci¢c Sites,beginning about 2 Myr prior to major NHG(which is dated to 2.9 Ma at the Iceland Plateau[42]). The underlying cause of this global coolingis unclear.

5. Conclusions

The occurrence of IRD at high northern lati-tudes in the late Miocene coincides with a markeddivergence between Atlantic and Paci¢c benthicN13C records and a lowering of deep water benthicforaminiferal Mg-temperatures by about 2‡C atCR ODP Site 926. This early NHG was likelyrelated to the initiation of a proto-NADW that

was colder, fresher and nutrient-depleted relativeto competing southern-sourced waters. Subse-quent restriction of the Panama Gateway between6 and 5 Ma coincided with proto-NADW becom-ing warmer and saltier, and a consequent decreasein Northern Hemisphere IRD. The extra heat re-leased at high northern latitudes through forma-tion of this ‘warm’ proto-NADW perhaps prohib-ited Northern Hemisphere ice growth in the earlyPliocene. However, the increased moisture £uxto the Arctic associated with formation of the‘warm’ NADW, coupled with the subsequentglobal temperature decline through the Plioceneprobably contributed to the major NHG in thelate Pliocene. This study demonstrates the needfor a multi-site approach when determining var-iations in global ice volume from benthic Mg/Caand N

18O records.

Acknowledgements

We thank Nick Shackleton for helpful discus-sions. We thank Mike Hall and Nick Shackleton(Godwin Laboratory, University of Cambridge)for running 70 Site 926 samples for stable iso-topes. We also thank two anonymous reviewerswhose thoughtful comments greatly improvedthis paper. This research used samples providedby the Ocean Drilling Program (ODP). ODP issponsored by the U.S. National Science Founda-tion (NSF) and participating countries underthe management of Joint Oceanographic Institu-tions (JOI), Inc. This research was supported by apost-doctoral fellowship at IMCS, Rutgers Uni-versity (C.H.L.) and NSF grant 9819675 (Y.R.).[BOYLE]

References

[1] W.S. Broecker, G.H. Denton, The role of Ocean-atmo-sphere reorganization in glacial cycles, Quat. Sci. Rev. 9(1990) 305^341.

[2] F. Woodru¡, S.M. Savin, Miocene deepwater oceanogra-phy, Paleoceanography 4 (1989) 87^140.

[3] J.D. Wright, K.G. Miller, R.G. Fairbanks, Early andMiddle Miocene stable isotopes: Implications for deep-

EPSL 6629 5-5-03


water circulation and climate, Paleoceanography 7 (1992)357^389.

[4] Y. Rosenthal, E.A. Boyle, N. Slowey, Environmental con-trols on the incorporation of Mg, Sr, F and Cd intobenthic foraminiferal shells from Little Bahama Bank:prospects for thermocline paleoceanography, Geochim.Cosmochim. Acta 61 (1997) 3633^3643.

[5] C.H. Lear, H. Elder¢eld, P.A. Wilson, Cenozoic deep-seatemperatures and global ice volumes from Mg/Ca inbenthic foraminiferal calcite, Science 287 (2000) 269^272.

[6] K. Billups, D.P. Schrag, Paleotemperatures and ice vol-ume of the past 27 Myr revisited with paired Mg/Ca and18O/16O measurements on benthic foraminifera, Pale-oceanography 17 (2002) 3-1^3-11.

[7] P.A. Martin, D.W. Lea, Y. Rosenthal, N.J. Shackleton,M. Sarnthein, T. Papenfuss, Quaternary deep sea temper-ature histories derived from benthic foraminiferal Mg/Ca,Earth Planet. Sci. Lett. 198 (2002) 193^209.

[8] Shipboard Scienti¢c Party, Site 806, Proc. ODP Init. Re-ports 130 (1991) 201^367.

[9] Shipboard Scienti¢c Party, Explanatory Notes, Proc.ODP Init. Reports 199 (2002) chapter 2, 1^70.

[10] N.J. Shackleton, S. Crowhurst, Sediment £uxes based onan orbitally tuned time scale 5 Ma to 14 Ma, Site 926,Proc. ODP Sci. Results 154 (1997) 69^82.

[11] E.A. Boyle, L.D. Keigwin, Comparison of Atlantic andPaci¢c paleochemical records for the last 250,000 years:Changes in deep ocean circulation and chemical invento-ries, Earth Planet. Sci. Lett. 76 (1985/6) 135^150.

[12] C.H. Lear, Y. Rosenthal, N. Slowey, Benthic foraminif-eral Mg/Ca-paleothermometry: A revised core-top cali-bration, Geochim. Cosmochim. Acta 66 (2002) 3375^3387.

[13] Y. Rosenthal, M.P. Field, R.M. Sherrell, Precise determi-nation of element/calcium ratios in calcareous samplesusing Sector Field Inductively Coupled Plasma MassSpectrometry, Anal. Chem. 71 (1999) 3248^3253.

[14] C.H. Lear, H. Elder¢eld, P.A. Wilson, A Cenozoic seawater Sr/Ca record from benthic foraminiferal calciteand its application in determining global weathering£uxes, Earth Planet, Sci. Lett. 208 (2003) 69^84.

[15] T.B. Coplen, New IUPAC guidelines for the reporting ofstable hydrogen, carbon and oxygen isotope-ratio data,J. Res. Natl. Inst. Technol. 100 (1995) 285^285.

[16] N.J. Shackleton, M.A. Hall, The Late Miocene stableisotope record, Site 926, Proc. ODP Sci. Results 154(1997) 367^373.

[17] D.W. Lea, T.A. Mashiotta, H.J. Spero, Controls on mag-nesium and strontium uptake in planktonic foraminiferadetermined by live culturing, Geochim. Cosmochim. Acta63 (1999) 2369^2379.

[18] H. Elder¢eld, G. Ganssen, Past temperature and N18O of

surface ocean waters inferred from foraminiferal Mg/Caratios, Nature 405 (2000) 442^445.

[19] Y. Rosenthal, G.P. Lohmann, Accurate estimation of seasurface temperatures using dissolution-corrected calibra-

tions for Mg/Ca paleothermometry, Paleoceanography17 (2002) Paper 1044, 10.1029/2001PA000749.

[20] T. Bickert, W.H. Berger, S. Burke, H. Schmidt, G. Wefer,Late Quaternary stable isotope record of benthic forami-nifera: Sites 805 and 806, Ontong Java Plateau, Proc.ODP Sci. Results 130 (1993) 411^420.

[21] Shipboard Scienti¢c Party, Site 926, Proc. ODP Init. Re-ports 154 (1995) 153^232.

[22] D.C. McCorkle, P.A. Martin, D.W. Lea, G.P. Klinkham-mer, Evidence of a dissolution e¡ect on benthic forami-niferal shell chemistry - Delta-C-13, Cd/Ca, Ba/Ca, andSr/Ca results from the Ontong Java Plateau, Paleocean-ography 10 (1995) 699^714.

[23] H. Elder¢eld, C.J. Bertram, J. Erez, A biomineralizationmodel for the incorporation of trace elements into fora-miniferal calcium carbonate, Earth Planet. Sci. Lett. 142(1996) 409^423.

[24] P.A. Baker, J.M. Gieskes, H. Elder¢eld, Diagenesis ofcarbonates in deep-sea sediments - Evidence from Sr/Caratios and interstitial dissolved Sr2þ data, J. Sediment.Pet. 52 (1982) 71^82.

[25] G.S. Mountain, B.E. Tucholke, Mesozoic and Cenozoicgeology of the U.S. Atlantic continental slope and rise, in:C.W. Poag (Ed.), Geologic Evolution of the United StatesAtlantic Margin, Van Nostrand Reinhold, New York,1985, pp. 293^341.

[26] C.N. Wold, Cenozoic sediment on drifts in the northernNorth Atlantic, Paleoceanography 9 (1994) 917^941.

[27] J.D. Wright, K.G. Miller, R.G. Fairbanks, Evolution ofmodern deepwater circulation: Evidence from the lateMiocene Southern Ocean, Paleoceanography 6 (1991)275^290.

[28] J.D. Wright, K.G. Miller, Control of North Atlantic deepwater circulation by the Greenland-Scotland Ridge, Pale-oceanography 11 (1996) 157^170.

[29] D.G. Roberts, Marine geology of the Rockall Plateau andTrough, Philos. Trans. R. Soc. London A 278 (1975) 447^509.

[30] K.G. Miller, B.E. Tucholke, Development of Cenozoicabyssal circulation south of the Greenland-ScotlandRidge, in: M.H.P. Bott (Ed.), Structure and Developmentof the Greenland-Scotland Ridge, Plenum Press, NewYork, 1983, pp. 549^589.

[31] D.W. Murray, L.C. Peterson, Biogenic carbonate produc-tion and preservation changes between 5 and 10 Ma fromthe Ceara Rise, western equatorial Atlantic, Proc. ODPSci. Results 154 (1997) 375^388.

[32] T.A. King, W.G. Ellis, D.W. Murray, N.J. Shackleton, S.Harris, Miocene evolution of carbonate sedimentation atthe Ceara Rise: A multivariate data/proxy approach,Proc. ODP Sci. Results 154 (1997) 349^365.

[33] W.H. Berger, R.M. Leckie, T.R. Janecek, R. Stax, T.Takayama, Neogene carbonate sedimentation on OntongJava Plateau: Highlights and open questions, Proc. ODPSci. Results 130 (1993) 711^744.

[34] E. Maier-Reimer, U. Mikolajewicz, T. Crowley, OceanGeneral Circulation Model sensitivity experiment with

EPSL 6629 5-5-03


an open Central American Isthmus, Paleoceanography 5(1990) 349^366.

[35] U. Mikolajewicz, T.J. Crowley, Response of a coupledocean/energy balance model to restricted £ow throughthe central American isthmus, Paleoceanography 12(1997) 429^441.

[36] K.W. Burton, H.F. Ling, R.K. O’Nions, Closure of theCentral American Isthmus and its e¡ect on deep-waterformation in the North Atlantic, Nature 386 (1997)382^385.

[37] F. von Blanckenburg, T.F. Nagler, Weathering versus cir-culation-controlled changes in radiogenic isotope tracercomposition of the Labrador Sea and North AtlanticDeep Water, Paleoceanography 16 (2001) 424^434.

[38] P. Tchernia, Descriptive Regional Oceanography, Perga-mon Press, Oxford, 1980, pp. 128^170.

[39] G.H. Haug, R. Tiedemann, R. Zahn, A.C. Ravelo, Roleof Panama uplift on oceanic freshwater balance, Geology29 (2001) 207^210.

[40] M.A. Kaminski, F.M. Gradstein, D.B. Scott, K.D. Mack-innon, Neogene benthic foraminifer biostratigraphy anddeep-water history of Sites 645, 646, and 647, Ba⁄n Bayand Labrador Sea, Proc. ODP Sci. Results 105 (1989)731^756.

[41] K.H. Nisancioglu, M.E. Raymo, P.H. Stone, Reorganiza-tion of Miocene deep water circulation in response to theshoaling of the Central American Seaway, Paleoceanog-raphy 18 (2003) 1006, 10.1029/2002PA000767.

[42] T. Fronval, E. Jansen, Late Neogene paleoclimates andpaleoceanography in the Iceland-Norwegian Sea: Evi-dence from the Iceland and Voring Plateaus, Proc. ODPSci. Results 151 (1996) 455^468.

[43] N.W. Driscoll, G.H. Haug, A short circuit in thermoha-line circulation: A cause for Northern Hemisphere glaci-ation?, Science 282 (1998) 436^438.

EPSL 6629 5-5-03


the closing of a seaway: ocean water masses and global climate change

Documents