Last Glaciald 13C Distribution and Deep-Sea Circulation in ... › ~apau › publications ›...

From WEFER G, MULITZA S, RATMEYER V (eds), 2003, The South Atlantic in the Late Quaternary: Reconstruction of MaterialBudgets and Current Systems. Springer-Verlag Berlin Heidelberg New York Tokyo, pp 695-722

Last Glacial δ13C Distribution and Deep-Sea Circulationin the Atlantic Ocean: A Model - Data Comparison

M. Matthies*, T. Bickert and A. Paul

Universität Bremen, Fachbereich Geowissenschaften, Klagenfurter Straße, 28334 Bremen, Germany

* corresponding author (e-mail): [email protected]

Abstract: We used a carbon cycle model (HAMOCC2) coupled to a general ocean circulation model(LSG) to explore the δ13C distribution in the glacial Atlantic Ocean. We compared the simulated δ13Cpattern with a new data set of benthic carbon isotopes of the Western and Eastern Atlantic from theLast Glacial Maximum (18,000 to 20,000 14C years or 21,000 - 23,500 calendar years before present).The model output fits the δ13C distribution derived from sediment samples, when the glacial export ofNADW to the Southern Ocean was reduced by 50 % and the inflow of glacial AABW was heldconstant. In most cases, the modeled δ13C pattern matched the paleodata within a range of ±0.2 ‰.Furthermore, the asymmetry between the glacial NADW distribution in the South Atlantic basinswas reproduced by the coupled ocean circulation and carbon cycle models. No additional increaseof the nutrient inventory in the deep ocean was necessary to reproduce the paleodata. Hence weconclude that a significant increase in biological pumping during glacials may not be necessary toexplain the reconstructed δ13C distribution in this region. The results are discussed with respect toother scenarios for the decrease of global atmospheric pCO2.

Introduction

The partial pressure of atmospheric carbon diox-ide (pCO2) in the in the Last Glacial Maximum(LGM), about 21,000 years ago is estimated toabout 200 ppm, about 80 ppm less compared to thepreindustrial value. The cause of this variation incarbon dioxide (CO2) has not yet been identified(Broecker and Henderson 1998; Archer et al.2000a; Sigman and Boyle 2000). In the recent lit-erature, there are many hypotheses trying to ex-plain this strong reduction in atmospheric pCO2.These hypotheses involve changes in the nutrientinventory, in the oceanic pH, in the ocean circula-tion, or a combination of these factors.

tivity in the surface ocean, whereby carbon is trans-ferred from the surface ocean to the deep-sea inthe form of sinking particles. Either an increase inthe ocean inventory of the nutrients phosphate(PO4

3-) and nitrate (NO3-) or a change in the ratio

of nutrients to carbon in phytoplankton could havestimulated the biological pump in this way. The ob-servation that iron availability limits phytoplanktongrowth in some parts of the ocean such as the South-ern Ocean provides a mechanism by which the bio-logical pump in high latitudes could have intensifiedin a dustier, more iron-rich glacial climate. Broeckerand Henderson (1998) suggested that an enhance-ment of glacial iron supply to the surface oceanstimulates the rate of nitrogen fixation, causing anincrease in the NO3

-: PO43- ratio of the deep-sea

and an increase in the effective nutrient reservoirof the ocean.

Biological Pump Scenarios

One group of mechanisms to lower the glacialpCO2 is to increase the rate of biological produc-

696 Matthies et al.

Ocean pH Scenarios

A second class of mechanisms to lower the glacialpCO2 is to change the pH or alkalinity of the wholeocean, converting seawater CO2 into bicarbonate(HCO3

-) and carbonate (CO32-) (e.g. Archer et al.

2000a). The pH in the ocean is controlled by themechanism of CaCO3 compensation. Any imbal-ance between the influx of dissolved CaCO3 origi-nating from chemical weathering on land and theremoval of CaCO3 by burial in the deep-seasediments will act to change the ocean pH until theflux balance is restored. It is more difficult to pre-serve CaCO3 in an acidic ocean. If the glacial rateof weathering were higher and the CaCO3 deposi-tion currently occurring in shallow waters wereshifted to the deep-sea, or the rate of CaCO3 pro-duction decreased, the CaCO3 burial efficiencywould increase and the ocean would become morebasic. If organic carbon production increased, itsdegradation in sediments would also promote cal-cite dissolution, further increasing the pH of theocean. CaCO3 compensation may also affect thepCO2 response to the biological pump scenariosdescribed above.

Another mechanism to reduce the atmosphericCO2 concentration without changing productivity oralkalinity has been proposed by Stephens and Keel-ing (2000). They suggest that a ventilation changewas driven by limitations of air-sea gas-exchangedue to an increase of sea-ice in the Southern Ocean.The mechanism that is proposed to explain the ob-served synchronisation between Antarctic tempera-ture and atmospheric CO2 calls for dramaticchanges in the structure of the deep ocean duringcold periods, including a collapse of NADW for-mation, an expansion of the volume of the deepocean filled with AABW, and the cooling of thisdeep layer virtually down to the freezing point(Keeling and Stephens 2001).

However, a study by Yu et al. (1996), based onmeasurements of the 231Pa/230Ta ratios in glacialand modern sediments, indicates that the export of231Pa from the Atlantic into the Southern Oceancontinued during the LGM at roughly the modernrate. They drew the conclusion that the export ofdeep water formed in the North Atlantic also con-tinued to move into the Southern Ocean at a com-parable rate during the LGM. If this were true, itwould falsify the hypothesized collapse of NADWformation during glacial cold periods, as first sug-gested by Duplessy et al. (1988) and also stated asa requirement by Keeling and Stephens (2001).

In the study presented here, we used a model-data comparison of δ13C changes in the deep At-lantic, in order to test the role of deep water over-turning as a main potential driving mechanism forglacial pCO2 reduction. A new data set of benthicstable isotopes from sediment cores in the SouthAtlantic (Bickert and Mackensen this volume) al-lows a more detailed reconstruction of the glacialnutrient distribution in this region. We will discussthe results in view of the various hypotheses de-scribed above.

Model Descriptions

Circulation Changes

Toggweiler (1999) reduced the deep water venti-lation in a box model that treats the boundary be-tween mid-depth and deep water masses as achemical barrier, separating the water with a lowCO2 content above from the water with high con-tents below. He generated reduced ventilation bydecreasing the vertical exchange between deep andAntarctic surface water. This mechanism invokesonly a physical process or a set of related physicalprocesses, that stratify the interior while reducingthe ventilation of the ocean's deepest water. Theseprocesses bottle up remineralised CO2 at depthsbelow 3,000 m and reduce the CO3

2- in the deepwater in contact with most of the ocean's CaCO3sediments. Through their control of the burial ofCaCO3, these ventilation and stratification proc-esses are able to drive atmospheric CO2 up anddown on timescales of several thousands up to tenthousand years. The mechanism works without anychanges in biological activity and requires only mini-mal changes in the distribution of nutrients.

Ocean General Circulation Model

The physical ocean model used in this study isknown as Hamburg Large-Scale GeostrophicOcean Model (LSG) (Maier-Reimer et al. 1993).The glacial and modern fields of the velocity, tem-

Last Glacial δ13C Distribution and Deep-Sea Circulation in the Atlantic Ocean 697

perature, salinity and convective adjustments werederived from the global model runs of Schäfer-Nethand Paul (2001) (labeled GM for the modern andG1 for a slightly modified version of the glacialsimulation) and represented the input fields for thecarbon cycle model. The modifications with respectto Maier-Reimer et al. (1993) include:• The air temperature advection has been removed.This procedure was originally used to bring thesimulated North Atlantic Deep Water (NADW)production to an appropriate magnitude of thepresent-day climate (see Schäfer-Neth and Paul2001).• The advection routine for temperature and salin-ity is formulated in accordance with the QUICKscheme of Leonard (1979), in order to compensatefor the strong reduction of NADW formation dueto the neglect of air temperature advection. Thisscheme results in a larger fraction of newly formedNADW that flows across the equator into the SouthAtlantic. The price to be paid is a smaller time-step,which is about four times less than the monthly time-step used by Maier-Reimer et al. (1993) and ne-cessitates a linear interpolation of the monthly meansea-surface temperature and salinity.• The sea-ice model of the original version of theLSG has not been included. Instead, the grid cellscovered with sea-ice are set to the freezing point(-1.8°C). Thus the model sea-ice cover has ratherbeen driven by monthly mean sea-surface tempera-tures.• A convection scheme, which makes the watercolumn completely stable after every time-step, hasbeen added.• Depth-dependent horizontal diffusivity valuesbetween 8 x 106 cm2 s-1 in the top layer and 4 x

Table 1. Ocean model forcing data for present-day and LGM boundary conditions.

106 cm2 s-1 in the bottom layer are employed. Valuesbetween 0.3 cm2 s-1 and 1.1 cm2 s-1 respectivelyhave been assumed for the vertical diffusivity.

The forcing data for the simulations underpresent-day and LGM boundary conditions arespecified in Table 1. The modified LSG ocean modelhas the same resolution, land-sea mask andbathymetry as the original model, except for theglacial experiment in which the Bering Strait isclosed.

The zonally integrated meridional overturning inthe modern Atlantic Ocean is shown in Figure 1,left. NADW is produced at a rate of 14 Sv, 8 Sv ofwhich is exported across the equator to the South-ern Ocean at depths below 1,500 m. There is a deepinflow of Antarctic Bottom Water (AABW) whichamounts to 4 Sv. The total quantity of deep water(NADW and AABW) which flows into the South-ern Ocean is 12 Sv, approximately 30 % less thanthe value of 18 Sv given by Schmitz (1995). Thewater mass boundary between NADW andAABW at a depth of 2,500 m is shallower thanestimated from observations by approximately1,500 m, which corresponds to one model layerthickness. The reason for this is probably that thenewly formed NADW is too warm (1 to 2°C)above 3,000 m in the northern North Atlantic Oceanand thus not dense enough to sink to an appropri-ate depth. However, this is a general problem ofcoarse resolution models. For a detailed descriptionsee Schäfer-Neth and Paul (2001).

In the glacial Atlantic (Fig 1 b), the NADW ex-port to the Southern Ocean is 4 Sv, a reduction of50 % compared to the modern; the glacial NADWproduction amounts to 6 Sv. The inflow of glacialAABW remains unchanged at 4 Sv and the bound-

698 Matthies et al.

Fig. 1. Annual mean of zonally averaged Atlantic meridional circulation. Contour interval is 2 Sv (1 Sv =106 m-3 s-1). Negative contour lines indicate anti-clockwise circulation, whereas positive values indicate clockwisecirculation. a) Simulation corresponds to the present-day boundary conditions. b) Simulation corresponds to theLGM boundary conditions.

ary between the Deep Current System and theBottom Current System is only slightly shallowerthan in the modern ocean.

oxygen, POC (particulate organic carbon, plank-ton soft tissue), and particulate CaCO3 (planktoncalcareous shells). For all tracers that include car-bon, the three carbon isotopes (the stable 12C and13C and the radiocarbon 14C) were considered. Theorganic matter (Corg and CCaCO3) produced in theuppermost model layer, depending on the nutrientand light availability, is immediately redistributedwithin the water column because the time-step of1 year is long relative to the time scale of sinkingparticulate organic matter (about 100 m day-1, e.g.Suess 1980). The prescribed profiles for the or-ganic carbon flux through the water column are dif-ferent for POC and CaCO3 and decrease ex-ponentially with depth (with an e-folding depth of800 and 4,000 m, respectively).

After each model run, the total inventory of 13Cwas calibrated to a preanthropogenic atmosphericδ13C value of -6.5 ‰, according to Friedli et al.(1986). The model configuration of Heinze et al.(1991) was modified in several respects:

Carbon Cycle Model

The model used here is the Hamburg Model of theOceanic Carbon Cycle (HAMOCC2) (Maier-Reimer and Hasselmann 1987; Heinze et al. 1991).Only a brief outline and modifications to Heinze etal. (1991) are given in the following. The model runson a global 72 x 72 E-grid (Arakawa and Lamb1977), with a horizontal resolution of approximately3.5° x 3.5° and uses 11 vertical layers with layerinterfaces at 0, 50, 112.5, 200, 350, 575, 850, 1,500,2,500, 3,500, 4,500, and 6,000 m, and has a time-step of 1 year. A quasi steady state is reached af-ter 20,000 model years. The model tracer variablesinclude atmospheric pCO2, ΣCO2 (total dissolvedinorganic carbon), total alkalinity (including the con-tribution of borate), PO4

3- (as biolimiting nutrient),

a) b)


• A modified profile of the vertical distribution ofPOC was employed. Originally, about one-third ofthe new production was assumed to fall immedi-ately to the bottom layer to account for organicmaterial coated by hard shells. This formulation waschanged in such a manner that no fraction of thenew production falls immediately to the bottomlayer, leading to some improvements in the tracerdistributions. First, this formulation yielded a greaternew production (from 6.5 to 6.7 GtC year-1) andlower CaCO3 production (from 1.3 to 1.1 GtCyear1), accompanied with a corresponding declinein the atmospheric pCO2 from 304 ppm to thepreindustrial value of 281 ppm. Second, it reducedthe models overestimation of the POC water in-ventory from 223 to 177 GtC year-1 and that of thesediment pool (from 90 to 24 GtC year-1). Third,this profile led to a better tracer distribution espe-cially in the bottom layer of the model, where theoriginal profile yielded an overestimation of the con-centrations in the northern parts of the Atlantic andPacific basin, as compared to the GEOSECS data.By use of this profile, the simulated δ13C values inthe western Atlantic basin were reduced from 0.6to 0.4 ‰ in the source areas of the AAIW, and thewater mass characteristics were more pronouncedfor AAIW, NADW and AABW. Consequently, abetter quantitative comparison with the GEOSECSdata was accomplished.• An explicit vertical diffusion is included with thesame slight depth dependency as in the ocean model(see above), to make the carbon cycle model moreconsistent with the ocean model. In general thevertical diffusion tends to reduce the vertical gra-dients, especially in the deep water column. Byusing the explicit vertical diffusion, the global newproduction increased from 6.1 to 6.7 GtC year-1,and CaCO3 production increased from 0.9 to 1.1GtC year-1 in the control run. The atmosphericpCO2 rose from 278 to 281 ppm. This leads to ad-vantages for some tracers, but others show aworse distribution than without explicit vertical dif-fusion. The advantages were most obvious in thezone of the phosphate maximum and oxygen mini-mum in the north and equatorial Pacific, where aconsistence with GEOSECS could be reached.However, beneath this zone, the tracer values werehigher in δ13C and phosphate and lower in oxygen,

as compared to GEOSECS. Implications of theδ13C concentration in the Atlantic consisted in asmall rise of 0.1 ‰ in the northern part (north of30°N) of the bottom water, and a same enhance-ment in the region between 30°S and 30°N abovea water depth of 800 m.• A constant Redfield stoichiometry for POC isused, with C:N:P:O equivalent to 117:16:1:-170according to Anderson and Sarmiento (1994).• A rain ratio (CCaCO3 to Corg) is prescribed whichis lower than the commonly used value of 0.20 -0.25. Here, we used a maximum rain ratio valueof 0.15, following Bacastow and Maier-Reimer(1990), cf. Yamanaka and Tajika (1996). Due to thischange in the rain ratio, all tracers which includedCaCO3 were influenced, especially the CaCO3production which is reduced from 1.7 to 1.1 GtCyear-1, if a value of 0.24 is applied. As a result,surface alkalinity and the carbonate ion concentra-tion increased and the atmospheric pCO2 de-creased from 313 to 281 ppm. The effects on theδ13C distribution in the Atlantic were negligible, theextension of the δ13C minimum to the south wasenlarged. The effect on the carbonate ion concen-tration was a better match of the calcite lysoclinedepth in the Southern Ocean with GEOSECS, andresulted in a deeper lysocline of about 400 m in thisregion.

The effects of these parameterisations on themodel inventories and mean values are summarizedin Table 2 and 3, respectively.

Controls on δ13C of Marine Carbon

In the marine carbon cycle, two fractionation proc-esses are important. First, during photosynthesis,organisms preferentially take up the lighter carbonisotope 12C, increasing δ13C in the surface oceandissolved as inorganic carbon (DIC). When iso-topically light organic matter is remineralised, δ13Cof marine ΣCO2 decreases. This fractionation ofabout -20 ‰ of marine photosynthesis leaves thesurface ocean depleted in nutrient with a high con-tent of δ13C, whereas the nutrient-rich deep wa-ters display low δ13C values. Through this mecha-nism, known as biological pump, carbon and nutri-ents are transported to the deep-sea and therebyreduce the CO2 in the surface and atmosphere.

700 Matthies et al.

Table 2. Implications of sensitivity experiments under various model parametrisations on the tracer inventories forthe modern ocean after 20,000 integrations.

Table 3. Global mean values of the modern ocean for the sensitivity experiments above.

Second, seawater δ13C can also be altered viaexchange with the atmospheric CO2, without anyassociated changes in nutrient concentrations. Iso-tope equilibrium of surface ocean waters with at-mospheric CO2 will cause higher δ13C in cold sur-face waters, and lower δ13C in warm surfacewaters (Broecker and Maier-Reimer 1992). Thedifference between the equilibration of CO2 and theisotopes is that the isotopes require more time toreach equilibrium (about 10 years compared toabout 1 year, depending on the depth of the mixedlayer and the gas-exchange rate). Because surfacewaters move about and are replaced on fastertimescales than this, there is no region of the oceanwhere surface water carbon is in complete isotopicequilibrium with the atmosphere (Broecker andMaier-Reimer 1992). To extract the effect of thegas-exchange signature from biological induced

alterations, the artificial tracer (δ13Cas) could beconstructed according to Broecker and Maier-Reimer (1992). If there were no air-sea exchange,the relationship between δ13C and PO4

3- in theocean would be

δ δ13 134C C

COCP

PO POmeanphoto

2 mean4 mean− = ⋅ ⋅ −

∆

Σ( ).

(1)

The suffix mean denotes average ocean con-centrations and ∆photo and C/P describe the meanδ13C in particulate organic matter and the Redfieldratio between organic carbon and phosphorus, re-spectively. When reasonable values are substituted(δ13Cmean = 0.3 ‰, ∆photo = -20 ‰, ΣCO2 mean =2200 µmol kg-1, C/P = 128 and PO4

3-mean = 2.2 µmol

kg-1), the predicted relationship between δ13C and


PO43- closely matches the relationship for waters

of the deep Indian and Pacific Oceans (13C =2.7 - 1.1 x PO4

3-). This is to be expected, as theeffect of air-sea exchange should be constant fortheses deep waters on account of the homogene-ity of source waters of the deep Indian and PacificOceans. The degree to which air-sea exchangeprocesses have affected the surface ocean δ13Ccan determined by subtracting the δ13C value pre-dicted from biological cycling exclusively from theactual δ13C,

δ δ13 134Cas = C - (2.7 -1.1 PO ).⋅ (2)

By definition, water with δ13Cas of 0.0 ‰ hasthe same air-sea exchange signature as the meanocean deep water. A positive value of δ13Cas meansthat the water reveals more of an influence of air-sea exchange at cold temperatures and less atwarmer temperatures, whereas a negative δ13Casvalue implies less δ13C enrichment due to air-seaexchange than for average deep ocean water.However, this formulation will only represent thetrue effects of air-sea exchange if the assumptionsabout constant C/P ratios and constant δ13C oforganic material and reasonable biologically in-duced changes in oceanic carbon content are suf-ficiently accurate. For example, Antarctic surfacewater, where δ13C of organic matter is substan-tially lower than in the rest of the ocean, δ13Cas willlead to an overestimation of the effects of the air-sea exchange. For the glacial ocean, we used thesame formulation according to Winguth et al.(1999), but assume the same Redfield slope as forthe modern ocean,

δ δ13Cas = C - (2.5-1.1 PO ).134⋅ (3)

A further effect on the marine δ13C came fromthe biosphere where the organic δ13C is lower(about -25 ‰) than the marine δ13C (about -20 ‰).Changes in the vegetation, soil composition andtransport of carbon via continental fluxes to theocean could influenced the concentration of δ13C.Estimations of this effect on the LGM δ13C rangedfrom 0.3 to 0.4 ‰ (Curry et al. 1988; Duplessy etal. 1988).

Carbon Cycle Modelling Results

The effect of the ocean circulation on the δ13Cdistribution, appart from the advection of tracers,depends on the residence time of a water mass.The longer a water mass resides in the deep-sea,the more organic material degrades within it, con-suming oxygen and depleting 13C. Circulationchanges were accompanied with changes in thetemperature and salinity which influences the solu-bility and dissociation constants of CO2.

Modern Atlantic (Control Run)

The reference run (control run) of the carbon cy-cle model yielded a global primary production rateof 8.62 GtC year-1 and a new production rate (de-fined as the amount of newly formed organic mat-ter which is transferred from the surface layer intodeeper layers each year and is thereby removedfrom the seasonal ocean-atmospheric exchangecycle) of 6.7 GtC year-1 (see Table 2). Themagnitudes of these values were within the rangeof results of other global biogeochemical models(e.g. Yamanaka and Tajika 1997; Marchal et al.1998; Murnane et al. 1999). Estimates from ob-served data of primary and new production weredifficult to compare directly with model outputs,because they included different methods for deter-mining the various kinds of production. Estimationsof the global new production ranged between 5.0and 22.0 GtC year-1 (Murnane et al. 1999). Thepreindustrial atmospheric pCO2 became stationaryat 281 ppm and was in good agreement with esti-mates from ice cores . The global mean δ13C valueamounted to 0.38 ‰ (see Table 3), and was slightlylower than the value of 0.50 ‰ reported byBroecker and Maier-Reimer (1992).

The simulated present-day δ13C distributionsalong the Geochemical Ocean Section Study(GEOSECS) sections in the Western and EasternAtlantic are shown in Figure 2. The GEOSECS dataset has been corrected according to Kroopnick(1985). For the Eastern Atlantic transect, we ex-tended the GEOSECS data set to the south usingthe data set of Mackensen et al. (1996). TheGEOSECS transect in the Western Atlantic re-mained unchanged. The main features of the dif-

702 Matthies et al.

Fig. 2. Meridional distribution of δ13C in the modern Atlantic Ocean. Contour interval is 0.2 ‰, see small panel forlocation of the section. a) and b) Simulated distribution in the Western and Eastern Atlantic, respectively. c) Ob-served δ13C distribution in the Western Atlantic (GEOSECS), corrected according to Kroopnick (1985). d) GEOSECSdistribution of δ13C in the Eastern Atlantic. The GEOSECS transect is extended to the south according to Mackensenet al. (1996).

ferent water masses are adequately reproduced bythe model. NADW and AABW can be clearly dis-tinguished, with high and low δ13C values, respec-tively. A shortcoming in the simulated δ13C distri-bution was the shallower level of the NADW wa-ter mass. The core of this water mass was aboutat 2,500 m depth in the Western Atlantic, whereasin GEOSECS it was about 500 m deeper, but asmentioned above, this represents one model layerat this depth level. This discrepancy was primarilya result of the lower than observed meridional over-turning. The simulated minimum of δ13C in theAntarctic Intermediate Water (AAIW) compareswell with data, but the values in the AABW seemedto be up to 0.2 ‰ higher, especially in the WesternAtlantic.

In Table 4 we calculated the root mean squareerror between the control run and GEOSECS datafor the entire Atlantic. Table 4 reveals a greaterdiscrepancy of the simulated δ13C distribution in themodern northern Atlantic compared to the south.This was a further implication for the underestima-tion of the simulated NADW production.

The contribution of the gas-exchange signature(δ13Cas) in the eastern Atlantic is shown in Figure3. At the ocean surface, the contribution of theδ13Cas signature is mainly due to the temperaturedependence of the gas-exchange formulation (seeHeinze et al. 1991). Cold regions like the SouthernOcean show higher δ13Cas values than the warmwaters of the equator and subtropical gyres, withmostly negative values relative to the deep-sea.

a) b)

d)c)


Table 4. Root mean square error between simulated δ13C values for the control run, glacial experiment 1, 2, 3 andobserved data in the entire Atlantic Ocean. Modern data: From GEOSECS, corrected according Kroopnick (1985)and Mackensen et al. (1996). LGM data: From Sarnthein et al. (1994) north of the equator and south of the equator,see Bickert and Mackensen this volume.

Fig. 3. Meridional section of the gas exchange signature (δ13Cas) in the Eastern Atlantic, see small panel for loca-tion of the section. a) Latitude-depth section of δ13Cas for the control run. Contour interval is 0.1 ‰. b) Anomalybetween glacial experiment 1 and control run. c) Anomaly between glacial experiment 2 and control run. d) Anomalybetween glacial experiment 3 and control run. Contour interval for the anomalies are 0.05 ‰.

a) b)

c) d)

704 Matthies et al.

Thus the Southern Ocean surface water is enrichedin δ13C from the gas-exchange with the atmosphereat cold temperatures, whereas the warm waterregions are reduced in δ13C. However, in the sur-face water of the Southern Ocean, upwelling ofdeep water (δ13Cas = 0.0 ‰) influences the gas-exchange signature and the observed low δ13C inorganic matter would tend to overestimate the con-tribution of the gas-exchange signature in this area.The high δ13Cas values of about 0.5 ‰ in the sur-face water of the Weddell Sea - according toLynch-Stieglitz et al. (1995) - were not reproducedby the model. One reason could be the use of aconstant gas-exchange rate, but Lynch-Stieglitz etal. (1995) have shown that its influence was low.It is most likely that the simulated surface δ13Cvalues were underestimated, whereas the PO4

3-

concentration remained in the range of the obser-vations. The deep water column in the north re-flected the transport and mixing of NADW carriedthe gas-exchange signature to the south. TheNADW should have a δ13Cas of about -0.4 ‰ ac-cording to Broecker and Maier-Reimer (1992),whereas our simulation reached a value of about -0.2 ‰.

duction of 0.2 ‰ as compared to the modern state,whereas the upper ocean north of the equatorshowed an increase in δ13C. A possible cause ofthe simulated decrease of δ13C in the deep oceanis aging of the deep water masses due to the re-duced NADW formation and export to the South-ern Ocean. The influence of the gas-exchange sig-nature on the δ13C, due to the lower glacial tem-peratures, showed only minor changes in the deepwater column compared to the control run (Fig. 3).Only the surface ocean is significantly affectedby the air-sea exchange, especially in the South-ern Ocean near the sea-ice edge. The effect ofchanges in the pattern of the remineralisation oforganic matter on the decrease of the δ13C shouldbe minor, since the global new production is lowerin this simulation.

In summary, the simulated distribution of δ13Cin the Atlantic Ocean showed a lowering in theδ13C of 0.2 ‰ in the deep water column, whichwas only due to a changed thermohaline circula-tion field. However, the effect of this circulation onthe pCO2 in the atmosphere was too small to ex-plain the concentrations measured in ice cores. Thissupports previous findings of Heinze et al. (1999)and Winguth et al. (1999). Therefore, we setup amodel run with an intensified biological pump.Glacial Atlantic (Experiment 1)

To investigate the response of the δ13C distributionin the Atlantic Ocean to climate change, we per-formed three model experiments for the glacialAtlantic. Experiment 1 used a modified tempera-ture, salinity, circulation field and more extended icedistribution as driven by glacial ocean circulation.Experiment 2 introduced, in addition, an initial nu-trient inventory which was increased by 30 %. Toreduce the pCO2 in the atmosphere without signifi-cant changes in the δ13C values we setup an modelrun with a chemical glacial ocean according toSanyal et al. (1995). As a result, experiment 1showed a reduced new production rate of 5.9 GtCyear-1 compared to 6.7 in the control run, and thechange in the global mean of δ13C amounted to 0.15‰. The average atmospheric pCO2 was loweredby 1 ppm and attains a value of 280 ppm. The dif-ferences in δ13C in the Atlantic are shown in Fig-ure 4 for the Western and in Figure 5 for the East-ern Atlantic. In the deep Atlantic, there was a re-

Glacial Atlantic (Experiment 2)

In this experiment, we increased the initial nutrientinventory of the glacial ocean by 30 %. Togetherwith the glacial circulation field, this yields a newproduction rate of 7.5 GtC year-1. As a result, theatmospheric pCO2 was reduced to 221 ppm. Thisapproaches the observed pCO2 value of approxi-mated 200 ppm for the LGM (e.g. Barnola et al.1987; Petit et al. 1999). The global mean δ13C valueis reduced to -0.47 ‰. The enhancement of thenew production is accompanied by an additionalreduction of the δ13C in the whole water columnof the Atlantic Ocean due to the remineralisationof organic matter (Fig. 6 and 7). Under these con-ditions, large parts of the deep water column re-veal negative δ13C values. The gas-exchange sig-nature shows no significant changes compared toexperiment 1, except in the upper ocean of theSouthern Ocean, but this has no influences on the


Fig. 4. Meridional distribution of δ13C in the Western Atlantic for glacial experiment 1. Contour interval is 0.2 ‰.a) Anomaly between experiment 1 and control run. b) Simulated δ13C distribution.

a)

b)

706 Matthies et al.

Fig. 5. Meridional distribution of δ13C in the Eastern Atlantic for glacial experiment 1. Contour interval is 0.2 ‰.a) Anomaly between experiment 1 and control run. b) Simulated δ13C distribution.

a)

b)



a)

b)

708 Matthies et al.


a)

b)


deep water column in this region (compare Fig. 3bwith 3c).

shows a strong depression relative to the sourcewater (Fig. 3d).

Glacial Atlantic (Experiment 3)

Experiment 3 was based on the observations ofSanyal et al. (1995) that the pH values of the gla-cial deep waters in the Atlantic and Pacific are 0.3±0.1 higher as compared to the Holocene period.If the observed higher glacial pH was the result ofexcess calcite dissolution, then 0.157 x 10-3 molesof CaCO3 must have been added to each kilogramof sea water. This corresponds to 0.161 x 10-3 moll-1, if a mean density of 1.025 kg l-1 is assumed. Wetranslated these conditions to an increase of theinitial model concentrations of 161 µmol l-1 ΣCO2and 2 x 161 µequiv l-1 alkalinity. To mirror an in-crease of Corg in the sediment we halved the maxi-mal rain ratio to 0.075.

It must be considered that this experiment doesnot take into account the complex interactions be-tween the sediment and the overlying water. In thisexperiment we introduced the effects of these in-teractions according to the assumptions of Sanyalet al. (1995), because the model has only a verysimple sediment module with a constant exchangerate between the bottom water and sediment layer.The new production is the same as in experiment1, only the CaCO3 production is lower because ofthe reduced rain ratio (from 0.15 to 0.075). Theimplication on the δ13C distribution is minimal (Fig.8 and 9) and the pCO2 decrease to 250 ppm due tothe weakening of the carbonate pump. The pro-posed rise in the pH value of 0.3 in the glacial deepAtlantic according to Sanyal et al. (1995) is notpredicted by the model. A maximum elevation of0.2 in the pH value is found in the northern IndicOcean at a water depth of 1,500 m, however, inthe Atlantic we only reached a maximal increaseof 0.1 at about 1,000 m. The reduction of the at-mospheric pCO2 is purchased at the expense of adeepening of the lysocline in the glacial Atlantic,which contradicts observations that show an up-ward movement of the lysocline. The simulatedgas-exchange signature shows roughly the sameanomaly patterns as in the other experiments, withan exception in the Southern Ocean, where δ13Cas

Comparison of Model Results withPaleoclimate Data

In Figures 10 through 13 we compare the distribu-tions of δ13C in glacial simulations and in thepaleoclimate records of the Western and EasternAtlantic. A new data set of benthic stable isotopesfrom sediment cores in the South Atlantic (Bickertand Mackensen this volume) combined with thedata of Sarnthein et al. (1994) for the eastern NorthAtlantic, allows a more detailed reconstruction ofthe glacial nutrient distribution in this region mostsensitive to circulation changes. For the South At-lantic, only values of sites outside of upwelling ar-eas have been chosen for the reconstruction of δ13Cdistributions in the past. Several studies from dif-ferent locations in the South Atlantic have shownthat an additional depletion in the δ13C of epibenthicforaminifer calcite occurs in areas characterizedby high surface water productivity and hence highorganic matter supply to the seafloor (e.g.Mackensen et al. 1993; Bickert and Wefer 1999).This depletion is most likely explained to be causedby the decay of organic matter, reducing 13C/12Cratio in the pore water, which influences to somedegree the carbon isotopic composition of the C.wuellerstorfi shells of high-productive areas.Therefore, we removed the values of such loca-tions prior to the model-paleodata comparison (seeBickert and Mackensen this volume, for a detaileddiscussion).

As a result, there is a reasonable fit of the modeloutput of experiment 1 to the δ13C distribution de-rived from sediment samples, when the glacialexport of NADW to the Southern Ocean is re-duced by 50 % and the inflow of glacial AABW isheld constant (Fig. 10 and 11, Table 4). The obtainedδ13C pattern from the model experiment matchesthe paleodata within a range of ±0.2 ‰ for mostof the data. Furthermore, the asymmetry betweenthe glacial NADW distribution in the South Atlan-tic basins is reproduced by the coupled ocean cir-culation and carbon cycle model. However, somedata points in the mid-depth Western Atlantic be-tween 15° and 30° south would suggest the glacial

710 Matthies et al.


a)

b)



a)

b)

712 Matthies et al.

Fig. 10. Comparison between the glacial δ13C distribution of experiment 1 and observations (data set of SFB 261)in the Western Atlantic. Contour interval is 0.1 ‰, see small panel for locations and data points. The colours of thedata points (cycles) corresponds to the colour values in the simulation. If possible, the observed δ13C values arewritten next to the corresponding data point.

NADW to extend deeper and further south than themodel does. This might be related to the underesti-mation of the production of NADW as alreadydiscussed for the control run of the modern ocean.

The much larger differences between the modeloutput of experiment 2 and the paleodata (Fig. 12and 13) indicate that there is no additional increaseof the nutrient inventory of the deep ocean neces-sary to reproduce the paleodata. From this obser-vation we derive that no significant increase in bio-logical pumping occurred during glacials. As a con-sequence, the drop-down of atmospheric CO2 couldnot be explained by the mechanism of an intensi-fied biologic pumping related to an enhanced ex-port production. This is in accordance with recentobservations that the export production in the gla-cial Southern Ocean shows no significant changes

(François et al. 1997; Nürnberg et al. 1997). How-ever, our results would even not support the recentlypresented hypothesis by Stephens and Keeling(2000) that the decrease in atmospheric CO2 wasdriven by a reduced outgassing of the upwellingdeep water due to an extended sea-ice in the gla-cial Southern Ocean. Using a box model with deepwater upwelling confined to south of 55°S, theysuggested that low glacial atmospheric CO2 levelsmight result from reduced deep water ventilationassociated with either year-round Antarctic sea-icecoverage, or wintertime coverage combined withice-induced stratification during the summer. As aresult, their box model reproduces 67 ppm of theobserved glacial-interglacial CO2 difference and isgenerally consistent with the additional observa-tional constraints.


Fig. 11. Comparison between the glacial δ13C distribution of experiment 1 and observations (data set of SFB 261and Sarnthein et al. 1994) in the Eastern Atlantic. Contour interval is 0.1 ‰, see small panel for locations and datapoints. The colours of the data points (cycles) corresponds to the colour values in the simulation. If possible, theobserved δ13C values are written next to the corresponding data point.

Figure 14 shows the pCO2 difference betweenocean and atmosphere for our control run and forthe glacial experiment 1 as well as the anomalybetween both simulations. Negative values indicatethat the ocean is a sink for atmospheric CO2 andpositive values indicate that the ocean is a CO2source. In both simulations the Southern Oceanrepresents a CO2 sink.

In Table 5 we used the same area-weightedzones for the average ∆pCO2 according toTakahashi et al. (1997). For the control run theSouthern Ocean represented a strong CO2 sink andthe equatorial region a strong source. The simulatedsource and sink ∆pCO2 pattern is comparable toTakahashi et al. (1997). Quantitative differencesbetween simulations and observations are in thehigh latitudes of both hemispheres and primarily due

to the fact that the observations are normalized tothe year 1990 and the data coverage in the South-ern Ocean is scarce. On the other side, the controlrun represented the preindustrial ocean. The simu-lated strong CO2 sink in the Southern Ocean issupported by the observations of Schlitzer (2002)who predicted a significant stronger export produc-tion in this region as satellite observations indicated.This should result in a stronger CO2 input to thesurface ocean. In the Atlantic sector of the South-ern Ocean this sink seems to be stronger in theglacial run compared to the control run. The mini-mum value in this region amounts to -80 ppm,whereas in the control run it amounts to -60 ppm.However, the area of the pCO2 sink is reducedcompared to the modern conditions due to the ex-tended sea-ice coverage. This is also taken from

714 Matthies et al.

Table 5 where the average ∆pCO2 values show areduction in the CO2 sink between 50°S and 90°Sdue to the sea-ice expansion compared to the con-trol run. An intensification is predicted in the zonebetween 14°S and 50°S. Together with the biologi-cal new production which is only slightly strongerin the vicinity of the sea-ice edge in this region, thecontribution to lowering the global mean atmos-pheric pCO2 appears to be small. Based on ourmodel simulations the extended sea-ice cover in theLGM could not have been responsible for the strongreduction of the global atmospheric CO2.

The reason for the obvious discrepancy betweenthe results of Keeling and Stephens (2001) and ourmodel output might be that box models and gen-eral circulation models (GCM) show a strong dif-ferent behaviour to changes in the high latitudes

(Archer et al. 2000b). As an example the GCMsimulation of Archer et al. (2000a) under conditionsof a more efficient nutrient utilization in the glacialSouthern Ocean as a result of a higher iron supplyshows no sufficient atmospheric CO2 reduction,whereas box models (see references in Archer etal. 2000b) reaches the glacial CO2 concentration.This different behaviour of the models to changesin high latitudes was discussed in Archer et al.(2000b) and they concluded that box models over-estimate high latitude sensitivity of the real ocean.This uncertainty should be examined in more de-tail.

The dramatic changes in the circulation andassociated exchange processes in the SouthernOcean as proposed by Toggweiler (1999) are al-ways accompanied with changes in the tempera-

Fig. 12. Comparison between the glacial δ13C distribution of experiment 2 and observations (data set of SFB 261)in the Western Atlantic. Contour interval is 0.1 ‰, see small panel for locations and data points. The colours of thedata points (cycles) corresponds to the colour values in the simulation. If possible, the observed δ13C values arewritten next to the corresponding data point.


Fig. 13. Comparison between the glacial δ13C distribution of experiment 2 and observations (data set of SFB 261and Sarnthein et al. 1994) in the Eastern Atlantic. Contour interval is 0.1 ‰, see small panel for locations and datapoints. The colours of the data points (cycles) corresponds to the colour values in the simulation. If possible, theobserved δ13C values are written next to the corresponding data point.

Table 5. Global annual mean values of sea-air pCO2-difference (∆pCO2, in ppm) for the control run and three glacialexperiments for five latitude zones (area-weighted averages). The five latitude belts were selected according toTakahashi et al. (1997).

ture and salinity. Such rearrangements in the wa-ter column required, in particular, a change in thesalinity that is unfortunately poorly documented fora few locations in the glacial Southern Ocean.

Ultimately, the influence of increased sea-ice inthe Southern Ocean and/or a northward shift of

glacial wind field on the reduction of the atmos-pheric CO2 concentration, as was suggested byElderfield and Rickaby (2000) and Sigman andBoyle (2000), is related to the different behaviourof the two model types. Winguth et al. (1999) haveshown that an increasing glacial wind in the high

716 Matthies et al.

Fig. 14. Annual mean sea-air difference in the CO2 partial pressure (∆pCO2). Contour interval is 20 ppm. Positivevalues indicate that the ocean is a source for atmospheric CO2, negative values indicate CO2 sinks. a) Horizontaldistribution of ∆pCO2 for the control run. b) Distribution of ∆pCO2 for the glacial experiment 1. c) Difference be-tween experiment 1 and control run. In the two upper panels, black filled areas denote the simulated annual meansea-ice distribution.

a)

b)

c)


southern latitudes has no significant effect on theδ13C pattern and the atmospheric pCO2.

In our opinion, the complex mechanism of car-bonate compensation represents a common way ofreducing the atmospheric pCO2, without produc-ing a strong reduction of the δ13C. However,Archer et al. (2000a) have shown in a model simu-lation which included a complex sediment modulethat this process works in the right direction, butviolates other paleoevidences from sediments, e.g.changes in the depth of the lysocline and theCaCO3 distribution on the seafloor.

In experiment 3, we mimiced the mechanism ofthe carbonate compensation according to Sanyalet al. (1995) and could reduce the atmosphericpCO2 to 250 ppm with a δ13C pattern like in ex-periment 1 (see Figs. 8, 9 and Table 4). However,the advantage of the pCO2 reduction in the atmos-phere and the reasonable reproduction of the δ13Cdistribution is obtained at the expense of carbon-ate distributions which are not in agreement withobservations. For example, the glacial lysocline inthe Western Atlantic is deeper (500 to 1,000 m) ascompared to the modern situation, whereas obser-vation indicated a rise of the glacial lysocline(Henrich et al. this volume). The estimated pH re-duction according to Sanyal et al. (1995) in the gla-cial Atlantic was not reproduced by the model.

Figure 15 shows another comparison betweenmodelled results and observations in terms of stand-ard deviation and correlation coefficient.

After interpolating modelled data to the observeddata levels, the standard deviation and the correla-tion coefficient were calculated for both data sets(model experiments and observations). These datapairs were transferred as polar coordinates into ascatter diagram according to Taylor (Gates et al.1998; Taylor 2001). The procedure of the conver-sion of the value pair in the Taylor diagram wasdone in three steps:1. Normalization of the standard deviation of themodel simulation and observation was defined asσnorm= σ2

model x σ-2observation.

2. The polar angle, α, was calculated as follows:α = arccos r. The parameter r denotes the corre-lation coefficient.3. The determination of the values for the x- andy-axis (the polar coordinates) was calculated by us-

ing the equations x = σnormx cos α and y = σnorm xsin α.

The Taylor diagram showed the degree of cor-respondence between simulated and observed δ13Cfields in the Atlantic Ocean. By definition, all ob-servations showed a normalized standard deviationof 1 and an angle of zero degrees (i.e. the correla-tion coefficient is equal to 1). This point was ac-cepted as reference point which determines thelocality of the various observed data sets used inthis diagram and always have the same position.The distance of the simulated tracer point from theorigin of the coordinate system is equal to the stand-ard deviation (normalized by the observed stand-ard deviation), whereas the distance from the ref-erence point (observations) is equivalent to the rms(root mean square) pattern difference between theobserved and modeled data fields (again normal-ized by the observed standard deviation). The co-sine of the polar angle corresponds to the correla-tion between the simulated and observed data sets.Thus, a model that is relatively accurate would lienear the dashed arc (indicating it had the correctvariance) and close to the observed reference pointindicating a small rms error and a high correlationcoefficient. The outer circular arc in Figure 15 de-scribes the correlation coefficient. In Figure 15 a),we show the comparison between the model simu-lations (control run, glacial experiment 1 and 2.Please note that experiment 3 shows the sameposition as experiment 1 and remains unmentionedin the legend of Figure 15) and the observed datasets (modern: GEOSECS, glacial: Sarnthein et al.1994). The glacial modeled δ13C fields are moreclosely resembled the observation fields (in termsvon standard deviation, correlation coefficient andrms error) than the modern modeled δ13C field. Thereason may probably be attributed to the previousfinding of low NADW production. The glacial simu-lation of experiment 2 substantially presents a bet-ter agreement between the observed δ13C patternof Sarnthein et al. (1994) than the simulation ofexperiment 1 (and experiment 3).

As shown in Figure 15 b), we combined the ob-served δ13C data set of GEOSECS with the dataof Mackensen et al. (1996). The observed glacialdata set include data of Sarnthein et al. (1994),relating to the north of the equator, otherwise, δ13C

718 Matthies et al.

Fig. 15. Comparison between simulated δ13C pattern with observed δ13C in the entire Atlantic Ocean. The graphicalrepresentation is according to Taylor (Gates et al.1998, Taylor 2001), see text for details. a) Simulated δ13C patternfor the control and the both glacial runs (experiment 1 and 2) in comparison with the GEOSECS observations andthe data of Sarnthein et al. (1994). Note that experiment 3 in this diagram shows the same position as experiment 1and remains unmentioned in the legend. b) The same as left, but for extended data sets. For the modern case theGEOSECS data are replenished to the south according to Mackensen et al. (1996). For the comparison with theglacial simulations, we used the data of Sarnthein et al. (1994) north of the equator, otherwise the SFB 261 data setis used.

data measured by the research program of SFB 261were used. For the modern situation, this led to aslightly improved conformity with these data setsat least as far as the correlation coefficient wasconcerned. The situation of the glacial simulationshowed an opposite behaviour in the δ13C patternbetween experiment 1 (and 3) and 2. The simula-tion in experiment 1 (and 3) now reveals a betterconformity with the observation than could bearchieved in experiment 2.

circulation conditions have changed. As comparedto the modern ocean, the data in the glacial East-ern Atlantic show generally very low δ13C concen-trations in the deep water column between 20°Nand the equator. In the Eastern Atlantic, the simu-lated δ13C concentrations correspond well to theobservations. In general, deviations are in the rangeof ±0.2 ‰, with an exception around 20°N wherethe model produced 0.4 ‰ higher concentrationsas compared to the observations. However, in spiteof the good correlation between glacial simulationand observation the reduction of the atmosphericCO2 partial pressure is too low. Thus, the modelcould reproduce the observed δ13C pattern, but itwas not able to simulate the glacial/interglacialpCO2 change of about 80 ppm. The distribution ofthe new production in the Atlantic, not shown, in-dicated a slight increase in the productivity in theice free polar regions, whereas the tropical and sub-tropical ocean showed no significant changes com-

a) b)

Summary and Conclusions

The new paleodata set of the LGM shows - espe-cially in the South Atlantic - that the reduction ofδ13C in the deep water column is less pronouncedthan previously assumed. In particular, the West-ern Atlantic shows no negative δ13C values. Thiscorresponds favourably with the glacial simulationof experiment 1, in which only the thermohaline


pared to the modern distribution. This simulatedpattern is supported by observations that new orexport production showed no significant changesin the glacial Southern Ocean (François et al. 1997;Nürnberg et al. 1997) and oligothrophic open oceanregions (Sarnthein et al. 1988; Mix 1989). How-ever, evidences of higher paleoproductivity wereobserved in the glacial Atlantic, particularly inthe eastern boundary upwelling regions (Sarntheinet al. 1988; Schneider et al. 1996) and in theequatorial zones (Mix 1989; Lyle et al. 1992;Schneider et al. 1996). A decrease of paleo-productivity was found in the Artic Ocean(Sarnthein et al. 1988; Schubert and Stein 1996;Knies and Stein 1998).

In model experiment 2 (implications of astronger biological pump), simulated δ13C concen-trations revealed that the reduction of carbon iso-topes was highly overestimated as compared to theobservations in the whole Atlantic. However, withrespect to the atmospheric CO2 concentration, thedecrease amounted to 60 ppm as compared to thecontrol run and was in relative good accordancewith observations from ice cores. The export pro-duction in this simulation showed a strong increasein the whole ice free ocean. This was supportedby observations made at the glacial subantarcticAtlantic Ocean (Kumar et al. 1995), equatorialAtlantic (Sarnthein et al. 1988; Mix 1989; Lyle etal. 1992) and upwelling regions of the eastern At-lantic (Sarnthein et al. 1988; Schneider et al. 1996).

Experiment 3 displayed a relative strong reduc-tion of the glacial atmospheric pCO2, in combina-tion with a good reproduction of the δ13C distribu-tion in the Atlantic. However, the result does notagree with observations of changed lysoclines. Animproved model performance could be archievedby accounting for the following considerations:• As mention above, the physical ocean modelshowed a too shallow southward extension of theNADW. A higher vertical resolution of the modelcould result in a better resolution of the variousmain water masses.• In the carbon cycle model, the δ13C of particulateorganic matter was calculated with a constant car-bon isotope fractionation factor of –20 ‰. How-ever, observations from the Southern Oceanshowed values lower than –30 ‰ in δ13C of or-

ganic matter in the surface water. These low or-ganic carbon isotope values were not reproducedby the model. Rau et al. (1991) attribute the lowδ13C of organic material in this region to the in-creased pool of aqueous CO2 in these cold waters.A CO2-dependent carbon isotope fractionationcould possibly correct this discrepancy betweensimulation and oberservation in the SouthernOcean.• A more complex production formulation includ-ing more different biological species could alsoimprove model performance. However, the func-tional relationships between the various species arenot well understood yet.• A more complex sediment module should be in-cluded which would enhance the performance inthe carbonate chemistry.• A module of the land biosphere should be includedwhich would supply more information on the effectof the marine carbon isotopes.

However, these recommendations are only help-ful, if the functional relationships are sufficientlyknown, ideally, on a global or basin wide scale.

We conclude on the base of our model simula-tions that no additional increase of the deep ocean'snutrient inventory is necessary to reproduce theδ13C distribution in the glacial Atlantic, which indi-cate no significant increase in biological pumpingduring the LGM. This would exclude all hypothesesthat involve changes in the nutrient inventory and/or Redfield ratios to explain the atmospheric gla-cial-interglacial CO2 reduction. Our results wouldrather support other scenarios for the decrease ofglacial atmospheric pCO2, such as the mechanismof carbonate compensation.

Acknowledgments

The models LSG and HAMOCC have been de-veloped and put at our disposal by the Max-Plank-Institut für Meteorologie. Special thanks go toAndreas Manschke for his technical support. Thecareful comments of A. Winguth and R. Zahnwere considerably helpful in improving the manu-script. We kindly acknowledge financial supportfrom Deutsche Forschungsgemeinschaft (Sonder-forschungsbereich 261). Data are available underwww.pangaea.de/Projects/SFB261.

720 Matthies et al.

References Gates WL, Boyle JS, Covey CC, Dease CG, DoutriauxCM, Drach RS, Fiorino M, Gleckler PJ, Hnilo JJ,Marlais SM, Phillips TJ, Potter GL, Santer BD,Sperber KR, Taylor KE, Williams DN (1998) An Over-view of the Results of the Atmospheric ModelIntercomparison Project (AMIP). Report 45, TheProgram for Climate Model Diagnosis and Inter-comparison, Livermore, http://www-pcmdi.llnl.gov/pcmdi/pubs/ab45.html

Heinze C, Maier-Reimer E, Winn K (1991) Glacial pCO2reduction by the world ocean: Experiments with theHamburg Carbon Cycle Model. Paleoceanography6: 395-430

Heinze C, Maier-Reimer E, Winguth AME, Archer D(1999) A global oceanic sediment model for long-termclimate studies. Glob Biogeochem Cycl 13:221-250

Johns TC, Carnell RE, Crossley JF, Gregory JM, MitchellJFB, Senior CA, Tett SFB, Wood RA (1997) The sec-ond Hadley Centre coupled ocean- atmosphereGCM: Model description, spinup and validation. ClimDyn 13: 103-134

Keeling RF, Stephens BB (2001) Antarctic sea ice andthe control of Pleistocene climate instability.Paleoceanography 16: 112-131 and (corrections) 330-334

Knies J, Stein R (1998) New aspects of organic carbondeposition and its paleoceanographic implicationsalong the northern Barents Sea margin during thelast 30,000 years. Paleoceanography 13: 384-394

Kroopnick PM (1985) The distribution of 13C of CO2 inthe world oceans. Deep-Sea Res 32: 57-84

Kumar N, Anderson RF, Mortlock RA, Froelich PN,Kubik P, Dittrich- Nannen B, Suter M (1995) In-creased biological productivity and export produc-tion in the glacial Southern Ocean. Nature 378:675-680

Leonard BP (1979) A stable and accurate convectivemodeling procedure based on quadratic upstreaminterpolation. Comp Meth Appl Mech Eng 19:59-98

Levitus S, Russell B, Boyer TP (1994) World Ocean At-las 1994. Volume 3: Salinity. Technical report, Na-tional Oceanic and Atmospheric Administration,Washington DC

Lorenz S, Grieger B, Helbig P, Herterich K (1996) Inves-tigating the sensitivity of the atmospheric generalcirculation model ECHAM 3 to paleoclimatic bound-ary conditions. Geol Rundsch 85: 513-524

Lyle MW, Prahl FG, Sparrow MA (1992) Upwelling andproductivity changes inferred from a temperaturerecord in the central equatorial Pacific. Nature 355:812-815

Lynch-Stieglitz J, Stocker T, Broecker WS, Fairbanks RG

Anderson LA, Sarmiento JL (1994) Redfield ratios ofremineralization determined by nutrient data analy-sis. Glob Biogeochem Cycl 8: 65-80

Arakawa A, Lamb VR (1977) Computational design ofbasic dynamical process of the UCLA general circu-lation model. Meth Comput Phys 16: 173-283

Archer D, Winguth A, Lea D, Mahowald N (2000a) WhatCaused the Glacial/Interglacial Atmospheric pCO2Cycles? Rev Geophys 38: 159-189

Archer DE, Eshel G, Winguth A, Broecker WS,Pierrehumbert R, Tobis M, Jacob R (2000b) Atmos-pheric pCO2 sensitivity to the biological pump in theocean. Glob Biogeochem Cycl 14: 1219-1230

Bacastow R, Maier-Reimer E (1990) Ocean-circulationmodel of the carbon cycle. Clim Dyn 4:95-125

Barnola JM, Raynaud D, Korotkevich YS, Lorius C (1987)Vostok ice core provides 160,000-year record of at-mospheric CO2. Nature 329: 408-414

Bickert T, Wefer G (1999) South Atlantic and benthicforaminifer δ13C-deviations: Implications for recon-structing the Late Quaternary deep-water circula-tion. Deep-Sea Res 46: 437-452

Broecker WS, Maier-Reimer E (1992) The influence ofair and sea exchange on the carbon isotope distri-bution in the sea. Glob Biogeochem Cycl 6: 315-320

Broecker WS, Henderson GM (1998) The sequence ofevents surrounding Termination II and their impli-cations for the cause of glacial-interglacial CO2changes. Paleoceanography 13: 352-364

Curry WB, Duplessy JC, Labeyrie LD, ShackletonPNJ(1988) Changes in the distribution of δ13C of DeepWater CO2 between the Last Glaciation and theHolocene. Paleoceanography 3: 317-341

Duplessy JC, Shackleton NJ, Fairbanks RG, Labeyrie L,Oppo D, Kallel N (1988) Deepwater source variationsduring the last climatic cycle and their impact on theglobal deepwater circulation. Paleoceanography 3:343-360

Elderfield H, Rickaby REM (2000) Oceanic Cd/P ratio andnutrient utilization in the glacial Southern Ocean.Nature 405: 305-310

François R, Altabet MA, Yu EF, Sigman DM, Bacon MP,Frank M, Bohrmann G, Bareille G, Labeyrie LD (1997)Contribution of Southern Ocean surface-water strati-fication to low atmospheric CO2 concentrations dur-ing the last glacial period. Nature 389: 929-935

Friedli H, Lötscher H, Oeschger H, Siegenthaler U,Stauffer B (1986) Ice core record of the 13C/12C ratioof atmospheric CO2 in the past two centuries. Na-ture 324: 237-238


(1995) The influence of air-sea exchange on the iso-topic composition of oceanic carbon: Observationsand modeling. Glob Biogeochem Cycl 9: 653-665

Mackensen A, Hubberten HW, Bickert T, Fischer G,Fütterer DK (1993) The δ13C in benthic foraminiferaltests of Fontbotia Wuellerstorfi (Schwager) relativto the δ13C of dissolved inorganic carbon in South-ern Ocean Deep Water: Implications for glacial oceancirculation models. Paleoceanography 8: 587-610

Mackensen A, Hubberten HW, Scheele N, Schlitzer R(1996) Decoupling of δ13CTCO2 and phosphate inRecent Weddel Sea deep and bottom water: Impli-cations for glacial Southern Ocean paleoceano-graphy. Paleoceanography 11: 203-215

Maier-Reimer E, Hasselmann K (1987) Transport andstorage of CO2 in the ocean - an inorganic ocean-circulation carbon cycle model. Clim Dyn 2: 63-90

Maier-Reimer E, Mikolajewicz U, Hasselmann K (1993)Mean circulation of the Hamburg LSG OGCM andIts Sensitivity to the Thermohaline Surface Forcing.J Phys Oceanogr 23: 731-757

Marchal O, Stocker TF, Joos F (1998) Impact of oceanicreorganizations on the ocean carbon cycle and at-mospheric carbon dioxide content. Paleoceano-graphy 13: 225-244

Mix AC (1989) Influence of productivity variations onlong-term atmospheric CO2. Nature 337: 541-544

Murnane RJ, Sarmiento JL, Le Quéré C (1999) Spatialdistribution of air-sea CO2 fluxes and the interhemi-spheric transport of carbon by the oceans. GlobBiogeochem Cycl 13: 287-305

Nürnberg CC, Bohrmann G, Schlüter M, Frank M (1997)Barium accumulation in the Atlantic sector of theSouthern Ocean: Results from 190,000-year records.Paleoceanography 12: 594-603

Petit JR, Jouzel J, Raynaud D, Barkov NI, Barnola JM,Basile I, Bender M, Chappellaz J, Davis M, DelaygueG, Delmotte M, Kotlyakov VM, Legrand M, LipenkovVY, Lorius C, Pépin L, Ritz C, Saltzman E, StievenardM (1999) Climate and atmospheric history of the past420,000 years from the Vostok ice core, Antarctica.Nature 399: 429-436

Rau GH, Takahashi T, Des Marais DJ, Sullivan CW (1991)Particulate organic matter δ13C variations across theDrake Passage. J Geophys Res 96: 15131-15135

Rau GH, Riebesell U, Wolf-Gladrow D (1997) CO2aq-dependent photosynthetic 13C fractionation in theocean: A model versus measurements. GlobBiogeochem Cycl 11: 267-278

Sanyal A, Hemming NG, Hanson GN, Broecker WS (1995)Evidence for a higher ph in the glacial ocean fromboron isotopes in foraminifera. Nature 373: 234-236

Sarnthein M, Winn K, Duplessy JC, Fontugne MR (1988)Global variations of surface ocean productivity inlow and mid latitudes: Influence on CO2 reservoirsof the deep ocean and atmosphere during the last21,000 years. Paleoceanography 3: 361-399

Sarnthein M, Winn K, Jung SJA, Duplessy JC, LabeyrieL, Erlenkeuser H, Ganssen G (1994) Changes in eastAtlantic deepwater circulation over the last 30,000years: Eight time slice reconstructions. Paleoceano-graphy 9: 209-267

Schäfer-Neth C, Paul A (2001) Circulation of the GlacialAtlantic: A Synthesis of Global and RegionalModeling. In: Schäfer P, Ritzrau W, Schlüter M,Thiede J (eds) The Northern North Atlantic: AChanging Enviroment. Springer, Berlin, pp 441-462

Schlitzer R (2002) Carbon export fluxes in the SoutherrnOcean: Results from inverse modelling andcompaison with satellite-based eastimates. Deep-SeaRes 49: 1623-1644

Schmitz WJ (1995) On the Interbasin-Scale ThermohalineCirculation. Rev Geophys 33: 151-173

Schneider RR, Müller PJ, Ruhland G, Meinecke G, SchmidtH, Wefer G (1996) Late quaternary surface tempera-tures and productivity in the east-equatorial southatlantic: Response to changes in trademonsoonwind forcing and surface water advection. In: WeferG, Berger W, Siedler G, Webb DJ (eds) The South At-lantic: Present and Past Circulation. Springer, Berlin,pp 527-551

Schubert CJ, Stein R (1996) Deposition of organic car-bon in Arctic Ocean sediments: Terrigenous supplyversus marine productivity. Org Geochem 24:421-436

Shea DJ, Trenberth KE, Reynolds RW (1990) A globalmonthly sea surface temperature climatology. NCARTechnical Note NCAR/TN 345, NCAR, Boulder,Colorado

Sigman DM, Boyle EA (2000) Glacial/interglacial varia-tions in atmospheric carbon dioxide. Nature 407:859-869

Stephens BB, Keeling RF (2000) The influence of Ant-arctic sea ice on glacial-interglacial CO2 variations.Nature 404: 171-174

Suess E (1980) Particulate organic carbon flux in theoceans-surface productivity and oxygen utilization.Nature 288: 260-263

Takahashi T, Feely RA, Weiss R, Wanninkhof R,Chipman DW, Sutherland SC, Takahashi TT (1997)Global air-sea flux of CO2: An estimate based onmeasurements of air-sea pCO2 difference. Proc NatlAcad Sci 94: 8292-8299

Taylor KE (2001) Summarizing multiple aspects of model

722 Matthies et al.

performance in a single diagram. J Geophys Res 106:7183-7192

Toggweiler JR (1999) Variation of atmospheric CO2 byventilation of the ocean's deepest water. Paleoceano-graphy 14: 571-588

Weinelt M, Sarnthein M, Pflaumann U, Schulz H, JungS, Erlenkeuser H (1996) Ice-free Nordic Seas duringthe Last Glacial Maximum? Potential sites ofdeepwater formation. Paleoclimatology 1: 283-309

Winguth AME, Archer D, Duplessy JC, Maier-ReimerE, Mikolajewicz U (1999) Sensitivity of paleonutrienttracer distributions and deep-sea circulation to gla-cial boundary conditions. Paleoceanography 14:304-323

Yamanaka Y, Tajika E (1996) The role of the vertical fluxesof particulate organic matter and calcite in the oce-anic carbon cycle: Studies using an oceanbiogeochemical general circulation model. GlobBiogeochem Cycl 10: 361-382

Yamanaka Y, Tajika E (1997) Role of dissolved organicmatter in the marine biogeochemical cycle: Studiesusing an ocean biogeochemical general circulationmodel. Glob Biogeochem Cycl 11: 599-612

Yu EF, François R, Bacon MP (1996) Similar rates ofmodern and last-glacial ocean thermohaline circula-tion inferred from radiochemical data. Nature 379:689-694

Last Glaciald 13C Distribution and Deep-Sea Circulation in ... › ~apau › publications ›...

Documents

Transcript of Last Glaciald 13C Distribution and Deep-Sea Circulation in ... › ~apau › publications ›...