The Southern Ocean at the Last Glacial Maximum a Strong Sink For

GLOBAL BIOGEOCHEMICAL CYCLES, VOL. 14, NO. 1, PAGES 455-475, MARCH 2000

The SouthernOcean at the last glacial maximum- A strong sink for atmospheric carbon dioxide

J. Keith Moore,• Mark R. Abbott,James G. Richman,and David M. Nelson

OregonState Unive•ity, Collegeof Oceanicand AtmosphericSciences

Abstract.Analysis of satelliteocean color, sea surface temperature, and sea ice cover datareveals consistent patterns between biological production, iron availability, and physicalforcings in the SouthernOcean. The consistencyof thesepatterns, in conjunctionwith informationon physicalconditions during the lastglacial maximum (LGM), enablesestimates of exportproduction at the LGM. The LGM SouthernOcean experiencedincreased wind speeds, colder sea surface and atmospheric temperatures, increaseddeposition of atmosphericdust, and a greatlyexpanded winter sea ice cover. Thesevariations had strong effects on SouthernOcean ecology and on air-seafluxes of CO2. The seasonalice zone(SIZ) wasmuch larger at the LGM (30 millionkm 2) thanat present(19 millionkm2). The Antarctic Polar Front (PF) likely marked the northern boundaryof thisexpanded SIZ throughoutthe SouthernOcean, as it doestoday in the DrakePassage region. A largenorthward shift in theposition of thePF duringglacial timesis unlikely due to topographicconstraints. North of the PF, the increasedflux of aeoliandust during glacial times altered phytoplankton species composition and increased exportproduction, and as a resultthis region was a strongersink for atmosphericCO2 than in the modem ocean. South of the PF, interactionsbetween the biota and sea ice stronglyinfluence air-sea gas exchange over seasonaltimescales. The combined influenceof meltingsea ice andincreased aeolian dust flux (with its associatediron) increasedboth primary and export production by phytoplanktonover daily-monthly timescalesduring austral spring/summer, resulting in a strongflux of CO2into the ocean. Heavyice coverwould have minimized air-sea gas exchange over much of the restof the year. Thus,an increasednet flux of CO2into the oceanis likely duringglacial times, evenin areaswhere annual primary production declined. We estimatethat export productionin the SouthernOcean as a wholewas increased by 2.9 - 3.6 Gt C yr-• at the LGM, relativeto the modemera. Alteredseasonal sea ice dynamicswould further increasethe net flux of CO2into the ocean. Thus the Southern Ocean was a strongsink for atmosphericCO2 and contributed substantially to the lowering of atmosphericCO2 levelsduring the last ice age.

1. Introduction 1990]. Several analysesof the sedimentaryrecord have arguedfor lower SouthernOcean primary productionat the Martin [1990] argued that the Southem Ocean LGM and concluded that the Southern Ocean was not a surroundingAntarctica was a much stronger sink for strong sink for CO2 at that time [Mortlock et al., 1991; atmosphericcarbon dioxide (CO2) at the last glacial Charles et al., 1991; Shemeshet al., 1993; N•irnberg et al., maximum (LGM) due to an increasein phytoplankton 1997]. In this paper, we presentnew data and arguments, primary productiondriven by elevatedfluxes of iron from which suggestthat the SouthernOcean was indeeda strong continentaldust sources to surfacewaters. This processwas net sink for atmosphericCO2 at the LGM, due to increased termed the iron hypothesisfor explaining the observed phytoplanktonproduction during spring/summer months and loweringof atmosphericCO2 during the lastice age [Martin, reduced outgassingof CO2 during winter monthsdue to heavy sea ice cover. Our analysisis basedlargely on the currentunderstanding of modemSouthern Ocean ecosystems 1Nowat NationalCenter for AtmosphericResearch, Boulder, supplementedwith satellitemeasurements of oceancolor and Colorado. sea ice distributions.Recent analysesof satellitesea surface temperaturedata allows new insightsinto the dynamicsand Copyright2000 by theAmerican Geophysical Union. location of the Antarctic Polar Front (PF) in the modem and glacial oceans[Moore et al., 1997; 1999a]. The arguments Papernumber 1999GB900051 presentedhere are consistentwith the known sedimentary 0886-6236/00/1999GB900051 $12.00 record from the Southern Ocean. Interactions between sea

455 456 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK ice and the biota over seasonal timescales are critical to zooplankton,i.e., krill and copepods,which have long reconcilingthe iron hypothesiswith the sedimentaryrecord. generationtimes comparedto the phytoplankton.These The physicalenvironment of the SouthernOcean at the bloom regions tend to have high f ratios and export LGM, -21,000 yearsago (earlier references cite 18,000years production[Trdguer and Jacques, 1992; Banse, 1996]. The agobased on •4Cdating, see Bard et al. [1990]),was very f ratio, the ratio of new productionto total production, different from that in today'socean. The ice age Southern representsthe fraction of phytoplanktonproduction available Oceanwas markedby coolersea surface temperatures (SST), for exportfrom the surfacelayer over annualtimescales an expanded winter sea ice extent, colder surface air [Dugdaleand Goering,1967; Eppley and Peterson,1979]. temperatures(*-9øC degrees lower on the Antarctic In contrastto thesespatially restricted bloom regions, continent),and an increasein the depositionof atmospheric large portionsof the SouthernOcean have perpetuallylow dust (-15-20 fold) [Hays, 1978; Climate: Long-Range chlorophyll concentrationswhere smaller nano-sizedand Investigation,Mapping, and Prediction (CLIMAP), 1981; pico-sizedphytoplankton dominate [Tregudr and Jacques, Petit et al., 1981, 1990; Jouzel et al., 1987; Pichon et al., 1992; Banse, 1996]. These areas have lower f ratios, low 1992]. In addition,atmospheric CO2 levels were at-200 export production, and perpetually high macronutrient ppm, -80 ppm lower than the modempreindustrial average, concentrations(phosphate 1.5 > •M and nitrate > 20 •M and sea level was 120-150 m lower [Barnola et al., 1987; south of the PF) [Comiso et al., 1993]. The Crowley, 1988]. microzooplankton that graze upon these smaller A role for the high latitudeoceans (in particularthe phytoplankton size have generationtimes comparableto SouthemOcean) in the ice age loweringof atmosphericCO 2 thoseof phytoplanktonand, thus,maintain a tight grazing has been proposedby numerousauthors [e.g., Knox and controlon phytoplanktonbiomass [Sherr and Sherr, 1994]. McElroy, 1984;Siegenthaler and Wenk,1984; Sarmiento and A similar situationexists in the equatorialPacific, another Toggweiler, 1984; Keir, 1988; Broeckerand Peng, 1989; high-nutrient,low-chlorophyll (HNLC) region[Landry et al., Martin, 1990]. Model results demonstratethat primary 1997]. production levels in the Southern Ocean can strongly The bloom regions have elevated dissolved iron influence atmosphericCO2 levels [i.e., Sarmientoand Orr, concentrations relative to other areas of the Southern Ocean, 1991; Joos and Siegenthaler,1991; Sarmientoand Le Qudrd, at leastpart of the time [Martin et al., 1990a,b; de Baar et 1996]. Martin [1990, 1992] and Martin et al. [1990a, al., 1995; Sedwick and DiTullio, 1997; Sedwick et al., 1997]. 1990b] suggestedrelease from iron limitation by increased Bottle incubation experiments indicate that offshore atmospheric deposition as a mechanism for increased phytoplankton respond to iron additions with increased SouthernOcean productivityat the LGM. growth rates and biomass and with a shift in species Atmosphericdust levels were muchhigher at the LGM compositiontoward larger diatoms [Martin et al., 1990b; [Petit et al., 1981, 1990]. This increaseddust deposition was Hebling et al., 1991; de Baar et al., 1995; Van Leeuweet al., a global phenomenonwith greatly increaseddust inputs 1997; Scharek et al., 1997; Takeda, 1998]. These results relative to modem times recorded in Greenland and imply that increasing iron levels in situ would increase Antarctica ice cores [Petit et al., 1981; De Angelis et al., export production, as happens in the other HNLC areas 1997]. Petit et al. [1981] found continental aerosolflux on [Coale et al., 1996; Gordon et al., 1997; Hutchins et al., the Antarctic continentwas - 20 times higher at the end of 1998; see also Sunda and Huntsman, 1997]. the last glacial period. Similarly, Petit et al. [1990] Recently,the SouthernOcean Iron ReleaseExperiment estimated that mean dust flux to the Antarctic continent at (SOIREE) providedfurther evidence that waterssouth of the the LGM was 15 times higherthan Holocenelevels. Kumar PF are indeed iron limited [Boyd et al., 1999]. In this et al. [1995] found that iron flux to the sediments in the experimenta large patch of seawater(at 141øE,61øS) was South Atlantic increasedby more than a factor of 5 at the fertilized with iron additions over a 13-day period, and a LGM. suite of measurements was made both within the Severalstudies have examinedchlorophyll distributions iron-fertilizedpatch and in adjacentwaters. Over this period, in the SouthernOcean using the previousgeneration ocean chlorophyll levels increasedmarkedly, and the pCO2 of color sensor, the Coastal Zone Color Scanner (CZCS) surfacewaters decreased in the iron-enrichedpatch [Boyd et [Comiso et al., 1993; Sullivan et al., 1993]. The CZCS al., 1999]. This experimentprovides dramatic evidence that studies,new data from the Sea-viewingWide Field-of-view the availability of iron limits phytoplankton biomass Sensor (SeaWiFS) and in situ measurementsindicate that accumulationwithin the SouthernOcean today. phytoplanktonblooms where chlorophyllconcentrations > Martin [1990] suggestedthat atmosphericdust that 1.0 mg m-3 are restricted primarily to threeregions. These accumulates on sea ice, releases substantial amounts of iron "bloom regions"include coastal and shelf waters,areas near to the water column as the ice melts, thus enhancing the recedingedge of the seasonalice sheet,and watersin the phytoplanktongrowth. Sedwickand DiTullio [1997] recently vicinity of the major hydrographicfronts [Lutjeharmset al., reportedevidence of suchiron fertilizationby meltingsea ice 1985; Trdguer and Jacques,1992; Laubscheret al., 1993; in the Ross Sea. The input of low-salinity meltwater also Comiso et al., 1993; Sullivan et al., 1993; Banse, 1996; increases vertical stratification of the water column, Moore et al., 1999b; this study]. improvingthe irradiance/mixingregime for phytoplankton Ecological structureis fundamentallydifferent in the [Smith and Nelson, 1986]. bloomingand nonbloomingregions of the SouthernOcean. An importantconcept for the argumentspresented here In the bloom regions,phytoplankton biomass is dominated is the effect of iron limitation on uptake ratios of by larger diatom species, although Phaeocystis (a macronutrients by diatoms. Recent studies have prymnesiophyte)can be an importantcomponent in Antarctic demonstrated that uptake ratios of silica/nitrate and coastal,shelf, and ice edgewaters. Thesespecies are able to silica/phosphorusare 2-3 times higherunder iron-limitation escapegrazing control by their predators,typically larger than underiron-replete conditions [Takeda, 1998; Hutchins MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 457 and Bruland, 1998; Hutchins et al., 1998]. Since the to calculatemonthly sea ice climatology.The climatological phytoplankton maintain Redfield ratios between carbon, monthly data for August and February were used to nitrogen,and phosphorus[RedfieM et al., 1963], silica/carbon determine the modem (maximum and minimum) extent of ratios within cells are also 2-3 times higher under the seasonal ice sheet. A minimum of 5% sea ice cover was iron-limitation, a fact noted previouslyby Martin [1992]. used in determiningmodern maximum (winter) sea ice These results suggestthat percent opal and biogenic opal extent. This low ice concentration was chosen to facilitate accumulationrates may not be reliable proxies of export comparisonwith the CLIMAP (Climate: Long-Range productionin the SouthernOcean over glacial-interglacial Investigation,Mapping, and Prediction) LGM seaice data, periods,as the amountof iron inputto surfacewaters varied whichdid not distinguishice concentrations,only maximum drastically [Takeda, 1998]. Such paleoproxies may extent. A minimum of 70% ice cover was used to determine underestimateglacial diatom productionby more than 50% the minimum (summer) sea ice extent. Significant [Takeda, 1998]. Thus increasesin exportcarbon of 100% or phytoplanktonbiomass can accumulatein the watercolumn more could go unrecordedin the opal sedimentrecord at ice concentrationsbelow -70% [Smithand Gordon, 1997]. [Martin, 1992]. Monthly seaice datafrom 1997-1998were combinedwith In this paper we discuss the likely ecological the SeaWiFS ocean color data to examine consequencesof increasedatmospheric dust flux and the chlorophyll/phytoplanktonbiomass patterns within different greatlyexpanded seasonal sea ice extentat the LGM and the ecologicalregions of the SouthernOcean. All seaice data subsequenteffects on air-seaCO2 fluxes. Our analysesare were remappedto a 9-km equalangle grid. basedlargely on satellite-derivedestimates of sea ice cover, The August (maximum) LGM sea ice extent was chlorophylldistributions, and frontaldynamics in the modem obtained from the CLIMAP project [CLIMAP Project Southern Ocean. We use a broad definition for the Southern Members, 1976, 1981]. The data obtained from CLIMAP Ocean, encompassingthe area from the North Subtropical were on a 2ø x 2ø grid. Thus the spatialresolution of the Front to the Antarcticcontinent. Our methodsfor processing CLIMAP ice extent is low. The CLIMAP sea ice boundary the satellite data and some assumptionsare outlined in wasremapped to a 9-km resolutionequal angle grid, and the section2. In section3 we examinephytoplankton biomass boundarywas smoothedwith a 50 point boxcarfilter on and primary productivitypatterns in the modem Southern latitude and longitude. Ocean incorporatingnew satellite ocean color data from The mean path of the modernPF presentedis from SeaWiFS. In section4, we review air-seaCO2 fluxes in the Moore et al. [1999a]. This mean path was determinedby modern Southern Ocean. Seasonal'sea ice dynamics are mapping the PF's position from satellite sea surface critical to our arguments.We examinesea ice distributions temperaturedata over a 7-yearperiod, 1987-1993 [Moore et in the modernocean and at the LGM and their relationship al., 1999a]. The PF marks the boundary between cold to the Antarctic Polar Front in sections 5 and 6. Antarctic Surface Water (ASW) to the south and warmer We examinethe ecologyof the SouthernOcean at the Subantarctic waters to the north. LGM in section 7. We proposethat the expansionof the We assumethroughout this paperthat the physiological seasonal ice sheet and the increased aeolian iron flux at the properties of Southern Ocean phytoplanktonhave not LGM resultedin an ecologicalregime shift within the whole changedsignificantly between modern and glacial times. area southof the PF towarda patternsimilar to that observed Thus, while latitudinal range and biomass/numerical in the southern Weddell and Ross Seas of the modem ocean. dominance within assemblages may change, the This would lead to increasedexport production during austral phytoplanktonspecies present at the LGM respondedto spring and summer, and reducedoutgassing of CO2 during chemicaland physicalforcings in essentiallythe sameway winter. In section8 we discussthe implicationsof this ice as phytoplanktonin the modernday SouthernOcean. This age SouthernOcean ecology for net air-seaCO 2 fluxes and, assumptionis justified because,in terms of evolutionary finally, the role of the Southern Ocean in the observed timescales, the LGM was a recent event. glacial atmosphericCO 2 decrease.

3. PhytoplanktonBiomass and PrimaryProduction 2. Methods and Materials in the Modern Southern Ocean

SeaWiFS-derived estimates of surface chlorophyll The SouthernOcean is by far the largestHNLC region concentrationfor the periodOctober 1997 throughSeptember in the modern ocean [Martin, 1990]. In Plate 1, mean 1998 were obtainedfrom the GoddardSpace Flight Center surfacechlorophyll concentrations from the SeaWiFSsatellite [McClain et al., 1998]. Daily level 3 standardmapped for the SouthernOcean during austral summer (December images of chlorophyll concentration(Version 2) were 1997 through February 1998) are displayed. Surface averagedto produce monthly and seasonalmeans for the chlorophyllconcentrations are indeedquite low (typically < Southern Ocean. 0.3-0.4 mg m-D) despite the high concentrations of available Monthly mean sea ice concentrations(NASA Team macronutrientsin the Southern Ocean (Plate 1). algorithm) from the Nimbus-7 Scanning Multichannel The highest chlorophyll concentrationsare seen in Mircrowave Radiometer (SMMR) and DMSP-F8,-F11,-F13 coastal waters, especially in the southernportions of the satellitesensors for the years 1978-1998 were obtainedfrom Weddell and Ross Seasand abovethe Patagoniashelf on the the National Snow and Ice Data Center (NSIDC) Distributed eastern coast of South America (Plate 1). Coastal waters Active Archive Center, at the University of Colorado at surroundingand downstreamof SouthernOcean islands also Boulder [National Snow and Ice Data Center, 1998]. have higher chlorophyll concentrations;note the high Monthly data were averagedover the 1978-1996time period chlorophyll values near Kerguelen Island along 70øE. 458 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK

Elevated chlorophyll levels likely associated with the the seasonalexpansion and contraction of the seaice. It is retreatingseasonal ice cover can be seenin the Weddell Sea in these regions of meltwater input that phytoplankton (in the region *-30ø-0øW,55ø-67øS) and in the northernRoss blooms most often occur within the SIZ [Smith and Nelson, Sea (*-150ø-170øW, 65ø-75øS) [see also Moore et al., 1986]. We have operationallydefined the MIZ as areaswith 1999b]o > 70% sea ice cover in the precedingmonth and < 70% sea The remaining areas of elevated chlorophyll are ice cover in the current month. A wide range of cutoff associatedwith the major fronts of the SouthernOcean. values for determining the MIZ (*-30-70%) give similar High chlorophyll values in the. vicinity of the North and results as sea ice typically melts rapidly in the Southern South SubtropicalFronts (*-40ø-45øS) are seeneast of South Oceangoing from high to low concentrationsin lessthan a America and southof Africa from *-15ø-70øE(see Belkin and month. Gordon [ 1996] for definitionsand locationof the subtropical Note that there is some overlap between ecological fronts). High chlorophylllevels in the vicinity of the PF can regions. Near Drake Passageand in the SouthwestPacific, be seen in several regions, including along the the SIZ overlapswith the PFR, and we have includedthese Pacific-AntarcticRidge (--145ø-160øW)[see also Moore et areasonly within our MIZ and SIZ regions.Similarly, we al., 1999b], approachingand crossingthe Mid-AtlanticRidge have excludedcoastal areas with depths< 500 m (shownin (*-20ø-5øW), and from *-20ø-30øE(Plate 1, see Plate 4 for the black in Plate 2) from the SWR, the POOZ, and the PFR to meanlocation of the PF). Thereis a strongtopographic better define open oceanpatterns. Areas in black near the influenceon thedynamics of thePF in eachof theseregions Antarctic continenthad > 70% ice cover during February [Mooreet al., 1999a]. Mesoscalephysical processes may 1998 and are not includedin this analysisbecause they were result in increasednutrient inputs to surfacewaters where heavily covered with sea ice throughout the year. SouthernOcean fronts interactwith large topographic Comparisonwith our sea ice climatologyindicates that the features[Moore et al., 1999b]. 1997-1998 seasonhad higherthan normal summerice cover In general,these patterns of chlorophylldistribution and in the RossSea and well below averagesummer ice coverin phytoplanktonbiomass correlate strongly with the availability the southernWeddell Sea (seePlate 4 for the climatological of iron for phytoplanktongrowth. Coastalwaters including Februaryice cover> 70%). the southern Weddell and Ross Seas and near Southern The seasonalchlorophyll cycles of these ecological Ocean islands typically have elevated dissolved iron regionsas estimated with SeaWiFSare displayed in Plate3. concentrations(relative to offshore waters) in surfacewaters Note the significantlyhigher chlorophyll concentrations in due to nearbyshelf sourcesfor iron [Martin et al., 1990a,b; the WRR comparedwith otherSouthern Ocean areas. The Westerlundand Ohman, 1991; Nolting et al., 1991]. Nolting southern Weddell and Ross Seas are areas of high et al. [1991] found dissolved iron concentrationsover the phytoplanktonbiomass and export production during austral SouthOrkneys shelf fully an orderof magnitudehigher than springand summer[Jennings et al., 1984;Smetacek et al., in offshore waters. Elevated iron concentrations are also 1992; Smith and Gordon, 1997; Smith and Nelson, 1990; sometimesassociated with melting sea ice and the Southern Bateset al., 1998; Arrigo et al., 1998]. In the southernRoss Ocean fronts [Nolting et al., 1991; de Baar et al., 1995; Sea,primary production often exceeds 1 g C m-2 d -• [Smith Ldischer et al., 1997; Sedwick et al., 1996; Sedwick and et al., 1996; Arrigo et al., 1998; Bateset al., 1998]. Bates DiTullio, 1997]. et al. [ 1998] calculatedmean export production at *- 0.5 g C We can examine these patterns •n chloropl•yll m-2 d 4 with an f ratio between0.55-0.6. Nelsonet al. [1996] distributionsmore quantitativelyif we dividethe Southern calculateda seasonallyintegrated f ratio of 0.42. Arrigo et Oceaninto ecologicalregions (Plate 2). We have largely al. [1998] estimatedprimary productionon the Ross shelf followed the review of Trdguer and Jacques [1992] in duringDecember at 3.94g C m-2 d 4. determining these ecological boundaries. The high While elevated primary production is seen in the chlorophyllareas of thesouthern Weddell and Ross Seas are Weddell and Ross Seas, macronutrientsare usually not consideredas a single region (south of 73øS, hereafter completelydepleted (nitrate concentrations typically remain termed the Weddell-Ross Region, WRR). The WRR > 10/xM, [e.g., Smithet al., 1996]). These areasmay correspondsapproximately with what Trdguerand Jacques experiencesome iron limitation,particularly toward the end [1992] termed the Coastal and Continental Shelf Zone of the summerperiod [Sedwicket al., 1996]. Sedwickand (CCSZ). The SeasonalIce Zone (SIZ) is defined as areas DiTullio [1997] arguedthat a diatombloom in the southern with > 5% sea ice duringAugust 1997 and < 70% ice cover RossSea was ultimately stoppedby iron limitation. Smith duringFebruary 1998. The Polar Front Region (PFR) is the et al. [1999] also argued that micronutrient availability area surrounding the PF, which we have operationally (notablyiron) limited the growthrates of phytoplanktonin defined as the area within 1ø of latitudeof the mean path of the RossSea during late summer. Iron limitation in these the PF (see Plate 4 and Moore et al. [1999a]). The areas would probably not be as severe as in offshore PermanentlyOpen OceanZone (POOZ) is the regionnorth SouthernOcean watersdue to iron inputsfrom sedimentand of the SIZ and south of the PFR. The Subantarctic Water melting sea ice; thus,the argumentsof Hutchinset al. [1998] Ring (SWR) is defined as open ocean areasfrom the PFR for differentdegrees of iron limitationin marineecosystems north to the SubtropicalFront [Banse, 1996]. We marked likely apply in the SouthernOcean. the northernboundary of the SWR as slightly north of the Lowest chlorophyll concentrationsare seen in the North SubtropicalFront as defined by Belkin and Gordon POOZ and SWR areaswith monthlymean valuesranging [1996]. between0.14 and0.32 mg m-3 (Plate 3). Thesemean values Not depictedin Plate 2 is the ecologicallyimportant comparewell with the in situ measurementscompiled by Marginal Ice Zone (MIZ), which is a subsetof the SIZ, Banse[1996]. Theseareas are markedby persistentlyhigh where there has been recentmelting of sea ice [Smithand nitrate and phosphateconcentrations, low chlorophyll Nelson, 1986]. The size and locationof the MIZ changewith concentrations,and low primary and export production,and 459

50'W

7p'w 70 'E

90*W 90 'E

110'W 110'E

150'W 0.05 0.2 0:4 1.0 3:0 Chlorophyll(m9/m 5)

Plate 1. Shownare the meansurface chlorophyll concentrations in the SouthernOcean during austral summer(December 1997 to February 1998) from SeaWiFS.

they are dominatedby smallerphytoplankton species, except summer (Plate 3). It should be noted that the PF is a in coastalwaters and in the vicinity of SouthernOcean fronts dynamic feature, and our definition for the PFR (which is [Trdguer and Jacques, 1992; Banse, 1996]. basedon mean PF location)by necessityincludes substantial Within the SIZ there is often a short bloom dominated areas that should be classified as either POOZ or SWR by large diatoms following sea ice retreat, with elevated depending on the location of the PF. Thus chlorophyll primary and export production[Wilson et al., 1986; Nelson concentrations are underestimated for the PFR, while areal and Smith, 1986; Smith and Nelson, 1985, 1986; Smith and extent of the area directly influenced by the front is Sakshaug,1990]. These ice edge bloomsprobably account overestimated. for the elevated mean chlorophyllconcentrations within the The areas of these ecological regions are compared MIZ and SIZ areas (Plate 3). After the initial bloom, there with the correspondingvalues from Trd•uer and Jacques is typically a shift from a new productionto a regenerated [ 1992] in Table 1. Differencesare due to variationsin the productionregime, with a correspondingshift from larger definitionsof the regionsand due to a differentplacement of diatomsto smallerpico-sized and nano-sizedphytoplankton, the PF. Smith. and Nelson [1986] estimated the SIZ to be similarto the POOZ and SWR pattern[Trdguer and Jacques, -16.4 millionkm 2, similarto our valuedespite substantially 1992; Banse, 1996]. different SIZ definitions. Recall that the SIZ area in Table Sporadicblooms of largerdiatoms also occur at the PF 1 does not include the WRR. Our estimate of the POOZ [Lutjeharms et al., 1985; Trdguer and Jacques, 1992; area is substantiallylower than that of Trdguer and Jacques Laubscher et al., 1993; de Baar et al., 1995; Banse, 1996]. [1992], largely due to a different placementof the mean PF These sporadicblooms probably accountfor the slightly location. The SWR is by far the largestregion (48.7 million highermean chlorophyll levels within the PFR durin• austral km2, Table 1). The MIZ reachesits maximumsize in 460 10'W 30'"' 10'E

30 'E

50'W • 50 'E

,o

704// 70 'E

90'W 90 'E

110"W * "'-- 110'E

130•/ 0-• 130 'E

SubontorcUcWater Ring • ___•---

PolarFront Recjion / -'"-7...... • .....•- • 1,50 'E Permanentl)/ Open Oceg:!nZone Seasonal Ice Zone 170 'W 170 øE Weddell Ross Region

Plate 2. Displayedare ecologicalregions within the SouthemOcean (see text for details).

Table 1. Displayedare the SpatialExtent of EcologicalRegions within the Southern Oceanfor the 1997-1998Season Compared with Valuesfrom Tr•guer and Jacques[ 1992].

.R.egion Area Region Area MIZ Area

Subantarcticwater ring 48.7 October 0.89

Polar Front region 4.89 Polar Front zone 3 November 2.82

Permanentlyopen ocean Permanentlyopen December 6.05 zone 8.70 ocean zone 14 January 2.27

Seasonalice zone 16.2 Seasonalice zone 16 February 0.85

Weddell-Rossregion 0.95 Coastaland March 0.23 Continental shelf 0.9

Columnone gives the areaof theecological regions shown in Figure2. Column2 presentsareal extent of theecological regions of Trdguerand Jacques [1992]. The third column shows the area of theMarginal Ice Zone (thisstudy). Areasare givenare in unitsof millionsof km2. MOORE ET AL.' SOUTHERN OCEAN CARBON SINK 461

10-

--h- SIZ POOZ

0.1 I I I ...... I I I '" I i i I '1......

Plate 3. Shown are seasonalpatterns of mean chlorophyllconcentration (1997-1998) within different ecologicalregions of the SouthernOcean (WRR, Weddell and RossSeas south of 73øS, SIZ, Seasonal Ice Zone, POOZ, PermanentlyOpen OceanZone, PFR, Polar Front Region,MIZ, Marginal Ice Zone, and SWR, SubantarcticWater Ring). Since variability is generallylow and typically thousandsof data point go into each average,error bars aroundthese means would be smallerthan the symbols shown.

December,with mostof its areain the Weddell Sea (Table the rate at which gas transfers across the sea surface, is 1). known to increasewith increasingwind speed,but the exact relationship is still in question [i.e., Murphy et al., 1991]. The partial pressure of CO2 in seawater is a function of 4. Air-Sea Flux of CO 2 in the Modern Southern temperature, total CO2, total alkalinity, and salinity, with Ocean temperature and total CO2 being the most important in surface oceanic waters [Takahashi et al., 1993]. Primary production lowers total CO2 and pCO 2 in The net air-sea flux of CO2 is a function of the surface waters. The SubtropicalFront has been noted as a differencein partialpressure of CO2 (pCO2)between air and very strong sink for atmosphericCO2 [Metzl et al., 1991; sea(ApCO2), and the gasexchange coefficient. The gas Takahashi et al., 1993, 1997]. Takahashi et al. [ 1993] note exchangecoefficient is the productof the solubilityof CO2 that this effect is strongerin the westernSouth Atlantic and and the gas transfer velocity, which is a function of wind attribute it to higher biological productivity in Subantarctic speed[Murphy et al., 1991]. The gastransfer velocity, or waters, combined with a cooling (lowering pCO2) of 462 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK

50'•

70'W

90'W

130 'E CLIM P LGM Max. Ic 5 's AntarcticModemMax.Polar Ice FrontExtent eExt'••• I Modem Min. Ice Extent 150 'E OceanDepth < 3000 m --4•'s 70'E

Plate 4. Shown are the maximum (August) and minimum (February) sea ice extent for the modern era from our sea ice climatology (1978-1996). Also shown is the CLIMAP project estimate of maximum ice extent at the last glacial maximum [CLIMAP Project Members, 1976, 1981], and the mean path of the AntarcticPolar Front [Moore et al., 1999a].

Subtropicalwaters moving southward. The global database mixing. The balance between this carbon sourceand of ApCO 2 presentedby Takahashi et al. [ 1997] reveals photosyntheticproduction largely determines the direction of strongsinks east of Australia and Africa at theselatitudes as the net air-seaflux of CO2 in Antarcticsurface waters during well. In general, temperatureand biota effects dominate the austral growing season. Areas of high primary pCO2 in SouthernOcean waters, although in the far souththe productionare strongsinks for atmosphericCO 2, while areas influenceof salinity (i.e., from meltwaterinput) can alsobe of low productionare typically regions of low net flux, significant [Goyet et al., 1991; Metzl et al., 1991, 1995; which can be in either direction[Bouquegneau et al., 1992; Takahashi et al., 1993, 1997; lshii et al., 1998; Sabine and Metzl et al., 1991, 1995; Takahashi et al., 1993, 1997; Rubin Key, 1998]. et al., 1994; Poisson et al., 1994; Robertson and Watson, Net air-sea CO 2 fluxes southof the PF are controlled 1995; lshii and Sugimura,1998; Bateset al., 1998; lnoue et largely by the biota and mixing processes,as seasonal al., 1998]. temperatureeffects are relatively small. Much of this region The southernWeddell and Ross Seasbecome strong is a weak net sourceor sink of CO2 for the atmospherein sinks for atmosphericCO 2 during summer,due to high the modern SouthernOcean [Murphy et al., 1991; Robertson biologicalproduction [Takahashi et al., 1993, 1997;Bates et and Watson, 1995; Bakker et al., 1997; Takahashi et al., al., 1998]. Bateset al. [1998] foundpCO 2 valuesranging 1997]. Relatively carbon-richwaters upwell at the Antarctic from 50-150/•atm (mean-100/•atm) belowatmospheric Divergence (AD) and enter surfacewaters through vertical levelsalong 76.5øS in the RossSea during austral summer. MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 463

EstimatedCO 2 flux into the oceanranged from 4-26 mmol 5.2. Summer Sea Ice Extent at the LGM C m-2 d -• [Bateset al., 1998].They suggested that this area was an annualnet sinkfor atmosphericCO2 due to strong Several studies have suggesteda greatly expanded biologicalproduction during austral summer and little net summer sea ice extent at the LGM, well north of the flux the rest of the year becauseof persistentice cover maximum winter ice extent of today [CLIMAP, 1981; Hays [Bates et al., 1998]. et al., 1976; Cookeand Hays, 1982]. This boundaryis only Along 6øW in the australspring of 1992, the PF was -2ø-5 ø of latitude south of the CLIMAP winter sea ice a strongsink for atmosphericCO 2 (due to phytoplankton boundary shown in Plate 4. The CLIMAP summer ice production), while most of the area farther southwas a net boundary was drawn along the sedimentary boundary source(due to the seasonalwarming of surfacewaters and betweendiatom ooze to the north and diatomace.usclay to very low phytoplanktonproduction) [Bakker et al., 1997]. the south. Burckle and coworkers showed that this Thus the seasonaltemperature effect on pCO 2 can be sedimentaryboundary does not mark minimum ice extentin significant for net air-sea flux in areas of very low the modern ocean and that the CLIMAP LGM summer ice phytoplanktonbiomass/production. boundary did not mark minimum extent, but rather a late spring/earlysummer sea ice extent, slightly north of the modem mid-December position [Burckle et al., 1982; Burckle, 1984; Burckle and Cirilli, 1987]. Burckle and 5. Sea Ice at the LGM Cirilli [1987] noted that waters south of this ice edge experienceonly a few monthsof ice-free time each year, resulting in a depressedannual flux of 'diatoms to the 5.1. Winter Sea Ice Extent at the LGM sediments that results in the silty, diatomace.us clay sedimenttype. The maximum sea ice extent at the LGM covered a Crostaet al. [1998b]combined a detailedstudy of much larger area than it doesin the modernocean [CLIMAP modem and LGM diatom speciesdistributions in the Project Members, 1976, 1981; Cooke and Hays, 1982; sedimentto estimatesummer sea ice cover. They noted the Crosta et al., 1998a]. The estimatedmaximum LGM sea ice presenceof openocean diatom species in all of their LGM extent from the CLIMAP project is shown in Plate 4 sedimentsamples, indicating at leastsome ice-free time each [CLIMAP Project Members, 1976, 1981]. Plate 4 also shows summer[Crosta et al., 1998b].The datapresented by Crosta the mean locationof the modem-dayPF and the maximum et al. [ 1998b]argue strongly that summersea ice extentwas and minimum sea ice extent in the modem Southern Ocean. south of 60øS in the Atlantic and Indian sectorsand south of Several studies have arrived at similar boundaries for 65øSin the Pacific. They concludedthat summersea ice maximum sea ice extent at the LGM [Hays et al., 1976; extentat the LGM wasvery similarto today'sminimum sea Cooke and Hays, 1982; Crosta et al., 1998a] (see Crosta et ice extent,south of all their core locations[Crosta et al., al. [1998a] for a comparisonof theseice extentboundaries). 1998b]. Therefore,for thepurposes of thispaper, we will Crostaet al. [1998a]based their estimateon the presenceof assumethat the minimum(summer) sea ice extentat the ice-associatedspecies of diatomsin the sediment. They LGM was the same as in the modem SouthernOcean. This argued that the maximum extent of sea ice was several assumptionrepresents a currentbest guess of summerice degreesof latitudefarther north than the CLIMAP position extent at the LGM. Crosta et al. [1998b] examinedfew in the southwestAtlantic and severaldegrees farther south cores from the southwest Pacific sector and none from the than the CLIMAP estimatein the southeastPacific [Crosta Weddell and Ross Seas. There is evidencethat summer sea et al.• 1998a]o Very few sedimentcores from the Pacific iceextent in theWeddell Sea was somewhat greater than in sectorof the SouthernOcean were used in constructingthe the modemera [Jordonand Pudsey,1992; Shimmield et al., CLIMAP boundary•In section6 we arguethat the CLIMAP 1994]. In addition much of the Ross Sea shelf was covered winterice boundaryis placedtoo far northover much of the by groundedice sheetsat the LGM [Kellogg,1987; Pacific sector. DeMaster et al., 1996]. The SIZ at,the LGM was thus much A number of factors could have contributed to this ice largerthan in the modemera, due to a greatlyexpanded ageexpansion of the seasonalice sheet.Cooler atmospheric wintersea ice area and summer ice cover similar to today. and sea surfacetemperatures would promote ice formation A summer sea ice distribution at the LGM similar to [Martins,n, 1990; Crosta et al., 1998b]. A decreasein the today'sseems counterintuitive, given the colderatmospheric inputof warmsalty North Atlantic Deep Water (NADW) to and sea surfacetemperatures. However, much of the heat theACC duringglacial times has also been suggested to play requiredfor seasonalmelting of the pack ice comesfrom the a role [i.e., Crowley and Parkins,n, 1988]. Wind-driven relativelywarm CircumpolarDeep Water below the surface advectionwas probablyan importantfactor. The increased layer through upwelling at the AD and vertical mixing wind speedsof the LGM [Petit et al., 1981] would advect [Gordon, 1981]. Comiso and Gordon [1996] attributed the ice formedin southernregions farther north at theLGM; the persistentformation of the Cosmonautpolynya during austral resultingleads (openings within the pack ice) wouldrapidly winter to the upwelling of warmer subsurfacewaters. freezeover due to the coldatmospheric temperatures [Smith Enhancedadvection due to the increasedmean wind speeds et al., 1990]. Cavalieriand Parkinson[1981] suggesteda also probably played an importantrole, as ice advected strongatmospheric cyclonic low pressuresystem led to rapid northwardinto openocean would melt more rapidly than in seaice expansionin theWeddell Sea by movingcold air and heavy ice covered areas [Martins,n, 1990]. Likewise sea ice equatorward. Comisoand Gordon [1998] found increasedlead and polynyaformation due to wind advection maximum winter ice extent in the Weddell Sea to be would enhanceice melting in spring,as theseopen water positivelycorrelated with meridionalwind speeds. areasabsorb solar radiation more efficientlythan ice-covered 464 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK areasdue to their lower albedo [Smith et al., 1990]. Comiso Persistent ice cover over the AD could also have and Gordon [1998] foundthat unusuallylow summersea ice slowed Ekman upwelling during winter monthsand, thus, extentsin the Weddell Sea follow winterswith higher than reducednutrient and carboninputs to the surfacelayer. Ice normal sea ice extent. They attributed this surprising would still advect away from the AD, and new ice would phenomenonto a combinationof wind effects including form rapidly in the leads owing to the very cold air increased northward advection of sea ice [Comiso and temperatures. However, some of the wind momentum Gordon, 1998]. imparted to surfacewaters in the modem SouthernOcean (where the AD is ice-free much of the year) would be 5.3. Sea Ice Seasonality absorbedby sea ice at the LGM. Wamserand Martinson [1993] concludedthat typical winter pack ice cover in the The greatlyexpanded SIZ of the last glacialperiod has Weddell Sea reduces the momentum flux from the profound implications for Southern Ocean ecology and atmosphereto the ocean by -33%. The momentumflux air-sea fluxes of CO2. The springice retreatduring glacial from atmosphereto oceanis little affectedby thin ice or at times would have begunlater in the seasonand muchfarther low ice concentrations,so any reductionin upwellingwould to the north,because of the cooleratmospheric temperatures be restrictedto winter months. This postulatedreduction in and expanded winter ice cover. Likewise the fall ice upwellingmight also be counterbalancedto someextent by expansionwould have begun earlier in the season. Indeed, the increased mean wind speeds (and thus northward the CLIMAP summer ice boundary and the boundary of advection) at the LGM [Petit et al., 1981]. Cooke and Hays [1982], which likely record the late spring/early summer (approximately mid-December) ice extent, are well north of today's maximum winter sea ice 6. Sea Ice and the Antarctic Polar Front extent [CLIMAP, 1981; Cooke and Hays, 1982; Burckle et al. 1982; Burckle, 1984]. This indicatesa substantialdelay 6.1o The Antarctic Polar Front as a Boundaryfor Sea Ice in the seasonalretreat of the pack ice at the LGM. Extent Primary production beneath heavy pack ice is negligibledue to the poor transmissionof light throughsnow Through much of the Drake Passage region, the and ice [Fischer et al., 1988; Wefer and Fischer, 1991; maximum extent of sea ice in the modem ocean is coincident Bouquegneauet al., 1992]. Thus, the effective growing with the mean location of the PF (Plate 4). The warmer sea seasonfor phytoplanktonsouth of the PF was shorterat the surfacetemperatures in SubantarcticSurface Waters north of LGM, with progressivelydecreased ice-free time as one the PF rapidly melt sea ice entering the front. For this moved poleward [Burckle and Cirilli, 1987; Crosta et al., reason,there is sometimesa pulseof ice-rafteddebris at the 1998b]. Crosta et al. [1998a] estimated months of sea ice PF [Watkins et al., 1982]. The location of the PF thus sets cover per year at the LGM based on diatom species the maximum extent of the seasonal sea ice cover in this area assemblagesin the sediments. They estimatedthat in areas of the SouthernOcean today. Similarly, in the southwest northof the modern-daymaximum ice extentthere was -3-5 Pacific (- 175øE), the PF also likely restrictsthe maximum monthsper year of substantialsea ice cover in the Atlantic extent of sea ice (Plate 4). While SST values were lower at Sector and 1-2 months in the Indian Sector at the LGM the LGM, Subantarctic waters north of the PF were still [Crostaet al., 1998a]. If the CLIMAP summerice boundary abovethe freezingpoint of seawater[CLIMAP, 1976, 1981] does mark a mid-December ice extent, these estimateswould and would thusrapidly melt seaice in areasnorth of the PF. seem to underestimate the duration of seasonal ice cover. We suggestthat the PF markedthe maximum extentof the Crosta et al. [1998a] note that their low values in the Indian seasonal sea ice over the entire Southern Ocean at the LGM sector are likely underestimatesdue to advection and (Plate 4). focusing of sedimentsor perhapsdue to diatom frustule The CLIMAP sea ice boundaryis north of the present dissolutionwithin the sediments.Thus, within the expanded mean locationof the PF in many areas(Plate 4). However, SIZ, the effective growing seasonfor phytoplanktonat the it is within the latitudinalrange of meandersof the PF over LGM was much reduced relative to the modem ocean, almost the entire Southern Ocean [Moore et al., 1999a]. perhapsby 50% or more in areas north of the modem Only in the southwestPacific near 180øW and in the maximum ice extent. southeastPacific around 100øW is the CLIMAP boundary north of the envelopeof PF pathsdocumented by Moore et al. [1999a]. It is also in these areas that the CLIMAP 5.4. PhysicalEffects of the ExpandedSeasonal Ice Zone on boundaryis mostsuspect due to a paucityof sedimentcores. Air-Sea CO2 Fluxes Cookeand Hays [1982] place the maximum ice extent-5 ø latitudefarther southalong 180øWbased on ice-rafteddebris Expanded seasonal ice cover would reduce the distributions in the sediments. Thus the CLIMAP maximum outgassingof CO2 to the atmosphereduring winter months ice extent could be recorded in the sediments without by actingas a barrierto air-seaexchange as happenstoday transportof ice north of the PF or a northwardmigration of in the Weddell Sea duringwinter [Weiss,1987; Bakker et al., the PF mean location. 1997; Crosta et al., 1998a]. This barrier effect can lead to supersaturationof CO2 beneath heavy ice cover. The accumulatedCO2 may not outgasduring the springice melt 6.2. A Glacial Northward Shift of the Antarctic Polar Front? becausethe input of meltwater lowers TCO2 and pCO 2in surface waters and spring phytoplanktonproduction also A number of authorshave suggestedthat the PF was decreasespCO 2 [Hoppemaet al., 1995; Ishii et al., 1998; shiftednorthward as much as 10ø of latitude during the last Sabine and Key, 1998]. ice age, possiblydue to a similar shift in the winds [Climap MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 465

Project Members, 1976; Hays, 1978; Dow, 1978; Prell et al., 2.5øC using a diatom transferfunction. These SST values 1980; Defelice and Wise, 1981; Morley, 1989; Howard and are similar to those found south of the PF today (mean Prell, 1992; see alsoKlinck and Smith, 1993]. Thesepapers summer SST at the poleward edge of the PF is ~2.8øC relate modem-day species assemblagesin the sediments [Moore et al., 1999a]). Yet Labeyrie et al. [1996] noted that (typically diatoms, radiolaria, or foraminifera are used) to the low instance of the "ice-related" factor in this core currentSST and water massdistributions then estimatepast indicatesthat it was alwaysnorth of the PF. Weaveret al. SST and frontal locations using the modem analogs. A [1998] note the replacementof "subpolar"by "polar" crucialassumption, either implicit or explicit, of all of these assemblagesof Foraminiferain Subantarcticwaters during studies is that the PF retains its role as an ecological glacial times. SST values north and southof the PF were boundary over glacial/interglacialtimescales. Thus if a cooler at the LGM [Pichon et al., 1992]; thus estimates of predominantly "polar" assemblage(an assemblagefound LGM SST basedon speciesassemblages are likely accurate, south of the PF today) is found farther north in sediments but no northwardmigration of the PF is requiredto account from the LGM, it is assumed that the PF has shifted for theseSST/species assemblage shifts. northward. The warmer, saltier Subantarctic waters found north of This interpretationoverlooks the possibilitythat today's the PF would rapidly melt sea ice in modern and glacial polar assemblagesmay have extendedtheir range north of times. Thus estimatesof maximumsea ice extentmay mark the PF duringglacial periods. We suggestthat while surface the LGM PF locationbetter than analogsof modern species isotherms(and polar biota assemblages)shift equatorward assemblages.However, as noted previously,estimates of duringglacial periods, the locationof the PF, which is tightly maximum sea ice extent are consistent with little or no controlledby topography,remains unchanged. There was northwardmigration of the PF. The maximumice extentof still a stronggradient in surfacetemperatures across the PF Cooke and Hays [1982] (based on ice-rafted debris) is during glacial times, but watersnorth and southof the front virtually identical to their positionfor the modem day PF, were colder. The PF marks the boundary between cool, except in the Atlantic sector,where the ice extends~1ø-2ø fresh Antarctic Surface Water to the south and warmer, farther north. This is consistent with little or no northward saltier Subantarctic Surface Water to the north. However, it migrationof the PF as meandersof the front couldbring sea is not tied to a particular isotherm. In fact there is ice north of the mean position. considerableregional and latitudinalvariability in mean SST values at the poleward edge of the PF within the modern The PF extends from the sea surface to the ocean floor, SouthernOcean [Moore et al., 1999a]. Thusthe PF todayis and, thus, its location is strongly influenced by bottom a physicalwater mass boundary and an ecologicalboundary. topography[Deacon, 1937; Moore et al., 1997; 1999a]. For However,the two likely becameuncoupled at the LGM with this reason,the winds can only indirectly affect the location of the PF. In several areas of the Southern Ocean, the the ecologicalboundary between polar and subpolarspecies shifting north of the PF. location of the PF is tightly controlledby the topography The PF todayforms an ecologicalboundary primarily [Moore et al., 1997, 1999a]. In these areas of strong due to the change in SST acrossthe front. Subantarctic topographiccontrol, a northwardmigration of the PF at the waters north of the PF reach much warmer temperatures LGM is unlikely. Strong topographiccontrol of the PF during summer than the cooler Antarctic SurfaceWaters to occursthrough much of the Drake Passageregion, along the the south. Some nutrients,notably silica, can have strong easternflank of Kerguelen Plateau (-75øE), and where the cross-frontalgradients that may also influence species PF crossesthe Pacific-Antarctic (~145øW), the Mid-Atlantic composition. The much cooler SST values in Subantarctic (~8øW), and SoutheastIndian Ridges(-145øE) [Moore et al., waters at the LGM (~3ø-6øC colder at the LGM [Pichon et 1997, 1999a]. In each of these areas the CLIMAP LGM al., 1992; Nelson et al., 1993; Ikehara et al., 1997]) may maximumsea ice extentis at or closeto the presentlocation have allowedtoday's polar speciesto thrive in Subantarctic of the PF (Plate 4). This indicatesstrongly that the PF did waters. Burckle[1984] arguedthat the northernlimit of high not migrate northwardin these areas,especially when the productivityby diatomsis controlledby a temperatureeffect low spatialresolution of the CLIMAP data set is considered. on the metabolic processesof diatoms. Neori and Moore et al. [1999a] showed that seasonal and interannual Holm-Hansen[1982] foundthat the photosyntheticrates of variability in the location of the PF is negligible in these Antarcticdiatoms dropped off sharplyat SST valuesabove areasof strongtopographic control; thus, variationsin wind 7øC. Fiala and Oriol [1990] found that Antarctic diatoms forcing over thesetimescales have no effect on the location of the front. could not survive in temperaturesabove 6-9øC. Thus the cooler SST valuesat the LGM may have allowed Antarctic Away from largetopographic features, above the deep diatomsto extendtheir latitudinal range equatorward. Silica ocean basins, the PF is able to meander over a broader concentrationsmay also limit diatomgrowth in Subantarctic latitudinal range [Moore et al., 1999a]. In these areas, a and northernAntarctic Surface Waters [Tr•guer and Jacques, northwardshift in the meanlocation of the PF is perhaps 1992]. Under iron-replete conditions, silicic acid possible. Interannualand seasonalvariability is higherin requirements decline relative to nitrate [Hutchins and theseareas, but temporalcoverage is poor [Moore et al., Bruland, 1998; Hutchins et al., 1998; Takeda, 1998]. 1999a]. Thus it is difficult to say whetherthe meanpath Hutchins et al. [1998] found that under iron-limitation shifts over interannualtimescales or is merely not well diatomswere more highly silicifiedthan underiron-replete resolved. It is in these areas that the CLIMAP ice extent conditions. Thus the increasedatmospheric flux of iron boundaryis farthestnorth of the modernPF location(e.g., may also have promoteddiatom production in Subantarctic ~ 110ø-90øW, ~80 ø-140øE, in Plate 4). watersby lowering cellular silica quotas. Thus the PF likely delimited the northwardextent of Labeyrie et al. [1996], studyinga sedimentcore from the SIZ throughoutthe SouthernOcean at the LGM (Plate ~96øE,46øS, estimated glacial summer SST valuesof-2.0 ø- 4). We suggestthat on average,the maximumsea ice extent 466 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK at the LGM was -1 o north of the mean location of the PF 7. SouthernOcean Ecology at the LGM depictedin Plate 4. In the areaswith strongtopographic control on the PF discussed above, maximum ice extent could not be very far north of the mean PF position (as 7.1. Antarctic Coastal Waters happenstoday in Drake Passage). In areas where the PF meandersmore extensively,northward meanders could allow Much of what constitutes coastal Antarctic waters in sea ice to reach areasseveral degrees of latitudenorth of the today'socean was eitherabove sea level or beneathgrounded mean PF location. In the modem Southern Ocean, the PF ice sheetsat the LGM [Kellogg, 1987]. This would lead to meanders over a latitudinal range that typically spans lower primary productionin coastalregions. Mackensenet between 2 ø and 5 ø of latitude [Moore et al., 1997, 1999a]. al. [1994] suggested reduced productivity along the Estimates of sea ice cover in the modem ocean and at continentalmargin at theLGM basedon sediment8 C data. the LGM are summarized in Table 2. We calculated the Due to the small areal extent of coastal waters, this would maximum ice extent at the LGM as 1 o north of the modem have little effect on the total productivity of the Southern day mean PF position. Thus the LGM SIZ encompassesthe Ocean [Arrigo et al., 1998]. However, any reductionin the PFR, POOZ, SIZ, and WRR regionsof the modem Southern productivity of coastal waters could have had dire Ocean (see Table 1). Recall that for Table 1 we used sea ice consequencesfor highertrophic levels that dependon these extent from the 1997-1998 seasonin calculatingregional areas for foraging, such as the Adelie penguins, which area, while in Table 2 we have used the climatologicalice reproduceonly on the Antarcticcontinent [Nicol and Allison, cover from the years 1978-1996. The LGM sea ice 1997]. The fact that theseshore-breeding birds survived the distributionis quite different from the modem ocean. The lastice age is evidencefor substantialhigh-productivity open POOZ covers -9.7 million km 2 in the modem Southern water near the Antarcticcontinent during australgummer at Ocean but did not exist at the LGM (Table 2). The SIZ was the LGM, likely owing to a strongsummertime retreat of the thusmuch larger in the glacialSouthern Ocean, coveting -30 seasonalice sheet as proposedby Crosta et al. [1998b]. million km2, comparedwith --19 millionkm 2 whichwe Persistent,large coastalpolynyas could alsohave provided calculateas the size of the SIZ today (Table 2). suitableforaging grounds. The ocean area south of our smoothed CLIMAP maximumice extent(Plate 4), covers37.5 millionkm 2. We suggestthat the sparse core data used by the CLIMAP 7.2. The Seasonal Ice Zone - From the Antarctic Polar Front project overestimatedmaximum ice extent in someregions, to the Antarctic Continent notablyin the Pacific Sector,where few coreswere available and the CLIMAP boundaryis well northof the PF (Plate 4). Over most of the Southern Ocean our estimate of maximum Much of the primary productionin the modem SIZ ice extent basedon PF locationis in rough agreementwith occursduring a short(several weeks), intensephytoplankton previousestimates based on the geologicalrecord. Gloersen bloom following the retreatof the seasonalice cover [Smith et al. [ 1992] calculatedmodem maximum ice extent (all ice and Nelson, 1986; Trdguer and Jacques, 1992]. concentrations)as 19 millionkm 2 with some3.5 millionkm 2 Macronutrientsare typically not completelyused up in ice of open water within this region. Likewise they calculated edgeblooms today [Jenningset al., 1984; Nelsonand Smith, summer ice cover as -4-4.5 million km2, which includes 1986; Smith and Sakshaug,1990]. Iron availability may some1.5 millionkm 2 of openwater and substantial areas ultimately limit macronutrient drawdown [Sedwick and with < 70% ice cover [Gloersen et al., 1992]. DiTullio, 1997]. Thus the increased aeolian iron fluxes at

Table 2. Areal Extent of Sea Ice Coverageand EcologicalRegions in the Modern SouthernOcean and at the Last Glacial Maximum.

Modern Southern Ocean LGM Southern Ocean

Winter maximum ice extent 20.1 31.0

Summer minimum ice extent 1.2 1.2

Seasonal ice zone 18.9 29.8

Permanentlyopen ocean zone 9.7 0.0

Modem day winter seaice extentwas calculatedas areaswith > 5% seaice coverduring the monthof Augustin our sea ice climatology(1978-1996). Winter seaice extent at the LGM is estimatedto extendon average1 degreeof latitudenorth of the meanlocation of the AntarcticPolar Front as definedby [Moore et al., 1999a]. Modem summersea ice extentis calculatedas areaswith > 70% seaice coverduring February in our seaice climatology. LGM summersea ice extentis assumedidentical to the modemera, basedon the data of Crosta et al. [1998b]. The PermanentlyOpen OceanZone (POOZ) is calculatedas the area southof the meanposition of the AntarcticPolar Front [Moore et al., 1999a]with < 5% ice coverduring August in our seaice climatology. The SeasonalIce Zone (SIZ) is calculatedas the differencein area betweenthe wintermaximum and the summerminimum ice extent.All valuesgiven are millions of km2. MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 467 the LGM probably led to more intensespring blooms of sediments [Anderson et al., 1998]. Farther south at core phytoplanktonfollowing the retreatingice edge.The greatly RC13-259, organiccarbon accumulation was lower at the expandedsize of the SIZ at the LGM also meansthat ice LGM by nearly 50% [Andersonet al., 1998]. High edge blooms would have occurredover a much larger area accumulation rates of ice-rafted debris in this core indicate than in the modernocean. In addition,the elevatedice age substantialice cover at the LGM [Cooke and Hays, 1982]. aeoliandust flux may have allowedlarger diatoms to bloom Thus if the growingseason was reduced by 50%, the carbon within the SIZ throughout the austral growing season, accumulationdata would argue that daily export fluxes did without the postbloom species shift toward smaller not changebetween the LGM and the Holoceneat this core phytoplanktonspecies (with lower exportproduction) often location,while the other coresjust to the northhad increased seentoday. As notedpreviously, this is basicallythe pattern export productionat the LGM. The ecologicalregime shift observedin the southernWeddell and Ross Seastoday. at the LGM suggestedhere (with increasedproduction by Sedimentaryestimates of paleoproductivityfor the area larger phytoplanktonspecies) could have resultedin a more southof the PF tend to indicatelower export productionat efficient transportof organic carbon to the deep ocean as the LGM comparedwith the modem era [Mortlock et al., larger cells (especiallydiatoms) sink fasterthan small cells. 1991; Charles et al., 1991; Shemesh et al., 1993; Shimmield Furthersupport for this hypothesiscan be found in the et al., 1994; Kumar et al., 1995]. Francois et al. [1997] dataof Francoiset al. [1997],who analyzed the bulk/• N foundexport fluxes in the easternIndian sectorwere similar signal in modern and glacial age sediments and found to modern values. As previously mentioned, consistentlyhigher valuessouth of the PF at the LGM. This paleoproductivityestimates that rely on biogenic opal are indicates more efficient nitrate utilization at the LGM probably too low becauseof changingC/Si ratios over (-70%) than in the modem SouthernOcean (-30% [Francois glacial-interglacialtimescales [Takeda, 1998]. et al., 1997]). Francois et al. [1997] also noted that annual Bottom current redistribution of sediments must also be exportproduction (as measured by excess23•Pa0/excess23ørh0 takeninto account, and normalizing opal flux to excess23øTh particle flux method) and preserved opal flux to the activitycan correctfor this resuspensionproblem [Francois sedimentsdeclined (South Atlantic and centralIndian Ocean) et al., 1997]. For example, Charles et al. [1991] and or remained at similar levels to modern values (East Indian Mortlocket al. [1991] arguedfor lower glacialproductivity sector).Export fluxes were similar over mostof the region in core RC13-271 locatedjust southof the PF (51ø59'S, south of the PF (averages south of PF of 0.13 (Holocene) 04ø31'E). Kumar et al. [1995] showedevidence of increased and0.14 (LGM) for excess23•Pa0/excess23øTh0,anda range of particleand organic carbon flux at this siteduring the LGM 0.4-1.0preserved 23øTh-normalized opal fluxes in cm'2 kyr -• based upon different paleoproxies with the Th flux duringboth periods [Francois et al., 1997, Figure 1]). correction. Andersonet al. [1998] documentedhigher From these two facts, increased nitrate utilization and organiccarbon accumulation at the LGM in this core relative a similaror lower annualexport flux, Francoiset al. [ 1997] to the Holocene. concludedthat nutrientinputs to the surfacelayer musthave The cores farthest south of the PF seem to show the beenmuch lower duringthe LGM, which they attributedto greatestLGM reductionin flux to the sediments[Charles et increasedvertical stratificationof the water column. They al., 1991; Mortlock et al., 1991; Shimmield et al., 1994; estimatedthat nutrientinput to the surfacelayer would have Kumar et al., 1995; Anderson et al., 1998]. This is to have been 10 times lower at the LGM due to increased consistentwith increasedwinter ice cover(shorter growing stratification(water supply by verticalmixing and upwelling season)at higher latitudes [Burckle, 1984; Charles et al., of-30 m yr-• for the Holoceneand -3 m yr'• at theLGM 1991; Shimmield et al., 1994]. All of the cores in these [Francois et al., 1997]). This representsa drasticdecrease studies(except for ShimmieMet al. [1994]) are north of the in nutrientinputs relative to the modernera. Hoppemaet al. maximumice extentof today'sSouthern Ocean (Plate 4). [ 1995], for example, estimated entrainment rates in the Thusthe growingseason for phytoplanktonat the LGM was WeddellSea today of 4-5 m month-1, and Gordon and Huber much shorter,as theseregions were likely ice-coveredfor [ 1990] estimatedupwelling rates of WeddellDeep Water at much of the year. This raisesthe possibilitythat while 45 m yr-• Thesetwo estimatescome from areas where the annualfluxes to the sedimentsmay have beenlower at the influenceof the seasonalice coveris very strong. LGM, primaryproduction and exportfluxes over daily to We suggest an alternate interpretation of the data monthly timescalesmay have been similar or increased presentedby Francois et al. [ 1997] that doesnot require a relative to the modern Southern Ocean. largereduction in nutrientinputs to the surfacelayer. Nearly There is evidencefor this hypothesisof increased all of the cores analyzed in their study are north of the spring/summerprimary productionat the LGM. In their maximumsea ice extentof today and southof the maximum analysisof core RC13-271 (51ø59'S,04ø31'E), Anderson et ice extent at the LGM (Plate 4). In the modern Southern al. [1998] comparedmean Holoceneand LGM opal and Ocean, these regions are marked by low phytoplankton organiccarbon (C-org) accumulationrates. Ratesof opal biomassand low exportproduction over a 12 monthgrowing accumulationwere similar (123 mmol m-2 yr-• for the season. Thus nitrate concentrationsremain high, and the Holoceneand 117 mmol m-2 yr -• for the LGM), but the sediment 6 •SN data indicate low nitrate utilization. At the accumulationof organiccarbon was substantially higher at LGM, the effective (ice-free) growing seasonwas much theLGM (5.56mmol m -2 yr -• LGM, and3.77 mmol m-2yr -• shorter,with perhaps only 3-6 months of productionperyear. Holocene)[Anderson et al., 1998]. This siteis slightlysouth Relativelyshort, intense phytoplankton blooms stimulated by of the PF and northof the maximumice extenttoday but thesea ice retreat and increased atmospheric iron deposition was ice coveredmuch of the year at the LGM (Plate 4). would have resultedin much more efficient utilization of Thusdespite the shortereffective growing season, the annual surfacenitrate (as happens in coastaland ice edge blooms in organicC flux increased. Four other cores southof 50øSin themodem Southern Ocean). This could explain the results thisregion had higheraccumulations of authigenicU at the presentedby Francoiset al. [1997]without decreasing LGM, alsoimplying higher fluxes of organiccarbon to the nutrientinputs to the surfacelayer at theLGM. 468 MOORE ET AL.: SOUTHERNOCEAN CARBON SINK

The large reductionin annual nutrient inputs to the extracellularmucilage during the colonial life stageand that surface layer due to increasedwater column stratification low molecularweight carbohydratesaccumulate in the proposedby Francoiset al. [1997] appearsat variancewith mucilage.Thus sinking Phaeocystis colonies can be efficient other observations. An increase in water column exportersof carbonfrom surface waters with C/N production stratificationwould have to be salinity driven and would be ratios well abovethe typical Redfield ratio seenin most controlledlargely by the balancebetween ice formation and phytoplankton[Smith et al., 1996, 1999]. Phaeocystisspp. melting, as the cooler sea surfacetemperatures of the LGM overwinter within the pack ice, and the ice age SIZ would tend to destabilizethe water column. Areas with high expansionwould likely extend their range northward with the inputof meltwaterduring •ustral spring would become more seaice [Gibsonet al., 1990]. This "seedstock" from melting stratified, but this effect would be restrictedto a relatively seaice and high photosyntheticefficiencies at low light short period. In these northern areas where ice melting levels often allows Phaeocystisto dominateearly spring exceededice formation,the strongwinds over the Southern phytoplanktonblooms in polarregions today [Gibson et al., Ocean would erode any meltwater-inducedstability by the 1990; Smith et al., 1991; 1999; Crocker et al., 1995; Smith following fall (as happensin the SIZ today), when deep and Gordon, 1997]. In the Antarctic, densePhaeocystis mixing (and thus high nutrient injection) would occur. bloomsoccur mainly in coastal/shelfwaters, which may be Severalauthors have suggestedopen ocean deep convection relatedto iron availability [Smithet al., 1999]. Surfaceto at the LGM due to a decrease in the stratification of surface volumeratio considerationssuggest that the colonial form of waters [Martinson, 1990; Rosenthal et al., 1997]. Areas Phaeocystisin particularmay have difficulty obtaining farthersouth near the AD, wheremuch of the seasonalpack necessaryiron over much of theSouthern Ocean [Sunda and ice is formed, would be less stratified due to increased brine Huntsman,1997]. Thus the combinationof elevatedaeolian injectionduring ice formation. Much of the nutrientinput to ironinputs and expansion of the SIZ duringthe last ice age Southern Ocean surface waters occurs at the AD, and the probablyled to a rangeexpansion and increased production higher mean wind speeds[Petit et al., 1981] and increased by Phaeocystissp. [Martin,1990; Gibson et al., 1990]. brine input arguethat the upwellingrate and nutrientinput There is someevidence for this rangeexpansion in the was likely higher at the LGM (at least during austral ice corerecord. Phaeocystisspp. are important producers of summer). Thus the large increasein stratificationnecessary dimethylsulfoniopropionate(DMSP), whichis a majorsource to accountfor the -10 fold reductionin nutrientinput over of cloud condensationnuclei, after conversionto dimethyl annual timescalessuggested by Francois et al. [1997] is sulphide(DMS) [Charlsonet al., 1987;Gibson et al., 1990]. unlikely. The proposed stratification increase and Duringglacial stages, the atmosphericcontent of non-seasalt correspondingreduction in CO2 outgassingwould have sulphate(NSS sulphate)recorded in ice cores from the occurredonly seasonallyduring the springice melt. Antarcticwas -20-46% higherthan in the modemera, likely In contrast,current understanding of the role of iron in due to increasedproduction of DMSP by marine biota Southern Ocean ecosystems argues strongly that the [Legrand et al., 1988; 1991]. Phaeocystis forms an combinedinfluence of meltwaterinput and elevatedaeolian importantcomponent of the ice algaein the SouthernOcean, dust flux would have led to higherprimary production and where it producessubstantial amounts of DMSP, much of exportfluxes duringaustral spring and summer. It couldbe whichis releasedto the water columnduring spring ice melt argued that the higher wind speedswould deepenmixed [Kirst et al., 1991]. Phaeocystisblooms result in high layers, leading to light limitation and thus preventing oceanicDMS concentrationsand consequentlyhigh DMS productionincreases during summer. However, the high fluxesto the atmosphere[Gibson et al., 1990; Crockeret al., nitrate utilization efficienciesreported by Francois et al. 1995; Turneret al., 1995]. Thusan ice agerange expansion [1997] arguethat this did not happen. In addition,because and increasein primary productionby Phaeocystiscould light harvesting efficiency is greater in iron-repletecells accountfor the LGM increasein NSS sulphatein Antarctic [Sunda and Huntsman, 1997], the increased iron flux from ice core data reportedby Legrand et al. [1988, 1991]. The the atmosphereduring the last ice age may haveimproved primary oxidation productsof DMS in the atmosphereare the photosyntheticefficiencies of Antarcticphytoplankton at SO2 (which is then converted to NSS sulphate) and low light levels,allowing them to adaptto any deepeningof methanesulphonate(MSA) [Legrand et al., 1991]. In the mixed layers. modern ocean, the molar ratio of MSA/NSS sulphate increaseswith latitude in the SouthernHemisphere, likely due to air temperatureeffects on DMS oxidationprocesses 7.3. PrimaryProduction and DMS Releaseby Phaeocystisat [Bateset al., 1992]. Thus the significantlyhigher MSA/NSS the LGM sulphate ratios in glacial Antarctic ice cores are also consistent with increased high latitude oceanic DMS Shiftsin speciescomposition could also complicate the productionby Phaeocycstis[Legrand, 1997]. picturesouth of the PF. As pointedout by Martin [1990], Phaeocystisspp. are commonphytoplankton in Antarctic coastalwaters and may have increasedtheir abundanceand 7.4. The SubantarcticWater Ring: From the PF to the North range at the LGM. Thesephytoplankton contain no hard SubtropicalFront parts(i.e., opal, CaCO3)and wouldlikely leavelittle in the way of a sedimentarysignal [Martin, 1990]. Phaeocystishas The primaryfactors affecting primary productivity in a complexlife cycle,alternating from a colonialform with this regionat the LGM were the increaseddeposition of non-motilecells embeddedin a mucilagematrix severalmm atmosphericiron and the substantiallycooler SST. The in diameter,to single,motile cellsfrom 3 to 9/•m in increasediron input to thesewaters would likely increase size [Matrai et al., 1995]. Matrai et al. [1995] found that > primaryand exportproduction [Martin, 1990;Sunda and 50% of carbon fixed by Phaeocystiscan be allocated to Huntsman, 1997]. As discussedpreviously, cooler SST MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 469 values and possibly lower silica requirementsmay have of heavy sea ice cover. This is the pattern recently enableda northwardrange expansionby Antarctic diatoms. suggestedfor the southernRoss Sea [Bates et al., 1998].Net There is substantialevidence that Subantarcticprimary CO2flux into the oceanwould thus increase, even if annual productionwas increasedat the LGM. The sedimentary export production remained at similar levels or even record for Subantarcticwaters consistently shows increased declined. A significantfraction of this productionincrease biogenicopal and/ororganic carbon flux to the sedimentsat duringaustral spring/summer was likely due to Phaeocystis the LGM [Mortlock et al., 1991; Charles et al., 1991; Kumar and would leave little in the way of sedimentaryrecord et al., 1993, 1995; Nelson et al., 1993; Rosenthal et al., [Martin, 1990]. 1995, 1997; Francois et al., 1997; Anderson et al., 1998]. In addition, physical factors associatedwith the Francois et al. [1992] argue for increased nutrient expandedSIZ wouldhave decreased the outgassingof CO2 consumptionin Subantarcticwaters of the Indian sector to the atmosphereat high southernlatitudes during austral during glacial times. Kumar et al. [1995] noted that the fall/winter. Persistentice-cover would act as a physical increasedflux of organic carbon to the sedimentsat the barrier to air-sea gas exchangeand may have reducedthe LGM exceededthe increasein biogenicopal over much of upwellingrate of carbonrich waters at theAD by decreasing the Subantarctic. Increased C/Si ratios in diatoms due to the momentumflux to surfacewaters during winter months. relief of iron stressmay have played a role [Takeda, 1998]. Increased vertical stratificationduring the spring ice melt Kumar et al. [1995] also estimatediron flux to the sediments could also reducethe mixing of carbon-richwaters into the in the Atlantic sectorwas -5 times higher at the LGM. This surfacelayer [Francoiset al., 1997]. elevated iron flux was attributed to increased aeolian A schematic summarizing these suggestedice age deposition of iron from terrestrial sourceson the South primary productionpatterns and their effect on air-seafluxes American continent [Kumar et al., 1995]. The flux of of CO2 is presentedin Figure 1. By necessityFigure 1 organiccarbon to the sedimentswas also greatlyincreased in representsa simplification;for example,within the SWR and this region, by a factor of 4 over the area from 42ø-54øS PFR there are regions of elevated export productionand [Anderson et al., 1998]. strong CO2 drawdown. However, mean chlorophyll concentrationsindicate an overall pattern similar to that presentedhere (comparePlate 3 and Figure 1). The LGM pattern'for the SWR (not shown) would be similar to the 8. Air-Sea CO2 fluxes at the LGM modernpattern but with a strongerseasonal cycle, peaking duringsummer. As notedpreviously, the POOZ did not exist As noted in section 7.4, available evidence indicates at the LGM. The entire area from the PF south to the that export productionin Subantarcticwaters was higher at Antarcticcontinent was shiftedtoward the patternseen in the the LGM and that this increase was stimulated by the Weddell-RossRegion todayø There would be a strongflux increasedatmospheric deposition of iron as postulatedby of CO2 into the oceansduring the ice-free growing season Martin [ 1990] [Mortlock et al., 1991; Charles et al., 1991; and little net flux the restof the year due to heavyice cover Kumar et al., 1993, 1995; Nelson et al., 1993; Rosenthal et (Figure 1). The lengthof the ice-freegrowing season would al., 1995, 1997; Francois et al., 1992, 1997; Anderson et al., vary with latitude, with a progressivelydecreased ice-free 1998]. This would increasethe flux of atmosphericCO2 into seasonas one moved poleward. Thus the four ecological Subantarctic waters particularly during spring/summer regionsfrom the PF southto the Antarcticcontinent outlined months,when productionwould be at a maximum. by Trdguerand Jacques[1992] were replacedby one large Changesin ecosystemstructure could also affect air-sea SIZ at the LGM. The type of ecologicalregime shift fluxesof CO2 if they alteredthe meanCaCO3/organic C ratio suggestedhere, from a small phytoplanktondominated of sinking matter, which would alter surface alkalinity assemblageto one dominatedby larger diatomspecies, has [Howard and Prell, 1994]. Lowering this ratio would result been documentedfor the glacial North Pacific [Sancetta, in increasedalkalinity and reducedpCO2 in surfacewaters 1992]ø Sancetta[1992] suggestedrelief from iron-limitation [Howardand Prell, 1994]. Lower CaCO3 accumulationrates as a possiblecause of the glacialshift towardlarger diatoms [Nelson et al., 1993; Howard and Prell, 1994] and increased and a higher productivity regime. The SubarcticNorth opal and organic carbon fluxes [Mortlock et alo, 1991; Pacific is also a HNLC regionin the modemocean [Martin, Charles et al., 1991; Kumar et al., 1995; Anderson et al., 1992]. 1998] to Subantarcticsediments imply that the ratio of Increasing primary and export productionover the CaCO3/Corgin sinking matter was lower at theLGM. This region south of the PF to values approachingwhat is would have furtherenhanced the flux of CO: into the ocean observedin the Weddell-RossRegion today would resultin [Howard and Prell, 1994]. This regiontherefore would have a globally significantsink for atmosphericCO 2. Evidence been a significantlystronger sink for atmosphericCO2 at the for this is seen in the many modeling studies of such LGM. production increases in Southern Ocean waters [i.e., South of the PF, the situation becomes more Sarmientoand Toggweiler,1984; Sarmientoand Orr, 1991; complicated,and interactionsbetween the sea ice and the Joosand Siegenthaler,1991]. Some simplecalculations can biota over seasonaltimescales would strongly affect net illustrate the potential impact of the proposedecological air-seafluxes of CO:. Air-sea fluxes of CO: throughthe regime shift on export production. Nelson et al. [1996] pack ice are negligible compared with open water areas estimateda seasonallyintegrated f ratio of 0.42 and a total [Weiss, 1987; Augstein,1994]. There was likely a strong annualprimary production of 142 g C m-2 yr -• for theRoss flux of CO: into the oceans during austral spring and Sea Shelf area. This indicatesan annualexport production summer (due to relatively intense phytoplanktonblooms of-60 g C m-2 yr4. This productionoccurs over an -6 driven by sea ice retreatand increasediron deposition)and month growing seasonand typically does not result in full a negligiblenet flux for much of the rest of the year because depletion of macronutrients[Nelson et al., 1996]. If a 470 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK

High export production HighCO 2 flux,I, iLargeFe -not phytoplanktonlimiting High export production HighCO 2 flux LargeFe-not phytoplanktonlimiting Fe -limiting Low export production Smallnet CO2flux ,I, 1' Small phytoplankton negligible heavysea ice ne?gibleGO2 flux hea.vyI CO flux

winter summer winter summer A) ModernWRR B) Modern SlZ Fe - not limiting HighCO2 flux I HighLargeexport phytoplanktonproduction

Low export production Lownet CO 2 flux,I, 1' ISmallFe -limiting phytoplankton negligible

winter summer winter summer C) ModernPOOZ, PFR, SWR D) Last GlacialMaximum WRR, SIZ, PFR Figure1. Displayedis a schematicof proposedseasonal patterns in primaryand export production, dominantphytoplankton size class, and air-sea carbon dioxide flux in differentecological regions of the SouthernOcean during the modemera andat the last glacialmaximum. Air-seaflux through similarexport production occurred over the entire SIZ during Subantarcticwaters (SWR is 48.7 million km-2 from Table the last ice age (30.7 millionkm -2, total areaof the WRR, 1). Currentestimates for totalprimary production over most SIZ, POOZ, and PFR from Table 1), the annual carbon of thisregion today are in therange of 50-150g C m-2 yr -• exportfrom surface waters would have been 1.84 x 10•5 gC [Antoine et al., 1996; Behrenfeldand Falkowski, 1997]. yr4, or 1.84 Gt C yr-•. Smith [1991] estimatesnew Banse [1996] noted that the f ratio in the SWR ranges productionfor this regiontoday as 0.47 Gt C yr'• Minas between0.3-0.5. Taking a meanannual production of 100 and Minas [ 1992]estimate nitrate-based new productionfor g C m-2 yr -• anda meanf ratioof 0.4gives an annual export Antarcticwaters as 30.2 g C m'2 yr 4 overthe growing season productionof-40 g C m-2 yr-•o Export production from the from winterto late summer,which would correspond to 0.93 modernSWR wouldthen amount to •-2.0 Gt C yr'•. The Gt C yr4 over30.7 million km '2. Weferand Fischer [1991] data of Kumar et al. [1995] suggesta large increasein estimatedexport flux at 100 m depthfor the SouthernOcean exportproduction (by at leasta factor of 2) at the LGM [see from sedimenttrap data at 0.17 Gt C yr4. Thuscarbon alsoRosenthal et al., 1995]. Andersonet al. [1998] founda exportsouth of the PF couldhave increasedby -0.9-1.6 Gt four-fold LGM increasein organiccarbon accumulation in C yr4 atthe LGM. Thislevel of increasedexport production Subantarctic sediments. A doublingof exportproduction would still not have fully utilized availablemacronutrients. would increaseexported carbon from thesewaters to 4.0 Gt Much of this exportedcarbon would be replacedby C yr-•. atmosphericCO 2 enteringthe ocean. Increased production by Thus, if exportproduction doubled within the SWR and Phaeocystissp. andhigher C/Si ratiosin diatomsat theLGM increasedby -0.9-1.6 Gt C yr-• in watersfarther south, would both tend to minimizethe sedimentarysignal from export production from the entire Southern Ocean would this increasedexport production. havebeen some -2.9-3.6 Gt C yr-• greaterat theLGM than Increasedexport production north of the PF couldalso at present. Martin [1990] estimated that the observed haveglobal significance because of thelarge area covered by drawdownof atmosphericCO2 during the last ice age MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 471 representeda -170 Gt C flux out of the atmosphere.The and sea ice dynamicscould provide valuable insights and estimated increase in Southern Ocean export production greatlyenhance our interpretationof the geologicalrecord. calculatedabove could have rapidlyremoved this amountof CO2. Thesecalculations represent conservative estimates of Acknowledgements.The authorswould like to thank Ricardo carbon flux as they do not require full utilization of Letelier,Charlie Miller, Sam Laney, and Luz Delgadofor comments macronutrients in Antarctic waters and do not include the and suggestionson early drafts of this paper. Thanks also to two effectsof an expandedwinter sea ice extenton outgassingof anonymousreviewers. The authorswould also like to thank the CO2. SeaWiFS project (Code 970.2) and the DistributedActive Archive Center (Code 902) at the GoddardSpace Flight Center, Greenbelt, MD 20771, for the productionand distributionof the oceancolor data,respectively. These activities are sponsoredby NASA's Earth Science EnterpriseProgram. We would also like to thank the 9. Conclusions NationalSnow and Ice Data Centerfor the modem-daysea ice data. This work was fundedby a NASA EarthSystem Science Fellowship (J.K.M.), NASA/EOS grant NAGW-4596 (M.R.A.), and by NSF In summary,the SouthernOcean was a significantly grant OCE 9204040 (J.G.R.). strongersink for atmosphericCO2 during the lastice ageand contributed substantially to the observed lowering of atmosphericCO2 levels. Availableevidence indicates that References the increasedatmospheric deposition of iron led to increased exportproduction and consequentlya stronger drawdown of atmosphericCO 2 in Subantarcticwaters [Kumar et al., 1995; Anderson, RF, No Kumar, R.A. Mortlock, P.N. Froelich, P. Anderson et al., 1998]. South of the PF, increased iron Kubik, B. Dittrich-Hannen, and M. Suter, Late-Quaternary depositionand the effectsof the greatlyexpanded seasonal changesin productivityof the SouthernOcean, J. Mar. Syst.,17, ice sheet appear to have led to an intensificationof 497-514, 1998. Antoine, D., A. Jean-Michel, and A. Morel, Oceanic primary phytoplanktonblooms, increasing nitrate utilization from -30 production2. Estimationat global scale from satellite(coastal to -70% as estimatedby Francoiset al. [1997]. This would zonecolor scanner)chlorophyll, Global Biogeochem.Cycles, 10, resultin a strongsink for atmosphericCO2 due to biological 57-69, 1996. productionduring the shortenedgrowing season of the LGM, Arrigo, KR, D. Worthen, A. Schnell, and M.P. Lizotte, Primary as happenstoday in the southernWeddell and Ross Seas productionin SouthernOcean waters, J. Geophys.Res., 103, [Takahashi et al., 1993, 1997; Bates et al., 1998]. With 15587-15600, 1998. minimal air-seafluxes of CO2 the restof the year becauseof Augstein, E, Future Ocean-atmosphereresearch in the Antarctic persistentsea ice cover, the entire region southof the PF region, in Antarctic Science- Global Concerns,edited by G. Hempel, pp. 278-282, Springer-Verlag,New York, 1994. would havebeen a significantlystronger sink for atmospheric Bakker, D.C.E., H.J.W. De Baar, and U.V. Bathmann,Changes of CO2 than in the modernocean. We have arguedthat the carbondioxide in surfacewaters during springin the Southern total biologicalexport of carbonfrom surfacewaters south Ocean, Deep Sea Res. Part II, 44, 91-127, 1997. of the PF was higherat the LGM than at present. However, Banse, K, Low seasonality of low concentrationsof surface evenareas where annual export production d•lined couldbe chlorophyllin the Subantarcticwater ring: Underwater irradiance, strongernet sinksfor atmosphericCO2 than at presentdue to iron, or grazing?,Prog. Oceanogr.,37, 241-291, 1996. the interactions between the sea ice and biota over seasonal Bard, E., B. Hamelin, R.G. Fairbanks, and A. Zindler, Calibration timescales• of •4C timescaleover the past 30,000 years using mass spectrometricU-Th ages from Barbadoscorals, Nature 345, Additionalanalysis of sedimentcores will improveour 405-410, 1990: understandingof SouthernOcean ecology at the LGM. New Bamola, J.M., D. Raynaud, Y.S. Korotkevich,and C. Lorius, paleoproductivitystudies in the Pacific sectorand southof Vostok ice core provides 160,000-yearrecord of atmospheric -60øS shouldbe a high priority, as there is relatively little CO2, Nature, 329, 408-413, 1987. data available at present. Further examination of the Bates, N.R., D.A. Hansell, C.A. Carlson, and L.I. Gordon, geologicalrecord will also improveestimates of summersea Distribution of CO2 species, estimates of net community ice extent at the LGM and could be used to teat our production,and air-seaCO2 exchangein the RossSea polynya, J. Geophys.Res., 103 2883-2896, 1998. hypothesisthat maximum winter ice extent was at or only Bates, T.S., J.A. Calhoun, P.K. Quinn, Variations in the slightly north of the modem mean PF location. methansulfonate to sulfate molar ratio in submicrometer marine The aeolian flux of iron to the oceans is much lower in aerosalparticles over the SouthPacific Ocean, J. Geophys.Res., the Pacific sector than in the Atlantic and Indian basins 97, 9859-9865, 1992. [Duce and Tindale, 1991]. Thus it is in the Pacific basin Behrenfeld,M.J., and P.G. Falkowski, Photosyntheticrates derived that the strongestbiological response to the elevatedLGM from satellite-based chlorophyll concentration, Limnol. Oceanogr., 42, 1-20, 1997. dust fluxes shouldhave occurred. New paleoceanographic Belkin, I.M., and A.L. Gordon, Southern Ocean fronts from the techniqueshold promise for testing the hypothesisthat Greenwich meridian to Tasmania, J. Geophys. Res., 101, phytoplanktongrowth rates were increasedin glacial times 3675-3696, 1996. due to this increasediron flux from the atmosphere[Bidigare Bidigare,R.R., et al., Iron-stimulatedchanges in •3Cfractionation et al., 1999]. and export by equatorial Pacific phytoplankton:Toward a Sedimentaryrecords integrate over long time periods, paleo-growthrate proxy, Paleoceanogr.,(in press),1999. however,and providelimited informationabout the seasonal Bouquegneau,J.M., W.W.Co Gieskes, G.W. Kraay, and A.M. interactions between sea ice and the biota discussed here. Larsson,Influence of physical and biological processeson the concentrationof 02 and CO2 in the ice-coveredWeddell Sea in Ultimately, our best hope for defining the role of the the springof 1988, Polar Biol., 12, 163-170, 1992. SouthernOcean in the glacial lowering of atmosphericCO2 Boyd, P., A. Watson, and C. Law, Spotlight is on iron at may lie in the realm of modeling. A general circulation international SOIREE in Southern Ocean, U.S. JGOFS model of the LGM climate that incorporatesrealistic biota Newsletter, 9, 10-11, 1999. 472 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK

Broecker,W.S., and T.H. Peng, The causeof the glacial to planktonblooms and carbondioxide drawdownin the Southern Interglacialatmospheric CO2 change:A polar alkalinity Ocean. Nature, 373, 412-415, 1995. hypothesis,Global Biogeochem.Cycles, 3, 215-239, 1989. Deacon,G.E.R., The hydrologyof the SouthernOcean, Discovery Burckle, L.H., Diatom distributionand paleoceanographic Reports 15, 1-124, 1937. reconstructionin the SouthernOcean: Present and last glacial Defelice, D.R., and S.W. Wise Jr, Surface lithofacies,biofacies, and maximum,Marine Micropaleontol.,9, 241-261, 1984. diatom diversitypatterns as modelsfor delineationof climatic Burckle,L.H., D. Robinson,and D. Cooke,Reappraisal of sea-ice changein the southeastAtlantic Ocean,Mar. Micropaleontol.,6, distribution in Atlantic and Pacific sectorsof the Southern Ocean 29-70, 1981. at 18,000 yr BP, Nature, 299, 435-437, 1982. DeMaster, D.J., O. Ragueneau,and C.A. Nittrouer, Preservation Burckle,L.H., and J. Cirilli, Origin of diatomooze belt in the efficiencies and accumulationrates for biogenic silica and Southern Ocean: Implications for late Quaternary organicC, N, and P in high-latitudesediments: The RossSea, J. paleoceanography,Micropaleontol., 33, 82-86, 1987. Geophys.Res., 101, 18501-18518, 1996. Cavalieri, D..J, and C.L. Parkinson,Large-scale variations in Dow,R.L., Radiolarian distribution and the late Pleistocene history observedAntarctic sea ice extentand associatedatmospheric of the southeasternIndian Ocean,Mar. Micropaleontol.3, circulation,Mon. WeatherRev., 109, 2323-2336, 1981. 203-227, 1978. Charles,C.D., P.N. Froelich,M.A. Zibello,R.A. Mortlock,RA, and Duce,R.A., andN.W. Tindale,Atmospheric transport of ironand J.J.Morely, Biogenicopal in SouthernOcean sediments over the its depositionin the ocean,Limnol. Oceanogr., 36, 1715-1726, last450,000 years: Implications for surfacewater chemistry and 1991. circulation,Paleoceanography, 6, 697-728, 1991. Dugdale,R.C., andJ.J. Goering, Uptake of newand regenerated Charlson,R..J, J.E. Lovelock,M.O. Andreae,and S.E. Warren, formsof nitrogenin primaryproductivity, Limnol. Oceanogr., Oceanicphytoplankton, atmospheric sulfur, cloud albedo,and 12, 196-206, 1967. climate, Nature, 326, 655-661, 1987. Eppley,R.W., andB.J. Peterson, Particulate organic matter flux and CLIMAP (Climate: Long-RangeInvestigation, Mapping, and planktonicnew productionin the deep ocean,Nature, 282, Prediction)Project Members, The surfaceof the ice ageEarth: 677-680, 1979. Quantitativegeologic evidence is usedto reconstructboundary Fiala,M., andL. Oriol,Light-temperature interactions onthe growth conditionsfor the climate 18,000 years ago, Science191, of Antarcticdiatoms, Polar Biol., 10, 629-636, 1990. 1131-1137, 1976. Fischer,G., D. Ftitterer,R. Gersonde,S. Honjo,D. Ostermann,and CLIMAP ProjectMembers, Seasonalreconstructions of the Earth's G. Wefer,Seasonal variability of particleflux in theWeddell Sea surfaceat the Last glacialmaximum, Geol., Soc.Am. Chart Ser., andits relationto ice cover,Nature, 335, 426-428, 1988. MC-36, Geol. Soc. Am., Boulder, Colo., 1981. Francois,R., M.A. Altabet,and L.H. Burckle,Glacial to interglacial Coale,K.H., et al., A massivephytoplankton bloom induced by an changesin surfacenitrate utilizationin the Indian sectorof the ecosystem-scaleiron fertilizationexperiment in the equatorial SouthernOcean as recorded by sediment/•N, Pacific Ocean, Nature, 383, 495-501, 1996. Paleoceanography,7, 589-606, 1992. Comiso,J.C., and A.L. Gordon,Cosmonaut polynya in theSouthern Francois,R., M.A. Altabet,E.F. Yu, D.M. Sigman,M.P. Bacon,M. Ocean: Structure and variability, J. GeophysRes., 101, Frank, G. Bohrmann, G. Bareille, and L.D. Labeyrie, 18297-18313, 1996. Contribution of Southern Ocean surface-water stratification to Comiso,J.C., and A.L. Gordon,Interannual variability in summer low atmosphericCO2 concentrations during the last glacial seaice minimum,coastal polynyas and bottomwater formation period,Nature, 389, 929-935, 1997. in the WeddellSea, in Ocean,Ice, andAtmospere: Interactions Gibson,J.A.E., R.C., Garrick, H.R., Burton,and A.R. McTaggart, at theAntarctic Continental Margin, editedby S. Jacobsand R. Dimethylsulfideand the algaPhaeocystis pouchetii in Antarctic Weiss,pp. 293-315,AGU, WashingtonD.C., 1998. coastalwaters, Mar. Biol., 104, 339-346, 1990. Comiso,J.C., C.R. McClain,CoW. Sullivan, J.P. Ryan,and C.L. Gloersen, P., W.J. Campbell, D.J. Cavalieri, J.C. Comiso, C.L. Leonard,Coastal zone color scannerpigment concentrations in Parkinson, and H.J. Zwally, Arctic and Antarctic Sea Ice, the SouthernOcean and relationshipsto geophysicalsurface 1978-1987: Satellite Passive-Microwave Observations and features,J. Geophys.Res. 98, 2419-2451, 1993. Analysis.NASA, WashingtonDC, 1992. Cooke,D.W., and J.D. Hays, Estimatesof AntarcticOcean seasonal Gordon,A.L., Seasonalityof SouthernOcean sea ice, J. Geophys. sea-icecover during glacial intervals, In: AntarcticGeoscience, Res., 86, 4193-4197, 1981. edited by C. Craddock,pp. 1017-1025, Univ. of Wisc. Press, Gordon, R.M., Coale, K.H., and K.S. Johnson,Iron distributionsin Madison, 1982. the equatorialPacific: Implications for new production,Limnol. Crocker, K., M.E. Ondrusek,R.L. Petty, and R.C. Smith, Oceanogr., 42, 419-431, 1997. Dimethlysulfide,algal pigmentsand light in an Antarctic Gordon,A.L., andB.A. Huber,Southern Ocean winter mixed layer, Phaeocystissp. bloom, Mar. Biol., 124, 335-340, 1995. J. Geophys.Res., 95, 11655-11672, 1990. Crosta,X., J.J.Pichon, and L.H. Burckle,Application of modern analogtechnique to marineAntarctic diatoms: Reconstruction of Goyet, C., C. Beauverger,C. Brunet, and A. Poisson,Distribution maximum sea-ice extent at the Last glacial maximum, of carbon dioxide partial pressure in surface waters of the Paleoceanography,13, 284-297, 1998a. southwestIndian Ocean, Tellus Ser. B, 43B, 1-11, 1991. Crosta,X., J.J.Pichon, and L.H. Burckle,Reappraisal of Antarctic Hays, JD, A review of the late Quaternaryclimatic history of seasonalsea-ice at thelast glacial maximum, Geophys. Res. Lett., Antarctic seas, In: Antarctic Glacial History and World 25, 2703-2706, 1998b. Paleoenvironments.Edited by E.MoV.Z. Bakker, and A.A. Crowley, TJ, Paleoclimate modelling, In: Physically-Based Balkema,pp. 57-71, BrookfieldVt., 1978. Modelling and simulationof Climate and Climatic Change- Hays, J•D., J.A. Lozano, N.J. Shackleton, and G. Irving, Part II, editedby M.E. Schlesinger,p. 883-949,Klywer Acad., Reconstruction of the Atlantic and western Indian Ocean sectors Norwell, Mass., 1988. of the 18,000 B.P. Antarctic Ocean,in Investigationof Late Crowley, T.J., and C.L. Parkinson,Late Pleistocenevariations in Quaternarypaleoceanography and paleoclimatology, edited by Antarcticsea ice II: effectof interhemisphericdeep-ocean heat R.M. Cline, andJ.D. Hays,Geol. Soc.Am., Washington,D.C,, exchange,Clim. Dyn., 3, 93-103, 1988. 1976. De Angelis,M., J.P. Steffensen,M. Legrand,H. Clausen,and C. Hebling, E.W., V. Villafie, and O. Holm-Hansen,Effect of iron on Hammer,Primary aerosol (sea salt and soil dust)deposited in productivityand size distribution of Antarcticphytoplankton, Greenlandice duringthe last climaticcycle: Comparison with Limnol. Oceanogr.,36, 1879-1885, 1991. east Antarctic records,J. Geophys.Res., 102, 26681-26698, Hoppema, M., E. Fahrbach,M. Schr6der,A. Wisotzki, and H.J.W. 1997. de Baar, Winter-summer differences of carbon dioxide and de Baar,H.J.W., J.T.M., de Jong,D.C.E. Bakker,B.M. L6scher,C. oxygen in the Weddell Sea surface layer, Mar. Chem., 51, Veth, U. Bathmann,and V. Smetacek,Importance of iron for 177-192, 1995. MOORE ET AL.' SOUTHERN OCEAN CARBON SINK 473

Howard, W.R., and W.L. Prell, Late quaternarysurface circulation Legrand, M., C. Feniet-Saigne,E.S. Saltzman, C. Germain, N.I. 0f the Southern Indian Ocean and its relationshipto orbital Barkov, V.N. Petrov, Ice-core record of oceanic emissions of variations,Paleoceanography, 7, 79-117, 1992. dimethylsulphideduring the last climate cycle, Nature, 350, Howard,W.R., and W.L. Prell, Late QuaternaryCaCO 3 production 144-146, 1991. andpreservation in the SouthernOcean: Implications for oceanic L6scher, B.M., H.J.W. de Baar, J.T.M. de Jong, C. Veth, and F. and atmosphericcarbon cycling, Paleoceanography, 9, 453-482, Dehairs, The distributionof Fe in the Antarctic Circumpolar 1994. Current, Deep Sea Res Part H, 44, 143-187, 1997. Hutchins,D.A., and K.W. Bruland,Iron-limited growth and Si:N Lutjeharms, J.R.E., N.M. Walters, and B.R. Allanson, Oceanic uptake ratios in a coastal upwelling regime, Nature, 393, frontal systems and biological enhancement,In: Antarctic 561-564, 1998. Nutrient Cyclesand Food Webs,edited by W.R. Siegfried,P.R. Hutchins,D.A., G.R. DiTullio, Y. Zhang, and K.W. Bruland, An Condy, and R.M. Laws, pp. 11-21, Springer-Verlag,Berlin, iron limitation mosaic in the California upwelling regime, 1985. LiranoloOceanogr., 43, 1037-1054, 1998. Mackensen, A., H. Grobe, H.W. Hubberten, and G. Kuhn, Benthic Ikehara, M., K. Kawamura, N. Ohkouchi, K. Kimoto, M. foraminiferalassemblages and the 8• C-signalin the Murayama,T. Nakamura,T. Oba,T, andA. Taira, Alkenonesea Atlantic Sector of the SouthernOcean: Glacial-to-interglacial surfacetemperature in the SouthernOcean for the last two contrasts,In:Carbon Cycling in the Glacial Ocean: Constraints deglaciations,Geophys. Res. Lett., 24, 679-682, 1997. on the Ocean's Role in Global Change,edited by R. Zahn et al., Inoue, H.Y., and Y. Sugimura, Distribution and variations of Springer-Verlag,Berlin Heidelberg,p. 105-135, 1994. oceanic carbon dioxide in the western North Pacific, eastern Martin, J.H., Glacial-interglacialCO2 change: The Iron hypothesis. Indian, and Southern Ocean south of Australia, Tellus Ser. B, Paleoceanography,5, 1-13, 1990. 40B, 308-320, 1988. Martin, J.H., Iron as a limiting factor in oceanicproductivity, In: Ishii, M., H.Y. Inoue, H. Matsueda,and E. Tanoue,Close coupling Primary Productivity and Biogeochemical Cycles in the Sea, between seasonal biological production and dynamics of edited by P.G. Falkowski, and A.D. Woodhead, pp. 123-137, dissolvedinorganic carbon in the Indian Ocean sectorand the Plenum Press, New York, 1992. western Pacific sector of the Antarctic Ocean, Deep Sea Res. Martin, J.H., S.E. FitzWater, and R.M. Gordon, Iron in Antarctic Part I., 45, 1187-1209, 1998. waters. Nature, 345, 156-158, 1990a. JenningsJr, J.C., L.I. Gordon,and D.M. Nelson,Nutrient depletion Martin, J.H., S.E. Fitzwater, and R.M. Gordon, Iron deficiency indicateshigh primaryproductivity in the Weddell Sea,Nature, limits phytoplankton growth in Antarctic waters, Global 308, 51-54, 1984. Biogeochern.Cycles, 4, 5-12, 1990b. Joos,F., and U. Siegenthaler,Possible effects of iron fertilizationin Martinson, D.G., Evolution of the Southern Ocean winter mixed the SouthernOcean on atmosphericCO2 concentration,Global layer and sea ice' Open ocean deepwater formation and Biogeochem.Cycles, 5, 135-150, 1991. ventilation,J. Geophys.Res., 95, 11641-11654,1990. Jordon,R.W., and C.J. Pudsey,High-resolution diatom stratigraphy Matrai, P.A., M. Vernet, R. Hood, A. Jennings, E. Brody, S. of Quaternary sediments from the Scotia Sea, Mar. Saemundsd6ttir, Light-dependence of carbon and sulfur Micropaleontol., 19, 201-237, 1992. productionby polar clonesof the genusPhaeocystis, Mar. Biol., Jouzel, J., C. Lorius, J.R. Petit, C. Genthon, N.I. Barkov, V.M. 124, 157-167, 1995. Kotlyakov, and V.M. Petrov, Vostok ice core: a continuous McClain, C..R, M.L. Cleave, G.C. Feldman, W.W. Gregg, S.B. isotopetemperature record over the last climaticcycle (160,000 Hooker, and N. Kuring, Sciencequality SeaWiFS data for global years), Nature 329, 403-407, 1987. biosphereresearch, Sea Tech.,September, 10-16, 1998. Keir, R.S., On the late Pleistoceneoocean geochemistryand Metzl, N., C. Beauverger, C. Brunet, C. Goyet, and A. Poisson, circulation,Paleoceanography, 3, 413-445, 1988. Surface water carbon dioxide in the southwest Indian sector of Kellogg, T.B., Glacial-interglacialchanges in global deepwater the SouthernOcean: a highly variableCO2 source/sinkregion in circulation,Paleoceanography, 2, 259-271, 1987. summer, Mar. Chern., 35, 85-95, 1991. Kirst, G.O., C. Thiel, H. Wolff, J. Nothnagel,M. Wanzek, and R. Metzl, N., A. Poisson, F. Louanchi, C. Brunet, B. Schauer, and B. Ulmke, Dimethylsulfoniopropionate(DMSP) in ice-algaeand its Bres,Spario-temporal distributions of air-sea fluxes of CO2in the possiblebiological role, Mar. Chem.,35, 381-388, 1991. Indian and Antarctic Oceans, Tellus Ser. B, 47B, 56-69, 1995. Klinck, J.M., and D.A. Smith, DA, Effect of wind changesduring Minas, H.J., and M. Minas Net communityproduction in •High the last glacial maximum on the circulation in the Southern Nutrient-LowChlorophyll" waters of the tropicaland Antarctic Ocean, Paleoceanography,8, 427-433, 1993. Oceans'grazing vs. ironhypothesis, Oceanol. Acta, 15, 145-162, Knox, F., and M.B. McElroy, Changes in atmosphericCO2: 1992. Influenceof the marine biota at high latitude,J. Geophys.Res., Moore, J.K., M.R. Abbott,J.G. Richman,Variability in the location 89, 4629-4637, 1984. of the AntarcticPolar Front (90ø-20øW)from satellitesea surface Kumar, N., R. Gwiazda, R.F. Anderson, and P.N. Froelich, temperaturedata, J. Geophys.Res. 102, 27825-27833,1997. 231pa]23øThratios in sedimentsas a proxyfor pastchanges in Moore, J.K., M.R. Abbott, and J.G. Richman, Location and SouthernOcean productivity. Nature, 362, 45-48, 1993. dynamicsof the AntarcticPolar Front from satellitesea surface Kumar, N., R.F. Anderson,R.A. Mortlock, P.N. Froelich, P. Kubik, temperaturedata, J. Geophys.Res., 104, 3059-3073,1999a. B. Dittrich-Hannen, and Mo Suter, Increased biological Moore, J.K., M.R. Abbott,J.G. Richman,W.O. Smith,T.J. Cowles, productivityand exportproduction in the glacialSouthern Ocean. K.H. Coale, W.D. Gardner,and R.T. Barber,SeaWiFS satellite Nature, 378, 675-680, 1995. oceancolor data from the SouthernOcean, Geophys. Res. Lett., Labeyrie, L, et al., Hydrographicchanges of the SouthernOcean 26, 1465-1468, 1999b. (southeast Indian sector) over the last 230 kyr, Morley,J.J., Variations in high-latitudeoceanographic fronts in the Paleoceanography,11, 57-76, 1996. SouthernIndian Ocean:An estimationbased on faunal changes, Landry, M.R., et al., Iron and grazing constraintson primary Paleoceanography,4, 547-554, 1989. productionin the centralequatorial Pacific: An EqPacsynthesis, Mortlock, R.A., C.D. Charles, P.N. Froelich, M.A. Zibello, J. Limnol. Oceanogr.,42, 405-418, 1997. Laubscher,R.K., R. Perisinotto,and C.D. McQuaid, Phytoplankton Saltzman, J.D. Hays, and L.H. Burckle, Evidence for lower productivityin the Antarctic Ocean during the last glaciation, productionand biomass at frontalzones in the Atlanticsector of Nature, 351, 220-223, 1991. the Southern Ocean, Polar Biol., 13, 471-481, 1993. Legrand,M.R., Ice-corerecords of atmosphericsulphur, Phil. Trans. Murphy, P.P., R.A. Feely, R.H. Gammon,K.C. Kelly, and L.S. Roy. Soc.Lon. B, 352, 241-250, 1997. Waterman,Autumn air-seadisequilibrium of CO2 in the South Legrand,M.R., R.J. Delmas,and R.J. Charlson,Climate forcing Pacific Ocean, Mar. Chern., 35, 77-84, 1991. implicationsfrom Vostok ice-coresulphate data, Nature, 334, National Snow and Ice Data Center, Passive Microwave Derived 418-420, 1988. Monthly Polar Sea Ice ConcentrationTime Series,Digital data 474 MOORE ET AL.: SOUTHERN OCEAN CARBON SINK

available from [email protected],Boulder, Colorado, Sarmiento,J.L., andJ.R. Toggweiler,A new modelfor the role of NSIDC Distributed Active Archive Center, Univ. of Colo. at the oceans in determiningatmospheric Pco2, Nature, 308, Boulder, 1998. 621-624, 1984. Nelson,C.S., P.J. Cooke C.H. Hendy, and A.M. Cuthbertson,AM, Sarmiento,J.L., and J.C. Orr, Three-dimensionalsimulations of the Oceanographicand climatic changes over the past 160,000 years impactof SouthernOcean nutrient depletion on atmosphericCO2 at deep sea drilling project site 594 off southeasternNew andocean chemistry, Limnol. Oceanogr., 36, 1928-1950,1991. Zealand, southwest Pacific Ocean, Paleoceanography,8, Sarmiento,J.L., and C. Le Qutrt, Oceaniccarbon dioxide uptake 435-458, 1993. in a model of century-scaleglobal warming, Science,274, Nelson, D.M., D.J. DeMaster, R.B. Dunbar, and W.O. Smith Jr, 1346-1350, 1996. Cycling of organiccarbon and biogenicsilica in the Southern Scbarek,R., M.A. Van Leeuwe,and H.J.W. De Baar,Responses of Ocean:Estimates of water-columnand sedimentary fluxes on the SouthernOcean phytoplankton to the additionof tracemetals, RossSea continentalshelf, J. Geophys.Res., 101, 18519-18532, Deep-SeaRes. Part II, 44, 209-228, 1997. 1996. Sedwick,P.N., P.R. Edwards,D.J. Mackey,F.B. Griffiths,and J.S. Nelson,D.M., andW.O. SmithJr., Phytoplanktonbloom dynamics Parslow,Iron andmanganese in surfacewaters of the Australian of the westernRoss Sea ice edge - II. Mesoscalecycling of subantarcticregion, Deep-Sea Res. I., 44, 1239-1253,1997. nitrogenand silicon,Deep Sea Res.Part I., 33, 1389-1412, 1986. Sedwick,P.N., and G.R. DiTullio, Regulationof algal bloomsin Neori, A., and O. Holm-Hansen,Effect of temperatureon rate of AntarcticShelf waters by the releaseof ironfrom melting sea ice, photosynthesisin Antarcticphytoplankton, Polar Biol., 1, 33-38, Geophys.Res. Lett., 24, 2515-2518, 1997. 1982. Sedwick,P..N, G.R. DiTullio, and D. Mackey, Dissolvediron and Nicol, S., I. Allison, The Frozen skin of the SouthernOcean, Am. manganesein surfacewaters of the RossSea, australsummer Sci., 85, 426-439, 1997. 1995-1996,Ant. J. of the U.S., 31, 128-130, 1996. Nolting, R.F., H.J.W. de Baar, A.J. Van Bennekom,and A. Masson, Shemesh,A., S.A. Macko, C.D. Charles,and G.H. Rau, Isotopic Cadmium,copper and iron in the ScotiaSea, Weddell Sea, and evidencefor reducedproductivity in the GlacialSouthern Ocean, Weddell/Scotia Confluence (Antarctica), Mar. Chem., 35, Science, 262, 407-410, 1993. 219-243, 1991. Sherr, E.B., and B.F. Sherr, 1994, Bacterivoryand herbivory:Key Niirnberg,C..C, G. Bohrmann,Mo Schliiter,and M. Frank,Barium rolesof phagotropicprotists in pelagicfood webs,Microb. Ecol., accumulation in the Atlantic sector of the Southern Ocean: 28, 223-235, 1994. Results from 190,000-year records, Paleoceanography,12, Shimmield,G., S. Derrick, A. Mackensen,H. Grobe,and C. Pudsey, 594-603, 1997. The history of barium, biogenic silica and organic carbon Petit, J.R., M. Briat, and A. Royer, Ice age aerosolcontent from accumulation in the Weddell Sea and Antarctic Ocean over the East Antarcticice core samplesand past wind strength,Nature, last 150,000 years, in Carbon Cycling in the Glacial Ocean: 293, 391-394, 1981. Constraintson the Ocean'sRole in Global Change,edited by R. Petit,J.R., L. Mounier,J. Jouzel,Y.S. Korotkevich,V.I. Kotlyakov, Zahn et al., pp. 555-574, Springer-Verlag,New York, 1994. and C. Lorius, Palaeoclimatologicaland chronological Siegenthaler,U., and T. Wenk, Rapidatmospheric CO2 variations implicationsof the Vostok core dustrecord, Nature, 343, 56-58, and ocean circulation, Nature, 308, 624-626, 1984. 1990. Smetacek, V., R. Scharek, L.I. Gordon, H. Eicken, E. Fahrbach, G. Pichon,J.J., M. Labeyrie,G. Bareille, M. Labracherie,J. Duprat, Rohardt,and S. Moore, Early springphytoplankton blooms in ice and J. Jouzel, Surfacewater temperaturechanges in the high plateletlayers of the southernWeddell Sea,Antarctica, Deep Sea latitudes of the southern hemisphere over the last Res. Part I, 39, 153-168, 1992. glacial-interglacialcycle, Paleoceanography, 7, 289-318, 1992. Smith, S.D., R.D. Muench, and C.H. Pease,Polynyas and leads:An PoissonA., N. Metzl, X. Danet, F. Louanchi,C. Brunet,B. Schauer, overview of physicalprocesses and environment,Jo Geophys. B. Br6s, and D. Ruiz-Pino,Air-sea CO2 fluxesin the Southern Res., 95, 9461-9479, 1990. Ocean Between 25øE and 85øE, in The Polar Oceans and Their Role in Shapingthe GlobalEnvironment, Geophys. Monogr. 85, Smith Jr, W.O., Nutrient distributionsand new productionin polar edited by O.M. Johannessenet al., pp. 273-284, AGU regions:Parallels and contrastsbetween the Arctic andAntarctic, WashingtonD.C., 1994. Mar. Chem., 35, 245-257, 1991. Prell, W.L., W.H. Hutson, D.F. Williams, A.W.H. B6, K. SmithJr., W.O, and L.I. Gordon,Hyperproductivity of the RossSea Geitzenauer,and B. Molfino, Surface circulationof the Indian (Antarctica)Polynya during australspring, Geophys. Res. Lett., Oceanduring the last glacialmaximum, Approximately 18,000 24, 233-236, 1997. yr B.P., Quat. Res., 14, 309-336, 1980. Smith Jr., W.O., and D.M. Nelson, Phytoplanktonbloom produced Redfield, A.C., B.H. Ketchum,and F.A. Richards,The Influenceof by a recedingice edge in the RossSea: Spatialcoherence with organismsof the compositionof seawater,in The Sea,Vol. 2, the densityfield, Science,227, 163-166, 1985. editedby M.N. Hill, pp. 26-77, Wiley Interscience,New York, Smith Jr., W.O., and D.M. Nelson, Importance of ice edge 1963. phytoplanktonproduction in the SouthernOcean, BioScience, 36, Robertson, J.E., and A.J. Watson, A summer-time sink for 251-257, 1986. atmosphericcarbon dioxide in the SouthernOcean between 88 Smith Jr., W.O., and D.M. Nelson,Phytoplankton growth and new øW and 80 øE, Deep Sea Res. Part II, 42, 1081-1091,1995. productionin the WeddellSea marginalice zonein the austral Rosenthal,Y., E.A. Boyle, L. Labeyrie, and D. Oppo, Glacial springand autumn,Limnol. Oceanogr., 35, 809-821, 1990. enrichmentsof authigenicCd and U in Subantarcticsediments: Smith Jr., W.O., L.A. Codispoti,D.M. Nelson, To Manley, E.J. A climatic control on the elements' oceanic budget?, Buskey,H.J. Niebauer,and G.F. Cota, Importanceof Phaeocystis Paleoceanography,10, 395-413, 1995. blooms in the high-latitudeocean carbon cycle, Nature, 352, Rosenthal,Y., E.A. Boyle, and L. Labeyrie,Last glacialmaximum 514-516, 1991. paleochemistryand deepwater circulation in theSouthern Ocean: Smith Jr., W.O., and E. Sakshaug,Polar phytoplankton,in Polar Evidence from foraminiferal cadmium, Paleoceanography,12, Oceanography,Part B: Chemistry,Biology, and Geology,edited 787-796, 1997. by W.O. Smith, AcademicPress, San Diego, Calif., 1990 Rubin, S., J. Goddard,D. Chipman,and T. Takahashi,Carbon dioxidepartial pressure in surfacewaters in the Pacificsector of Smith Jr., W.O., D.M. Nelson, G.R. DiTullio, and A.R. Leventer, the southernoceans during austral summers1992 and 1994, Temporal and spatialpatterns in the Ross Sea: Phytoplankton Antarct. J. U.S., 31, 113-115, 1994. biomass,elemental composition, productivity and growth rates, J. Sabine,C.L., and R.M. Key, Controlson fco2 in the SouthPacific, Geophys.Res., 101, 18455-18465, 1996. Mar. Chem., 60, 95-110, 1998. SmithJr., W.O., D.M. Nelson,and S. Mathot,Phytoplankton growth Sancetta,C., Primaryproduction in the gla•cialNorth Atlanticand rates in the Ross Sea, Antarctica determinedby independent North Pacific oceans, Nature, 360, 249-251, 1992. methods:temporal variations, in press,J. Plank. Res., 1999. MOORE ET AL.: SOUTHERN OCEAN CARBON SINK 475

Sullivan, C.W., K.R. Arrigo, C.R. McClain, J.C. Comiso, and J. Weaver, P.P.E., L. Carter,and H.L. Neil, Responseof surfacewater Firestone.Distributions of phytoplanktonblooms in the Southern massesand circulationto late Quaternaryclimate change east of Ocean. Science, 262, 1832-1837, 1993. New Zealand, Paleoceanography,13, 70-83, 1998. Sunda, W.G., and S.A. Huntsman, Interrelated influence of iron, Wefer, G., and G. Fischer,Annual primaryproduction and export light, and cell size on marinephytoplankton growth, Nature, 390, flux in the SouthernOcean from sedimenttrap data,Mar. Chem., 389-392, 1997. 35, 597-613, 1991. Takahashi,T., J. Olafsson,J.G. Goddard,D.W. Chipman,and S.C. Weiss, R.F., Winter Weddell Sea Project 1986: Trace gas studies Sutherland,Seasonal variation of CO2 and nutrientsin the duringlegs ANT V/2 and V/3 of Polarstern,Antarct. J. of the high-latitude surface oceans: A comparativestudy, Global U.S., 22, 99-100, 1987. Biogeochem.Cyc., 7, 843-878, 1993. Westedund,S., and P. Ohman,Iron in the water columnof the Takahashi,T., R.A. Feely, R.F. Weiss, RH. Wanninkhof, D.W. Weddell Sea, Mar. Chem., 35, 199-217, 1991. Chipman, S.C. Sutherland,and T.T. Takahashi,Global air-sea Wilson, D.L., W.O. Smith Jr., and D.M. Nelson, Phytoplankton flux of CO2:An estimatebased on measurementsof sea-airpCO2 bloom dynamicsof the westernRoss Sea ice edgeI. Primary difference, Proc. Natl. Acad. Sci., 94, 8292-8299, 1997. productivityand species-specific production, Deep Sea Res., 33, Takeda,S., Influenceof iron availabilityon nutrientconsumption 1375-1387, 1986. ratio of diatoms in oceanicwaters, Nature, 393, 774-777, 1998. Tr6guer, P., and G. Jacques, Dynamics of nutrients and phytoplankton,and fluxesof carbon,nitrogen, and siliconin the Antarctic Ocean, Polar Biol., 12, 149-162, 1992. Turner, S.M., P.D. Nightingale,W. Broadgate,and P.S. Liss, The distribution of dimethyl sulphide and Mark R. Abbott, JamesG. Richman, and David M. Nelson, dimethylsulphoniopropionatein Antarctic waters and sea ice, Collegeof Oceanicand Atmospheric Sciences, Oregon State Deep Sea Res. Part II, 42, 1059-1080, 1995. University,104 OceanAdmin Building,Corvallis, OR, 97331- Van Leeuwe,M..A, R. Scharek,H.J.W. De Baar,J.T.M. De Jong, 5503 andL. Goeyens,L, Iron enrichmentexperiments in the Southern Ocean: physiological responsesof plankton communities, J. K. Moore, NationalCenter for AtmosphericResearch, Deep-Sea Res. II, 44, 189-208, 1997. AdvancedStudies Program, P.O. Box 3000, Boulder,CO, 80307. Wamser, C., and D.G. Martinson,Drag coefficientsfor winter (ikm oore @ c ed. u car. edu • Antarcticpack ice, J. Geophys.Res., 98, 12431-12437, 1993. Watkins,N.D., M.T. Ledbetter,and T.C. Huang,Antarctic glacial historyusing spatial and temporal variations of ice-rafteddebris in abyssal sedimentsof the Southern Ocean, in Antarctic Geoscience,edited by C. Craddock,pp. 1013-1016,Univ. of (ReceivedFebruary 2, 1999; revisedJuly 1, 1999; Wisc. Press, Madison, Wisc., 1982. acceptedJuly 22, 1999.)