Holocene paleoceanography and glacial history of the West ...

ARTICLE IN PRESSQuaternary Science Reviews 23 (2004) 2075–2088Holocene paleoceanography andglacial history of the WestSpitsbergen area, Euro-Arctic marginMorten Hald a, , Hanne Ebbesen a , Matthias Forwick a , FredGodtliebsen b ,Liza Khomenko c , Sergei Korsun c , Lena RingstadOlsen b , Tore O. Vorren aa Department of Geology, University of Tromsoe, Tromsoe, Norwayb Department of Mathematics and Statistics, University of Tromsoe, Norwayc Department of Biology, St. Petersburg University, RussiaAbstractTwo sediment cores from the West Spitsbergen area, Euro-Arctic margin, MD99-2304 and MD99-2305, have been investigatedfor paleoceanographic proxies, including benthic and planktonic foraminifera, benthic foraminiferal stable isotopes and ice rafteddebris. Core MD99-2304 is located on the upper continental margin, reflecting variations in the influx of Atlantic Water in the WestSpitsbergen Current. Core MD99-2305 is located in Van Mijenfjord, picturing variations in tidewater glacier activity as well as fjordoceancirculation changes. Surface water warmer than today, was present on the margin as soon as the Van Mijenfjord wasdeglaciated by 11,200 cal. years BP. Relatively warm water invaded the fjord bottom almost immediately after the deglaciation. Arelatively warm early Holocene was followedby an abrupt cooling at 8800 cal. years BP on the continental margin. Another coolingin the fjord record, 8000–4000 cal. years BP, is documented by an increase in ice rafted debris and an increase in benthicforaminiferal d 18 O: The IRD-record indicates that central Spitsbergen never was completely deglaciated during the Holocene.Relatively cool andstable conditions similar to the present were establishedabout 4000 cal. years BP.r 2004 Elsevier Ltd. All rights reserved.1. IntroductionThe West Spitsbergen continental margin is aclimatically sensitive area. The main reason for this isthe present oceanographic regime characterizedbynorth-flowing warm Atlantic Water in the WestSpitsbergen Current along the continental margin withsharp gradients to cool, partly sea-ice covered PolarWater (Fig. 1). Small geographical displacements in theocean fronts in this area may impose large climaticchanges affecting the society, life habitats andearthprocesses in the area. Climate models predict that theglobal warming will have its largest impact in the highnorthern latitudes (e.g. Hadley Centre (2004)). However,the instrumental recordof climate variability is too shortandspatially incomplete to reveal the full range of Corresponding author. Tel.: +47-776-44412; fax: +47-776-45600.E-mail address: morten.hald@ig.uit.no (M. Hald).seasonal to millennial-scale climate variability, or toprovide empirical examples of how the climate systemresponds to large changes in climate forcing. Longertime series may be provided with well dated and highresolution proxy records.We have investigatedtwo proxy records from theWest Spitsbergen area, one from the open ocean uppercontinental margin andone from a silledfjordsetting,Van Mijenfjord. The core from the continental margin islocatedunder the axis of Arctic Water. This core shouldreflect changes andvariability in heat flux to the highnorthern latitudes, representing the northern limb of theNorth Atlantic conveyor circulation. The fjordrecordon the other handrepresents more of a continentalclimate signal, linkedto processes such as iceberg raftingfrom tidewater glaciers and exchange between fjordwater andopen oceanic water. Our knowledge of theHolocene climatic recordof the Svalbardarea is sparsedue to lack of high resolution and well-dated proxy0277-3791/$ - see front matter r 2004 Elsevier Ltd. All rights reserved.doi:10.1016/j.quascirev.2004.08.006

2076ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–20888° 12° 16° 20° 24° 28°2000100081°ARCTICOCEANKara SeaSiberiaYeniseyriver50020010080°80°SvalbardWSCBearIslandBarentsSeaObriver379°60°Norwegian Sea70°NCScandinaviaPOLAR WATERARCTIC WATERCOASTAL WATERATLANTIC WATERSEA-ICEBORDER, APRILWESTWSC SPITSBERGENCURRENTNCNORWEGIANCURRENTSURFACECURRENTSMD99-2304200050010002IsfjordMD99-23051100200Van Mijenfjorden78°77°(A)(B)Fig. 1. Location maps: (A) index map showing the surface water masses andsurface currents in the Norwegian Sea andadjoining seas; (B) close upof Svalbard and surrounding continental margin. Location of the two investigated sediment cores MD99-2304 and MD99-2305 are shown.1=Akseløya; 2=Linnèvannet and3=Billefjorden.records. The purpose of the present paper is to (1)present well-dated paleoclimate data for the WestSpitsbergen area during the Holocene and (2) investigatecentury to decadal scale climate changes and variability.2. Physical settingSpitsbergen is the largest islandof the Svalbardarchipelago, situatedbetween 761 and801N, andbordered by the Arctic Ocean to the north, the BarentsSea to the south andeast, andthe Norwegian–GreenlandSeato the west (Fig. 1). The archipelago isdominated by Phanerozoic sedimentary rocks (SteelandWorsley, 1984).The continental margin west of Spitsbergen ischaracterizedby a relatively narrow shelf with a typicalglacial morphology representedby shallow banksbetween glacial troughs, the latter foundin thecontinuation of the east-west trending fjords. Thecontinental slope has a relatively steep gradient of4–51. Quaternary sediments rest on Tertiary sediments,andhave accumulatedin well-definedsubmarinetroughs on the shelf andin submarine fans on the slope(Andersen et al., 1996).Spitsbergen is cut by several fjords. Van Mijenfjord(Fig. 1) is the secondlargest of these, 50 km long andca10 km broad. It has restricted oceanic communicationdue to the island Akseløya in the outer part. The fjord isdivided into three basins. Maximum depth of the outerbasin is 112 m as compared to 74 m in the middle basinandca 30 m in the inner basin. Most of the fjordislocated within Paleogene sedimentary bedrock ofvarious types such as sandstones, siltstones, shales,coals andcoals pebbles (Steel andWorsley, 1984).Early Cretaceous sedimentary rocks are found in theinnermost part (Salvigsen andWinsnes, 1989). The innerpart of the fjord is dominated by sediments producedduring a surge AD 1300 (Punning et al., 1976; Rowan etal., 1982; Haldet al., 2001). The outer basin of the fjordis characterizedby a maximum of 25 milli-seconds (ms)of acoustically transparent sediments representing the

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 2077last deglaciation and the Holocene (Dahlgren, 1998).The sea-floor surface sediments are dominantly silty clay(HaldandKorsun, 1997).The catchments area of Van Mijenfjordis approximately2.8 10 3 km 2 , and50% of this area is coveredbyglaciers (Hagen et al., 1993). A tidewater glacier(Fridtjofbreen) is located in the northern outermostparts of Van Mijenfjord. The fjord head is occupied bythe large tidewater glacier system of Paulabreen (PGS),of which two glaciers (the Paulabreen andBakininbreen)are the only glaciers that today calve into VanMijenfjord. Several tributary glaciers feed the mainglaciers of the PGS. All the glaciers in the area haveexperienceda significant loss of volume during the last100 years (Hagen et al., 1993). The climate of the VanMijenfjordis coldanddry. Annual temperature of thewest coast of Spitsbergen is 6 1C andmean annualprecipitation varies between 180 and440 mm (Hanssen-Bauer et al., 1990). The equilibrium line altitudes of theglaciers vary from 200 m above sea level along the coastto 4600 m in the central-eastern parts of Spitsbergen(Hagen et al., 1993).2.1. OceanographyThe Spitsbergen margin andfjords are influencedbyrelatively warm Atlantic Water transportedfrom southwith the West Spitsbergen Current andcoldPolar Waterfrom north (Fig. 1). The sea-ice conditions varyseasonally (Dowdeswell and Dowdeswell, 1989). Acover of land-fast ice forms in the major fjord systemsandin other shelteredcoastal areas by about lateNovember andis usually retaineduntil late May or June(Wadhams, 1981). Arctic pack ice is usually presentaroundnorthern Spitsbergen throughout the year, withthe lowest flow densities in August and September(Vinje, 1985). Between November andApril the pack icemay surroundthe entire Svalbardarchipelago, withminimum densities on the west coast of Spitsbergen. Theasymmetric distribution of polar pack ice results fromoceanic circulation in which polar waters transport packice southwardon the eastern side of SvalbardandtheWest Spitsbergen Current brings warm Atlantic Waternorthwardto the western coast of Svalbard(Wadhams,1981; Vinje, 1985) (Fig. 1).The mouth of Van Mijenfjordis almost closedfromthe open ocean by the islandAkseløya andthus allowsonly for a very modest inflow of Atlantic Water. Thewater column is dominated by cold local water, withbottom water temperatures of around 1 1C andaseasonally warmedsurface layer (cf. Haldet al., 2001).Gulliksen et al. (1985) reporteda minimum July bottomwatertemperature of 1.5 1C. Bottom water salinity isaround34 PSU in the two outer basins anda pycnoclineis observedin the upper 10–15 m. Tidewater rangesare o1m.3. Material and methodsWe investigatedtwo sediment cores, core MD99-2304is locatedon the upper West Spitsbergen continentalmargin 771 37. 26 0 N and091 56. 90 0 E, at 1315 m waterdepth, and core MD99-2305 located in Van Mijenfjord771 46.87 0 N and151 17.81 0 E at 110 m water depth. Thesampling sites were determined by several high resolutionseismic surveys during the 1990s using 3.5 kHzpenetration echo sounder and Sparker. Both cores weresampledduring the IMAGES-1999 cruise with ‘‘R/VMarion Dufresne’’ using a modified piston coringsystem, a ‘‘Calypso-corer’’, especially designed tosample long cores. Multi sensor track (MST) loggingwas appliedto the sealedcore MD99-2305. The sealedcores, with an inner diameter of 10 cm, were opened bysplitting them longitudinally in two equal parts. Onehalf was studied with regard to various geotechnical andsedimentological analyses including MST logging (onlyMD99-2304), X-radiography, and colour determinationby using the Munsell Colour Charts. The whole lengthof MD99-2305, 18 m, was analysed. MD99-2304 is 23 mlong, but only the upper 1.5 m were usedin the presentstudy. In core MD99-2305 ice rafted debris (IRD)41 mm was countedfrom X-radiographs.Preparation of the foraminiferal samples mainlyfollowedthe methods of Feyling-Hanssen (1958) andMeldgaard and Knudsen (1979). Between 300 and400individuals of benthic and/or planktonic foraminiferawere identified from the 4100 mm fraction using abinocular microscope. The planktonic foraminiferalcensus data were used to reconstruct sea surface summertemperatures (SST) in core MD99-2304. Three differentreconstructions were performed. One reconstructionrepresents quantitative estimations of SST basedonthe Modern Analogue Technique in the data analysisprogram C2 version 1.3 (Juggins, 2002). The moderndata set used in the estimations is from ATL 947core-top database of modern planktonic foraminiferain the 150 mm fraction (Pflaumann et al., 2003).The general deviation of SST is about 0.3–1.3 1C.The two other SST reconstructions were basedlinear regression using smaller modern data setsfrom the Nordic Seas of respectively 125 mm fraction(Johannessen et al., 1994) and100 mm fractions (Burhol,1994).d 18 O measurements were carriedout at the NorwegianGMS laboratory at the University of Bergen, usingFiningan MAT 251 mass spectrometer. The reproducibilityat this laboratory is 0.07% for d 18 O: The sampleswere prepared following the procedures described byShackleton andOpdyke (1973), Shackleton et al. (1983)and Duplessy (1978). In MD99-2305, d 18 O was analysedon the benthic foraminifer Cassidulina reniforme, aninfaunal species that appears to precipitate its shell indisequilibrium with an offset of 0.13% d 18 O(Austin and

2078ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088Kroon, 1996). The plotted d 18 O are correctedfor theglobal ice volume effect (Fairbanks, 1989).Accelerator Mass Spectrometry (AMS) radiocarbondates (Table 1) were performedon bivalve tests (MD99-2305) andforaminifera (MD99-2304) in Trondheim,Norway andUppsala, Sweden (TUa), Arizona, US(AA) andKiel, Germany (KIA). For the TUa-dates, thetargets were preparedat Radiocarbon Laboratory inTrondheim, Norway and were measured at the SvedbergLaboratory in Uppsala, Sweden.In order to check the statistical reliability and trendsof proxy data variations and to identify a lower limit ofreliable time resolution we appliedthe SiZer methodologydeveloped by Chaudhuri and Marron (1999). Thismethodwas appliedonly on the most high resolutiondata sets in MD99-2305 (i.e. IRD, foraminiferal flux and% Elphidium. excavatum f. clavatum) (Fig. 3). The SiZermethodology has not been applied on the lowerresolution data sets in MD99-2304 and the d 18 O andd 13 C data in MD99-2305. For these proxy sets we willonly discuss millennial scale changes. The SiZer methodis explainedin the following.A realistic statistical model for this situation is given byy i ¼ mðx i Þþ i ; 1 ¼ 1; ...; n; (1)where m(x) is the underlying smooth target curve, hererepresenting the ‘‘mean behaviour’’ of the proxy data attime x. The i are independent Gaussian variables withmean 0 andvariance s 2 i ; and y i represents the observedproxy value in data point number i.In smoothing approaches, one frequently uses a locallinear kernel estimator; see Fan (1991), to produce agoodestimate of m from the observations y 1 ,y,y n .Tobe specific, this means that, at time x, ^m h ðxÞ equals the fit^a 0 where ^a 0 ; ^a 1 minimizesX ni¼1ðy i a 0 a 1 ðx i xÞÞ 2 K h ðx i xÞ: (2)Table 1Radiocarbon ages for MD99-2304 and MD992305, showing the core location, depth in core (cm), reference to the dating laboratory, Radiocarbondating laboratory of Trondheim and the Tandem Accelerator Laboratory in Uppsala (Tua); Kiel Germany (KIA) and Arizona, US (AA)Core Core depth, Lab ref Material 14C age 1-sigma Cal.age 2-sigma RangesMD99-2305 8–9 Tua-4168 Nucula tenuis 345 750 410 468 294MD99-2305 100–101 TUa-3249 N. tenuis 1115 755 1056 1176 944MD99-2305 156–157 Tua-4169 Acanthocardia echinata 1300 745 1260 1325 1167MD99-2305 212–213 Tua-4170 Thyasira sp. 1930 745 1937 2050 1838MD99-2305 408–409 TUa-4171 Yoildia sp. 3940 7105 4443 4788 4209MD99-2305 540–541 TUa-3250 N. tenuis 5400 755 6208 6316 6316MD99-2305 625–626B Tua-4173 Thyasira equalis 6475 7105 7403 7563 7211MD99-2305 625–626A Tua-4172 Nuculana minuta 6610 765 7497 7604 7404MD99-2305 786–787 Tua-4174 Yoldiella lenticula 7780 755 8637 8866 8519MD99-2305 853–854 Tua-4175 Y. lenticula 8040 765 8929 9059 8784MD99-2305 926–927 Tua-4176 Y. lenticula 8430 765 9189 9648 9042MD99-2305 1106–1107 Tua-4177 N. tenuis 8850 755 9839 10,278 977MD99-2305 1132–1135 Tua-3144 Yoldia hyperborea 8670 765 9788 9950 9417MD99-2305 1135–1136 TUa-3251 N. tenuis 8910 765 9929 10,292 9790MD99-2305 1389–1390 TUa-3252 Y. hyperborea 9350 765 10,497 10,808 10,248MD99-2305 1390–1391 TUa-3253 N. tenuis 9375 770 10,564 10,816 10,274MD99-2305 1479–1481 TUa-3254 Y. hyperborea 9470 785 10,617 10,853 10,312MD99-2305 1480–1481 TUa-3255 N. tenuis 9535 780 10,801 10,864 10,333MD99-2305 1535–1536 TUa-3256 N. tenuis 9615 775 10,990 10,880 10,354MD99-2305 1569–1570 TUa-3257 N. tenuis 9765 775 11,118 11,348 10,791MD99-2305 1580 TUa-4319 N. tenuis 9835 745 11,141 11,357 10,834MD99-2305 1595–1596 TUa-4318 N. tenuis 9920 745 11,171 11,639 11,074MD99-2305 1763–1767 TUa-3258 2 bivalve fragments 45,285 +31402250MD99-2304 2,5 Tua-4421 Plan+bent. forams 580 730 560 627 494MD99-2304 28,5 TUa-3911 Planktonic forams 7855 755 8730 8905 8555MD99-2304 38,5 TUa-3912 Planktonic forams 8105* 760 8942* 9080 8804MD99-2304 56,5 TUa-3913 Planktonic forams 8010 765 8855 9023 8688MD99-2304 80 AA 36609 Shell fragments 8965 785 10,044 10,303 9784MD99-2304 130 KIA9346 Shell fragments 9670 755 11,404 11,725 11,082MD99-2304 156,8 KIA9526 Shell fragments 10,030 750 10,607 10,866 10,349MD99-2304 186 AA 36610 Shell fragments 12,170 7180 13,945 14,386 13,503MD99-2304 215 KIA9863 Benthic forams 12,660 770 14,835 15,473 14,197Asterisk denotes age excluded from the age model. The 14 C dates were calibrated (cal. ages) using the INTCAL-98 calibration data set (Stuiver et al.,1998).

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 2079In Eq. (2), K h ( )=(1/h) K( /h), where K is a kernelfunction symmetric aroundzero. Typically, we haveKðxÞ ¼1 x 2 pffiffiffiffiffiexp : (3)2p 2The degree of smoothness in the estimate ^m h iscontrolledthrough the bandwidth h. Note that whereastraditional smoothing methods seek an ‘‘optimal’’ valueof h, SiZer exploits ‘‘all’’ values of h to findsignificantfeatures. This is crucial in climatological applicationssince important features typically change as a functionof scale. Hence, the whole scale-space is needed in orderto findall significant features in a data set. By the SiZerapproach one also avoids the bias difficulty that appearsin traditional smoothing. This is caused by the fact that^m h is an unbiasedestimator of the true curve at theresolution corresponding to h.MD99-2304CoredepthStratigraphy(C yr)Datings14(cm)1020±308295±558545±608450±65100 8965±859670±5510030±5020012170±18012660±70300MD99-2305Coredepth(cm) Stratigraphy100200300400500600700Datings14(C yr)1115±551300±453940±1055400±556475±1056610±654. ChronologyThe chronology of MD99-2305 is basedon 21 AMSradiocarbon dates performed on mollusc bivalves (Table1). Most of these bivalves were identified to species level.All the dates were corrected for a reservoir effect of 464years (including a DR value of 64 +/ 35 years)(MangerudandGulliksen, 1975) andshow increasingages versus depth in core (Fig. 2). An age model wasproduced by converting the dates to calendar yearsfollowing the calibration model of Stuiver et al. (1998)(Fig. 4F). Sedimentation rates (mm/cal. years) werecalculatedby applying a fourth order polynomial fitthrough the dates. The age model implies that sedimentationrates vary from 66 cm/1000 cal. year to 4500 cm/1000 cal. years. The highest rates occurredduring theearly Holocene, the lowest during mid-Holocene andthen a slight increase the last 1700 years. Thesedimentation rates in the core allow a high stratigraphicresolution in the various proxies. Averageresolution for the IRD counts is 6.8 years; foraminiferalcounts 42 years andstable isotope analyses 175 years.The age model for MD99-2304 is based on 8 out of 9AMS dates performed on mixed benthic/planktonicforaminifera; planktonic foraminifera only, or unidentifiedfragmentsof bivalve shells (Fig. 5H,F, Table 1).One date, TUa 3912 (8105 +/ 60 yrs BP) appears toooldandwas excludedfrom the age model. This may bedue to reworking. However, based on the low relief ofthe core location (Labeyrie et al., 2003), the finesediment texture, and lack of structures indicatingsediment disturbances we assume reworking of foraminiferagenerally to be low. A reservoir age of 464 yearswas appliedalso for these dates, although we are awarethat the reservoir age may be higher for the benthicforaminifera at sea bottom (1315 m), comparedto theplanktonic foraminifera reflecting the upper water1800LEGEND:PebblesMacrofossils45285+31402250MudSandy peliteDiamictonmasses. The reservoir age may also have variedthroughtime (Waelbroeck et al., 2001). However, at presentthere is no data from the study area to confirm this. Thedates were calibrated to calendar years (Stuiver et al.,1998) andsedimentation rates were calculatedby linearinterpolation between the dated levels. The results revealrelatively high sedimentation during early Holocene (40to 4200 cm/1000 cal. years) dropping abruptly around8800 cal. years BP to 3.2 cm/1000 cal. years for the800900100011001200130014001500160017007780±558040±658430±658850±558670±658910±659350±659375±709470±859535±809615±759765±759835±459920±45Fig. 2. Lithostratigraphy and radiocarbon dates in Core MD99-2304on the continental margin (left) andMD99-2305 in Van Mijenfjord(right), West Spitsbergen. The legendat bottom right applies to bothrecords.

2080ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088remaining part of the Holocene. Due to low carbonatecontent it was only possible to retrieve one date from theuppermost part of the Holocene. Stratigraphical analyseswere performedon sediment samples from each cmin the core. This implies a mean time resolution of 27years prior to 8800 cal. years BP and314 years after thisdate.5. Results5.1. The SiZer analysisThe results of the SiZer analysis for core MD99-2305are shown in Fig. 3. In the top panels of Fig. 3 A–C, afamily of smoothes (h) is given. The curves plottedhavesmooth ranges as given in the figure. The thick solidlinecorresponds to a choice of h one typically woulduse ifonly one scale was to be used. This ‘‘optimal’’ h is a datadriven bandwidth and a best choice of the bandwidthfrom a purely mathematical point of view (Ruppert etal., 1995). In the lower panels of Fig. 3A–C, SiZer plotsare given. The plot shows a feature map as a function ofscale (controlledby h) andlocation (given by x) for thesignal. For each (x,h) position, SiZer tests whether,a 1 a0; i.e. whether the curve at this level of resolutionhas a derivative different from zero which means thatthere is a true increase or decrease of the curve. Readersinterestedin more details about this are referredtoChaudhuri and Marron (1999).A significantly positive derivative in the SiZer map isflaggedas black, while a significantly negative derivativeis flaggedas light grey. This means that black showsareas where there is a significant increase, while lightgrey shows where there is a significant decrease. Atlocations where the derivative is not found to besignificantly different from zero, the colour grey is used.Hence grey means that there is no significant change.The horizontal black line (around h=100) correspondsto the smoothing obtainedby the solidblack line in theupper panel of Fig. 3A. Dark grey is usedto indicatethat too few data points are available to do inference.Typically, dark grey colour occurs at very small scalesfor SiZer. It shouldbe notedthat when the SiZer mapshow an, e.g. positive derivative of the curves, thiscorresponds to a rise in the actual proxy value, and viceversa for negative derivatives. In the following wedescribe from too (modern) to bottom (Younger Dryas)andexplain the variability andtrends in the curves asshown by the SiZer map.The results show that for the IRD variations the‘‘optimal’’ h is 109 years (Fig. 3A). Following thehorizontal line for this h, there was a markeddecreasefollowedby an increase between 800 and1000 years ago.Further a markedincrease is shown around3000 cal.years BP anda significant increase/decrease about7500 cal. years BP. Finally the large amount of IRD inthe older part of the record is depicted by two markedincreases older than 10,000 cal. years BP.For the foraminiferal flux the ‘‘optimal’’ time horizonis 311 years (Fig. 3B). It shows a short decrease about2000 cal. years BP andan increase between 3500 and4000 cal. years BP. In addition, significant decreases aredepicted at ca 7000 cal. years BP and 8500 cal. years BP.A reduction in foraminiferal flux is evidenced for thelowermost part of the stratigraphy, older that 10,500 cal. years BP.The ‘‘optimal’’ h for % E. excavatum is 161 years (Fig.3C). At this scale the only significant changes areincreases at about and11,000, 10,200 and8500 cal. yearsBP. For the periodfrom 8400 to 6200 cal. years BP thereare too few data to do any analysis. Both foraminiferalflux and E. excavatum have too few observations to giveany information on variability on time scales less than100 years throughout most of the Holocene (Figs. 3BandC).5.2. Proxy data variations in MD99-2305The lithology of MD99-2305 is characterizedby twounits (Fig. 2). The lower 2 m of the core comprise amassive andfirm diamicton with a high content of clastsanda low content of fossils. One AMS date performedon shell fragments from this unit gave a close to infiniteage of 45385 +/ 3140/2250 14 C years (Table 1),assumedto result from reworking. The diamicton has asharp boundary to the upper unit characterized by a softmudwith scatteredclasts. The acoustic data show thatboth units are widespread in the fjord; Dahlgren (1998)and Haldet al. (2001). We interpret the diamicton torepresent a basal till andthe upper unit to result fromglaciomarine fjord sedimentation. The boundary betweenthe two units is dated to 11,170 cal. years BP. Weconsider this to be a minimum age for the deglaciationof Van Mijenfjord.The proxy data for IRD, foraminiferal flux and E.excavatum are representedby the ‘‘optimal’’ curve asgiven in Fig. 3. The early deglaciation of Van Mijenfjordaround11,200 cal. years BP is characterizedby high icebergrafting decreasing rapidly reaching relatively lowvalues from ca 10,000 cal. years BP (Fig. 4E). A peak inIRD occurredaround7500 cal. years BP, followedbyhigh flux values between 7000 and4000 cal. years BP.As IRD declined there was a rise in the flux offoraminifera showing relatively high values during earlyHolocene (11,000–7000 cal. years BP). The same timeinterval was characterizedby a high abundance ofbivalve shells. The benthic foraminiferal species E.excavatum shows peak values during the earliest partof the deglaciation, followed by a marked decline untilca 8600 cal. years BP. There was a markedrise of thisspecies after 7000 cal. years BP, reaching peak values

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 20811000IRD flux # grains > 1 mm/cm 2 /ka100No.10h (years)500020001000500200100(A)2 4 6 8 10kcal. years BP4030Foraminiferal flux (no/cm 2 /years)No.20100h (years)10000500020001000500200100(B)2 4 6 8 10kcal. years BP6050No.40302010100005000% Elphidium excavatumh (years)20001000500200100(C)2 4 6 8 10kcal. years BPLEGEND:SignificantlydecreasingNo significantchangeToo few observationsto do inferenceSignificantlyincreasingFig. 3. SiZer analysis of proxy data from core MD99-2305, Van Mijenfjord, West Spitsbergen (A) IRD grains 41 mm (B) Foraminiferal flux and(C)% E. excavatum. In the top panel of each figure a family of smoothes (h-values) is given. The dots shows all data, the thick line shows the ‘‘optimal’’smooth (cf. discussion in text) andthe thin lines show various smooth values, with maximum andminimum values given. In the lower panel a SiZerplot is shown depicting at which time the proxy show a significant increase, decrease, no change or where there are too few observations.

2082ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–20880% E.excavatum0 10 20Foram flux (#/cm2/kyr)0 10 2030 40 50 5 15δ 13 C C. reniforme-0.4 -0.8 -1.2 -1.6IRDflux# grains >1mm/cm 2 ka1 10 100(A) (B) (C) (D) (E) (F)1000200030004000Cal. years BP500060007000800090001000011000120004.5 4 3.5δ 18 OC. renif.0 400 800 1200 1600depth in core (cm)Fig. 4. Proxy record for sediment core MD99-2305, Van Mijenfjord, West Spitsbergen: (A) % E. excavatum smoothedat value h=161 years; (B)foraminiferal flux, smoothedby h=311 years; (C) d 13 C measuredon the benthic foraminifer C. reniforme; (D) d 18 O measuredon the benthicforaminifer C. reniforme; (E) ice rafteddebris (IRD) smoothedat h=109 years; (F) age (cal. years BP) vs. cm depth in core plotted as a fourthpolynomial fit through the dated levels in the core. Cf. discussion in text of smoothing technique in (A), (B) and (E).around2800 cal. years BP. E. excavatum shows aincreasing trend(Fig. 4A) andforaminiferal flux (Fig.4B) a decreasing trend throughout the Holocene. Thed 13 C(Fig. 4C) and d 18 O(Fig. 4D), measuredon thebenthic foraminifer C. reniforme, show a parallel trendthroughout most of the Holocene. Both show a markeddepletion during the earliest part of the deglaciationleading to peak low values between 10,500 and 8300 cal.years BP. Maximum values occurredbetween ca 8000and4000 cal. years BP followedby a slight depletionduring the last 3000 years.5.3. Proxy data variations in MD99-2304The lithology of the upper 150 cm of MD99-2304 ischaracterizedby a brownish-grey homogenous mudwith scatteredclasts. The planktonic foraminifera aredominated by the polar species Neogloboquadrinapachyderma (sinistral) (Fig. 5A), followedby thesubpolar species Globigerina quinqueloba (Fig. 5B). Inaddition, the following subpolar species are recorded: N.pachyderma (dextral) (Fig. 5C), Globigerinita glutinata(not shown), G. bulloides (Fig. 5D), G. scitula andG. uvula (not shown). The SST reconstruction showshighly varying temperatures during earliest Holoceneuntil ca 10,800 cal. years BP, followedby stable,relatively warm temperatures until a rapidcooling ofca 4 1C around8800 cal. years BP. This cooling step isbracketedby two AMS dates (Table 1, Fig. 5F). Theremaining part of the Holocene appears to be characterizedbystable, cool sea surface temperatures.However, possible short-livedchanges may be obscuredby the low stratigraphic resolution for this part of thecore. Reconstruction of SST for the core-top sample (ca4 1C) is in agreement with instrumental observations ofsummer temperatures in the area.6. Paleoceanographic implications6.1. The proxy indicatorsPlanktonic foraminifera are reliable tracers of surfaceandsub-surface ocean temperatures (e.g. Be´ andTolderlund, 1971) andquantitative reconstructions ofSST basedon various statistical approaches have been

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 2083% N.pachyderma (s) % G. quinq % N.pachyderma (d) %G. bulloides SST-1(°C) SST-2 (°C) SST-3 (°C) cm in core0 50 100 0 50 100 0 10 20 30 0 1 2 3 4 3 4 5 6 7 8 9 3 4 5 6 7 8 9 3 4 5 6 7 8 9 0 40 80 120 16001000(A) (B) (C) (D) (E) (F) (G) (H)200030004000Age (cal. years BP)50006000700080009000100001100012000Fig. 5. Proxy recordfor sediment core MD99-2304, western Svalbardcontinental margin. (A–D) Percent frequencies for the most importantplanktonic foraminiferal species. (E) Sea surface temperature SST-1 reconstructed by the modern analogue technique, using the modern database ofPflaumann et al. (2003). (F) Sea surface temperature (SST-2) reconstructed by linear regression applying a modern data set from Johannesen et al.(1994). (G) Sea surface temperature (SST-2) reconstructed by linear regression applying a modern data set from Burhol (1994). (H) Age (cal. yearsBP) vs. cm depth in core plotted by linear interpolation between the dated levels in the core.applied over the last three decades. In the present studythe planktonic foraminiferal fauna was countedin the100 mm fraction, were as the modern data set ofPflaumann et al. (1996, 2003) usedfor the SST-1reconstruction (Fig. 5E) represents the 150 mm fraction.This adds an uncertainty to the reconstruction, since oneof the subpolar species, G. quinqueloba, frequently issmaller than 150 mm (Carstens andWefer, 1992). Thus,this species may be overrepresentedin the proxy recordcompared to the modern data set and lead to too highreconstructedSSTs. Thus, in order to elucidate thisuncertainty, we additionally reconstructed SST byapplying two smaller modern data sets in the 125 mmfraction (Johannessen et al., 1994) (Fig. 5F) andthe100 mm fraction (Burhol Fig. 5G), respectively. Theplanktonic foraminiferal fauna in the coldest watermasses (0–3 1C) is often completely dominated by thesingle species N. pachyderma (sinistral). Estimates in thistemperature interval are therefore considered to beuncertain (Pflaumann et al., 2003). In general, the mainstructure of three SST reconstructions compare well. Allshow a markedSST reduction at 8800 cal. years BP andfairly stable cool temperatures for the remaining part ofthe record. The SST-2 appears to be offset towardswarmer temperatures by 1–2 1C comparedto SST-1. Theoxygen andcarbon isotopes measuredon benthicforaminifera reflect both the ambient bottom watertemperature as well as the water mass isotopic compositionduring the time of precipitation of the carbonatetest. The benthic foraminiferal species E. excavatum iscommon in Arctic fjords and occurs with high frequenciesclose to tide-water glacier termini (HaldandKorsun, 1997; Korsun andHald1998).6.2. Early HoloceneThe SST-reconstruction from the continental marginoff West Spitsbergen (MD99-2304, Figs. 5E–G) showthat significantly warmer water than at present wasstarting to develop around 11,000 cal. years BP. Stablewarm conditions prevailed from 10,800 cal. years BPuntil 8800 cal. years BP. This relatively warm periodsuggests that Atlantic Water heat flux in the WestSpitsbergen Current was stronger than at present. Thiswarm periodis reflectedin the Van Mijenfjordbydepletion in benthic d 18 O of 0.8% between 10,900 and10,300 cal. years BP (Fig. 4D), suggesting a warming ofthe bottom water in the fjord. Further, warming issupportedby the rise in foraminiferal flux, depletion inbenthic d 13 C(Fig. 4C) andhigh abundance of macrofossils(Fig. 2), indicative of more productive watersthan today. Influx of relatively warm oceanic water fromthe continental margin across the fjordsill probablycausedthis warming.6.3. Middle and Late HoloceneThe rise in d 18 O; d 13 C and E. excavatum anddepletionin foraminiferal flux in the fjordrecord(Fig. 4) indicatesa gradual cooling starting around 8800 cal. years BP.

2084ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088This gradual cooling in the fjord record differs from theabrupt cooling observedon the continental margin atthis time (Fig. 5). We assume that the abrupt cooling onthe margin may partly be an indication of the slowsedimentation rates and hence low stratigraphic resolutionafter 8800 cal. years BP as documented for this site(Fig. 5). The fall in SST on the margin indicates reducedinfluence of warm Atlantic Water andincreasedinfluence of Polar Water and/or runoff from land.This may be due to a change in the drift patternand/or intensity of the West Spitsbergen Current. Theconditions on the continental margin apparently remainedstableandcool for the remaining part of theHolocene. However, neither dating nor time resolutionfor this part of the record does not permit any detailedreconstructions.We suggest that the following three factors may havecontributed to the marked reduction in sedimentationrates on the continental margin around8800 cal. yearsBP: (1) Early Holocene sedimentation rates on continentalmargins that were glaciatedduring the lastglacial, are often foundto be high due to a large accessto easily erodable glaciogenic sediments from recentlydeglaciated areas (Haldet al., 1999); (2) increase inbottom currents due to a glacio-eustatic fall in sea level(Landvik et al., 1987) and(3) increasedbottom watererosion due to a deeper flow of the Atlantic Waterduring cooler conditions. Today, Atlantic Water attainsa gradually deeper flow as it flows northward along theSvalbard margin, due to cooling and increase of density(Hopkins, 1991).In Van Mijenfjord, the period 7500–4000 cal. years BPseems to represent a general cooling. This is supportedby high benthic d 18 O indicating cool bottom water, areduction in the biogenic production, rise in E.excavatum andenhancedIRD (Fig. 4). This generalcooling may partly be influencedby flux of cool watersfrom the continental margin to the fjord, and partly dueto a more restrictedfjord/margin water-mass exchangeacross the fjord sill due to glacio-isostatic rebound. Asthe isostatic rise continuedtowards the present time(Landvik et al., 1987), the sill of Van Mijenfjordgradually became shallower, and the fjord would likelydevelop more of a local climatic signal with reducedconnection to the open ocean.The proxy data in the fjord record suggest that coolconditions prevailed for the last 4000 years, as indicatedby continuedhigh frequencies of E. excavatum andlowforaminiferal flux (Fig. 4). The slight depletions in bothd 18 Oandd 13 C indicate changes of the source of thebottom water, rather than a rise temperature sincemodern bottom water temperatures in Van Mijenfjordtoday are already at a minimum between 1 and1.5 1C. One possible mechanism for depleting thebottom water with respect to d 18 O; without changing thetemperature, is increasedbrine formation. When sea-icefreezes during the fall, brine, characterized by very lowtemperatures andhigh salinity, sinks to a deeper level.This probably explains the very coldbottom water inVan Mijenfjordtoday (Skardhamar, 1998). As brineformation takes place in the surface water, it also attainsa light d 18 O composition. The brine rapidly sinksthrough the water column andthereby brings d 18 O-depleted water to the bottom. Current investigationsrecently revealed d 18 O-depleted and very cold bottomwater in silled fjords of Svalbard, including VanMijenfjord (Mikalsen and Hald, unpublished). Thereduction in iceberg rafting over the last 4000 years,except for the surge event around800 cal. years BP(Haldet al., 2001), may be a response to a morecontinental climate of the region. This may havedeveloped during an increasing isolation of VanMijenfjorddue to glacio-isostatic rebound(cf. Landviket al., 1987).7. Holocene glacial history of SpitsbergenThe content of grains 41 mm in the MD99-2305 (Fig.4) may reflect both rafting from icebergs andsea-ice. Inaddition, coarser grains may be induced by storminducedsediment suspensions in shallow water (Rumohret al., 2001). However, lack of turbidites or otherbottom current structures in the sediments, indicate thatthis latter process is of less importance in vanMijenfjord. To our knowledge, no systematic studieshave been undertaken with respect to the content ofdebris in sea-ice in the Svalbard fjords. However, weassume that most of the IRD reflects rafting from icebergsbasedon the following. At present, the fjordiscoveredby sea-ice 6–9 months per year, andhas threeglaciers that calve into the fjord, but there is a very lowcontent of IRD both at present andduring the last 1000years. This is well documented both in the present study(Fig. 4E) andby Haldet al. (2001). For two of theintervals with high IRD flux, there is very goodevidencefor increasedice berg rafting. One of these intervals isthe markedIRD-peak at ca 800 cal. years BP (Fig. 4E)that is correlatedto a surge event that is welldocumented both on land and in submarine sedimentsin the fjord(Punning et al., 1976; Rowan et al., 1982;Haldet al., 2001). Another interval with high IRD fluxwas the last deglaciation of Van Mijenfjord followingthe Younger Dryas. This deglaciation is well documentedbothin the present core andfrom studies on land(Landvik et al., 1987; Mangerudet al., 1992).Van Mijenfjordwas deglaciatedby 11,200 cal. yearsBP (corresponding to 9900 14 C years). The lowest date islocatedjust above a diamicton, interpretedto be a till(Fig. 2). This age is assumedto represent a minimum agefor the deglaciation of the fjord and is relatively youngcompared to the oldest dates of the deglaciation of van

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 2085Mijenfjordobtainedby Mangerudet al. (1992).They retrieved a number of radiocarbon datesfrom raised marine deposits underlain by till east ofthe sill in the Van Mijenfjordarea, ranging between10,650 and9500 14 C years BP. We cannot excludethat the oldest deglacial sediments in MD99-2305 ismissing, thus implying a younger minimum age for thedeglaciation compared to oldest dates retrieved byMangerudet al. (1992).The deglaciation phase of the fjord was characterizedby a very high IRD flux (Fig. 4E). A markedreductionof IRD occurredbetween 11,200 and11,000 cal. yearsBP (Figs. 3A and4E). We assume that this reductionreflects that a larger part of the glacier withdrew onland. Again, the present data gives somewhat youngerages for this deglaciation phase compared to resultsfrom Mangerudet al. (1992), inferring that the VanMijenfjordglacier broke up very fast, between 10,600and10,300 14 C years BP. However, the chronology ofthe IRD curve from 11,200 cal. years BP (9900 14 C yearsBP) andonwards is well supportedin the presentstudy. This we believe is a good indicator that an intensebreak-up of the glacier prevailedat least until 11,000 cal.years BP.The moderate to low IRD flux between 10,500 and8000 cal. years BP (Fig. 4E) corresponds partly to aperiod during which the other proxy data in MD99-2305(Figs. 4A–D) indicate a climatic optimum. Further,increasedrafting peaking around7500, as well asbetween 7000 and4000 cal. years BP, correlates to acooling of the bottom water of the fjord, and to lowSSTs on the continental margin. Thus, increasedraftingmay be a response to a reduction of the equilibrium linealtitude causing increased tide-water glaciation in VanMijenfjord. After 4000 cal. years BP, ice rating wasreduced. Today, 50% of the catchment area of VanMijenfjordis glaciated, andonly three glaciers calvedirectly into the fjord (Hagen et al., 1993). Thesignificant IRD-flux at ca 800 cal. years BP is, asmentionedabove, correlatedto a markedsurge event,namely that of the Paula Glacier System (PGS) in theinner part of the fjord(Haldet al., 2001; Dahlgren, inprep.). During this event the PGS advanced 35 kmbeyondits present glacier front, leaving an IRD spikethat is evident as a marker horizon in high resolutionacoustic profiles as well as in a number of sediment coresthroughout the fjord. We find no evidence in theacoustic record of any older surges during, e.g. earlyand middle Holocene. The presence of IRD throughoutthe entire Holocene indicates that central Spitsbergenwas never completely deglaciated during this interglacialperiod.Holocene glacial fluctuations on Svalbardhave beeninvestigatedin the Isfjordarea (Elverhøi et al., 1995;Svendsen and Mangerud, 1997). These studies inferreduced glacial activity during early/middle Holoceneandincreasedglaciation for the last 4000 years inBillefjordin the inner Isfjord, andthe last 2000 years atLinnévannet, outer Isfjord. Reduced glaciation duringearly Holocene correlates well to the present study.However for the midandlate Holocene the Isfjordreconstructions and the present study differ. The IRDrecord in Van Mijenfjord indicates enhanced tidewaterglaciation during mid-Holocene and somewhat reducedduring late Holocene. Svendsen and Mangerud (1997)present goodevidences that the Linne´ vannet glacier, inthe outer Isfjordarea, startedto form 4000–5000 yearsago reaching a maximum during the Little Ice Age.Thus, we cannot exclude that there are local differencesin glacial history on Svalbard, although this is not whatwe wouldexpect since the areas are only some50–100 km apart. The Holocene glacier variations ininner Isfjordare basedlargely on sedimentation rates(Elverhøi et al., 1995). In terms of sedimentation ratesall records show a similar pattern. However, we considerIRD to be a more reliable indicator of tidewaterglaciation, rather than sedimentation rates. Changes insedimentation rate are not only dependent on relativeglaciation in the catchment area, but also the conditionsin the accumulation basin.8. CorrelationsWe compare the proxy records from the Svalbardmargin andVan Mijenfjordwith a high resolutionmarine SST proxy recordfrom the Western Barents Sea,ca 751N (Sarnthein et al., 2003) (Fig. 6). Both SSTrecordsare reconstructed by applying the modernanalogue technique to planktonic foraminiferal faunaldata, and they are interpreted to reflect latitudinaltemperature gradients of Atlantic Water in the SpitsbergenCurrent during the Holocene. These SST-recordsare comparedto the Van MijenfjordIRD recordandsummer insolation at 781N (Berger andLoutre, 1991).All records have individual, independent, radiocarbonbasedtimescales. The absolute temperatures estimateson the Svalbardmargin are somewhat higher comparedto those from the Barents Sea. We assume this can beattributedto the difference in size fraction usedfor theforaminiferal analysis. In the present study we used the100 mm fraction that will include more of the smallsubpolar species, implying higher SST, comparedto thestudy of Sarnthein et al. (2003) using the150 mm fraction.However, both SST-records show a remarkable similarityduring the early Holocene, including the abruptcooling around8800 cal. years BP. The early Holoceneappears to represent the Holocene climatic optimum.This supports modelling experiments by Liu et al. (2003)predicting increased SST with increasing northernlatitude as a function of stronger insolation during theearly Holocene (Fig. 6D). The middle Holocene SST

2086ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–208801000SST(°C)SST(°C)W.Barents Sea marginW.Svalbard margin2 4 6 8 10 2 4 6 8 10IRD flux# grains > 1mm/cm 2 /ka1 10 100(A) (B) (C) (D)Insolation (W/m 2 78 − N)500 520 540 560 580200030004000Cal. years BP50006000700080009000100001100012000Fig. 6. Correlation between SST records from (A) Western Barents Sea (Sarnthein et al., 2003), (B) W. Svalbardmargin (SST-1, present study), (C)Van MijenfjordIRD (present study) and(D) insolation 781N (Berger andLoutre, 1991).depicts a decreasing trend in the Barents Sea. Superimposedonthis trendare millennial–centennial scalefluctuations. These fluctuations were not identified in theSvalbard record, probably due to low stratigraphicresolution. The overall cooling trend during the middleHolocene in the Barents Sea correlates well to increasedtidewater glaciation in Van Mijenfjord. The Barents Searecordis locatedat 751N in a sensitive boundary area(Fig. 1A), andabrupt coolings/warmings can be linkedto relatively small displacements of the Arctic oceanicfront system, separating cool Arctic Water from warmerAtlantic Water (Sarnthein et al., 2003).An early Holocene warm periodis well documentedina number of proxy records from the Svalbard andwestern Barents Sea region. In addition to high SSTs ofthe West Spitsbergen Current (HaldandAspeli, 1997;Sarnthein et al., 2003), warmer surface waters in thefjords are indicated by thermophilic molluscs (Salvigsenet al., 1992). The conclusion of Svendsen and Mangerud(1997) that the glacier Linne´ breen in outer Isfjorddidnot exist during early Holocene, may imply atmospherictemperatures of at least 1.5–2.5 1C higher than atpresent, if winter precipitation was similar to thepresent. In a study of lacustrine sediments on the islandBear Island, (741N, western Barents Sea), Wohlfarth etal. (1995) concluded that a climatic optimum occurredbetween 9500 and8000 14 C years. This is in fairly goodagreement with the present study (Fig. 6). However,there are indications of a relatively warm middleHolocene which contradict the present study. Thisincludes both reduced glaciation of the glacier Linnébreen,as well as presence of thermophilic bivalves(Salvigsen et al., 1992). However, the frequency of thethermophilic bivalves appears to decline during theHolocene, implying a cooling trendin line with thepresent study. In a study of plant macrofossils in asediment core from a lake in the outer coastal Isfjordarea Birks (1991) suggests that mean July temperatureswere 2 1C warmer than today between 8000 and 500014 C years BP. But the periodprior to 8000 years BP wasnot investigatedin that study.9. Conclusions Sea surface temperatures, significantly warmer than atpresent, were recorded on the continental margin offWest Spitsbergen from ca 11,200–8800 cal. years BP. As the Van Mijenfjordwas deglaciatedafter theYounger Dryas, ca 11,200 cal. years BP, waterwarmer than present almost immediately invadedthe fjordbottom. The presence of IRD throughout the entire Holoceneindicates that central Spitsbergen was never completelydeglaciated during this interglacial period.After the glaciation during the Younger Dryas, therewas an intense deglaciation phase until ca 11,000 cal.years BP. Moderate to low tidewater glaciationcharacterizedthe early Holocene, 11,000–7500 cal.years BP, followedby enhancedglaciation around7500 cal. years BP andbetween 7000 and5000 cal.years BP. The glacial history is supportedby the quantitativereconstructions of SST on the continental margin and

ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–2088 2087relative bottom water temperature in the VanMijenfjord record, indicating that the warm earlyHolocene, 11,200–8800 cal. years BP, was followedbya cooling, 8800–4000 cal. BP andrelatively stable,cool conditions the last 4000 years.AcknowledgementsFunding was provided by the VISTA programme, theResearch Council of Norway to the Strategic UniversityProgramme SPONCOM, andby the RoaldAmundsenCentre, University of Tromsø. Technical assistance wasprovided by the crews onboard R/V Jan Mayen and R/V Marion Dufresne, Jan P. Holm (computer drawings),Edel Ellingsen, Trine Dahl, Odd Aasheim (laboratorywork), Elsebeth Thomsen Hanken (macrofossil identification).Carin Andersson Dahl and Kari-Lise Rørvikhelpedwith the C2-SST reconstructions. Stable isotopeswere measuredat the University of Bergen, Norway,Radiocarbon dates were run by the RadiocarbonLaboratory in Trondheim, Norway, Uppsala, Sweden,Kiel, Germany andArizona, USA. To all these personsandinstitutions we offer our sincere thanks.ReferencesAndersen, E.S., Dokken, T.M., Elverhøi, A., Solheim, A., Fossen, I.,1996. Late Quaternary sedimentation and glacial history of thewestern Svalbardcontinental margin. Marine Geology 133,123–156.Austin, W.E.N., Kroon, D., 1996. Late glacial sedimentology,foraminifera andstable isotope stratigraphy of the Hebrideancontinental shelf, northwest Scotland. Late Quaternary Paleoceanographyof the North Atlantic Margins, vol. 111. GeologicalSociety Special Publication, London, pp. 187–213.Bé, A.W.H., Tolderlund, D.S., 1971. Distribution and ecology of livingplanktonic foraminifera in surface waters of the Atlantic andIndian Oceans. In: Funnel, B.M., Riedel, W.R. (Eds.), TheMicropaleontology of Oceans. Cambridge University Press,London.Berger, A., Loutre, M.F., 1991. Insolation values for the climate of thelast 10 million years. Quaternary Sciences Review 10 (4), 297–317.Birks, H.H., 1991. Holocene vegetational history andclimate change inwest Spitsbergen—plant macrofossils from Skardtjørna, an Arcticlake. The Holocene 1 (3), 209–218.Burhol, A., 1994. Resent planktonic foraminifera in the NENorwegian Sea andwestern Barents Sea. UnpublishedMasterThesis, University of Tromsø (in Norwegian).Carstens, J., Wefer, G., 1992. Recent distribution of planktonicforaminifera in the Nansen Basin, Arctic Ocean. Deep-SeaResearch 39, S507–S524.Chaudhuri, P., Marron, S., 1999. SiZer for exploration of sturctures incurves. Journal of the Amerikan Statistical Association 94,807–823.Dahlgren, T., 1998. Holocene Sedimentation And Glacial history ofinner Van Mijenfjorden, Spitsbergen, University of Tromsø, 86pp.Dowdeswell, J., Dowdeswell, E.-K., 1989. Debris in icebergs and ratesof glaci-marine sedimentation; observations from Spitsbergen anda simple model. Journal of Geology 92 (2), 221–231.Duplessy, J.-C., 1978. Isotope studies. In: Gribbin, J.R. (Ed.),Climatic Change. Cambridge University Press, Cambridge,pp. 46–67.Elverhøi, A., Svendsen, J. I., Solheim, A., Andersen, E. S., Milliman,J., Mangerud, J., Hooke, R.L., 1995. Late Quaternary SedimentYieldfrom the High Actic SvalbardArea: The Journal of Geology,vol. 103, pp. 1–17.Fairbanks, R.G., 1989. A 17000-year glacio-eustatic sea level record:influence of glacial melting rates on the Younger Dryas event anddeep-ocean circulation. Nature 342, 637–642.Fan, J., 1991. Design-adaptive nonparametric regression. Journal ofthe American Statistical Association 87, 998–1004.Feyling-Hanssen, R.W., 1958. Mikropaleontologiens teknikk. NorwegianGeotechnical Institute 29, 1–14.Gulliksen, B., Holte, B., Jakola, K.-J., 1985. The soft bottom fauna invan Mijenfjord and Raudfjord, Svalbard. In: Gray, J.S., Christiansen,M.E. (Eds.), Marine Biology of Polar Regions and Effectsof Stress on Marine Organisms, pp. 199–215Hadley Centre, 2004. http://www.met-office.gov.uk/research/hadleycentre/models/modeldata.html.Hagen, J.O., Liestøl, O., Sollid, J.L., Wold, B., Østrem, G., 1993.Glacier atlas of SvalbardandJan Mayen. Norsk PolarinstituttMeddelelser(129), 141pp.Hald, M., Aspeli, R., 1997. Rapid climatic shifts of the northernNorwegian Sea during the last deglaciation and the Holocene.Boreas 26, 15–28.Hald, M., Korsun, S., 1997. Distribution of modern Arctic benthicforaminifera from fjords of Svalbard. Journal of ForaminiferalResearch 27, 101–122.Hald, M., Kolstad, V., Polyak, L., Forman, L., Herlihy, F.A., Ivanov,G., Nescheretov, A., 1999. Late-glacial andHolocene paleoceanographyandsedimentary environments in the Saint Anna Trough,Eurasian Arctic Ocean Margin. Palaeogeography, Palaeoclimatology,Palaeoecology 146, 229–249.Hald, M., Dahlgren, T., Olsen, T.-E., Lebesbye, E., 2001. LateHolocene paleoceanography in Van Mijenfjorden, Svalbard. PolarResearch 20, 23–35.Hanssen-Bauer, I., Kristensen, M.S., Steffensen, E.L., 1990. TheClimate of Spitsbergen. DNMI, Oslo.Hopkins, T.S., 1991. The GIN Sea—a synthesis of its physicaloceanography andliterature review 1972–1985. Earth-ScienceReviews 30, 175–318.Johannessen, T., Jansen, E., et al., 1994. The relationship betweensurface water masses, oceanographic fronts, andpaleoclimaticproxies in surface sediments of the Greenland, Iceland, NorwegianSeas. Nato ASI Series 117, 61–85.Juggins, S., 2002. C2 1.3 version. http://www.staff.ncl.ac.uk/stephen.juggins.Korsun, S., Hald, M., 1998. Modern benthic foraminifera off tidewater glaciers, Novaja Semlja, Russian Arctic. Arctic andAlpineResearch 30 (1), 61–77.Labeyrie, L., Jansen, E., Cortijo, 2003. Les rapports de campagnes a` lamer. MD 114/IMAGES V OCE/2003/02. Brest, Institut PolaireFrancais-Paul Emile Victor, 850pp.Landvik, J.Y., Mangerud, J., Salvigsen, O., 1987. The late WeichselianandHolocene shoreline displacement on the west-central coast ofSvalbard. Polar Research 5, 29–44.Liu, Z., Brady, E., Lynch-Stieglitz, J., 2003. Global ocean response toorbital forcing in the Holocene. Paleoceanography 18, 1041,doi:10.1029/2002PA000819.Mangerud, J., Gulliksen, S., 1975. Apparent Radiocarbon Ages ofRecent Marine Shells from Norway, Spitsbergen andArcticCanada. Quaternary Research 5, 263–273.Mangerud, J., Bolstad, M., Elgersma, A., Helliksen, D., Landvik, J.Y.,Lønne, I., Lycke, A.K., Salvigsen, O., Sandahl, T., Svendsen, J.I.,

2088ARTICLE IN PRESSM. Hald et al. / Quaternary Science Reviews 23 (2004) 2075–20881992. The Last Glacial Maximum on Spitsbergen, Svalbard. Boreas38, 1–31.Meldgaard, S., Knudsen, K.L., 1979. Metoder til indsamling ogoparbejding af prøver til foraminifer-analyser. Dansk natur Danskskole.Pflaumann, U., Duprat, J., Pujol, C., Labeyrie, L.D., 1996. SIMMAX:A modern analog technique to deduce Atlantic sea surfacetemperatures from planktonic foraminifera in deep-sea sediments.Paleoceanography 11, 15–35.Pflaumann, U., Sarnthein, M., Chapman, M., Ld’AZbreu, L., Funnell,B., Huels, M., Kiefer, T., Maslin, M., Schultz, H., Swallow, J., vanKreveld, S., Vautravers, M., Vogelsang, E., Weinelt, M., 2003.Glacial North Atlantic Sea-surface conditions reconstructed byGLAMAP 2000. Paleoceanography 18 (3) 10-1–10-21.Punning, J.-M., Troitsky, L., Rajamae, R., 1976. The genesis andageof the Quaternary deposits in the eastern part of Van Mijenfjorden,West Spitsbergen. Geologiska Fo¨reningens Fo¨rhandlingar 98,343–347.Rowan, D.E., Pewe, T.L., Pewe, R.H., 1982. Holocene glacial geologyof the Svea lowland, Spitsbergen, Svalbard. Geografiske Annaler64A, 35–51.Rumohr, J., Blaume, F., Erlenkeuser, H., Fohrmann, H., Hollender,F.-J., Mienert, J., Schäfer-Neth, C., 2001. Records and processes ofnear-bottom sediment transport along the Norwegian–GreenlandSea margins during the Holocene and Late Weichselian (TerminationI) times. In: Schäfer, P., Ritzrau, W., Schlüter, M., Thiede, J.(Eds.), The Northern North Atlantic, a Changing Environment.Springer, Berlin, Heidelberg, New York, 500pp.Ruppert, D., Sheather, S.J., Wand, M.P., 1995. An effective bandwithselector for local least square selections. Journal of the AmericanStatistical Association 90, 1257–1270.Salvigsen, O., Winsnes, T.O., 1989. Geological map of Svalbard1:100,000. Sheet C10G Braganzavågen (Map). Norsk PolarinstituttTemakart 4.Salvigsen, O., Forman, S.L., Miller, G.H., 1992. Thermophilousmolluscs on Svalbardduring the Holocene andtheir paleoclimaticimplications. Polar Research 11 (1), 1–10.Sarnthein, M., Van Kreveld, S., Erlenkeuser, H., Grootes, P.M.,Kucera, M., Pflaumann, U., Schulz, M., 2003. Centennial-tomillennial-scaleperiodicities of Holocene climate and sedimentinjections off the western Baretns shelf, 751N. Boreas 32, 447–461.Shackleton, N.J., Opdyke, N.D., 1973. Oxygen Isotope and PalaeomagneticStratigraphy of Equatorial Pacific Core V28-238: OxygenIsotope Temperatures andIce Volume on a 10 5 Year and10 6 YearScale. Quaternary Research 3, 39–55.Shackleton, N.J., Imbrie, J., Hall, M.A., 1983. Oxygen andcarbonisotope recordof The East Pacific core V19-30: implications for theformation of deep water in the late Pleistocene North Atlantic.Earth Planetary Science Letter 65, 233–244.Skardhamar, J., 1998. Sirkulasjonen i Van Mijenfjorden, en arktiskfjord (Circulation in Van Mijenfjorden, an Arctic fiord). MasterThesis, University of Bergen, Bergen, 79pp.Steel, R.J., Worsley, D., 1984. Svalbards post-Caledonian strata—anatlas of seidmentological patterns and paleogeographicevolution. In: Spencer, A.M.e.a.e. (Ed.), Petroleum Geology ofthe Northern European Margin. Graham andTrotman, London,pp. 109–135.Stuiver, M., et al., 1998. INTCAL98 Radiocarbon age calibration,24,000-0 cal BP. Radiocarbon 40 (3), 1041–1083.Svendsen, J.I., Mangerud, J., 1997. Holocene glacial and climaticvariations on Spitsbergen, Svalbard. The Holocene 7, 45–57.Vinje, T., 1985. The physical environment, the western Barents Sea.Drift, composition, morphology anddistribution of the sea icefields in the Barents Sea. Norsk Polarinstitutt Skrifter 179c p. 26.Wadhams, P., 1981. The ice cover in the Greenland and NorwegianSeas. Review Geophysical Space Physics 19, 345–393.Waelbroeck, C., Duplessy, J.-C., Miches, E., Labeyrie, L., Paillard, D.,Duprat, J., 2001. The timing of the last deglaciation in NorthAtlantic climate records. Nature 412, 724–727.Wohlfarth, B., et al., 1995. Early Holocene environment on Bjørnøya(Svalbard) inferred from multidisciplinary lake sediment studies.Polar Research 14, 253–275.