Variations in temperature and extent of Atlantic Water in the northern North Atlantic during the Holocene

Transcription

1 ARTICLE IN PRESS Quaternary Science Reviews 26 (2007) Variations in temperature and extent of Atlantic Water in the northern North Atlantic during the Holocene Morten Hald a,, Carin Andersson b, Hanne Ebbesen a,1, Eystein Jansen b, Dorthe Klitgaard-Kristensen c, Bjørg Risebrobakken b, Gaute R. Salomonsen a,2, Michael Sarnthein d, Hans Petter Sejrup e, Richard J. Telford b,f a Department of Geology, University of Tromsø, N-9037 Tromsø, Norway b Bjerknes Centre for Climate Research, Allégt. 55, N-5007 Bergen, Norway c The Norwegian Polar Institute, N-9296 Tromsø, Norway d Institut für Geowissenschaften, University of Kiel, Olshausenstrasse 40, D Kiel, Germany e Department of Earth Sciences, Allégt. 41, University of Bergen, 5007 Bergen, Norway f Department of Biology, University of Bergen, Norway Received 10 June 2007; received in revised form 6 October 2007; accepted 8 October 2007 Abstract We compare six high-resolution Holocene, sediment cores along a S N transect on the Norwegian Svalbard continental margin from ca 601N to 77.41N, northern North Atlantic. Planktonic foraminifera in the cores were investigated to show the changes in upper surface and subsurface water mass distribution and properties, including summer sea-surface temperatures (SST). The cores are located below the axis of the Norwegian Current and the West Spitsbergen Current, which today transport warm Atlantic Water to the Arctic. Sediment accumulation rates are generally high at all the core sites, allowing for a temporal resolution of years. SST is reconstructed using different types of transfer functions, resulting in very similar SST trends, with deviations of no more than 71.0/1.5 1C. A transfer function based on the maximum likelihood statistical approach is found to be most relevant. The reconstruction documents an abrupt change in planktonic foraminiferal faunal composition and an associated warming at the Younger Dryas Preboreal transition. The earliest part of the Holocene was characterized by large temperature variability, including the Preboreal Oscillations and the 8.2 k event. In general, the early Holocene was characterized by SSTs similar to those of today in the south and warmer than today in the north, and a smaller S N temperature gradient (0.23 1C/1N) compared to the present temperature gradient (0.46 1C/1N). The southern proxy records (60 691N) were more strongly influenced by slightly cooler subsurface water probably due to the seasonality of the orbital forcing and increased stratification due to freshening. The northern records ( N) display a millennial-scale change associated with reduced insolation and a gradual weakening of the North Atlantic thermohaline circulation (THC). The observed northwards amplification of the early Holocene warming is comparable to the pattern of recent global warming and future climate modelling, which predicts greater warming at higher latitudes. The overall trend during mid and late Holocene was a cooling in the north, stable or weak warming in the south, and a maximum S N SST gradient of ca 0.7 1C/1N at 5000 cal. years BP. Superimposed on this trend were several abrupt temperature shifts. Four of these shifts, dated to , and 1000 and 400 cal. years BP, appear to be global, as they correlate with periods of global climate change. In general, there is a good correlation between the northern North Atlantic temperature records and climate records from Norway and Svalbard. r 2007 Elsevier Ltd. All rights reserved. 1. Introduction Corresponding author. Tel.: ; fax: address: Morten.Hald@ig.uit.no (M. Hald). 1 Current address: GEUS, Ø. Voldgade 10, 1350 Copenhagen, Denmark. 2 Current address: Norconsult AS Post-box 110, 3191 Horten, Norway. Climate modelling experiments predict a global warming of about 3 1C during the next century and that warming of the high northern latitudes may be as much as twice the global mean depending on the future emissions of /$ - see front matter r 2007 Elsevier Ltd. All rights reserved. doi: /j.quascirev

2 3424 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) greenhouse gases (IPCC, 2007). To improve the predictions of future climates, particularly on regional scales, a better understanding of the natural climate changes based on records extending back beyond the instrumental records is needed. This paper contributes to this goal by investigating the Holocene climate changes in the northern North Atlantic (Nordic seas). Today, northwest Europe and the northern North Atlantic region have a climate C warmer than the zonal mean, in part because of the huge amount of heat transported northwards by the warm Atlantic Water in the Norwegian Current (NC) and West Spitsbergen Current (WSC). This relatively warm climate is more or less characteristic for the last years, the Holocene interglacial. However, there is evidence that climate varied considerably during the Holocene, although with a smaller amplitude than the extreme fluctuations of the glacials. To attain a more comprehensive view on the natural climate changes in the northern North Atlantic, we synthesise results from six previously published highresolution sea-surface temperature proxy records from the continental margin off Norway, the western Barents Sea and Svalbard. The data are treated in a coherent fashion in terms of transfer function methodology and age model methods. The core sites are located between ca 621N and 77.41N along the axis of route of Atlantic Water inflow. The study of Holocene climate in deep sea records has to some extent suffered from low resolution or poor preservation during coring: for the present study we focus on depo-center areas on the shelf and continental slope with high Holocene sediment accumulation rates. This enables at best a multi-decadal time resolution, allowing us to study relatively abrupt and short changes in natural climate on sub-centennial time scales. 2. Oceanography The hydrography along the Norwegian Barents Sea Svalbard margin is dominated by three main water masses: Atlantic Water; Arctic Water (mainly the northern part) and Coastal Water. At present, warm (7 13 1C) and salty (X35 PSU) Atlantic Water is carried into the northern North Atlantic by the NC and continues into the Arctic Ocean in the WSC and to the Barents Sea in the North Cape Current (Fig. 1). The NC and WSC are located along the Norwegian and western Svalbard margin, respectively. Atlantic Water reaches a depth of ca 600 m in the south and 4800 m in the north. North of Svalbard Atlantic Water sinks below Polar Water and continues as a Fig. 1. (A) Surface waters in the northern North Atlantic and adjoining seas modified from Mosby (1968). Definition of the water masses is according to Hopkins (1991). (B) Bathymetric map showing the location of the sediment cores along the Norwegian Svalbard margin BIT ¼ Bear Island Trough.

3 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) subsurface (4100 m deep) current into the Arctic Ocean. The warm Atlantic Water meets and mixes with Polar Water from the Arctic Ocean forming Arctic Water. This water mass enters the study area via the East Spitsbergen Current (ESC) from the north-eastern Barents Sea and continues northwards along the inner shelf of western Svalbard. Arctic Water is characterized by reduced temperatures and salinities compared to Atlantic Water, and it is seasonally covered by sea-ice. The boundaries between Arctic Water and Atlantic Water may form sharp climatic gradients in terms of ocean temperatures and seaice distribution, in particular in the western Barents Sea (Fig. 1) and is termed the Arctic Front (cf. Hopkins, 1991). The boundary between Polar Water and Arctic Water is termed the Polar Front (cf. Hopkins, 1991). Norwegian Coastal Water is transported along the Norwegian coast in the Norwegian Coastal Current (NCC). The Coastal Water is influenced by freshwater runoff from the Norwegian mainland and from the Baltic Sea and thus characterized by reduced salinities (o35 PSU). Generally the Coastal Water overlies the Atlantic Water as a westwards thinning wedge. Mixing of the two water masses increases northwards. Atlantic Water contributes to the present mild climate of north-western Europe because large amounts of heat are released into the atmosphere during winter time when this water cools before sinking to contribute to the North Atlantic Deep Water (NADW). This deep-water formation is a sink for atmospheric CO 2, and a large portion of the ventilated deep waters of the world oceans are formed in the northern North Atlantic. The inflow of Atlantic Water to the northern North Atlantic is balanced by surface outflow of the cold East Greenland Current, together with the deep-water formation. This circulation pattern is part of the North Atlantic thermohaline circulation (THC) during which heat (about W) is transferred from the South Atlantic to the North Atlantic. An initial climate change in the northern North Atlantic region can be transmitted to the world ocean by the NADW, or to the global climate system through the carbon cycle, and via ice albedo feedbacks. In addition, climatic changes on sub-orbital time scales may be triggered in the northern North Atlantic by abrupt changes in e.g. Atlantic Water heat flow, sea-ice distribution or deep-water formation. 3. Material and methods The material for this study was obtained mainly from previously published proxy data from six high-resolution sediment core sites along the Norwegian Barents Sea Svalbard continental margin (Fig. 1, Table 1). The proxy data include planktonic foraminiferal census down-core data, AMS 14 C dates (Table 2) and tephra time markers (Table 2). Some additional census data and AMS 14 C dates available since initial publication have been added to the database Planktonic foraminiferal census data Planktonic foraminifera were extracted from samples in the sediment cores using slightly different size fractions. The fraction 4100 mm was used in core T88-2 and MD , the fraction 4125 mm was used in core 8903 and T79-51/2, and the fraction 4150 mm was used in core MD and (Table 1). The latter fraction was used by CLIMAP to facilitate global comparisons. The fractions were repeatedly split until enough species remained to identify and count approximately 300 specimens. In a few cases samples with less (X100) specimens Table 1 Core location, water depth, length, geographical area and proxy data reference Core ID Latitude Longitude Water depth (m) Core length (cm) Location References Troll 8903/ N E North Sea Klitgaard-Kristensen et al. (2001) N E N E MD N E Mid Norwegian Margin (Vøring Plateau) JM97-948/2A BC N E T79-51/ N E North Norwegian margin (Andfjorden) Risebrobakken et al. (2003) and Andersson et al. (2003) Hald et al. (1996) T N E SW Barents Sea margin Hald and Aspeli (1997) and Ebbesen et al., in prep. JM N E N 141E North-western Barents Sea Sarnthein et al. (2003) margin BC 751N 141E MD N E W. Svalbard margin Hald et al. (2004) and Ebbesen et al. (2007)

4 3426 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Table 2 Radiocarbon datings, Vedde Ash horizons and calendar year calibrations for the proxy records on the Norwegian Barents Sea Svalbard margin, northern North Atlantic CORE ID Cm in core Lab ID Material dated 14 C age years BP 1s years DR years Max. cal. years BP 2s Min. cal. years BP 2s Mean cal. years BP References T79-51/2 2 TUa-1119 Yoldiella sp Hald and Hagen (1998) T79-51/2 60 TUa-948 Yoldiella sp T79-51/2 134 TUa-949 Yoldiella sp T79-51/2 148 TUa-950 Yoldiella sp T79-51/2 178 Tua-1705 Yoldiella sp T79-51/2 226 TUa-951 Yoldiella sp T79-51/2 252 TUa-952 Yoldiella lenticula T79-51/2 288 TUa-1121 Yoldiella sp T79-51/2 318 NSRL-2057 Nuculana sp JM97-948/2A KIA 6285 NPD Andersson et al. (2003) and Risebrobakken et al. (2003) JM97-948/2A KIA 4800 NPD MD Gif NPD MD KIA 3925 NPD MD KIA 5601 NPD MD KIA 3926 NPD MD KIA 6286 NPD MD KIA 6287 NPD MD Gif NPD MD KIA NPD MD KIA 463 NPD MD KIA 464 NPD MD Vedde Ash Tephra MD Tua 3316 NPS MD KIA 465 NPS T Gif Plankt forams Hald and Aspeli (1997) T Tua-3914 Plankt forams T Tua-3915 Plankt forams T Gif Plankt forams T Gif Plankt forams T Tua-116 Plankt forams T Tua-465 Plankt forams T Gif Plankt forams T Tua-464 Plankt forams T Vedde Ash Tephra Ebbesen et al., in prep. T Tua-466 Plankt forams Hald and Aspeli (1997) T Gif Plankt forams MD Tua-4421 Plankt forams Hald et al. (2004) MD Tua-3911 Plankt forams Ebbesen et al. (2007) MD Tua-3913 Plankt forams MD AA Shell fragm MD KIA9346 Shell fragm MD KIA9526 Shell fragm MD AA Shell fragm MD KIA9863 Benthic forams Box core KIA9191 NPS Sarnthein et al. (2003) KIA9192 NPS Kasten core 25 KIA7648 NPS KIA7649 NPS

5 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Table 2 (continued ) CORE ID Cm in core Lab ID Material dated 14 C age years BP 1s years DR years Max. cal. years BP 2s Min. cal. years BP 2s Mean cal. years BP References KIA7650 NPS KIA7651 NPS KIA11534 NPD KIA7653 NPS KIA7654 NPS KIA8553 NPS KIA11535 NPD KIA9193 NPS KIA8554 and NPS KIA KIA7657 NPS KIA7658 NPS KIA7659 NPS KIA8554 NPS KIA9354 NPS /03 5 Tua-759 NPD Klitgaard- Kristensen et al. (2001) 102 Tua-760 NPD Tua-761 NPD Tua-762 U. mediterranea Tua-1300 U. mediterranea Tua-1301 U. mediterranea Tua-759 U. mediterranea Tua-1302 U. mediterranea Tua-764 U. mediterranea TROLL Tua-387 U. mediterranea Tua-349 Yold. lenticulata Vedde Ash Tephra Tua 771 N. labradoricum For core locations and references confer Table 1 and Fig. 1. were included. In core between 400 and 1000 specimens were counted in order to reduce the statistical problem of almost mono-specific samples from subpolar and polar regions, where Neogloboquadrina pachyderma sinistral (s) may account for up to % of the total planktonic foraminiferal fauna (Sarnthein et al., 2003) Sea-surface temperature reconstructions In the present study, sea-surface temperatures (SST) have been reconstructed using transfer functions, statistical models to estimate past environmental conditions from the relationship between modern species distributions and their environment. Quantitative reconstructions of SST using transfer functions, based on various statistical approaches, have been applied for more than three decades (e.g. Imbrie and Kipp, 1971). Most methods for creating a transfer function are based on non-linear regression methods (Telford and Birks, 2005). Planktonic foraminifera are reliable tracers of surface and subsurface ocean temperatures (e.g. Be and Tolderlund, 1971). SST has been reconstructed using transfer functions from the planktonic foraminiferal data presented in the previous publications on which this paper is based (Table 1). However, these publications used substantially different statistical approaches, as well as different modern calibration data sets (for details see the references given in Table 1). To obtain a better basis for correlations, we have re-calculated SST using the modern calibration set from Pflaumann et al. (2003) and the same statistical approaches. This calibration set relates the planktonic foraminiferal fauna in the surface sediments to modern temperatures at 10 m depth. The statistical approaches used in the current paper are modern analogue technique (MAT), weighted averaging partial least squares (WAPLS), and maximum likelihood (ML). These methods are integrated in a computer program called C.2 version 1.3 (Juggins, 2002) as options. 4. Chronology and age-depth modelling The chronology of the sediment cores are based mainly on AMS 14 C dates discussed in detail in the original

6 3428 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) publications (Table 2). However, we have re-calibrated the dates using the Marine04 data base (Hughen et al., 2004) (Table 2). The DR in Table 2 shows the geographically dependant reservoir correction applied. Further, all dates that fall into the Younger Dryas (YD) period ( C years BP), corresponding to Greenland Stadial 1, GS-1 (Bjo rck et al., 1998), the marine reservoir age are assumed to be larger (Bondevik et al., 1999, 2006, Haflidason et al., 2000, Waelbroeck et al., 2001). For these dates a DR equal to years was applied. In addition to 14 C dates, two short box cores were dated by 210 Pb: core JM97-948/2A (Andersson et al., 2003) at the same location as MD , and core JM (M. Hald, unpublished) at the same location as core T88-2. The Vedde Ash tephra was identified and used as an age fix point in the cores 8903/TROLL 91-1, MD and T88-2. The earlier studies used a Vedde age of years BP based on the GRIP ice core chronology (Gro nvold et al., 1995). In this study we use the new Vedde age of years BP based on the North Grip Ice core chronology (Andersen et al., 2006). Age models for each proxy record were constructed using a linear interpolation between the dates. As 14 C-based fix points in the age model we used the calendar age represented by the mean for the 2s interval of highest probability. We adapt an informal use of chronostratigrahic zone names such as YD ( cal. years BP) and Preboreal (PB) ( cal. years BP) as well as the terms early Holocene ( to ca 8000 cal. years BP), middle Holocene (ca cal. years BP) and Late Holocene (ca 4000 cal. years BP recent). 5. Results 5.1. Planktonic foraminifera Thirteen species were identified in the sediment cores (Table 4). The planktonic foraminiferal fauna is dominated by the following three species, Neogloboquadrina pachyderma, the dextral (d) and sinistral (s) forms, Globigerina bulloides and Turborotalia quinqueloba (Figs. 2 5). Other species mainly occur with frequencies less than 1%. N. pachyderma (s) is a cold water indicator, dominant in Antarctic and Arctic seas, as well as Arctic and Polar Water in the North Atlantic (Be and Tolderlund, 1971; Johannessen et al., 1994; Pflaumann et al., 2003). Studies of d 18 O N. pachyderma (s) of indicate that this species lives in somewhat deeper water or reproduces in a colder season than N. pachyderma (d) (Johannessen et al., 1994). The percentage of this species dominated along the Norwegian Barents Svalbard margin during the YD PB transition (Fig. 2). N. pachyderma (s) rapidly declined in the North Sea and on the Norwegian margin and the southwestern Barents Sea after cal. years BP. In the two northernmost cores, in the north-western Barents Sea and the western Svalbard margin, a gradual decline started around cal. years BP (Fig. 2). In the North Sea and the north Norwegian margin the abundance of N. pachyderma (s) has remained relatively low during the last years. In contrast, the records on the middle Norwegian margin, the Barents Sea margin and the Svalbard margin displayed a marked increase of Fig. 2. Percent distribution of N. pachyderma (s) of total planktonic foraminiferal fauna vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. For location see Fig. 1 and Table 1.

7 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Fig. 3. Percent distribution of N. pachyderma (d) of total planktonic foraminiferal fauna vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. For location see Fig. 1 and Table 1. Fig. 4. Percent distribution of G. bulloides of total planktonic foraminiferal fauna vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. For location see Fig. 1 and Table 1. N. pachyderma (s) around 9000 cal. years BP. This was followed by a number of pronounced maxima and minima in the abundance of N. pachyderma (s) throughout the remaining Holocene. Some of these maxima and minima can be correlated between the proxy records, (e.g. the minima at ca 8000, 7000, 6000 and 2500 cal. years BP). N. pachyderma (d) is a warm water indicator of the Norwegian Barents Svalbard margin. It has it maximum

8 3430 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Fig. 5. Percent distribution of T. quinqueloba (s) of total planktonic foraminiferal fauna vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. For location see Fig. 1 and Table 1. distribution linked to the influx of temperate Atlantic Water in the south-eastern part of the northern North Atlantic and the relatively warm Irminger Current south and east of Iceland (Be and Tolderlund, 1971; Johannessen et al., 1994; Pflaumann et al., 2003). The species percentage had a marked rise after cal. years BP (Fig. 3). This was followed by an overall increasing trend in the cores in the North Sea, Norwegian margin and the south-western Barents Sea. Superimposed on the increasing trend are pronounced maxima and minima of which many can be correlated between the cores, in particular the minima around , 8000 and 3000 cal. years BP. In the two northernmost cores N. pachyderma (d) is rare during mid and late Holocene. G. bulloides is considered to be the most thermophilic species in the present foraminiferal dataset. Today its maximum distribution in the northernmost Atlantic is linked to the relatively temperate Atlantic waters of the Irminger Current southeast of Iceland where it reaches up to 60% (Pflaumann et al., 2003). Its peak distribution is in surface waters between 11 and 16 1C (Be and Tolderlund, 1971; Sautter and Thunell, 1991). The highest abundances of G. bulloides is in the two southernmost cores (Fig. 4) where it reached maximum abundances between and 9000 cal. years BP, followed by a broader maximum centred around 7000 cal. years BP. At these locations it is also evident that G. bulloides had an overall decline during the last 6000 years. Superimposed on this trend are abrupt maxima and minima with amplitude of maximum 10%. G. bulloides is generally rare on the continental margin off northern Norway, Barents Sea and Svalbard with maximum abundances during the early Holocene (Fig. 4.). T. quinqueloba is a subpolar species that is associated with the oceanic fronts especially the Arctic front separating the Atlantic and Arctic water masses in the central northern North Atlantic (Johannessen et al., 1994; Pflaumann et al., 2003) and western Barents Sea (Burhol, 1994). This species also responds rapidly to changes in nutrient supply (Reynolds and Thunell, 1985). A major rise of T. quinqueloba is observed at the YD PB transition (Fig. 5). The species percentage dominated the planktonic foraminiferal fauna during early Holocene in the three northernmost cores on Barents Sea and Svalbard margins. Peak values of 460% were reached between and 9000 cal. years BP, followed by an abrupt decline on the Svalbard margin and a stepwise decline on the Barents Sea margin. T. quinqueloba was less frequent on the Norwegian margin. The species abundance had a broad maximum during early middle Holocene on the Norwegian margin, whereas it was fairly stable (o20%) throughout the Holocene in the North Sea (Fig. 5). Comparing the largescale occurrence of T. quinqueloba along the core transect indicate a south to north time transgressive rise from to cal. years BP. This was followed by a north to south decline from 9000 to 4000 cal. years BP Temperature reconstructions Fig. 6 shows the reconstructed SSTs using the three different statistical approaches: MAT, WAPLS and ML.

9 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Fig. 6. Reconstructed sea surface summer temperatures (SST) vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. The SST reconstructions are based on planktonic foraminiferal transfer functions using various statistical approaches: ML ¼ maximum likelihood; MAT ¼ modern analogue technique; WAPL ¼ weighted averaging partial least square. For core location see Fig. 1 and Table 1. In general, the different reconstructions reveal similar trends, although the absolute values differ. The root mean squared error of prediction (RMSEP), a measure of the uncertainty in the predictions, and often used to select the best (in a statistical sense) transfer function model, is about 1 1C for MAT, 1.4 1C for WAPLS, and 1.6 1C for ML. These RMSEP estimates may be biased by spatial autocorrelation in the training set. Spatially close sites in the SST training set resemble each other more than randomly chosen sites; this lack of independence violates the assumptions of the statistical methods, and makes cross-validation estimates of the RMSEP overoptimistic (Telford and Birks, 2005). This problem is most severe for MAT, and to a lesser extent for WAPLS. ML is least affected. Below, we use the ML-based SST reconstructions (Fig. 7) The size fraction problem The various size fractions used for the foraminiferal census data add an uncertainty to the faunal interpretations and SST reconstructions. The subpolar species Turborotalia quinqueloba is usually smaller than 150 mm (Carstens and Wefer, 1992; Carstens et al., 1997), and therefore may be overrepresented in cores T88-2 and MD , with foraminiferal census data from the 4100 mm fraction, compared with the other cores using the larger fractions. Further, the calibration data set of Pflaumann et al. (2003) applied in this study contains census data in the fraction 4150 mm. To investigate this problem we compared the modern distribution of T. quinqueloba in the Pflaumann et al. (2003) data set to two other modern planktonic data sets, one with census data in size fractions 4100 mm (Burhol, 1994) and another with census data 4125 mm (Johannessen et al., 1994) (Fig. 8). In addition, we compared foraminiferal census data from the 150 mm-fraction and 100 mm-fraction from the same samples in core JM (Table 3). The modern data set of T. quinqueloba in the fraction 4100 mm is from the western Barents Sea (Fig. 8) and shows relative high values (20 40%) on the western shelf and continental slope (Burhol, 1994) (Fig. 8). Peak values (ca 40%) may be attributed to high production of this species close to the Arctic Front. The geographical overlap of this data set with that of Pflaumann et al. (2003) is restricted to nine samples on the continental slope off north-western Barents Sea, N. The frequencies of T. quinqueloba is about 20% larger in the 4100 mm data set. A similar trend is found in the central Norwegian Greenland Sea when comparing the 4125 mm data set (Johannessen et al., 1994) (Fig. 8) with the 4150 mm data set of Pflaumann et al. (2003). The percentage values are about 10% larger in the 4125 mm data set. To test the implication of this size-fraction related difference, we artificially increased T. quinqueloba with 20% in the census data in core and reduced all other species relatively.

10 3432 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) Fig. 7. Reconstructed sea surface summer temperatures (SST) vs. age (cal. years BP) in the sediment cores on the Norwegian Barents Sea Svalbard margin, northern North Atlantic. The SST reconstruction is based on a planktonic foraminiferal transfer function using the maximum likelihood statistical approach. A second order polynomial fit is plotted through SST data for core for the last cal. years. For core location see Fig. 1 and Table 1. The resulting ML SST curve showed almost a complete overlap to the original ML-reconstruction for SSTs 44 1C (Fig. 9), whereas for the colder intervals the artificial SSTs were between 1 and 2 1C warmer (Table 4). Comparing samples counted in the 4150 and 4100 mm fractions from five identical stratigraphic levels in core JM show some faunal differences, but no systematic trend (Table 3). The SST reconstructions are very similar, except from the lowermost samples that show lower temperatures in the 4100 mm fraction. In summary, we conclude that T. quinqueloba in the modern data sets in the northern North Atlantic is more frequent with 10 20% compared with the mm fractions. This may result in a slightly higher temperature for the colder (o4 1C) temperature end. However, further comparisons on this are needed, both from down core and modern data sets Holocene temperature history All records reveal a major temperature rise from the YD stadial into the PB interstadial (Fig. 7). The slight differences in timing of this transition probably arise from chronological uncertainty, partly linked to 14 C plateaus and changes in 14 C reservoir age (Bondevik et al., 1999, 2006; Waelbroeck et al., 2001). The largest amplitude of this temperature rise was 10 1C in the North Sea. This temperature shift occurred within less than 100 years, corresponding to an average annual warming about 0.1 1C/year. However, all absolute temperature values should be treated with caution. We consider this amplitude to be a maximum value since the modern calibration set (Pflaumann et al., 2003) relates the fauna to temperatures at 10 m water depth, while the foraminifera probably live somewhat deeper and hence cooler waters. It would have been interesting to test the Pflaumann et al. (2003) database with modern temperatures deeper than 10 m water depth. However, this is a large task and beyond the scope of the present paper. Further north, the YD/PB temperature transition occurred in a stepwise manner. The first step occurred at the end of the YD followed by a second temperature rise at the end of PB. A cooling between the warming steps occurred during the PB and is most pronounced on the northern Norwegian and SW Barents Sea margins. The two northernmost sites display a more gradual temperature rise starting at the end of the YD (core 23258) and middle PB (MD ) and peaking at ca cal. years BP. After the major temperature rise, the records in the North Sea and Norwegian margin stabilized from years BP. The North Sea record displayed remarkably stable SSTs to the present; whereas the Norwegian margin records show a weak warming trend of 2 1C over the last years. Core MD on the Norwegian margin displays a number of short temperature shifts superimposed on the general rising trend. The records on the Barents Sea and Svalbard margins reveal a more variable temperature pattern during the Holocene. The

11 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) early Holocene was characterized by relatively high temperatures, the duration of which diminishes northwards. On the SW Barents Sea margin (core T-88-2) a stable warm period occurred between and 8000 years BP. On the NW Barents Sea margin (core 23258) a temperature optimum occurred between and 9000 years BP, followed by a second optimum around 8000 years BP. On the Svalbard margin the early Holocene temperature optimum was restricted between and 9000 years BP. An overall temperature low occurred between 8000 and 3000 years BP in the three northernmost records. However, this period was interrupted by periods of warmer water, but again, the duration and/or frequency of these warmer periods diminished towards the north. A number of cooling events are recorded and some of them can be correlated between the records. However, we admit that both the age uncertainties as well as the variation in temporal resolution between the cores are limiting factors for correlation. A cooling period between 9000 and 8000 cal. years BP is shown in most of the cores, except T79-2 (northern Norway) and T-88-2 (Western Barents Sea). This cooling event also includes the 8,2-event. The latter can be observed in three of the cores: the North Sea, core MD on Norwegian margin and the Barents Sea (core 23258). Marked coolings Fig. 8. Modern surface sediment distribution of Turborotalia quinqueloba (A) Upper right: Western Barents Sea, samples counted in the size fraction 4100 mm (from Burhol, 1994); lower left: Norwegian Sea, samples counted in the size fraction 4125 mm (Johannessen et al., 1994) and (B) North Atlantic samples counted in the size fraction 4150 mm (Pflaumann et al., 2003). Fig. 9. Comparing SST reconstructed by the ML-statistical approach in core Red curve represents census data were T. quinqueloba was artificially increased with 20% on behalf of the other planktonic species. Black curve represents the original data set of Sarnthein et al. (2003). Table 3 Foraminiferal census data (%) in core JM , southern Barents Sea counted in the 100 mm and 150 mm fractions, respectively Depth in core cm NPS 100 mm NPS 150 mm NPD 100 mm NPD 150 mm T. quinq 100 mm T.quinq 150 mm G. bull 100 mm G.bull 100 mm G. uvula 100 mm G. uvula 150 mm SST (ML) 150 mm SST (ML) 100 mm For core location and reference confer Table 1 and Fig. 1.

12 3434 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) around 7500 cal. years BP and between 5500 and 3000 cal. years BP can be correlated between the three northernmost cores. The western Barents Sea cores (core T88-2 and 23258) also show a cooling around 6500 cal. years BP. The onset seems to be earlier and the duration longer for the cooling periods in the northern Barents Sea (core 23258) compared to the southern Barents Sea margin (core T88-2). Further, the onset and termination of the cal. BP cooling in the north can also be correlated to several short lasting cooling episodes on the Norwegian margin. This cold period is followed by a warming peaking around 2000 years. A secondary warm period, just after Table 4 Planktonic foraminiferal species identified in the Holocene proxy records on the Norwegian Barents Sea Svalbard margin, northern North Atlantic Berggrenia pumilio Globigerina bulloides Globigerina falconensis Globigerinella calida Globigerinita glutinata Globigerinita uvula Globorotalia hirsuta Globorotalia inflata Globorotalia scitula Neogloboquadrina pachyderma (d) Neogloboquadrina pachyderma (s) Orbulina universa Turborotalita humilis Turborotalita quinqueloba 1000 cal. BP, is recorded on the Norwegian margin (core MD ) and the Barents Sea margin (T88-2). Core MD also displays two abrupt cooling events of the recent past at around 400 and 100 cal. years BP, respectively. 6. Discussion The data and results present the paleoceanographic evolution of the upper water masses along the Norwegian Barents Sea Svalbard margins during the present interglacial, the Holocene. The location of the studied sediment cores are under the axis of the influx of Atlantic Water to the North Sea and along the continental margin off Norway and Svalbard, northern North Atlantic. Thus the proxy records reflect variations in temperature and extent of Atlantic Water during the Holocene in this area Early Holocene There was an abrupt transition from the cold YD and into the PB, with a major shift from a polar foraminiferal fauna and cool SST, to a subpolar fauna and temperatures similar or warmer than those of the present (Figs. 2 7). This transition is well known throughout the North Atlantic region from a number of studies of both marine and terrestrial records (e.g. Becker et al., 1991; Dansgaard et al., 1993; Koc et al., 1993). Based on the chronologies, it Fig. 10. Summer sea surface temperature vs. latitude (1N) for selected time-slices based on a planktonic foraminiferal transfer function using the maximum likelihood statistical approach (time slices from to 1000 cal. years BP) and instrumental observations, August 2006 and August 1986 (NOAA, 2007).

13 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) seems likely that the onset of this major warming occurred fairly simultaneously over the entire core transect (60 771N). However, the northernmost records display a more gradual change in both fauna and temperatures compared to the records further to the south. This may suggest a higher heat loss to the atmosphere or presence of a cold water pool linked to the remnants of the last glacial towards the north. During the transition form YD to PB the temperature gradient from 60 to 77 1N changed from close to 0 1C at the end of YD, cal. BP to 410 1C at cal. BP (Fig. 10). The latter gradient is close to the modern value (Fig. 10). The faunal shift at the YD/PB transition differs somewhat along the transect. In the North Sea the warm water indicator N. pachyderma (d) together with G. bulloides replaced the polar fauna that dominated during the YD. N. pachyderma (d) also dominated the early Holocene fauna along the Norwegian margin (Fig. 7), and clearly shows that Atlantic Water abruptly replaced Arctic Water at this time. However, in the three northernmost records the YD/PB faunal shift was characterized by a major decline in N. pachyderma (s) and a corresponding rise and dominance of T. quinqueloba. The slightly higher frequencies of T. quinqueloba in the cores T88-2 and MD , compared to core 23258, are probably due to the difference in the size fractions of the faunal samples studied (cf. discussion on page 9 10). A modern analogue to this T. quinqueloba dominated fauna can be found in the central Norwegian Sea (Johannessen et al., 1994) at the Arctic Oceanic Front (AOF), separating Arctic and Atlantic water masses in this area. This suggests that a strong frontal system prevailed during the early Holocene in the north. The apparent time-transgressive occurrence of T. quinqueloba may indicate a gradual northward displacement and strengthening of the AOF during the early Holocene from cal. years BP. The planktonic fauna in the northernmost cores during early Holocene ( cal BP) is also associated with the more thermophiclic species N. pachyderma (d) (Fig. 3) and G. bulloides (Fig. 4). The presence of these species together with T. quinqueloba explains the relatively high SST reconstructed for this period. The warm conditions in the north are also reflected by the low-temperature gradient between 60 and 77.41N at and 9000 cal. year BP of about 4 1C, compared to the present gradient of about 8 1C (Fig. 10). This corresponds to a gradient of C/1N compared to the present SST gradient of C/1N. The 8.2 kyr BP event falls within a longer cooling period described by Rohling and Pälike (2005). It is reflected by a percentage increase of the polar species N. pachyderma (s) and a corresponding reduction of T. quinqueloba and G. bulloides (Figs. 2, 4 and 5), causing a temperature shift of o2 1C. On the Norwegian margin the 8.2 k event is also reflected as a short lived d 18 O increase in N. pachyderma (s), corresponding to a cooling of about 3 1C (Risebrobakken et al., 2003) Middle and late Holocene Although there are some significant shifts in the frequencies of the various planktonic species, the overall temperature trend during middle and late Holocene reveals a rather stable pattern in the three southernmost cores. This is in particular the case for the two most near-shore records, the North Sea record core 8903 and T79-51/2 off Northern Norway. This stability, despite the faunal variability, can be explained by the fact that very often a percentage reduction in one subpolar is replaced by an increase of another subpolar species. For example, the reduction in N. pachyderma (d) around cal. BP in the North Sea and Norwegian margin records (Fig. 3) coincides with a rise in G. bulloides (Fig. 4). These apparently simultaneous changes over long distances (60 691N) suggest regional oceanographic changes affecting the planktonic marine ecosystems at this time. Rather than temperature, these shifts may be linked to other factors, e.g. biogenic productivity and nutrients (Knies et al., 2003). The three northernmost records from the Barents Sea and Svalbard margin display a larger faunal and temperature variability compared to the three southernmost records. The warm conditions during the early Holocene, with SST higher than those of the present, were followed by a number of multi centennial periods during the middle and late Holocene with cooler conditions. The cooling events ; around 7500 BP, around 6500, and around 1000 cal. years BP in the three northernmost records, can be correlated to the Norwegian margin (MD ), and thus indicate that these changes were regional events. We suggest that the cooling periods were caused by increased influence of Arctic Water. In the northernmost cores sited this could be forced by southward displacement of Arctic Water from the north-eastern Barents Sea, perhaps due to less heat advection from the south. Today, the western Barents Sea and Svalbard waters are characterized by sharp gradients between cold Arctic Water and warm Atlantic Water. Relatively small southward movements of these oceanic fronts may explain the shifts between the cooler and warmer periods. An enhanced displacement of colder Arctic Water starting in the north and moving southwards could also explain the fact that the onset was earlier and duration longer for the cooling periods further to the north (Fig. 7). The overall cooling during middle and late Holocene in the north resulted in an increasing S N SST gradient reaching a maximum of close to 12 1C at 5000 cal. years BP (Fig. 10). 7. Correlations 7.1. Long term change and forcing factors The overall temperature trend in the three northernmost cores (Fig. 7) to some degree follows the pattern of orbital forcing of insolation until about 3000 cal. years BP. This is

14 3436 ARTICLE IN PRESS M. Hald et al. / Quaternary Science Reviews 26 (2007) indicated by the second-order polynomial fit plotted through the SST data in core (Fig. 7). The total insolation was higher around cal. years BP at 60 1N compared to the present (Berger and Loutre, 1991). Summer insolation (June) at 601N was about 10% higher than today, while only slightly lower during winter (December). The high summer insolation gradually declined towards the present. The link between climate records in the North Atlantic region and this orbital forcing is well known (e.g. Koc et al., 1993; Kaufman et al., 2004). In contrast, the temperature trend in the three southernmost cores in our study does not follow the trend of the summer insolation. Rather the temperature curve displays a stable or weakly increasing trend (Fig. 7). Since planktonic foraminifera live in the upper ca 200 m of the water column their estimated SST may differ from true surface temperatures. This difference can be attributed to either depth stratification and/or seasonality. Depth stratification will reflect the temperature gradient in the uppermost water column. At location T51-1/2, northern Norwegian margin, SST is about 31 C lower at 100 m compared to the surface (Hald and Hagen, 1998). Another factor is that the various planktonic species may grow at different times, thus reflecting different seasonal temperatures (e.g. Field, 2004). Parallel measurements of d 18 Oon N. pachyderma (s) and N. pachyderma (d) in two of the cores in this study (T79-51/2 and MD ) (Hald and Hagen, 1998; Risebrobakken et al., 2003) reveal an oxygen isotopic offset between the two subspecies of ca %, reflecting a temperature difference of up to 5 1C. This difference can be attributed either to seasonality or depth stratification of the species. These factors may also vary geographically and seem to be less pronounced under the influence of Arctic Waters (Johannessen et al., 1994). This may explain why the northern sites being closer to Arctic Water, appear to be better related to true SST compared to the sites further south. Holocene SST has been reconstructed from core MD with four proxies: radiolarians (Dolven et al., 2002), diatoms (Birks and Koc, 2002; Andersen et al., 2004), alkenones (Calvo et al., 2002), and planktonic foraminifera (Andersson et al., 2003; Risebrobakken et al., 2003). All four proxies display the marked warming at the YD/PB transition, although with somewhat different amplitudes. The trends differ for the remaining of the Holocene. Diatom and alkenone SSTs display a marked Holocene climatic optimum, and , respectively, followed by a stepwise cooling during middle and late Holocene. Both proxies are believed to mainly reflect the uppermost layers of the ocean, compared with the somewhat deeper dwelling radiolarians (ca 50 m) (Dolven et al., 2002) and planktonic foraminifera (o200 m). SST based on the radiolarians display much the same stable trend as the planktonic foraminifera SST (Fig. 7). Risebrobakken et al. (2003) suggested that the apparent lack of a pronounced Early Holocene warming optimum on the middle Norwegian margin (core MD ) was due to increased influence on the planktonic fauna from subsurface Arctic Water driven by increased westerlies. Further, the subsequent slight warming during middle and late Holocene in this record was an effect of reduced westerlies. Another possible cause for the lack of a Holocene optimum may be that the foraminifera calcified beneath the summer thermocline where the stronger summer insolation created warm waters near the surface and stabilised the surface layer. Thus proxies derived from the near surface reacted to the insolation anomaly, whereas proxies derived from below, did not register the insolation anomaly (see discussion in Jansen et al. in print). However, for records in the North Sea (core 8903) and off North Norway (T51-1/2) a subsurface stratification of the planktonic fauna may also be related to increased freshwater flux from land linked to final deglaciation and increased precipitation in western Scandinavia during early Holocene (e.g. Nesje et al., 2000; Bakke et al., 2005). This may have enhanced the Coastal Current and subsequently forced the planktonic fauna into deeper and slightly cooler and saltier water. Both increased westerlies and increased precipitation can be related to a predominant high North Atlantic Oscillation (NAO) weather mode. NAO is associated with changes in the westerlies in the North Atlantic and NW Europe (Hurrell, 1995; Hurrell and Van Loon, 1997). Another factor that may have contributed to the north to south difference in SST pattern may be a weakening in the THC. In a climate modelling experiment by Renssen et al. (2005) the Holocene climate evolution in the northern North Atlantic was simulated using a coupled atmospheresea ice-ocean vegetation model. Simulated mean annual SST during the last 9000 years show a strong cooling (3 6 1C) off Svalbard and in the western Barents Sea, whereas the cooling further south is much smaller (1 1C or less). The cooling is explained by a weakening the THC combined with reduced insolation. The polar amplification of this cooling may be related to internal feedbacks to the overall cooling linked to the well known snow/ice albedo positive feedback (Renssen et al., 2005). This could involve increase in the sea ice distribution and the snow cover leading to higher surface albedo and thus to further cooling (Kerwin et al., 1999; Crucifix et al., 2002) Abrupt climate change and land-ocean correlations Abrupt climatic shifts are characterized by significant shift in SST occurring within a few centuries or less that are superimposed on the general trends. Some of these changes may lead into colder or warmer periods lasting less than a hundred years to more than a thousand years. The rate of the abrupt warming at the YD/PB transition, measured about 0.1 1C/year in the North Sea record (Fig. 7), is much larger than the rate of the recent global warming. The average global warming of the atmosphere is estimated to be ca 1 1C during the last 100 years (corresponding to an