Early Holocene temperature variability in the Nordic Seas: The role of oceanic heat advection versus changes in orbital forcing



[1] The separate roles of oceanic heat advection and orbital forcing on influencing early Holocene temperature variability in the eastern Nordic Seas is investigated. The effect of changing orbital forcing on the ocean temperatures is tested using the 1DICE model, and the 1DICE results are compared with new and previously published temperature reconstructions from a transect of five cores located underneath the pathway of Atlantic water, from the Faroe-Shetland Channel in the south to the Barents Sea in the north. The stronger early Holocene summer insolation at high northern latitudes increased the summer mixed layer temperatures, however, ocean temperatures underneath the summer mixed layer did not increase significantly. The absolute maximum in summer mixed layer temperatures occurred between 9 and 6 ka BP, representing the Holocene Thermal Maximum in the eastern Nordic Seas. In contrast, maximum in northward oceanic heat transport through the Norwegian Atlantic Current occurred approximately 10 ka BP. The maximum in oceanic heat transport at 10 ka BP occurred due to a major reorganization of the Atlantic Ocean circulation, entailing strong and deep rejuvenation of the Atlantic Meridional Overturning Circulation, combined with changes in the North Atlantic gyre dynamic causing enhanced transport of heat and salt into the Nordic Seas.

1. Introduction

[2] The Holocene Thermal Maximum (HTM) is seen throughout the Northern Hemisphere, in marine and terrestrial reconstructions, and is considered a response to the early Holocene orbital forcing [Renssen et al., 2009]. The timing and magnitude of the HTM are influenced by local and regional feedback mechanisms, e.g., from the atmosphere, ocean, sea ice and vegetation, as well as the impact from disintegrating ice sheets [Crucifix et al., 2002; Helmens et al., 2007; Kaufman et al., 2004]. Traditionally, all reconstructions of warm early to mid-Holocene temperatures at high northern latitudes have been related to the orbital forcing; however, we will show that in the case of the Nordic Seas this view represents an oversimplification.

[3] In the Nordic Seas and surrounding areas, the HTM has been found in a variety of climate archives. Sea surface temperature (SST) reconstructions based on alkenones and diatoms show warmer than present temperatures in the eastern Nordic Seas approximately 9–6 ka BP [Birks and Koç, 2002; Calvo et al., 2002; Kim et al., 2004; Koç et al., 1993; Rimbu et al., 2004; Risebrobakken et al., 2010]. Thermophilous marine mollusks from raised beach deposits at Spitsbergen show maximum temperatures ∼10–9 ka BP [Salvigsen et al., 1992]. Seppä et al. [2009] determined Holocene temperatures based on pollen records from 36 sites from two northern European transects. One reached from 57°N (southern Fennoscandia) to 70°N (boreal-arctic boundary in northern Fennoscandia), showing the HTM 8–6 ka BP. Contemporary with the warmest Scandinavian temperatures, the Norwegian glaciers reached their Holocene minimum positions 9–4 ka BP [e.g., Nesje, 2009]. At west Spitsbergen plant macrofossils indicate warmer than present temperatures 9–4 ka BP [Birks, 1991], and local glaciers disappeared 11–5 ka BP [Svendsen and Mangerud, 1997]. At Greenland, borehole temperatures and δ18O show a decreasing trend during the Holocene, with the warmest conditions 8–5 and 10–7 ka BP, respectively [Dahl-Jensen et al., 1998; Vinther et al., 2009].

[4] The occurrence of warm early Holocene surface water in the Nordic Seas has been associated with increased northward advection of Atlantic water [Kaufman et al., 2004; Koç et al., 1993]. This view has recently been challenged. Warmer than present temperatures are seen in diatom and alkenone-based SST reconstructions at the Vøring Plateau [Birks and Koç, 2002; Calvo et al., 2002]; however, at the same site there is no evidence for an HTM in temperature reconstructions based on foraminifera and radiolarians [Cortese et al., 2005; Dolven et al., 2002; Risebrobakken et al., 2003]. Jansen et al. [2008] argue that the lack of an HTM in the temperature reconstructions based on foraminifera and radiolarians rules out increased advection as an explanation for the HTM, and that the HTM is solely the response to the radiative forcing due to the orbital configuration at the time. This view has been supported by a comparison of foraminifera-based temperature reconstructions from the Nordic Seas and the North Atlantic [Andersson et al., 2010]. Andersson et al. [2010] show that foraminiferal-based temperature reconstructions, not just from the Vøring Plateau but also from the North Atlantic, record increasing temperatures through the Holocene, contrasting the decreasing summer radiation. They also show that their results are consistent with modeled subsurface mid-Holocene temperatures in the same region [Andersson et al., 2010].

[5] In contrast to the evidence from the North Atlantic and the Vøring Plateau, an early and short-lasting HTM has been reported in temperature reconstructions based on planktic foraminifera from NE Nordic Seas [e.g., Hald et al., 2004, 2007; Sarnthein et al., 2003]. This foraminiferal-based HTM is seen earlier than the HTM recorded by alkenones [Kim et al., 2004; Marchal et al., 2002; Risebrobakken et al., 2010]. Maximum advection of warm Atlantic water and polar amplification of the orbital forcing have been suggested to explain the strong early Holocene temperature signal in the NE Nordic Seas [Hald et al., 2004, 2007; Sarnthein et al., 2003]. These explanations, however, do not account for the discrepancy in timing between different proxies in the same region or the latitudinal discrepancy in timing.

[6] Hence, the cause of the discrepancy in timing of the early Holocene temperature development in the Nordic Seas, between different proxies and with latitude for the same proxies, is still unclear and details unresolved. Therefore it is important to resolve why the proxies disagree, in order to improve our understanding of how different forcing and dynamics, and the interaction between these, influenced the Nordic Seas and surrounding areas during the early Holocene.

[7] We hypothesis that the strong early Holocene summer insolation at high northern latitudes only influenced the summer mixed layer (SML) temperatures in the Nordic Seas, and that these changes cannot be seen as representative for the mean state of the Norwegian Atlantic Current (NwAC), reflecting the oceanic heat advection through the eastern Nordic Seas. Following this hypothesis, the separate roles of orbital forcing and oceanic heat advection must be understood, and the cause of variations in heat advection to the Nordic Seas during the early Holocene must be investigated. Accordingly, we will argue that by separating the response to changes in orbital forcing from the response to variable oceanic heat advection, the discrepancies between early Holocene temperatures in different proxies and at different locations in the Nordic Seas can be explained. Here we want to test this hypothesis by 1) using the 1DICE column model to test the effect of the early Holocene orbital configuration on the Barents Sea temperature, 2) reconstructing SML temperatures by compiling existing alkenone data and 3) reconstructing the mean state of the NwAC using proxies based on foraminifera. By testing this hypothesis, the link between the southeastern and northeastern Nordic Seas, and the seemingly inconsistent temperature reconstructions, is clarified. Understanding the relative importance of the different forcing and dynamics during HTM also have implications for the discussion on how increased radiative forcing is predicted to cause a future polar amplification and how polar amplification will affect high northern latitudes.

[8] We define HTM as directly related to the early Holocene orbital configuration, entailing stronger-than-present radiative forcing at high northern latitudes that again caused a temperature increase. Hence, in this study the term HTM will further be used only when insolation can be considered the responsible driver behind the recorded temperatures. The main focus of this paper will be on the time period 12–6 ka BP; however, full Holocene records are presented to provide a background reference.

2. Oceanographic Setting and Core Locations

[9] Warm and saline Atlantic water enters the Nordic Seas between Iceland and the Faroe Islands and through the Faroe-Shetland Channel [Orvik and Niiler, 2002]. This warm Atlantic water flows northward in two branches, where the eastern branch of the NwAC follows the shelf edge along Norway (Figure 1). The NwAC bifurcates when it reaches the Barents Sea opening, continuing in the West Spitsbergen Current (WSC) and the North Cape Current (NCaC). The WSC continues northward through the Fram Strait where it submerges underneath Arctic water and continues into the Arctic Ocean. Parts of this subsurface Atlantic water enter the Barents Sea from the Arctic through the Franz-Victoria Trough. The NCaC enters the Barents Sea through the Barents Sea opening (Figure 1). Polar water enters the Nordic Seas from the Arctic Ocean and moves southwards in the East Greenland Current (EGC). There is a strong topographical steering of the main currents in the Nordic Seas [Orvik and Niiler, 2002]. Arctic water, representing a mix of the Polar and Atlantic water, occupies central parts of the Nordic Seas. Atlantic water is separated from Arctic water found further west in the Nordic Seas and northeastward in the Barents Sea by the Arctic front (in the Barents Sea also known as the Polar front [Pfirman et al., 1994]). In the Nordic Seas, the Polar front separates the Polar and Arctic waters.

Figure 1.

(a) Map showing the location of the discussed core sites. The present-day pathway of Atlantic water through the eastern Nordic Seas and the pathway of Polar water in the western Nordic Seas are indicated. The black line indicates a schematic outline of the southern Barents Sea column used in the 1DICE model experiments. NwAC = Norwegian Atlantic Current. NCaC = North Cape Current. WSC = West Spitsbergen Current. FWT = Franz Victoria Trough. EGC = East Greenland Current. OWSM = Ocean Weather Station Mike. The light purple, stippled line indicates the location of the Nordic Seas Arctic front (NSAF) and the Barents Sea Arctic front (BSAF) (also called Polar front), representing the transition zone between Atlantic and Arctic water masses. The light blue, stippled line indicates the Polar front (PF) that represent the transition zone between Polar and Arctic waters in the western Nordic Seas. (b) A schematic picture of the cross-section showing the water mass structure from the Faroe-Shetland Channel (MD99-2284), past the Vøring Plateau (MD95-2011), the Barents Sea margin (M23258) through the Fram Strait toward the Franz-Victoria Trough (PSh-5157) at present and at 10 ka BP. The arrows indicate submergence of Atlantic water underneath Arctic water.

[10] Atlantic water (7.0–9.5°C; 35.1–35.3 psu) reaches down to a water depth of 400–700 m along the Norwegian and Barents Sea margins. Arctic intermediate water (from 400 to 600 to approximately 1100 m water depth) is seen as a salinity minimum zone (34.7–34.9 psu; −0.5–0.5°C) between the Atlantic water and the deep bottom water (34.91 psu; <0.5°C) [Blindheim and Østerhus, 2005]. The winter mixed layer (WML) reaches down to the interface between Atlantic and intermediate water and is characterized by temperatures around 6°C while the SML reach down to 20–40 m water depth and is characterized by temperatures around 12°C at OWSM (Figure 1) [Nilsen and Falck, 2006]. The interannual variability of the mixed layer temperatures as given from the instrumental time series at OWSM is greater during the summer months, ±1.5°C compared to ±1°C in winter. Furthermore, the depth of the SML is more stable then the depth of the WML [Nilsen and Falck, 2006]. Variations in the depth of the WML can be caused by the dynamics of the NwAC related to variable residence time, and degree of convection and entrainment. The NwAC also provide a constant source of heat to the WML, therefore the WML temperature is strongly influenced by the variability of the advective heat flux of the NwAC [Nilsen and Falck, 2006]. Most of the incoming radiation is absorbed in the upper approximately 20–40 m of the water column and the warmer temperature of the SML primarily results from atmospheric heating due to stronger insolation during summer than winter [Nilsen and Falck, 2006]. The thermocline gradually diminishes as mixing occurs during autumn and winter, and the annual mean temperature profiles will normally be comparable to the winter/spring situation. The heat loss from the Atlantic water to the atmosphere increases northward; instrumental records show a decrease in WML temperature by 3–4°C from 63°N to 76°N [Skagseth et al., 2008].

[11] In this paper we will present data from five sites from the eastern Nordic Seas and the Barents Sea: 1) MD99-2284 (Faroe-Shetland Channel, 62°22.48N, 0°58.81W, 1500 m water depth), 2) MD95-2011 (Vøring Plateau, 66°58.19N, 7°38.36E, 1048 m water depth), 3) PSh-5159N (SW Barents Sea, 71°21.80N, 22°38.77E, 422 m water depth), 4) M23258 (Barents Sea margin, 75°N, 14°E, 1768 m water depth) and 5) PSh-5157 (Franz-Victoria Trough, N Barents Sea, 78°55.48N, 41°52.97E, 461 m water depth) (Figure 1). All sites are located underneath the pathway of Atlantic water flowing through the NwAC, the WSC and the NCaC from the southernmost to the northernmost parts of the eastern Nordic Seas, including the inflow to the Barents Sea through the Barents Sea opening and through the Franz Victoria Trough. In the Faroe-Shetland Channel, at the Vøring Plateau and at the Barents Sea margin Atlantic water occupies the upper 400–700 m of the water column, overlaying the Norwegian Sea Arctic intermediate water [Blindheim and Østerhus, 2005]. Deep bottom water characterizes the water column bellow 1100 m. In SW Barents Sea, Atlantic water occupies the full water column, except for the upper 20–30 m where a mixture of costal and Atlantic water is found [Risebrobakken et al., 2010]. In the Northern Barents Sea, Arctic water occupies the upper 150–200 m of the water column, overlaying Atlantic derived water, and cold bottom water below 300 m. During summer a warm, fresh surface water layer forms in the upper approximately 20 m [Pfirman et al., 1994].

3. Methods

[12] The 1DICE column model was developed and described by Björk [1989]. Here, the southern Barents Sea model setup from Smedsrud et al. [2010] is used. The S Barents Sea area was defined as the sea between northern Norway (71°N) and Bear Island (75°N), and eastward from the Barents Sea margin (20°E) to Novaya Zemlya (Figure 1). The present-day setup of the model has been used; however, the insolation forcing was changed, using horizontally averaged monthly mean insolation at 71°N for the time slices 12, 11, 10, 9, 8, 7, 6 and 0 ka BP [Laskar et al., 2004]. A planetary albedo of 40% was used to account for the effect of clouds [Hartmann, 1994]. In 1DICE, a stable annual cycle is usually established within 3 years. For each time slice, the model was therefore run over 10 years with daily time steps to verify that a stable yearly cycle was established. Each run was initiated in August using horizontally averaged profiles of the summer temperature and salinity of the southern Barents Sea, based on available stations in work by Nilsen et al. [2008]. The 1DICE model reproduces the major features of the observed southern Barents Sea annual cycle [Smedsrud et al., 2010].

[13] New planktic foraminiferal census counts were done at the size fraction >150 μm in 276 samples from MD99-2284 (Faroe-Shetland Channel) and in 130 samples from MD95-2011 (Vøring Plateau), and relative abundances were calculated. The new counts from MD95-2011 were done to optimize the early Holocene resolution (every cm 12–7 ka BP), compared to the previously published early Holocene assemblage data from MD95-2011 (every 5 cm) [Risebrobakken et al., 2003]. The relative abundance of G. bulloides in PSh-5159N (SW Barents Sea) is presented for the first time. The samples from PSh-5157 (Franz Victoria Trough) did not contain enough foraminifera to calculate relative abundance. In addition to these new faunal data, previously published relative abundance data from MD95-2011 (supplemented by the new counts) [Andersson et al., 2003; Risebrobakken et al., 2003], M23258 (Barents Sea margin) [Sarnthein et al., 2003] and PSh-5159N (SW Barents Sea) [Chistyakova et al., 2010; Risebrobakken et al., 2010] is presented. In PSh-5159N the census counts are done on the size fraction >100 μm, not at >150 μm.

[14] The Maximum Likelihood (ML) transfer function method [ter Braak and Prentice, 1988; ter Braak and van Dam, 1989] was used to calculate foraminiferal temperatures. This was done for all cores to ensure that all presented foraminiferal temperature records have been calculated in exactly the same manner. The Maximum Likelihood method assumes a unimodal species-environment response model [Telford and Birks, 2005]. The composition of the main species and how each species may influence the temperature estimate is considered when interpreting the results. Furthermore, the counts done at 100 μm (PSh-5159N) may cause temperatures to warm in the cold end, as T. quinqueloba is overrepresented at 100 μm compared to 150 μm and the training set used is for 150 μm [Hald et al., 2007; Pflaumann et al., 2003].

[15] New δ18O and δ13C measurements performed on Neogloboquadrina pachyderma (sin) (241 samples) and Cibicides wuellerstorfi δ13C (134 samples) from MD99-2284 are presented, as well as N. pachyderma (sin) δ18O from PSh-5157 (216 measurements). All measurements were done in Bergen, using a Finnigan MAT 253 mass spectrometer equipped with an automatic preparation line (“Kiel device”). All foraminifera were crushed and cleaned in methanol, using an ultrasonic bath, before being measured. The foraminifera used for the stable isotope measurements were picked from the 150–500 μm and 106–500 μm fractions in MD99-2284 and PSh-5157, respectively, and the measurements was done on 4–8 N. pachyderma (sin) and >2 C. wuellerstorfi. The smallest N. pachyderma (sin) were avoided when picking from the PSh-5157 samples, as the smaller specimens might cause slightly higher δ18O [Hillaire-Marcel et al., 2004]. The PSh-5157 record is considered reliable even though a minor size effect cannot be excluded. The N. pachyderma (sin) δ13C record from MD95-2011 corresponds to the MD95-2011 N. pachyderma (sin) δ18O record published by Risebrobakken et al. [2003]. In addition, we present previously published N. pachyderma (sin) δ18O from MD95-2011 [Risebrobakken et al., 2003], and N. pachyderma (sin) δ18O and δ13C records from PSh-5159N [Risebrobakken et al., 2010] and from M23258 [Sarnthein et al., 2003]. All new and previously published δ18O records were corrected for the ice volume effect according to Fairbanks [1989]. The presented δ18O records hence reflect a combined temperature and salinity signal.

[16] Further, we have compiled previously published U37K and U37K′ SST records from the sites MD95-2011 [Calvo et al., 2002], M23258 [Marchal et al., 2002; Martrat et al., 2003; J.-H. Kim and R. R. Schneider, GHOST global database for alkenone-derived Holocene sea-surface temperature records, 2004,] and PSh-5159N [Risebrobakken et al., 2010]. The alkenone SST records are used as they were originally published. Accordingly, the calibration equation used to calculate SST follow Prahl and Wakeham [1987] for MD95-2011 (U37K), Rosell-Melé et al. [1995] for M23258 (U37K′) and Müller et al. [1998] for PSh-5159N (U37K′). Further details on the methodology and choice of calibration equation behind each U37K and U37K′ SST records can be found in the original publications [Calvo et al., 2002; Martrat et al., 2003; Risebrobakken et al., 2010]. In PSh-5157, there were not enough alkenones to get any signal. No alkenone work has been done on samples from MD99-2284.

[17] The age models of MD99-2284 and PSh-5157 are presented here for the first time. Additional radiocarbon dates are added to the original age model of MD95-2011 [Risebrobakken et al., 2003]. The age models of M23258 and PSh-5159N are based on radiocarbon dates presented by Sarnthein et al. [2003] and Risebrobakken et al. [2010], respectively. New age models were calculated for all cores to ensure a common chronological framework. The radiocarbon dates was calibrated using Calib 6.0.0 [Stuiver and Reimer, 1993] and the Marine04 calibration data set [Hughen et al., 2004]. For the northern cores (M23258, PSh-5159N and PSh-5157) ΔR = 71 ± 21 was used [Mangerud et al., 2006], while ΔR = 0 ± 0 was used for MD99-2284 and MD95-2011. ΔR = 200 ± 50 was used for all radiocarbon ages within the 11–10 14C ka interval [Hald et al., 2007]. Vedde Ash and Saksurnavatn Ash ages from Rasmussen et al. [2006] were used when these horizons were identified in the sediment. The age models are based on linear interpolation between the tie points given by the calibrated radiocarbon dates and the ash layers (Table 1). Ten dates have been omitted as they gave inverted ages, probably due to influence of resedimented material. Primarily, the dates that provide inverted ages are from the younger parts of the cores, hence, all chronologies are considered reliable for the focus interval of this study.

Table 1. Radiocarbon Dates, Ash Horizons and Calibrated Ages Used to Create the Age Modelsa
Lab IDCoreSample Depth (cm)Dated Material14C DateΔRCalibrated Age Range ±1σRel. Prob.bCalendar Age BP 1950 (Med. Prob.c)CommentReferences for Individual Dates
  • a

    NPD = N. pachyderma (dex), NPS = N. pachyderma (sin).

  • b

    Relative probability.

  • c

    Median probability.

Poz-20001PSh-51577,5Bulk foraminifera1460 ± 16071 ± 21767–11011942Not usedThis study
 PSh-515711,5Bulk foraminifera1170 ± 3071 ± 21629–6871659 This study
Poz-15134PSh-515715,5Mollusc7490 ± 40/7540 ± 5071 ± 217837–7932/7868–798217883/7930Not usedThis study
Poz-20002PSh-515731Bulk foraminifera3305 ± 3071 ± 212988–312313055Not usedThis study
Poz-36197PSh-515736.5Bulk foraminifera3135 ± 3571 ± 212767–287412830 This study
Poz-20003PSh-515761Bulk foraminifera3425 ± 3571 ± 213163–329213223 This study
Poz-20004PSh-515779Bulk foraminifera4180 ± 3571 ± 214090–422414163 This study
Poz-36198PSh-5157100.5Bulk foraminifera5240 ± 5071 ± 215472–558015528 This study
Poz-15136PSh-5157126Snail/mollusc6380 ± 4071 ± 216709–683616769 This study
Poz-12699PSh-5157143Yoldia6820 ± 4071 ± 217228–733217280 This study
Poz-15137PSh-5157168,5Mollusc7360 ± 4071 ± 217690–780217751 This study
Poz-15138PSh-5157220,5Mollusc9000 ± 5071 ± 219517–965119590 This study
Poz-15130PSh-5159N7.5Mollusc fragments, benthic foraminifera102.46_0.32pMC71 ± 21    Ivanova et al. [2008]
Poz-20399PSh-5159R14.17Lenticulina sp.635 ± 3071 ± 21174–2580.788197 Ivanova et al. [2008]
Poz-19995PSh-5159N21.5Bulk foraminifera1670 ± 3071 ± 211118–122311164 Ivanova et al. [2008]
Poz-19997PSh-5159N40.5Bulk foraminifera2845 ± 3071 ± 212426–260312508 Risebrobakken et al. [2010]
Poz-20568PSh-5159N45.5Bulk foraminifera4960 ± 4071 ± 215132–528715204 Risebrobakken et al. [2010]
Poz-15131PSh-5159N50.5Mollusc fragments6105 ± 3571 ± 216393–650016451 Risebrobakken et al. [2010]
Poz-19998PSh-5159N60.5Bulk foraminifera7040 ± 4071 ± 217423–750717472 Risebrobakken et al. [2010]
Poz-12701PSh-5159N69.5Brachiopod7500 ± 4071 ± 217844–793917892 Risebrobakken et al. [2010]
Poz-19999PSh-5159N86.5Bulk foraminifera8550 ± 5071 ± 219010–917119103 Risebrobakken et al. [2010]
Poz-15132PSh-5159N99.5Mollusc fragments, benthic foraminifera, ostracode9700 ± 5071 ± 2110444–10556110496 Risebrobakken et al. [2010]
Poz-19991PSh-5159R122.5Mollusc10010 ± 5071 ± 2110565–10779110693 Risebrobakken et al. [2010]; Chistyakova et al. [2010]
Poz-15133PSh-5159N133.5Mollusc fragments10290 ± 5071 ± 2111029–11196111100 Risebrobakken et al. [2010]
Poz-12629PSh-5159N148.5Astarte crenata10360 ± 5071 ± 2111118–11217111165 Risebrobakken et al. [2010]
Poz-16594PSh-5159R241Bulk benthic foraminifera12150 ± 7071 ± 2113405–13606113515 Risebrobakken et al. [2010]; Chistyakova et al. [2010]
Poz-19992PSh-5159R333Bulk benthic foraminifera13550 ± 7071 ± 2115486–162270.96115813 Risebrobakken et al. [2010]; Chistyakova et al. [2010]
KIA7648M2325825NPS1165 ± 3571 ± 21622–6891655 Sarnthein et al. [2003]
KIA7649M2325851NPS2555 ± 3071 ± 212071–221012145 Sarnthein et al. [2003]
KIA7650M2325867NPS3500 ± 3571 ± 213255–335813307 Sarnthein et al. [2003]
KIA7651M2325893NPS4825 ± 4071 ± 214889–506715002 Sarnthein et al. [2003]
KIA11534M23258118NPD6140 ± 7071 ± 216404–658116495 Sarnthein et al. [2003]
KIA7653M23258154NPS7660 ± 4571 ± 217988–811218050 Sarnthein et al. [2003]
KIA7654M23258177NPS8380 ± 4571 ± 218796–896918872 Sarnthein et al. [2003]
KIA8553M23258192NPS8760 ± 4071 ± 219312–942619368 Sarnthein et al. [2003]
KIA11535M23258207NPD8955 ± 5571 ± 219473–959919541 Sarnthein et al. [2003]
KIA9193M23258241NPS9330 ± 7071 ± 219996–101860.92710074 Sarnthein et al. [2003]
KIA8554M23258249NPS9235 ± 5071 ± 219880–1009519965Not usedSarnthein et al. [2003]
KIA9354M23258250NPS9435 ± 5571 ± 2110147–10264110210 Sarnthein et al. [2003]
KIA7657M23258315NPS10980 ± 70200 ± 5012050–12365112216 Sarnthein et al. [2003]
KIA7659M23258355NPS12010 ± 5571 ± 2113295–13409113359 Sarnthein et al. [2003]
KIA 9354M23258394NPS12390 ± 6071 ± 2113672–13837113751 Sarnthein et al. [2003]
Poz-8245MD95–20115NPD1020 ± 1000 ± 0519–6651600Not usedThis study
GifA96471MD95–201110.5NPD980 ± 600 ± 0526–6191573 Risebrobakken et al. [2003]
KIA 5600MD95–201124.5NPD1590 ± 400 ± 01102–121511153Not usedRisebrobakken et al. [2003]
KIA 3925MD95–201130.5NPD1040 ± 400 ± 0590–6490.789611 Risebrobakken et al. [2003]
KIA 5601MD95–201147.5NPD1160 ± 300 ± 0667–7271703 Risebrobakken et al. [2003]
Poz-8244MD95–201155.5NPD1530 ± 900 ± 0976–117411085Not usedThis study
KIA 3926MD95–201170.5NPD1460 ± 500 ± 0941–105711008 Risebrobakken et al. [2003]
KIA 6286MD95–201189.5NPD1590 ± 300 ± 0740–8381796Not usedRisebrobakken et al. [2003]
Poz-8246MD95–2011102NPD1790 ± 600 ± 01275–139011338 This study
KIA 3927MD95–2011130.5NPD2350 ± 400 ± 01912–202911970Not usedRisebrobakken et al. [2003]
KIA 6287MD95–2011154NPD2335 ± 250 ± 01888–200011953 Risebrobakken et al. [2003]
GifA96472MD95–2011170.5NPD2620 ± 600 ± 02190–236012301 Risebrobakken et al. [2003]
Poz-8242MD95–2011225NPD3000 ± 500 ± 02724–282212775 This study
Poz-8241MD95–2011250NPD3380 ± 700 ± 03162–333913246 This study
KIA 10011MD95–2011269.5NPD3820 ± 350 ± 03711–382313769 Risebrobakken et al. [2003]
Poz-8240MD95–2011300NPD4080 ± 700 ± 04006–422414122 This study
KIA 463MD95–2011320.5NPD4330 ± 500 ± 04396–452614464 Risebrobakken et al. [2003]
Poz-8238MD95–2011451NPD6420 ± 1600 ± 06729–711116906 This study
KIA 464MD95–2011520.5NPD7260 ± 600 ± 07657–778817725 Risebrobakken et al. [2003]
Poz-8237MD95–2011528.5NPD7690 ± 1100 ± 08035–827418154 This study
Poz-8236MD95–2011533.5?8530 ± 1600 ± 08988–937719154Not usedThis study
Poz-8235MD95–2011541.5?8280 ± 1400 ± 08618–899418822 This study
Poz-8234MD95–2011570.5?8700 ± 900 ± 08618–899419362 This study
TUa-3315MD95–2011703.5NPS10775 ± 85200 ± 5011629–120510.95811805 Risebrobakken et al. [2003]
 MD95–2011709.5Tephra    12170Vedde AshRisebrobakken et al. [2003]
TUa-3316MD95–2011730.5NPS11875 ± 1400 ± 013174–13429113320 Risebrobakken et al. [2003]
KIA465MD95–2011750.5NPS12220 ± 900 ± 013578–137520.78813635 Risebrobakken et al. [2003]
KIA3519MD95–2011813.5NPS13450 ± 900 ± 015247–159690.92615760 Dreger [1999]
KIA-10676MD99–22842.5NPS1690 ± 300 ± 01222–128311251 This study
Poz-10150MD99–228419.5NPS3515 ± 350 ± 03355–343813399 This study
Poz-10151MD99–228436.5NPS5295 ± 350 ± 05601–568615649 This study
Poz-10157MD99–228453.5NPS7300 ± 400 ± 07707–781417762 This study
Poz-33098MD99–228471.5NPS7940 ± 700 ± 08334–847618405 This study
TUa-3301MD99–2284100.5NPS8680 ± 850 ± 09261–945119343 Bakke et al. [2009]
Poz-33098MD99–2284165.5NPS9340 ± 900 ± 010074–10319110181 This study
 MD99–2284185.5Tephra    10350Saksurnavatn Ash 
TUa-3302MD99–2284213.5NPS10050 ± 95200 ± 5010584–10915110779 Bakke et al. [2009]
TUa-3304MD99–2284249.5NPS10700 ± 90200 ± 5011546–118740.72311654 Bakke et al. [2009]
 MD99–2284362.5Tephra    12170Vedde Ash 
Poz-29526MD99–2284423.5NPS11440 ± 800 ± 012799–129860.73312912 This study

4. Results

[18] The 1DICE column model is used to investigate the effect of changing insolation forcing at 71°N, at the time slices 12, 11, 10, 9, 8, 7, 6 and 0 ka BP, on temperature profiles representing average conditions in the S Barents Sea. 1DICE modeled SSTs are consistently higher during the early Holocene than at present (Figure 2a). At 10 ka BP, when the summer insolation forcing reached its Holocene maximum, the modeled August SSTs were 0.6°C higher than at present. The corresponding temperature increase below 20 m water depth is considered insignificant (Figure 2a). Autumn and winter mixing entails a homogeneous S Barents Sea water column. The higher temperatures of the modeled 10 ka SML causes diffusion of some extra heat downward to the deeper water column, but at a very modest level. The amplitude of the 10 ka warming below 60 m depth is a maximum of 0.05°C in July. The rest of the year the difference is smaller than 0.05°C. During the year, the minimum surface mixed layer depth in 1DICE is seen for June through August (∼12 m), while it rises to a maximum of 58 m in October. The difference in temperature below the mixed layer, between the present and 10 ka, also varies through the year. It is lowest at the end of winter mixing (March, 0.03°C, Figure 2b), and largest at the peak of the insolation (June, 0.05°C, Figure 2b).

Figure 2.

Vertical profiles of temperature for the southern Barents Sea column as calculated by the 1DICE model. (a) Mid August temperature profiles for the years 12, 11, 10, 9, 8, 7, 6 and 0 ka BP. The difference in temperature are solely caused by the difference in monthly mean solar radiation at 71°N for the respective years. The influence of higher increased summer insolation is close to zero for all years below the mixed layer. A blowup of the temperature differences seen in the upper part of the water column is shown in the black box. (b) Difference in temperatures between 10 and 0 ka BP for selected months. The increased insolation at 10 ka BP warms the mixed layer effectively down to a maximum depth of 90 m. The “spikes” at the bottom of the mixed layer are caused by differences in mixed layer depths between the years. A blowup of the upper 100 m is shown in the black box.

[19] The compiled Nordic Seas alkenone SST records provide a consistent picture of the main Holocene trends. The records from the Vøring Plateau (MD95-2011) [Calvo et al., 2002], the SW Barents Sea (Psh-5159N) [Risebrobakken et al., 2010] and the Barents Sea margin (M3258) [Marchal et al., 2002] all shows warm alkenone SSTs during the early to mid-Holocene followed by gradual cooling and maximum temperatures within the 9–6 ka BP interval (Figure 3a). Minor variability, in part of local character, is found superimposed on the main trends. Consistently colder conditions and a larger Holocene temperature decrease are seen at the Barents Sea margin compared to the Vøring Plateau and the SW Barents Sea (Figure 3a). The reconstructed temperature differences between these sites (Vøring Plateau – SW Barents Sea and Vøring Plateau – Barents Sea margin) at 9–6 ka BP are 0.6°C and 5.5°C respectively. These early Holocene temperature differences are within the range of present-day August SST differences between these sites (International Council for the Exploration of the Sea, public oceanographic database, 2010,

Figure 3.

(a) Alkenone SSTs from MD95-2011 (red) [Calvo et al., 2002], PSh-5159N (black) [Risebrobakken et al., 2010] and M23258 (blue) [Marchal et al., 2002]. (b) Alkenone SST of PSh-5159N [Risebrobakken et al., 2010] (black) and 1DICE modeled August temperatures (vertical mean 0–17 m water depth) (red brown) compared to the mean August insolation at the top of the atmosphere (yellow) [Laskar et al., 2004] and the mean August insolation after correcting for a planetary albedo of 40% (purple) [Hartmann, 1994]. (c) Foraminiferal temperatures calculated for this study using the Maximum Likelihood method. MD99-2284 (gray), MD95-2011 (red), PSh-5159N (black) and M23258 (blue). The light gray bar indicates HTM as given by the alkenone SST records.

[20] The main planktic foraminiferal species found in the studied cores during the early Holocene are N. pachyderma (sin), Neogloboquadrina pachyderma (dex), Globigerina bulloides and Turborotalita quinqueloba (Figure 4). Renaming N. pachyderma (dex) to Neogloboquadrina incompta has been suggested [Darling et al., 2006]; however, we use the old name to maintain consistency with previous publications of part of the data discussed. The calculated foraminiferal temperatures reflect the relative abundance of N. pachyderma (sin) at all sites (Figure 4); decreases in the relative abundance of N. pachyderma (sin) result in warmer estimated temperatures.

Figure 4.

Calculated foraminiferal temperatures, and relative abundance of N. pachyderma (sin), N. pachyderma (dex), T. quinqueloba and G. bulloides, from MD99-2284, MD95-2011, PSh-5159N and M23258 are shown. All MD99-2284 data is from this study. The relative abundance data from MD95-2011 represents a combination of new counts done for this study and previously published counts [Andersson et al., 2003; Risebrobakken et al., 2003]. With exception of the relative abundance of G. bulloides, the assemblage data from PSh-5159N were published by Risebrobakken et al. [2010] and Chistyakova et al. [2010]. The relative abundance data from M23258 is from Sarnthein et al. [2003]. All foraminiferal temperatures were calculated for this study. The gray bar indicates the interval within when early Holocene maximum advection took place, including the increase toward the absolute maximum at 10 ka BP and the following decrease toward the 9 ka BP minimum.

[21] Foraminiferal temperatures of the Faroe-Shetland Channel, the Vøring Plateau and the Barents Sea margin record stepwise warming from 12 to 10 ka BP, followed by a cooling toward 9 ka BP (Figures 3c and 4). The temperature peak at 10 ka BP represents the warmest early Holocene foraminiferal conditions. Contemporaneously with the temperature peak at 10 ka BP the relative abundance of G. bulloides reached the Holocene maximum level and N. pachyderma (dex) occurred in high numbers in the Faroe-Shetland Channel, at the Vøring Plateau and at the Barents Sea margin (Figure 4). At the Barents Sea margin [Sarnthein et al., 2003], more N. pachyderma (dex) is found in relation to the early Holocene foraminiferal temperature maximum than anytime later, in contrast to the Faroe-Shetland Channel and the Vøring Plateau where the early Holocene relative abundance of N. pachyderma (dex) is significantly lower than during the late Holocene (Figure 4). The relative abundance of T. quinqueloba also peaks in the Faroe-Shetland Channel and at the Barents Sea margin during this early Holocene maximum in foraminiferal temperatures. At the Barents Sea margin, T. quinqueloba is the most dominant species [Sarnthein et al., 2003]. Turborotalita quinqueloba is highly abundant also at the Vøring Plateau; however, at the Vøring Plateau the relative abundance of T. quinqueloba remained at the Holocene maximum level throughout the 11–6 ka BP interval (Figure 4). With the exception of T. quinqueloba at the Vøring Plateau, the relative abundance of G. bulloides, N. pachyderma (dex) and T. quinqueloba decreases 10–9 ka BP at the Faroe-Shetland Channel, the Vøring Plateau and the Barents Sea margin (Figure 4).

[22] In the SW Barents Sea, gradually increasing SSTs are indicated 11–7.5 ka BP, followed by cooling, instead of the temperature maximum at 10 ka seen at the Faroe-Shetland Channel, the Vøring Plateau and the Barents Sea margin (Figure 4). The differences in foraminiferal temperature between the SW Barents Sea and the other cores are reflected by the relative abundance data. During the early Holocene, the planktic foraminiferal fauna of the SW Barents Sea was dominated by T. quinqueloba (Figure 4) [Chistyakova et al., 2010; Risebrobakken et al., 2010]. Neogloboquadrina pachyderma (sin) and N. pachyderma (dex) first play a role from 7.5 ka BP, and G. bulloides is of minor influence.

[23] A pronounced low in N. pachyderma (sin) δ18O is seen in the Faroe-Shetland Channel, at the Vøring Plateau [Risebrobakken et al., 2003] and in the SW Barents Sea [Risebrobakken et al., 2010] between 11 and 9 ka BP (Figure 5), followed by comparable more stable conditions. The low δ18O signal is least pronounced in the Faroe-Shetland Channel (Figure 5). Low N. pachyderma (sin) δ18O is also recorded at the Barents Sea margin in the early Holocene [Sarnthein et al., 2003]; however, the signal is slightly delayed compared to the timing of the depleted signal in the other cores (Figure 5). At the Barents Sea margin, the early Holocene N. pachyderma (sin) δ18O depletion, are followed by increasing N. pachyderma (sin) δ18O values (Figure 5) [Sarnthein et al., 2003]. In the Franz-Victoria Trough a gradual N. pachyderma (sin) δ18O depletion is recorded from 10 ka BP to approximately 8 ka BP, followed by a gradually heavier N. pachyderma (sin) δ18O 8–4.2 ka BP (Figure 5).

Figure 5.

Ice volume corrected δ18O of N. pachyderma (sin). Smoothed records are shown in the lower panel: MD99-2284 (Faroe-Shetland Channel) (gray), MD95-2011 (Vøring Plateau) (red) [Risebrobakken et al., 2003], PSh-5159N (SW Barents Sea) (black) [Risebrobakken et al., 2010], M23258 (Barents Sea margin) (blue) [Sarnthein et al., 2003] and PSh-5157 (light blue). PSh-5159N [Risebrobakken et al., 2010] and MD95-2011 [Risebrobakken et al., 2003] record more depleted δ18O than MD99-2284 during the early Holocene. The gray bar indicates the interval within when early Holocene maximum advection took place, as seen from the foraminiferal temperature and relative abundance data (Figure 4).

[24] Between 12 and 11 ka BP, strongly depleted C. wuellerstorfi δ13C are recorded in the Faroe-Shetland Channel, followed by the heaviest Holocene values recorded 11 to 9.8 ka BP, while the N. pachyderma (sin) δ13C were higher 12–11 and lower 11–9.8 ka BP (Figure 6). A similar N. pachyderma (sin) δ13C development took place at the Vøring Plateau, and in the SW Barents Sea [Risebrobakken et al., 2010]. Low N. pachyderma (sin) δ13C is also recorded at the Barents Sea margin 11–9.8 ka BP, however, the decrease from 12 ka BP is less pronounced than in the other cores (Figure 6) [Sarnthein et al., 2003]. Following this early Holocene depletion, a gradual increase in N. pachyderma (sin) δ13C is seen at all sites until approximately 3 ka BP.

Figure 6.

Benthic (C. wuellerstorfi - red) δ13C from MD99-2284 and planktic (N. pachyderma (sin) - black) δ13C records from MD99-2284, MD95-2011, PSh-5159N [Risebrobakken et al., 2010] and M23258 [Sarnthein et al., 2003]. The light gray bar represents the interval characterized by extremely light benthic δ13C in MD99-2284, indicative of weak ventilation, while the dark gray bar shows the planktic depletion of δ13C seen as a response to an initial deep mixing of the water column due to an overshooting AMOC.

5. Discussion

5.1. Effect of Increased Summer Insolation on Surface and Subsurface Ocean Temperatures

[25] Today, the shallow SML experiences atmospheric-induced heating due to insolation forcing, while below the depth of the SML rather constant temperatures are seen year-round and the year-to-year temperature changes relates to the NwAC and the variability of its advective heat transport [Nilsen and Falck, 2006]. If today's conditions are representative of the past, the proxies representing the SML will provide information on long-term temperature changes caused by variable radiative forcing, while proxies of annual mean temperature conditions will provide information on changes in heat advection related to the mean state of the NwAC. However, summer insolation was stronger at high northern latitudes during the early Holocene than today (Figure 3b) [Laskar et al., 2004].

[26] The 1DICE results show that only SML temperatures were significantly affected by the higher summer insolation (Figure 2a). The influence of insolation on deeper ocean temperatures was within ±0.05°C of the present and is therefore considered to be insignificant throughout the year (Figure 2b). Similarly, Andersson et al. [2010] and Liu et al. [2003] investigated the effect of insolation on the water column in the Nordic Seas and in the North Atlantic using Community Climate System Model 3.0 (CCSM3) and a low-resolution general circulation model respectively. In agreement with our results, both studies suggest that seasonal summer warming was restricted to the upper ∼30 m in the Nordic Seas during the early Holocene.

[27] If the results from 1DICE are indicative of early Holocene conditions, we would expect proxy data for the SML to record a insolation driven HTM. Conversely we would expect proxy data for depths bellow the seasonal thermocline to be unaffected by early Holocene insolation. Within the study area, coccolithophores bloom during summer [Brown and Yoder, 1994; Smyth et al., 2004], and their depth habitat is restricted to the photic zone (10–15 m). The long-term decreasing trends in alkenone SSTs at the Vøring Plateau, the Barents Sea margin and in the SW Barents Sea are therefore interpreted to represent the Holocene temperature development of the SML. Hence, the maximum response to enhanced summer insolation on SML temperatures in the eastern Nordic Seas occurred within the 9–6 ka BP interval (Figure 3a). This interpretation of the alkenone SSTs is in line with argumentation given by Jansen et al. [2008] and Andersson et al. [2010].

[28] The C37:4 alkenone is more abundant in the Nordic Seas than at mid to low latitudes and resuspension and lateral advection of alkenone bearing material may take place [e.g., Bendle and Rosell-Melé, 2004]. However, an examination of the alkenone-SST relationship in the Nordic Seas shows that sediment from the Norwegian Sea provides reliable SSTs for the Holocene [Bendle and Rosell-Melé, 2004]. The consistency between the three independent alkenone SST records reported here further supports the use of alkenones in the Nordic Seas for SML reconstructions during the Holocene. Additionally, the warmer SSTs reconstructed for the HTM are in agreement with terrestrial records from the same region [e.g., Nesje, 2009; Seppä et al., 2009].

[29] Both 1DICE and SW Barents Sea alkenone SSTs shows an early Holocene temperature maximum followed by a gradual temperature decrease (Figure 3b). The absolute SML temperature maximum is seen 1 ka later in the SW Barents Sea than by 1DICE and the amplitude of change differs. Feedbacks related to disintegrating ice sheets, meltwater, vegetation, sea ice and albedo changes influence the timing and amplitude of the HTM [Kaufman et al., 2004; Otto et al., 2009; Renssen et al., 2009]. 1DICE gives the isolated effect of increased insolation forcing on the ocean temperature in terms of the seasonal and vertical structure for separate time slices compared with the present values, not taking into account any feedback mechanisms. The SW Barents Sea alkenone SSTs shows the temperature development with all potential effects from active feedback mechanisms implemented. The time lag and amplitude difference between the modeled and observed Holocene SML temperatures are therefore considered reasonable.

[30] Absolute 1DICE SSTs are lower than the SW Barents Sea alkenone SSTs (Figure 3b). This discrepancy in absolute temperatures reflects that PSh-5159N is located in the warmest part of the Barents Sea, while 1DICE represents mean conditions of the S Barents Sea (Figure 1). The “core top” temperatures of PSh-5159N and 1DICE, 10.9°C and 7.7°C respectively, are comparable with instrumentally recorded August temperatures for the representative location/area, 10.6°C and 8°C respectively [Nilsen et al., 2008; International Council for the Exploration of the Sea, public oceanographic database, 2010].

[31] The Holocene long-term trends and intracore relationships of alkenone SSTs and foraminiferal temperatures differ substantially (Figure 3): 1) None of the sites record higher early Holocene than late Holocene foraminiferal temperatures, with an exception at the Barents Sea margin. 2) An early Holocene maximum in foraminiferal-based temperatures are seen before the maximum in alkenone SSTs. It is therefore reasonable that the two proxies respond to different dynamical forcing. Differences between the proxies have previously been noted at the Barents Sea margin [e.g., Marchal et al., 2002] and at the Vøring Plateau [Andersson et al., 2010; Jansen et al., 2008; Risebrobakken et al., 2003]. Planktic foraminifers are found living over a wide range of water depths; N. pachyderma (sin) reflect water at 70–250 m [Simstich et al., 2003], N. pachyderma (dex) calcify at approximately 50 m [Nyland et al., 2006], G. bulloides dwells in the upper 60 m of the water column [Schiebel et al., 1997] and T. quinqueloba calcify at 25–75 m [Simstich et al., 2003]. Furthermore, the season of maximum production differ for different species [Chapman, 2010; Jonkers et al., 2010; Kucera, 2007]. Calculated foraminiferal temperatures integrate information from these different preferential conditions [Andersson et al., 2010]. The deeper living depths of foraminifera compared to coccolithophores causes the foraminifera to record temperatures biased toward annual mean conditions, independent of the calcification season, due the year-round relatively constant temperatures below the SML [Jansen et al., 2008]. A similar relationship is observed in other temperature-sensitive proxies based on foraminifera, e.g., Mg/Ca and δ18O reconstructions [Andersson et al., 2010; Leduc et al., 2010]. Since the foraminifera respond to annual mean conditions, these proxies can be used to detect changes in the mean state of the NwAC and accordingly its advection of heat.

[32] To summarize, the early Holocene orbital forcing only influenced SML temperatures in the Nordic Seas. Consequently, only proxies representing SML temperatures record the Nordic Seas HTM documented within the 9–6 ka BP interval. The foraminiferal temperatures record annual mean temperatures and does therefore not see the HTM, however, they do provide a means to detect changes in heat advection through the early Holocene.

5.2. Early Holocene Foraminiferal Assemblages and Temperature Estimates

[33] The high content of N. pachyderma (dex) and G. bulloides in the Faroe-Shetland Channel, at the Vøring Plateau and the Barents Sea margin at 10 ka BP demonstrates increased influence of warm Atlantic water (Figure 4). In the Nordic Seas G. bulloides and N. pachyderma (dex) reflects Atlantic water, with G. bulloides as the more warm-loving of the two species [e.g., Bé and Tolderlund, 1971; Johannessen et al., 1994]. For example Hald et al. [2007] shows warm foraminiferal temperatures at 10 ka BP close to the Barents Sea, at the Barents Sea margin and at the West Spitsbergen margin, but not at the Vøring Plateau. Here we document this 10 ka BP maximum in foraminiferal temperatures at the Vøring Plateau and in the Faroe-Shetland Channel, hence, the temperature maximum is not restricted to the northeastern Nordic Seas. Following the argumentation from above (4.1), the 10 ka BP warming was a result of intensified heat advection in the NwAC and not related to local heating due to the strong summer insolation anomaly. Hald et al. [2007] found that the relative abundance of G. bulloides and N. pachyderma (dex) was lower at the West Spitsbergen margin than at the Barents Sea margin throughout the 11–9 ka BP interval, indicating a sizable oceanic heat loss before the NwAC entered the Fram Strait.

[34] A minor increase in relative abundance of G. bulloides and N. pachyderma (dex) in the SW Barents Sea, compared to Barents Sea margin (Figure 4), shows that at 10 ka BP more of the advected heat followed the WSC rather than the NCaC, or that the core of Atlantic water in the NCaC entered the Barents Sea at deeper depths than recorded by planktic foraminifera. Benthic foraminiferal fauna and benthic δ18O from the SW Barents Sea document warm bottom water at 10 ka BP, supporting that Atlantic water did enter the Barents Sea [Aagaard-Sørensen et al., 2010; Chistyakova et al., 2010; Risebrobakken et al., 2010].

[35] Turborotalita quinqueloba dominates the planktic foraminiferal assemblages in the SW Barents Sea and at the Barents Sea margin 11–7 and 10.5–9 ka BP, respectively [Chistyakova et al., 2010; Risebrobakken et al., 2010; Sarnthein et al., 2003]. Turborotalita quinqueloba is a subpolar species; however, maximum relative abundance occurs near the Arctic front [Johannessen et al., 1994; Matthiessen et al., 2001]. The SW Barents Sea results exemplify how the dual character of T. quinqueloba may influence the foraminiferal temperatures when this species dominates the assemblage. The highest foraminiferal temperatures in the SW Barents Sea are recorded 9–8 ka BP, mirroring the relative abundance of T. quinqueloba, however, the maximum abundance of G. bulloides and N. pachyderma (dex) document warmer conditions 11–10 ka BP (Figure 4). Furthermore, warmer foraminiferal temperatures are calculated for the Vøring Plateau and the SW Barents Sea than upstream in the Faroe-Shetland Channel 9–5 ka BP (Figures 3c and 4). Heat is steadily released from the Atlantic water as it moves toward the Arctic [Skagseth et al., 2008]; hence, a northward temperature increase is not realistic. We argue that the too-high downstream temperatures 9–5 ka BP is an artifact of the continuously high content of T. quinqueloba at the Vøring Plateau and in the SW Barents Sea compared to the Faroe-Shetland Channel (Figure 4). In agreement, Risebrobakken et al. [2003, 2010] found that the Arctic front was located closer to the Vøring Plateau than today and the Barents Sea front was located in vicinity of the SW Barents Sea and the Barents Sea margin through the early to-mid Holocene.

[36] Independent of the front versus temperature discussion, the Holocene maximum in relative abundance of the warm loving G. bulloides and N. pachyderma (dex) provides clear evidence of NwAC temperatures, and hence northward heat advection, peaking at 10 ka BP (Figure 4). This maximum in heat advection corresponds in time with the early Holocene establishment of a warm mollusk fauna at Spitsbergen [Salvigsen et al., 1992]. Colder foraminiferal conditions are reflected by the reduced relative abundance of G. bulloides and N. pachyderma (dex) (Figure 4) from 9 to 6 ka BP, when the warmest alkenone SSTs are recorded (Figure 3) [Calvo et al., 2002; Marchal et al., 2002; Risebrobakken et al., 2010], underlining that the different proxies respond to different dynamical mechanisms.

[37] Knudsen et al. [2004] found the highest Holocene content of subpolar planktic foraminifera off North Iceland at ∼10 ka BP, reflecting enhanced advection of Atlantic water through the Irminger Current. Maximum in alkenone SST is recorded 10–9 ka off North Iceland, slightly earlier than at the Vøring Plateau [Bendle and Rosell-Melé, 2007], emphasizing the importance of local feedbacks in determining the exact timing of the HTM [Kaufman et al., 2004]; however, the long-term trends are comparable [Bendle and Rosell-Melé, 2007]. Hence, the relationship between early Holocene advection and insolation response off Iceland and in the eastern Nordic Seas is comparable.

5.3. Early Holocene Meltwater Influence

[38] Following a northward oceanic heat release [Skagseth et al., 2008], higher δ18O are expected northward if solely influenced by temperature. This is not the case 12–8.5 ka BP; lower δ18O are recorded in the SW Barents Sea and at the Vøring Plateau than in the Faroe-Shetland Channel (Figure 5). Accordingly, low-salinity water influenced the Vøring Plateau and SW Barents Sea 12–8.5 ka BP, with a gradually diminishing effect (Figure 5). Supporting the interpretation of reduced salinity, Moros et al. [2004] document ice rafting at the Vøring Plateau 12–8.5 ka BP. Hence, salinity did at times influence the Holocene δ18O at the Vøring Plateau, contrasting the argumentation of Risebrobakken et al. [2003]. Maximum heat advection at ∼10 ka BP corresponds with low δ18O at all sites (Figure 5). In view of that, temperature did play an additional role in defining the δ18O signature. In agreement with northward cooling, consistently higher Holocene δ18O is recorded at the Barents Sea margin and in the Franz-Victoria Trough than at the upstream sites (Figure 5).

[39] 11–9.5 ka BP the low δ18O, and hence the influence of low-salinity water, is larger in the SW Barents Sea than at the Vøring Plateau (Figure 5). The SW Barents Sea δ18O depletion is explained by melting and refreezing of sea ice after favorable preconditioning [Risebrobakken et al., 2010]. Such a mechanism cannot explain the low salinities at the Vøring Plateau, as the Vøring Plateau site is located further from land and at much deeper water depths than the SW Barents Sea site. The Fennoscandian Ice Sheet (FIS) covered most of Norway, Sweden and Finland at 12 ka BP [Andersen et al., 1995], while remnant ice covered smaller parts of mainland Norway than today's glaciers at 8 ka BP [Nesje, 2009]. Consequently, gradually less fresh water was released to the study area within the time interval 12–8 ka BP. We argue that the early Holocene low-salinity signal at the Vøring Plateau reflects a direct influence of meltwater, or an expansion of the coastal water, following the final disintegration of the FIS. Reduced early Holocene salinities is also documented in the North Atlantic following deglaciation of the northern Hemisphere ice sheets, with a primary influence from the Laurentide ice sheet [Came et al., 2007; Solignac et al., 2008; Thornalley et al., 2009]. Since both the Nordic Seas and the North Atlantic were influenced by meltwater during the early Holocene, it is reasonable that also the Faroe-Shetland Channel to some extent was affected by fresher surface water.

[40] No salinity response comparable to the Vøring Plateau and SW Barents Sea signal occurred at the Barents Sea margin [Sarnthein et al., 2003] (Figure 5). Today, Atlantic water submerges underneath Arctic water in the Fram Strait [Manley, 1995]. If the submerging zone was located further south the thermocline depth at the Barents Sea margin could have been deeper that today, at a time when today's circulation pattern in the Northern Nordic Seas was not yet established [Risebrobakken et al., 2010].The thermocline depth represents the upper calcification depth of N. pachyderma (sin) [Simstich et al., 2003]. Hence, the lack of fresh water influence on the Barents Sea margin δ18O can be explained by a deeper calcification depth of N. pachyderma (sin) along the Barents Sea margin than at the Vøring Plateau and in the SW Barents Sea, in combination with a distal location from the main source of fresh water to the Nordic Seas during the early Holocene. The Barents Sea Ice Sheet disintegrated faster than the FIS, and the melting-refreezing mechanism suggested for the shallower and more sheltered Barents Sea [Risebrobakken et al., 2010] would not influence the more open ocean Barents Sea margin.

5.4. Early Holocene Changes in Northward Oceanic Heat Advection Through the NwAC

[41] We have documented that the foraminiferal temperature peak at ∼10 ka BP can be followed from the Nordic Seas entrance to the Barents Sea margin (Figure 4), and we argue that it reflects intensified heat advection in the NwAC, independent of local heating due to stronger summer insolation. Furthermore, we will argue that this early Holocene northward propagating warm pulse occurred as a response to a reorganization of the region's climatic state during the last phase of deglaciation and the early Holocene.

[42] During deglaciation of the Northern Hemisphere ice sheets, large amounts of meltwater entered the Nordic Seas and the North Atlantic. Depending on when, where and at what rate meltwater hit the ocean the deglacial fresh water flux can weaken the Atlantic Meridional Overturning Circulation (AMOC) [e.g., Liu et al., 2009; Mignot et al., 2007]. Several studies show that tropical and north Atlantic subsurface water temperature increase in association with an AMOC slowdown [e.g., Barker et al., 2009; Manabe and Stouffer, 1997; Rühlemann et al., 1999]. In the North Atlantic/Nordic Seas, it is suggested that a temperature inversion built up and was maintained due to the influence of deglacial fresh water [Knorr and Lohmann, 2007; Liu et al., 2009]. A surplus reservoir of heat and salt in the tropical and subtropical North Atlantic subsurface were then available for northward advection when the ice sheets diminished, the meltwater flux decreased and the North Atlantic convection sites rejuvenated [Manabe and Stouffer, 1997]. A succeeding rapid salinity recovery in the Polar regions probably broke the temperature inverted density structure causing an destabilization of the Nordic Seas water column [Knorr and Lohmann, 2007; Mignot et al., 2007], resulting in an intensified AMOC, characterized by deeper and stronger circulation than today [Liu et al., 2009; Manabe and Stouffer, 1997]. The subsurface layer cooled rapidly through air-sea interactions, and the convection gradually weakened as the reservoir dampened [Mignot et al., 2007]. We argue that the 10 ka BP maximum in foraminiferal temperatures resulted from such an intensified northward advection of warm water, following an overshoot of the AMOC (Figures 3, 4, and 5). In the Faroe-Shetland Channel, the early Holocene heat pulse is preceded by an interval of extremely light benthic δ13C (Figure 6), implying weak ventilation and overturning circulation. Destabilization of the Nordic Seas water column initiated deep mixing that entailed lowering of the planktic δ13C, as some of the δ13C depleted bottom water from the Faroe-Shetland Channel was welled up (Figure 6). Following the NwAC the low N. pachyderma (sin) δ13C signal propagated northward, influencing all sites (Figure 6). At the same time, the δ13C of the bottom water in the Faroe-Shetland Channel increased due to the intensified mixing. Potentially, increased turbulence and upwelling of nutrients related to the intensified mixing also influenced the high early Holocene relative abundance of T. quinqueloba (Figure 4).

[43] As the Atlantic reservoir of surplus heat and salt rapidly emptied [Manabe and Stouffer, 1997], the strength of the overturning and the northward heat advection diminished, entailing cooling at all sites after the peak warmth at approximately 10 ka BP (Figure 4). The continuous flux of low salinity water that influenced the Nordic Seas during the early Holocene (Figure 5) was not strong enough to prevent overturning after initial AMOC rejuvenation.

[44] Bakke et al. [2009] hypothesized that the atmospheric jet flickered between a southerly location of the storm tracks, directing the westerlies toward central Europe, and a northern location directing the westerlies toward the Nordic Seas during the late Younger Dryas. Risebrobakken et al. [2003] argued for strong westerlies directed toward the Nordic Seas during the early Holocene. Episodic intrusions of warmer waters into the Nordic Seas coincided with the northward displacement of the storm tracks [Bakke et al., 2009], evidence further supported by a North Atlantic temperature reconstruction [Thornalley et al., 2010]. Fawcett et al. [1997] emphasize the importance of oceanic heat flux into the Nordic Seas in driving the Younger Dryas/Preboreal transition; reducing the sea ice extent moves the winter storm tracks northward. The effect of North Atlantic oceanic heat transport were larger for wintertime than summertime conditions, emphasizing the role the deeper WML and its influence on the regional climate [Alley et al., 1999].

[45] Meltwater freshened the North Atlantic surface water through the deglaciation and early Holocene, affecting interaction between the subpolar gyre (SPG) and the subtropical gyre (STG) [Thornalley et al., 2009]; weakening the SPG due to freshening strengthens the STG and intensifies the transport of heat and salt into the Nordic Seas [Hátún et al., 2005]. Accordingly, the deglacial to early Holocene gyre dynamics enhanced the transport of warmer and salty water into the Nordic Seas.

[46] Summarizing, stronger heat advection into the Nordic Seas at 10 ka BP than before and after took place following a reorganization of the AMOC after the deglaciation, combined with a gyre dynamic causing enhanced transport of heat and salt. A northward displacement of the storm tracks toward the Nordic Seas helped maintaining the enhanced poleward heat transport. As the effect of deglacial preconditioning and meltwater influence on the gyre dynamic diminished, the northward heat transport gradually decreased after ∼10 ka BP.

[47] In the Franz-Victoria Trough, N. pachyderma (sin) δ18O document a temperature maximum at ∼8 ka BP (Figure 5), corresponding in time with the HTM recorded by alkenones, not with the foraminiferal temperatures recorded at the other studied sites. In agreement with our record, Duplessy et al. [2001] show a low in Franz-Victoria Trough N. pachyderma (sin) δ18O 7.85–6.9 ka BP. Neogloboquadrina pachyderma (sin) prefer temperate Atlantic water in the NE Fram Strait [Carstens et al., 1997]. Accordingly, N. pachyderma (sin) δ18O from the Franz-Victoria Trough reflects conditions at approximately 180 m water depth, corresponding to today's Arctic-Atlantic water interface at the site. A fraction of the Atlantic water that submerges underneath Arctic water in the Fram Strait enters the Barents Sea from the Arctic Ocean through the Franz-Victoria Trough. We argue that the N. pachyderma (sin) δ18O temperature maximum in the Franz-Victoria Trough at 8 ka BP reflects a subsurface propagation through the Fram Strait and into the Arctic Ocean of the contemporaneously high eastern Nordic Seas SML temperatures, in line with arguments by Duplessy et al. [2001] and Lubinski et al. [2001].

6. Summary and Conclusions

[48] Strong early Holocene summer insolation at high northern latitudes increased the SML temperatures in the Barents Sea and the Nordic Seas. However, no significant temperature increase occurred below the SML. Consequently, only proxy records representing the near surface can be used to reconstruct the response to the orbital forcing and hence the HTM. The alkenone SST records document the HTM within the time interval 6–9 ka BP, in good agreement with the timing of the HTM observed in terrestrial records from the same region.

[49] Proxies providing information on conditions underneath the SML represent the mean state of the NwAC and thus changes in northward heat advection through the Nordic Seas. Warmer than, or as warm as, present conditions are recorded by foraminiferal-based proxies at 10 ka BP, all along the pathway of the NwAC from the Faroe-Shetland Channel in the south to the Barents Sea margin in the northern Nordic Seas. The temperature maximum at 10 ka BP shows intensified heat advection through the NwAC. Our findings emphasize that high latitude radiative forcing is not responsible for the overall conditions of the water column and ocean dynamics.

[50] A major reorganization of the ocean circulation following the deglaciation led to the intensified heat advection peaking at 10 ka BP. We hypothesize that strong meltwater discharge to sensitive areas periodically reduced the strength of the AMOC through the deglaciation, entailing weak ventilation of the Nordic Seas until 11 ka BP and a buildup of an Atlantic subsurface reservoir of heat and salt. Rejuvenating the AMOC entailed a strong and deep overturning circulation that intensified the early Holocene northward heat advection in the NwAC. As the surplus heat reservoir rapidly emptied, the overturning diminished, reducing the northward heat advection through the Nordic Seas. Predominant deglacial to early Holocene gyre dynamics and atmospheric forcing further enhanced the transport of warm and salty water into the Nordic Seas.

[51] Meltwater from the disintegrating ice sheets and glaciers in vicinity of the study area, as well as melting and refreezing of sea ice, freshened the upper part of the water column 12–8.5 ka BP. The salinity of the upper water column gradually increased, as the glaciers reached their minimum Holocene extent and less fresh water reached the study area. The fresh surface water had an additional stabilizing effect on the SML; however, the effect was not large enough to prevent the overturning circulation after its initial recovery.


[52] Dagfinn Bøe, Rune Søraas and Odd Hansen are thanked for technical assistance. All data from M23258 was obtained through the PANGAEA database. The MD sites were cored through the IMAGES program by R/V Marion Dufresne, while the PSh cores were collected during a joint SIO RAS/BCCR cruise on R/V Professor Shtokman in 2004. The Research Council of Norway has supported this study through the projects, POCAHONTAS and ARCTREC. This is publication A357 from the Bjerknes Centre for Climate Research. We thank Thomas Cronin and one anonymous reviewer for comments that significantly improved the manuscript.