Between ca. 12 and 11 ka BP, no data exists from PSh-5159N and MD99-2304. At all other sites the planktic foraminiferal assemblages were dominated by N. pachyderma, except in T-79-51/2 (Figure 2C and Table 3). In T-79-51/2, the first half of this time interval was also dominated by N. pachyderma, however, from ca. 11.5 ka BP the Atlantic Water species occupied ca. 50% of the assemblage (Figure 2C). Since T-79-51/2 was counted at ≥125 μm rather than ≥100 μm the relative abundance of T. quinqueloba might be underestimated. Turborotalita quinqueloba would, however, not be the dominant species even if the relative abundance is underestimated by up to 20% (Husum and Hald, 2012). Cores T-88-2 and M23258, located at greater depths along the western Barents Sea margin recorded more than 80% of N. pachyderma (Figure 2C and Tables 1, 3). Even if the relative abundance of T. quinqueloba is underestimated in M23258, N. pachyderma would be the dominant species. In JM09-KA11-GC, located just north of Bear Island, T. quinqueloba and N. incompta were both present in significant amounts, with T. quinqueloba as the second dominant species (Figure 2C and Table 3). Despite a higher content of T. quinqueloba and N. incompta in JM09-KA11-GC the overall fauna is representative of Arctic Water (Johannessen et al., 1994; Husum and Hald, 2012).

Several sites from the Fram Strait and west Spitsbergen shelf were barren in planktic foraminifera at this time, argued to result from a dominance of Polar Water, with a strong meltwater influence indicated by a high content of ice rafted debris (Rasmussen et al., 2014; Werner et al., 2016; Consolaro et al., 2018; Figure 2A: 11, 12, and 13). Information about relevant oceanographic conditions of the northern Barents Sea is scarce for this time interval. To our knowledge, no planktic foraminiferal assemblage record exists, however, Lubinski et al. (2001) argued for inflow of cold subsurface water. Sea ice reconstructions from the northern Barents Sea do not reach as far back as 12 ka BP (Belt et al., 2015; Berben et al., 2017; Figure 2A: 14), hence, no inference can be made regarding the relation between the sea ice margin and the location of the BSAF.

Based on the combined evidence from planktic foraminiferal assemblage data from the region, we argue that the Barents Sea was to a large extent bathed by cold Arctic Water masses between ca. 12 and 11 ka BP (Figure 2B₁). The very southernmost shelf site did see traces of Atlantic Water, even though the overall assemblage was representative of colder water conditions than today. With Arctic Water conditions at most sites, the BSAF was located west of the westernmost sites, at the western Barents Sea margin. However, the Atlantic Water masses bathing T-79-51/2 suggests that the BSAF turned eastwards at the northeastern margin of the Lofoten Basin, just north of T-79-51/2 (Figure 2B₁).

The main difference between the ca. 12 to 11 ka BP and the ca. 11 to 10.2 ka BP time interval is that the southwestern Barents Sea site PSh-5159N recorded a total dominance of T. quinqueloba (Figure 2C and Table 3). In T-88-2, west of PSh-5159N, the relative abundance of N. pachyderma is reduced with 38%, but the site was still dominated by this polar species. Neogloboquadrina incompta still dominated the planktic foraminiferal assemblage in T-79-51/2. Further north, N. pachyderma was the dominant species, however, somewhat more T. quinqueloba was present compared to the previous time interval, and it increased in relative abundance after ca. 10.5 ka BP, both in M23258 and MD99-2304. As for the ca. 12 to 11 ka BP interval, taking into account the potential underestimation of the relative abundance of T. quinqueloba would not change the dominant species at T-79-51/2 or at M23258. Hence, T-79-51/2 was bathed by Atlantic Water, Psh-5159N was close to the BSAF and all the other sites were bathed by Arctic Water between ca. 11 and 10.2 ka BP (Figure 2B₂).

The west Spitsbergen shelf was still barren in foraminifera, feeling the influence of icebergs and Polar surface Water, indicating rather cold conditions and a more extensive sea ice cover than at present until 9.6 ka BP (Rasmussen et al., 2014; Figure 2A: 11). Based on dinoflagellates, it was also argued that, just north of M23258, sea ice was present through parts of the year (Rigual-Hernández et al., 2017; Figure 2A: 8). A comparable increase in relative abundance of T. quinqueloba to the one seen in M23258 and in MD99-2304 from ca. 10.5 ka BP is also seen further north in the Fram Strait (Aagaard-Sørensen et al., 2014; Werner et al., 2016; Consolaro et al., 2018; Figure 2A: 10, 12, and 13). The area around Kveithola, close to JM09-KA11-GC, did no longer record the direct influence of ice sheet disintegration but was still under the influence of deglaciation processes (Lantzsch et al., 2017). At the PSh-5159N core site, it has been argued that a warm, fresh summer mixed layer was replaced by sea ice during winter (Risebrobakken et al., 2010; Figure 2A: 2). Furthermore, it has previously been argued that the southwestern Barents Sea was in close proximity to the Arctic front at this time, with relative abundances of T. quinqueloba reaching up to 80% (Aagaard-Sørensen et al., 2010; Risebrobakken et al., 2010; Figure 2A: 2 and 7). Information is scarce regarding the conditions in the northern Barents Sea, but it is argued that cold conditions prevailed at least until 8–7 ka BP (Lubinski et al., 2001). Again, existing sea ice reconstructions from the northern Barents Sea do not reach as far back in time (Belt et al., 2015; Berben et al., 2017; Figure 2A: 14). The relation between the sea ice extent and the BSAF is therefore unknown for this time interval.

Based on the information gained from our synthesis, supported by the existing literature as reviewed above, we infer that the BSAF during this time interval also followed the western Barents Sea margin, however, it turned further eastwards into the southern Barents Sea, reaching PSh-5159N (Figure 2B₂).

At ca. 10.2 ka BP, a significant and abrupt change which resulted in an overall dominance of the Arctic front indicator T. quinqueloba is observed in the planktic foraminiferal fauna at all sites except PSh-5159N and T-79-51/2 (Figure 2C). Turborotalita quinqueloba was already the dominant species in PSh-5159N whereas the Atlantic Water species N. incompta remained the most abundant in T-79-51/2 (Figure 2C). Even if the relative abundance of T. quinqueloba should be somewhat underestimated in T-79-51/2, the conclusion that T-79-51/2 was bathed by Atlantic Water (Figure 2B₃) would not change. This overall dominance of T. quinqueloba lasted until ca. 8.8 ka BP. In particular, MD99-2304, M23258 and T-88-2 recorded mean relative abundances of T. quinqueloba of 56, 42, and 64%, respectively (Figure 2C; Table 3). Since the planktic foraminiferal abundance in M23258 is counted at ≥150 μm instead of ≥100 μm, the relative contribution of T. quinqueloba is likely underestimated at this site compared to MD99-2304 and T-88-2 (Husum and Hald, 2012). In that case, T. quinqueloba would be more dominant than already indicated. In JM09-KA11-GC and PSh-5159N the increased abundances of T. quinqueloba were, relative to the previous time interval, more gradual, respectively from 30 to 45% and from 65 to 84% (Figure 2C and Table 3).

In the Fram Strait and west of Svalbard similarly high abundances of T. quinqueloba have been recorded for this time interval (Werner et al., 2013, 2016; Aagaard-Sørensen et al., 2014; Rasmussen et al., 2014; Consolaro et al., 2018; Figure 2A: 10, 11, 12, and 13). In these studies, it has been argued that the increased T. quinqueloba abundances were associated with enhanced influence of Atlantic Water and increased sea surface temperatures. The Polar front, and the highly productive surface waters associated with the front, moved northwards (Consolaro et al., 2018; Figure 2A: 12), and the increased absolute abundance of planktic foraminifera seen west of Svalbard reflected a change from Polar to Atlantic Water at ca. 9.6 ka BP (Rasmussen et al., 2014; Figure 2A: 11). Werner et al. (2016) observed a limited sea ice extent in the northern Fram Strait between 11 and 8.5 ka BP (Figure 2A: 13). In the northern Barents Sea, the reduced spring sea ice concentrations recorded between ca. 9.5 and 8.5 ka BP were associated with a, relative to historical times, retreated ice margin (Berben et al., 2017; Figure 2A: 14). The lack of a correspondence between the reduced spring sea ice extent and a shift in the BSAF seems to imply that the sea ice extent is no driver behind the location of the BSAF in this part of the Barents Sea at this time. Although T. quinqueloba remained the dominant species in the southwestern Barents Sea, the occurrence of G. uvula has been attributed to an enhanced influence of Coastal Water (Aagaard-Sørensen et al., 2010; Figure 2A: 7).

We argue that the pronounced and sharp increase in T. quinqueloba seen within the three northernmost investigated cores on the western Barents Sea margin (Figure 2C), and in other sites west of Svalbard and in the Fram Strait (Werner et al., 2013, 2016; Aagaard-Sørensen et al., 2014; Rasmussen et al., 2014; Consolaro et al., 2018; Figure 2A: 10, 11, 12, and 13), reflects a rather fast eastwards migration of the BSAF (Figure 2B₃). More specifically, the high mean abundances of T. quinqueloba imply that the core sites were actually at the true location of the BSAF (Johannessen et al., 1994; Husum and Hald, 2012). The subtler fauna changes seen in JM09-KA11-GC indicate that the BSAF was slowly moving toward the core site but had not reached it yet (Figure 2B₃). The high relative abundance of T. quinqueloba in both T-88-2 and PSh-5159N implies a location close to the BSAF for both sites, and more specifically a front location just above (Figure 2B₃).

At ca. 8.8 ka BP, the foraminiferal assemblage in MD99-2304 and M23258 changed abruptly from a T. quinqueloba to a N. pachyderma dominance (Figure 2C and Table 3). Within these records, the mean values of the cold Polar Water species N. pachyderma reached 68 and 84% between ca. 8.8 and 7.4 ka BP, respectively (Figure 2C and Table 3). Even if the potential T. quinqueloba underestimation in M23258 is taken into account, the site would be totally dominated by N. pachyderma. A smaller, but still significant, decrease in the relative abundance of T. quinqueloba was recorded for T-88-2, from 64% to 37%, associated with a small increase in N. incompta and other Atlantic species and a significant increase in N. pachyderma (Figure 2C and Table 3). Contrary, in JM09-KA11-GC and PSh-5159N, the relative abundance of T. quinqueloba continued to increase (Figure 2C). In T-88-2, JM09-KA11-GC and PSh-5159N the Arctic front indicator T. quinqueloba was still the dominant planktic foraminiferal species throughout this time interval (Figure 2C and Table 3).

A similar increased abundance of N. pachyderma was observed at ca. 8.8 ka BP in a record from the western Svalbard slope, located between the MD99-2304 and M23258 core locations (Rasmussen et al., 2007; Figure 2A: 9). However, farther north on the west Spitsbergen continental margin, the Arctic front indicator T. quinqueloba remained dominant until ca. 7 ka BP (Werner et al., 2016; Consolaro et al., 2018; Figure 2A: 12 and 13). The latter core sites are from locations closer to the continental margin and at more shallow water depths compared to the more southern located cores from the western Svalbard slope, respectively MD99-2304 and M23258 (Figure 2A). A PIP₂₅ index record from the west Spitsbergen continental margin documented a reduced sea ice cover between ca. 8.5 and 7 ka BP (Müller et al., 2012; Figure 2A: 13). Reduced spring sea ice conditions were also recorded in the northern Barents Sea (Berben et al., 2017; Figure 2A: 14), even though the northward heat advection reached a minimum during this interval (Risebrobakken et al., 2011; Eldevik et al., 2014). Both northern latitude spring and summer insolation was, however, high (Laskar et al., 2004) and may have influenced the spring sea ice retreat. In the southwestern Barents Sea, the relative contribution of T. quinqueloba has been used as an argument for a continued influence of Arctic front conditions (Aagaard-Sørensen et al., 2010; Risebrobakken et al., 2010; Figure 2A: 2 and 7).

Based on the data presented in Figure 2C, combined with the existing literature, we argue that the BSAF migrated further eastwards into the Barents Sea from ca. 8.8 ka BP (Figure 2B₄). More specifically, the BSAF has moved over the two northernmost core sites at the western Barents Sea margin, reaching a position close to where it is located at present (Figure 2B₄). This could be seen as contradicted by the remaining high relative abundance of T. quinqueloba at the site located at the continental margin slightly further north, from where it is argued that the front was still located in the vicinity of the site (Consolaro et al., 2018; Figure 2A: 12). However, as the site discussed by Consolaro et al. (2018) is from a shallower location, this record rather confirms the strong topographic steering of the BSAF location. The increased dominance of T. quinqueloba in JM09-KA11-GC and the continuously high relative abundance of the same species in PSh-5159N strongly argues for the BSAF to be at, or at least very close to, these core locations ca. 8.8 to 7.4 ka BP (Figure 2B₄). The BSAF was, however, moving away from T-88-2, compared to the previous time interval, as shown by the decreased relative abundance of T. quinqueloba (Figures 2B₄,C). The increase in N. pachyderma contemporary with the decrease in T. quinqueloba is interpreted as an indication that T-88-2 transfers toward present conditions, with a rather high content of N. pachyderma due to the strong heat loss from the Atlantic Water before reaching the site. Overall, the observed changes in the planktic foraminiferal fauna reflects a continuous eastwards migration of the BSAF (Figure 2B₄). In the northwestern areas, the BSAF location was directed by the topography of the area, more specifically the western Barents Sea and BIT margins.

Between ca. 7.4 ka BP and present day, N. pachyderma continued to dominate the foraminiferal assemblages in MD99-2304 and M23258 with mean relative abundances of 93 and 89%, respectively (Figure 2C and Table 3). Neogloboquadrina pachyderma was also the most abundant species in T-88-2, occupying 58% of the planktic foraminiferal fauna. A gradual increase in relative abundance of N. incompta took place in the more southern located core T-79-51/2. In JM09-KA11-GC the relative abundance of T. quinqueloba stayed high, at 62%. Although T. quinqueloba remained the dominant species in PSh-5159N, with a stable value around 42%, the relative abundance of the front indicator was significantly smaller than in JM09-KA11-GC as well as compared with the previous time (Figure 2C and Table 3) interval. More specifically for PSh-5159N, increased abundances of N. incompta and other species (19%), indicative of warmer Atlantic Water (Bé and Tolderlund, 1971; Johannessen et al., 1994), and G. uvula (3%) associated with Coastal Water (Husum and Hald, 2012) have been recorded (Figure 2C and Table 3). Underestimated relative abundance of T. quinqueloba would not change the dominant species, neither in M23258 nor in T-79-51/2. Common for all sites is that the planktic foraminiferal assemblages in general represent an equivalent of the late Holocene assemblage compositions at the sites (Figure 2C).

The dominance of N. pachyderma between ca. 7.4 ka BP and present day is also seen at other sites on the western Svalbard margin, as well as further northeast into the Fram Strait, indicating generally cold conditions throughout the last ca. 7 ka BP (Werner et al., 2013, 2016; Rasmussen et al., 2014; Consolaro et al., 2018; Figure 2A: 10, 11, 12 and 13). Increased IP₂₅ concentrations after ca. 7 ka BP point to an expanded sea ice extent in the Fram Strait (Müller et al., 2012; Figure 2A: 13), corresponding to overall decreasing summer mixed layer temperatures in the Nordic Seas realm (Risebrobakken et al., 2011). Furthermore, for this time interval the overall cold conditions documented by planktic foraminifera at the western Svalbard and Barents Sea margin correspond to colder bottom water, or increased influence of Arctic Water indicators, as documented by benthic foraminifera from northern Svalbard and the northern Barents Sea (Duplessy et al., 2001; Slubowska et al., 2005; Klitgaard-Kristensen et al., 2013; Figure 2A: 15, 16, and 17). As compared to the previous time interval, the increased relative abundance of planktic Atlantic Water indicators seen in the southwestern Barents Sea sites, T-79-51/2, T-88-2 and PSh-5159N, as well as the decreased relative abundance of N. pachyderma seen at JM09-KA11-GC, correspond to a reduced sea ice extent in the western Barents Sea (Berben et al., 2014; Figure 2A: 5). The warming in the southwestern Barents Sea, combined with the cooling along the Svalbard margin and with the Arctic bottom Water in the northern Barents Sea, implies that a larger fraction of the NwAC entered the Barents Sea through the NCaC from ca 7.4 ka BP.

Based on the results summarized in Figure 2C, combined with the knowledge from previous studies, we argue that the BSAF was located close to its present position over the last ca. 7.4 ka BP (Figure 2B₅). The overall dominance of N. pachyderma along the western Svalbard and Barents Sea margin, indicative of a stronger influence of cold Arctic Water masses (Bé and Tolderlund, 1971; Johannessen et al., 1994; Pflaumann et al., 2003; Husum and Hald, 2012), is here interpreted as evidence for the sites being located further away from the BSAF, under conditions comparable to present-day (Figure 2B₅). The increased relative abundances of N. incompta and other Atlantic species in T-79-51/2, T-88-2, and PSh-5159N indicate an increased influence of Atlantic Water into the southwestern Barents Sea (Bé and Tolderlund, 1971; Johannessen et al., 1994), and hence an eastward movement of the BSAF relative to the previous time interval (Figures 2B₄₋₅). This interpretation is in line with previous findings from individual studies from the area (Risebrobakken et al., 2010; Berben et al., 2014; Figure 2A: 2 and 5). The Arctic front indicator T. quinqueloba still dominated the fauna at PSh-5159N, however, the strongly reduced relative abundance relative to the preceding time interval supports that the BSAF moved away from the site (Figures 2B₄₋₅). The relative abundance of T. quinqueloba was significantly lower in PSh-5159N than in JM09-KA11-GC, a site that was located underneath the BSAF (Figures 2B₅,C), as it is also today. The overall similarity between the planktic foraminiferal fauna of today, or the late Holocene, and the general fauna composition in all sites over the last ca. 7.4 ka BP supports the interpretation of a BSAF close to its present-day location (Figure 2B₅), with a strong topographic steering in the west, and the eastern position more defined by variable inflow of Atlantic Water by the NCaC. Some variability did occur in the relative abundance data also over the last 7.4 ka BP, especially in T-88-2 and in PSh-5159N (Figure 2C). We argue that this variability is of a different character than seen before. The changes in T-88-2 are primarily driven by more or less N. pachyderma balanced by less or more N. incompta, hence, indicating variability driven by a stronger or weaker heat loss in the NwAC before reaching T-88-2. The continuous occurrence of G. uvula in PSh-5159N documents significantly different conditions at this site after 7.4 ka BP, suggesting that the fauna variability was related to changes at the interface between Coastal and Atlantic Water rather than Atlantic and Arctic Water (Husum and Hald, 2012). Hence, all results combined point to a northeastward shift of the BSAF at ca. 7.4 ka BP into the Barents Sea toward the present-day location (Figure 2B₅).

Presently, the BSAF is well constrained by the topography of the western Barents Sea, following the northern margin of the BIT (Loeng, 1991; Parsons et al., 1996; Harris et al., 1998). In the eastern Barents Sea, the BSAF location is less stationary and dependent on the strength of the Atlantic Water inflow (Loeng, 1991; Parsons et al., 1996; Loeng and Drinkwater, 2007). The above discussion sets the development of the BSAF from the deglaciation through the early Holocene. Following the deglaciation, not only the Barents Sea oceanography, but also the upstream oceanography of the eastern Nordic Seas experienced significant changes (e.g., Risebrobakken et al., 2011; Eldevik et al., 2014). Since the present BSAF is in part tightly linked to upstream changes in the NwAC, the relation between upstream changes in the NwAC taking place as the BSAF established should be evaluated.

The northward heat advection in the NwAC reached a post glacial maximum ca. 10 ka BP due to a major reorganization of the Atlantic circulation through the deglaciation (Risebrobakken et al., 2011; Eldevik et al., 2014). A maximum heat advection at ca. 10 ka BP is supported by the Norwegian Sea planktic foraminiferal assemblage data shown by Zhuravleva et al. (2017). Strong heat loss did, however, take place before the NwAC reached the Barents Sea margin, and the Atlantic Water submerged further south than at present until ca. 9 ka BP (Risebrobakken et al., 2011). This southern submerging of the northward flowing Atlantic Water is reflected by enhanced abundance of Atlantic Water indicating benthic foraminifera in JM09-KA11-GC and PSh-5159N (Risebrobakken et al., 2010; Groot et al., 2014), when the planktic foraminiferal fauna shows that Arctic Water, or the BSAF, was located over the same sites at the same time (Figures 2B₁₋₃). The eastern Nordic Seas and the Barents Sea experienced significant melt water influence until ca. 9 ka BP, in association with the final deglaciation (Risebrobakken et al., 2010, 2011; Rasmussen et al., 2014), supporting the Arctic Water and BSAF conditions inferred from the synthesized relative abundance data.

The eastward movement of the BSAF, toward its present location at the northwestern Barents Sea and western Spitsbergen margins (Figure 2B₄), corresponds to the northward movement of the zone of submergence and the diminishing influence of melt water (Risebrobakken et al., 2011). The inflow of Atlantic Water to the Barents Sea was, however, smaller than at present until ca. 7.4 ka BP, also in line with the diminishing northward heat advection that followed the ca. 10 ka BP advection maximum and the gradual return to normal salinities (Risebrobakken et al., 2011; Eldevik et al., 2014).

From around ca. 7.4 ka BP, the BSAF has likely been located within the range of its present location in the eastern Barents Sea (Figure 2B₅), with the absolute location influenced by variable strength of the Atlantic Water inflow (Loeng, 1991; Parsons et al., 1996; Loeng and Drinkwater, 2007). Our results show that the investigated sites along the western Barents Sea margin and at the Barents Sea opening were bathed by Arctic Water following the deglaciation. A gradual transition took place when the BSAF moved eastward toward, and eventually over, the sites, before the present day oceanographic conditions settled (Figures 2B₁₋₅). This transformation of the Barents Sea, from an Arctic Water dominated basin to the present-day situation, with the southern Barents Sea being bathed by Atlantic Water, is consistent with the overall oceanographic changes that took place in the eastern Nordic Seas throughout the early Holocene.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.