In this article, the Atlantic inflow with the North Cape Current to the Barents Sea, and potential recirculation of this water in the Bear Island Trough is investigated. This is resolved by combining continuous current meter and repeated hydrographic section data in the Barents Sea Opening. The estimated annual mean net fluxes, when excluding the contribution from the Norwegian Coastal Current, are 1.1 Sv and 39 TW (estimated relative to a reference temperature of 0°C). The estimated heat flux associated with the outflow in the Bear Island Trough is relatively large (−12 TW). Long-term current meter records do not show a close relationship between variations in the inflow and the outflow. One-year low-pass filtered variations in temperature and salinity of the outflowing waters in the Bear Island Trough are positively correlated with variations of the inflowing Atlantic Water with a lag of about seven months. On the basis of a simple mixing model, the outflow water consists of about 80% Atlantic Water that is additionally cooled by 2.0–2.5°C by heat loss to the atmosphere. By considering the relative weak mean currents in this region, it can be inferred that the fraction of the Atlantic Water that recirculates follows a relatively short pathway in the Bear Island Trough before returning to the Norwegian Sea.
 The Norwegian Atlantic Current (NwAC) is a two-branch flow of Atlantic Water (AW) through the Norwegian Sea and thus the major source of oceanic heat and salt to the Arctic Ocean [Orvik and Niiler, 2002; Orvik and Skagseth, 2005]. In the Barents Sea Opening (BSO), the shelf edge branch of the NwAC bifurcates and partly continues into the Barents Sea as the North Cape Current (NCaC). This flow is fundamentally important for the ecosystem of the Barents Sea, both with respect to advection of fish larvae and zooplankton [Skjoldal et al., 1992], and for climate conditions [e.g., Ottersen and Loeng, 2000]. Further, variations in the NCaC may trigger decadal scale climate variations through internal feedback mechanisms [Ikeda, 1990; Aadlandsvik and Loeng, 1991].
 The first detailed description of the NCaC was given by Blindheim , who, based on an array of current meter moorings in the fall of 1977 estimated a total inflow across the BSO of 3.1 Sv [1 Sv = 106m3s−1]. In the northern BSO (Bear Island Trough - BIT), a deep intensified outflow of 1.2 Sv was observed, which was interpreted as brine-enriched water [Knipowitsch, 1905; Midttun, 1985].
 On the basis of long-term current meter monitoring in the BSO, Ingvaldsen et al.  found that the strength of the NCaC was highly dependent on the local wind forcing, where divergence in the surface Ekman layer sets up barotropic currents. On the basis of this array, the estimated mean fluxes from 1997 to 2006 of the AW inflow are 1.8 Sv and 48 TW [1 TW = 1012 Js−1] with large inter-annual variations [Skagseth et al., 2008]. However, these estimates do not capture the contributions from the Norwegian Coastal Current (NCC) of the order 10 TW based on the mooring array of Blindheim , the water with T < 3°C, and the fluxes associated with the outflow in the BIT.
 To link fluxes across the BSO to the interior Barents Sea climate and ecosystem variations, the fate of the AW is crucial. It is generally accepted that after passing the BSO, the NCaC splits into two branches (Figure 1). One branch flows eastwards, whereas the other branch continues toward the southern flank of the Hopen Trench (HT). The latter branch has been described in two conceptually different models. Loeng  suggested that the AW in the Hopen Trench subducts below the Polar Water farther east into the Barents Sea where it leaves the area as a deep flow of dense brine-enriched water in the northeast into the Arctic Ocean [Loeng et al., 1993], and into the Norwegian Sea [Blindheim, 1989]. In contrast to this concept, Gawarkiewicz and Plueddemann  argued for a barotropic recirculation of AW in the BIT at bottom depths greater than the interior Barents Sea sill depth (about 260 m) with a return flow along the southern flank of the Spitzbergen Bank (SB). Anomalies from the southern part of the BSO are found to appear in the northern part with a time lag of the order months [Furevik, 2001; Ingvaldsen, 2005]. Because of sparse current measurements in the northern part it has to date been unresolved whether this relation is caused by variation in the width of the NCaC or by recirculation of AW.
 In this study, the focus is on the circulation of AW in the western Barents Sea. Advantage has been taken of a recent successful two-year current meter deployment that provides information about the flow on the flank of the Spitzbergen Bank. First, the improved coverage of these data is utilized in combination with the hydrographic monitoring in the BSO, to estimate the mean fluxes in BSO, including both inflow and outflow. Then the measured velocity variations in these cores are compared. Finally, to resolve the inter-annual variations a similar analysis is done for the hydrographic changes. On the basis of these analyses, inferences about the fate of the AW that enters the BSO can be made.
2. Data and Method
 The Institute of Marine Research has operated an array of current meter moorings in the Barents Sea Opening since 1997. In this study, a subsample of these data is used with focus on the period from 2003 to 2005 when an upward looking RDI 150 kHz ADCP was deployed on the flank of the Spitzbergen Bank (Figure 2). The combined time series of currents at 50 m depth and 15 m above bottom are used to capture the spatial and temporal variations in the BSO fluxes. The hydrographic variations are resolved using sections in the BSO, occupied four to six times per year, and more irregularly prior to the mid 1980s (Figures 2 and 3) .
 To obtain the velocity field needed for the flux estimates, the current meter and hydrographic data are combined. First, the vertical shear is calculated using the thermal wind relation. Then the current meter data are interpolated along the bottom, and horizontally at the 50 m depth into the positions of baroclinic currents. The geostrophic velocity field is then adjusted so that the offset from the measured current at 50 m and near the bottom is minimized in a least square sense (Figure 2). The heat flux is estimated relative to the temperature of 0°C that is close to the outflow from the Barents Sea to the Arctic Ocean in the northeast [Loeng et al., 1993]. Note that the hydrographic sections can be regarded as quasi-synoptic snapshots. To reduce the effect of high frequency variations, bi-annual means from 2003 to 2005 are used for both the hydrographic sections and the current meter data.
 Covariation between the measured Atlantic inflow and the outflow in the BIT is resolved by comparing three-month low-pass filtered observed current in the cores of the in- and outflow (Figure 4). The velocity of the inflow is represented by the spatial mean current on the southern flank of the BIT (72.5–73.0°N), and the outflow velocity is represented by the near bottom current at 73.5°N. The similar covariation in the water mass characteristics is resolved be considering hydrographic data since the 1960s (Figure 3). The properties of the inflow water are defined as the mean between 72–73°N over depths from 50 to 200 m, whereas the properties of outflow water are taken as the depth average between 300–450 m in the BIT at 73.5°N. The fraction of AW in the outflow water and the effect of local cooling are evaluated with a mixing model containing three water masses:
where (T, S)BIT are the properties of the outflowing water in the BIT, (T, S)A are the properties of the inflowing AW, and (T, S)P are the properties of the Polar water. HF is the ocean to air heat flux, t represents the time the AW is exposed to the atmosphere before gaining density, subducting, and sinking into the BIT, ρ is the density, z is depth of the AW, and cp is the specific heat capacity. Taking the water mass characteristics from the data, the fraction of AW in the outflowing water (α), and the AW cooling by heat loss from the ocean to the air (HFt/ρzcp) can be found.
 Averaged over the period from the fall of 2003 to the fall of 2005, there is an eastward flow into the Barents Sea of warm AW in the southern part, and a bottom intensified westward flow in the northern part (Figure 2). In the BIT, centered at 73.5°N, there is a shear zone associated with strong horizontal temperature (and density - not shown) gradients that lead to decreasing westward velocities toward surface. A substantial fraction of the outflow in the northern part can be interpreted as AW when assuming the regional definition of AW as water with T > 3°C [e.g., Ingvaldsen et al., 2004]. However, colder and denser water masses occupy the region near the bottom of the BIT and on the flank of the Spitzbergen Bank.
 Considering the whole water column (Figure 2) there is a net inflow through the section of 1.1 Sv and 39 TW, resulting from an inflow of 2.0 Sv and 51 TW in the southern part and an outflow of 0.9 Sv and 12 TW in the northern part.
 The current meter records show only a weak positive correlation (not statistically significant, r = −0.22, and neff = 21) between the inflow in the southern part and the outflow in the northern part from 1997 to 2006 (Figure 4).
 Despite some scatter for the individual samples, there is a close correlation between the low-pass filtered hydrographic variations (Figure 3). The series show prominent inter-annual variability, including the absolute minimum in the late 1970s with the return of the Great Salinity Anomaly to the Nordic Seas [Dickson et al., 1988]. The mean salinity of the Atlantic inflow is about 35.1, and about 35.0 for the outflow in the BIT. The maximum correlation is r = 0.83 at seven months lag, that is, the variability of the outflow trails that of the inflow. The mean temperature of the inflowing AW is about 5°C, and the outflow temperature is about 2°C. Also for temperature, there is a close correlation between the series with r = 0.84 at a lag of seven months. When taking the outflow properties higher up on the Bear Island Slope, a similar close relation to the AW was found, but the lag was reduced to about three months at minimum (not shown).
 In the solution of equations (1) and (2), the following water mass definitions are used; Bear Island Trough water: T = 2°C, S = 35.0, the inflowing AW: T = 5°C, S = 35.1; the Polar water: T = 1°C, S = 34.5. On the basis of these values, the outflow water consists of an 83% fraction of AW that has been cooled by 2.3°C through heat loss to the atmosphere.
 The estimated fluxes across the BSO put constraints on both the budgets of the Barents Sea and the circulation in the western Barents Sea. Thus it is of interest to evaluate some of the major uncertainties. In the method, there is an uncertainty in combining data with different sampling that is, mean currents from continuous records with mean hydrographic fields based on snap-shots. Though hard to quantify, this error is reduced by averaging over a number of hydrographic sections and considering time scales longer than a month. Then the effect of eddies that otherwise is prominent on the shorter time scales is reduced [Ingvaldsen et al., 2004]. The neglected fluctuating part of the heat flux; H′ = ρcp<v′T′>, based on the hourly sampled raw data, is estimated to be less than 1 TW. Ingvaldsen et al.  have found that the fluxes in the BSO depend heavily on the atmospheric forcing. However, as the mean sea level pressure (mslp) fields for the two years period, 2003–2005, are relatively similar to a normal year (not shown), the estimates are likely comparable to the long-term mean. An exception from this is in the summer of 2005 when seasonal anomalously strong southwesterly winds were prominent in the western Barents Sea leading to increased Atlantic inflow across the BSO. More important are long period variations in temperature that contribute substantially to variations in the heat flux. Assuming constant volume flux, the observed 1°C temperature increase of the AW flow into the Barents Sea from the 1970s to 2007 [Skagseth et al., 2008] leads to a 20% increase of the heat flux.
 On the basis of the above-mentioned analysis, the annual mean net flux estimate across the BSO is 39 TW with an uncertainty of ±5 TW. Including the contribution from the Norwegian Coastal Current of 10 TW, assuming a similar uncertainty of ±5 TW, gives a total net flux in the range of 49 TW ± 7 TW. The somewhat higher estimated inflow of 2.0 Sv compared with 1.5 Sv by Ingvaldsen et al. , is at least partly due to the additional contribution from regions where T < 3°C (Figure 2).
 The high correlation between the properties of the Atlantic inflowing water and the outflow water in the northern part of the BSO indicate that there is a relatively direct pathway. Harris et al.  argued for a relatively long pathway along the 260 m isobaths with mixing and cooling en route. Intuitively, the high correlation between the variations of the AW inflow and the deep outflow is more in accordance with a relatively short pathway and thus a relatively short time scale. Using mean current of ∼3–5 cm/s, and not 30 cm/s as used by Harris et al. , and a time scale of seven months based on the lagged correlation analysis (Figure 3), a pathway of the order 500–1000 km is obtained. This indicates that the major fraction of this water does not intrude very far into the Hopen Trench. Further, the analysis indicates that the deepest outflow water spends more time within the Barents Sea relative to the water leaving on the northern slope. Defining an area of 250 km by 250 km, based on a mean pathway of 750 km, the required ocean to air heat loss to obtain the 2.0–2.5°C temperature drop of the AW at a rate of = .75 Sv is 98 W/m2. In comparison the long-term annual mean heat loss based on the NCEP/NCAR reanalysis [Kalnay et al., 1996] in the BIT between 20–30°E is about 90 W/m2.
 This recirculation along a relatively short pathway suggests a significant topographic control given the strong barotropic component (or presence of flow at depth) to the flow, which is in accordance with Gawarkiewicz and Plueddemann . However, the weak correlation between velocities in the inflow and outflow (Figure 4) somewhat contrasts this idealized view, and probably emphasizes the importance of the large-scale wind forcing on the flow field in the western Barents Sea [Ingvaldsen et al., 2004].
 Because of the relative short time-lag of seven months, it is natural to question how the seasonal cycle of the ocean to air heat flux would affect the modification of the AW from the inflow en route the outflow in BIT. The temperature of the Atlantic Inflow (TA) shows a clear seasonality with maximum in September and minimum in April. Similarly, a seasonal cycle is clear in the outflow temperature (TBIT), but here the values are shifted about half-year forward in time. The atmospheric reanalysis show seasonal maximum/minimum ocean to air heat loss a few months after the maximum/minimum for TA. From this, the seasonal cycle in the outflow relative to the inflow temperature should be reduced, which is in fact in accordance with the observations (not shown). A more detailed analysis of the causal mechanisms and forcing of the circulation in the western Barents Sea are beyond the scope herein, but is the focus of ongoing projects.
 Annual mean fluxes in the BSO are estimated by combining hydrographic section data and long-term current meter data. Excluding the contribution from the Norwegian Coastal Current, the net fluxes across the BSO are 1.1 Sv and 39 TW. The heat flux associated with the outflow in the Bear Island Trough is considerable and needs to be included in heat budgets for the Barents Sea.
 The hydrographic changes in the inflowing AW and the outflowing water in the BIT can be explained by a simple mixing model that includes Polar waters and realistic ocean to air heat loss. On the basis of the high correlation including a short time lag, it can be inferred that the outflowing water in the BIT is modified AW en route a relatively short pathway into the Barents Sea.
 This work has received contributions from the Norwegian Research Council projects PROCLIM, NorClim, POCHAHONTAS, ESSAR, and IPY projects iAOOS and DAMOCLES. Thanks to K. Drinkwater, K. Orvik and P. Haugan for comments on earlier versions of the manuscript.