Mooring data indicate the Bering Strait throughflow increases ∼50% from 2001 (∼0.7 Sv) to 2011 (∼1.1 Sv), driving heat and freshwater flux increases. Increase in the Pacific-Arctic pressure-head explains two-thirds of the change, the rest being attributable to weaker local winds. The 2011 heat flux (∼5 × 1020J) approaches the previous record high (2007) due to transport increases and warmer lower layer (LL) temperatures, despite surface temperature (SST) cooling. In the last decade, warmer LL waters arrive earlier (1.6 ± 1.1 days/yr), though winds and SST are typical for recent decades. Maximum summer salinities, likely set in the Bering Sea, remain remarkably constant (∼33.1 psu) over the decade, elucidating the stable salinity of the western Arctic cold halocline. Despite this, freshwater flux variability (strongly driven by transport) exceeds variability in other Arctic freshwater sources. Remote data (winds, SST) prove insufficient for quantifying variability, indicating interannual change can still only be assessed by in situ year-round measurements.
 Pacific Waters (PW) dominate about half of the upper Arctic Ocean [McLaughlin et al., 1996]. They provide ∼1/3rd of all Arctic freshwater [Serreze et al., 2006]. Supplying 10–20% of the oceanic heat input to the Arctic, warm (summer) PW trigger sea-ice retreat in the western Arctic and provide an under-ice, upper water column heat source [Woodgate et al., 2010]. Cold (winter) PW maintain the western Arctic cold halocline [Aagaard et al., 1981]. In situ measurements from the Bering Strait, the only Pacific-Arctic gateway, showed extreme oceanic fluxes in 2007, which likely influenced that year's precipitous sea-ice retreat [Woodgate et al., 2010]. Given continued Arctic sea-ice retreat and increased Arctic freshwater variability, we consider changes in PW fluxes to 2011.
2. Estimating Bering Strait Oceanic Fluxes of Volume, Heat and Freshwater
2.1. Data Sources
 The only oceanic Pacific-Arctic gateway is the ∼50 m deep, ∼85 km wide Bering Strait (Figure 1). The flow in the strait is dominantly homogenous, with a surface-intensified, warm, fresh, ∼10 km wide coastal current (the Alaskan Coastal Current, ACC) present seasonally in the east [Woodgate et al., 2005a, and references therein]. Stratification (summer-only) is typically 2-layer with the upper layer warmer and ∼1 psu fresher [Woodgate and Aagaard, 2005]. Hourly current-meter and temperature-salinity sensor data from year-round moorings in the strait suggest mooring A3, ∼60 km north of the center of the strait (Figure 1), gives a reasonable average of lower layer flow properties and, with surface and ACC corrections, can be used to infer total fluxes [Woodgate et al., 2006, 2010]. We use these techniques to infer Bering Strait oceanic fluxes to 2011 (Figure 1). The strong coherence of velocity in the region allows volume transport also to be estimated from the strait moorings A1 or A2.
 Volume transports are calculated from hourly northward velocity (at ∼45 m depth) at A3 assuming flow homogeneity (laterally and vertically) and a cross-section area of 4.25 × 106m3. This neglects the ∼0.1 Sv of the ACC (1 Sv = 106m3/s) [Woodgate and Aagaard, 2005]. A 7-mooring array in the strait deployed from 2007 (not shown) shows that away from the ACC spatial flow variability is <10% and suggests this simplistic transport quantification is reliable to ∼0.1 Sv, possibly with a systematic bias of <20%.
 Heat transports are calculated hourly from a combination of transport and temperature data. Since PW are generally at freezing on leaving the Arctic [Steele et al., 2004], to assess the PW heat contribution to the Arctic we calculate heat fluxes relative to a typical freezing point of Bering Strait waters (−1.9°C). Lower layer (LL) temperatures are taken from mooring data at ∼45 m depth. Upper layer temperatures are estimated hourly from the NOAA daily (0.25° spatial resolution) Optimum Interpolation satellite sea surface temperature (SST) product [Reynolds et al., 2007] averaged over a 60 km × 80 km cross-strait region centered on A3. Heat transports are estimated assuming upper layer thickness of 10 m or 20 m and a constant annual contribution from the ACC of 1 × 1020 J [Woodgate et al., 2006, 2010].
 Freshwater transports are also calculated hourly from transport and mooring salinity data from ∼45 m depth, relative to an Arctic mean salinity of 34.8 psu [Aagaard and Carmack, 1989]. Lacking long-term sea surface salinity measurements, to account for stratification, the ACC and freshwater from sea-ice transport (∼140 ± 40 km3/yr [Travers, 2012]) we assume a constant annual contribution of 800–1000 km3/yr [Woodgate and Aagaard, 2005; Woodgate et al., 2006] for these terms.
2.2. Annual Means From 1991 to 2011
 Since 2002 (with one exception, 2005), annual mean (AM) transports (Figure 1) exceed the climatological transport value of 0.8 Sv [Roach et al., 1995]. The AM volume transport increases by over 50% from 2001 (∼0.7 Sv) to 2011 (∼1.1 Sv). The 2001–2011 trend is 0.03 ± 0.02 Sv/yr. (Trend errors in this paper are the 95% level of 1-sided Student's t-test, equivalent to 90% of 2-sided t-test.) There is no significant trend over the 1990–2011 period.
 Lower layer AM temperatures are warmer in 2002–2007 and 2010–2011, with 2002, 2003, 2007 and 2011 being the warmest on record. In contrast, AM SSTs are warmest in 2004 and 2007, but for 2008–2011 are cooler, comparable with the 1990s. ACC temperatures (from mooring site A4) are similar to the SST although in 2008 and 2009 ACC temperatures are colder, due at least in part to the current being present for less of the year. Lower layer AM salinities remain within 0.25 psu of the climatological value of 32.5 psu [Aagaard and Carmack, 1989] for all the record, except 1991.
 Heat and freshwater (FW) fluxes depend strongly (heat ∼50%, FW >90%) on volume flux changes.
 The 2007 heat flux, (5–6 × 1020 J/yr, ∼15–20 TW [Woodgate et al., 2010]) is still the highest recorded, although the 2011 heat flux is also ∼5 × 1020 J, despite the surface layer contributing only half as much heat as in 2007. The 2011 heat flux increase is due to increased transport and increased warming of the lower water column not the surface layer.
 The freshwater (FW) flux variability here, calculated only from lower layer changes, likely underestimates the real variability, which will reflect ACC and stratification changes. Nevertheless, this estimate suggests Bering Strait FW fluxes increase from 2000–2500 km3 in 2001 to 3000–3500 km3 in 2011. Within a decade, AM values vary by ∼1000 km3, roughly twice the variability in net precipitation (∼500 km3 [Serreze et al., 2006]) or river run off (Russian rivers, ∼400 km3 [Shiklomanov and Lammers, 2009]) in the same period.
3. What Are the Drivers of Interannual Variability?
3.1. Increased Transport due to Pressure-Head and Wind Changes
 The Bering Strait throughflow is attributed to two forcings – a Pacific-Arctic pressure-head term (PH) and the local wind velocity (Wvel) which in the annual mean opposes the northward flow (see, e.g., discussion inWoodgate et al. [2005a]). Following Woodgate et al. , we estimate the varying contributions from each by assuming a linear fit – “Transport = PH + C × Wvel”, where C is an unknown parameter representing wind-flow coupling. Taking for Wvel the NCEP (National Centers for Environmental Prediction) 10 m 6-hourly wind at various strait sites (Figure 1) in the direction best correlated with the water velocity (∼330°), we calculate C and PH as coefficients of a least squares fit of Wvel and the estimated transport for 1-year segments of data (Figures 1g and 1h). Results suggest both forcings increase northward flow from 2001 to 2011 – by increased PH and by (in the annual mean) weaker southward winds. Best fit trends for this period suggest the flow increase due to changes in the PH term (0.024 ± 0.011 Sv/yr) is almost twice that due to the weakening of the winds. However, neither term alone indicates all years with maximum flow.
 Our results suggest a ∼30% increase in the PH term from 2001 to 2011.
3.2. Oceanic Warming Greater Than SST Warming
 About 50% of the increased heat flux relates to ocean warming: years with large heat fluxes are years with warmer lower layer temperatures (LLTs) (Figure 1c) but not always high SSTs – e.g., 2011 AM SSTs are lower than in 2003–2007 and comparable with the 1990s (Figure 2). An inverse relationship between LLT and SST might reflect increased mixing. Although solar heating warms the upper layer, enhanced vertical mixing will move this heat downward, yielding lower layer warming and upper layer cooling.
 Thirty-day smoothed data indicates that recent LLTs rise above 0°C roughly 20 days earlier than in the late 1990s (1998–2011 trend: 1.6 ± 1.1 days/yr;Figure 2c). This may relate to an observed 3-fold increase in northward transport in recent springs (also seemingly pressure-head driven) speeding advection of heat from the south (not shown). The corresponding trend in SST is not significant (0.8 ± 1.1 day/yr). Although SST rises above 0°C earlier in the year than LLT (Figure 2c), cooling is coincident throughout the water column (Figure 2d) since the water column is less stratified in autumn [Woodgate et al., 2005b]. There is no significant trend in the onset of cooling. These changes do not contribute greatly to the AM heat flux, which is more strongly influenced by the higher temperatures of the later summer (July–October).
3.3. A Multidecadal Perspective From Remote Data?
 How remarkable are the recent oceanic flux increases? NCEP wind data since 1948 (Figures 2f and 2g) suggest that recent winds are uncommon, but not unprecedented. However, wind changes explain only ∼1/3rd of the recent transport increase, and there is, so far, no proven way of estimating past variability of the larger PH term prior to the mooring record. While we might expect higher flows in the low wind years of 1989 and the late 1940s, without a better understanding of the processes setting the PH, such extrapolations are dubious.
 SST data since 1982 (Figure 2) also show no significant trends, and SST changes are not coincident with LLT changes, showing SST to be of limited use in estimating total heat content. However, the timing of SST autumn cooling does coincide with LLT cooling, suggesting that SSTs may be used as a proxy for the timing of water column cooling.
 Thus, currently, in situ measurements remain the only method of estimating Bering Strait oceanic fluxes to the accuracy required to quantify recent interannual change.
4. Impacts on the Water Column of the Chukchi and the Arctic
 The heat flux through the Bering Strait, about 1/3rd of the Fram Strait heat flux, acts as a trigger for sea-ice retreat in the Chukchi and western Arctic and a subsurface source of heat to about half of the Arctic Ocean [Woodgate et al., 2010]. Although upper layers may be significantly modified during transit of the Chukchi, prior data suggest that lower layers are cooled, but undergo remarkably little salinity change [Woodgate et al., 2005a]. We thus consider lower layer water properties at A3 to investigate interannual variability of waters sequestered in the Arctic.
4.1. Two Distinct Pacific Water Contributions to the Arctic
 The temperature-salinity (TS) properties of the lower layer at A3 range from ∼29–34 psu and from the freezing point (−1.9°C) to ∼8°C in the period 1998–2011 (Figure 3, first column). Greater insight however comes from considering the volume of water transported through the strait in each TS-class (Figure 3, second column).
 In the multiyear mean, two modes are apparent – a Pacific Summer Water (PSW) mode dominantly 1–4°C, 32–33 psu, and a Pacific Winter Water (PWW) mode at the freezing temperature but with a greater salinity range ∼31.5–33.5 psu. (This analysis neglects the warmer, fresher upper layers which are 10%–20% of the volume – 10–20 m layer present half the year.) These modes are of roughly equal volume, consistent with the seasonal cycle of temperature (warm ∼ June to December) being offset from the seasonal cycle of transport (high transports ∼ April to September) [Woodgate et al., 2005b]. It is perhaps surprising that the salinities of these modes overlap (Figure 3, third column) – especially in recent warm years (2007 and 2011), warm waters are found at salinities greater than much of the PWW. There is significant interannual variability in the temperatures and TS distributions of the PSW. High salinities in PWW are rarer in recent times.
 Remarkably, for all the period 1998 to 2011, waters above 0°C exhibit a sharp cutoff in salinity, being always fresher than 33.1 psu. Summer waters may be fresher than this salinity, but higher salinities are only found in cold (<0°C) waters. Even then the salinity change is modest (<1 psu), consistent with formation of <2 m of ice in 50 m of water. (Mean Bering Strait ice thickness for 2007 is ∼1.5 m [Travers, 2012].) The 33.1 psu salinity cutoff suggests a quasi-constant maximum salinity source for PSW, likely relating to shelf break depth upstream in the Bering Sea – indeed monthly National Ocean Data Center climatological salinities show that at 50 m most of the Bering Sea Basin is ∼33–33.2 psu, with fresher waters on the shelf.
4.2. A Pacific Control on Temperatures and Salinities of the Arctic Halocline?
 The western Arctic Ocean water column TS is characterized by a contraction in temperature towards freezing at a salinity (∼33.1 psu) which is remarkably constant over recent decades [e.g., Woodgate et al., 2005c, Figure 2]. The sharp salinity cutoff for Bering Strait warm waters may explain this – there is no source of waters with significant heat at salinities ∼33.1 psu since Atlantic waters are too salty and, as found here, Pacific waters at these salinities (or higher) are always cold. Thus suggests the constancy of this Arctic property is set by conditions in the Bering Sea.
 Note that in the western Arctic, the term Pacific Summer Water (henceforth APSW) refers to the temperature maximum (Tmax) in salinity range 31–33 psu above this TS contraction [Steele et al., 2004], e.g., Figure 3 (temperature maximum in yellow lines). However, a comparison with TS from the Chukchi (Figure 3) indicates that the APSW Tmax is actually only a cooled remnant of the deeper part of the warm Bering Strait/Chukchi Sea PSW. Thus spatial and temporal variability of the APSW Tmax depends not just on source changes and likely changing pathways but also on the processes removing heat from the upper part of the Pacific Waters.
4.3. Pacific Water Contribution to Heat in the Arctic Water Column
 Finally, we consider where Bering Strait heat may be stored in the Arctic. Figure 3 (fourth column) shows the heat flux per density class entering through the strait. The upper layers (missing from this summation) contribute heat at lower densities. The summation presented indicates that between 0.2 to 0.6 × 1020 J/yr are contributed at sigma-0 values around 26 kg/m3, equating to a ventilation depth of 50–75 m in the Arctic Ocean.
 Mooring data indicate a >50% increase in the Bering Strait throughflow from 2001 (∼0.7 Sv) to the end of the present record in 2011 (∼1.1 Sv), and combined with NCEP wind data, suggest that 2/3rds of this increase is attributable to an increase in the Pacific-Arctic pressure-head. Predicting future changes in the Bering Strait throughflow thus requires understanding of the origin and variability of this pressure-head.
 Making a 2-layer assumption and assuming a constant Alaskan Coastal Current (ACC) contribution, we estimate heat and freshwater fluxes, both of which depend strongly (heat ∼50%; freshwater >90%) on transport changes. Freshwater fluxes rise from 2000–2500 km3 in 2001 to 3000–3500 km3 in 2011, and exhibit interannual variations of ∼1000 km3, which is larger than the interannual variability in other Arctic freshwater sources. Our assumptions likely underestimate the true variability, since stratification and the ACC contribute about 1/3rd of the flux. Heat fluxes increase from ∼3 × 1020 J in 2001 to ∼5 × 1020 J in 2011 (comparable to the record of 2007). Again this also probably understimates variability – ACC temperatures (A4 in Figure 1c) show more variability than Sea Surface Temperatures (SSTs). Lower layer temperatures (although not SSTs) suggest that Bering Strait waters warmed earlier from 2001 to 2011 (1.6 ± 1.1 days/yr), possibly relating to increased transport of warm water from the south in the spring. The onset of autumn cooling, coherent over the water column, shows no significant change over time.
 The 2011 Bering Strait heat flux is enough to melt ∼1.5 × 106 km2 of 1 m thick ice. A volumetric analysis of water properties shows that significant quantities of heat (0.2 to 0.6 × 1020 J/yr) are associated with surprisingly high densities (σ0 ∼ 26 kg/m3), suggesting an Arctic ventilation depth of 50–75 m. About equal volumes of cold (<0°C) and warm (>0°C) waters are supplied to the Arctic. Interestingly, despite large variations in temperature (∼2°C), Pacific Summer Water salinities never exceed 33.1 psu in the record available. This curiously stable maximum salinity, likely with origins in the Bering Sea, may explain the stable salinity of the western Arctic cold halocline.
 Local winds and SST miss much of the variability important for quantifying the Bering Strait oceanic fluxes, viz., the pressure-head forcing and temperature variability at depth. Thus, in situ observations remain currently the only viable method of assessing interannual change in the Pacific inflow to the Arctic.
 This work was funded by NSF (ARC-0632154 and ARC-053026) with ship-time from NOAA-RUSALCA. We thank J.Johnson, D.Leech and the scientists and crews of the Alpha Helix, the CCGC Sir Wilfrid Laurier, and the Russian vessels Khromov, Sever, and Lavrentiev for their dedication to fieldwork; M.Schmidt for MODIS imagery; NOAA/OAR/ESRL PSD Boulder Colorado for the NOAA-High Resolution SST data (http://www.esrl.noaa.gov/psd); and NODC for Bering Sea climatology data (http://www.nodc.noaa.gov/OC5/PACIFIC2009/index.html).
 The Editor thanks two anonymous reviewers for their assistance in evaluating this paper.