Department of Fisheries and Oceans, Bedford Institute of Oceanography, Dartmouth, Nova Scotia, Canada
Corresponding author: K. Azetsu-Scott, Department of Fisheries and Oceans, Bedford Institute of Oceanography, 1 Challenger Dr., PO Box 1006, Dartmouth, NS B2Y 4A2, Canada. (Kumiko.Azetsu-Scott@dfo-mpo.gc.ca)
 Freshwater transport through Davis Strait can supply additional buoyancy to the deep convection region of the Labrador Sea which influences the strength of the meridional overturning circulation and consequently the global climate. The freshwater contribution from local sea ice meltwater, meteoric water (fluvial, glaciofluvial and precipitation) and the Arctic outflow were quantified using oxygen isotope composition (δ18O), salinity and nutrient relationships in September–October, 2004. Freshwater transported by the Arctic outflow was isolated using a modified nutrient relationship method and further deconvoluted into sea ice meltwater, meteoric water and Pacific water. For the first time, fluxes of individual freshwater components were estimated using observations of the velocity field derived from mooring arrays and geostrophic currents from hydrography. The Arctic outflow dominated in western Davis Strait (>60%) and its influence extended eastward close to the Greenland Slope. The sea ice meltwater fraction was small (<2%) and limited to the surface layer of the central and western Strait. The meteoric water fraction was highest on the Greenland Shelf (>6%) and attributed to glacial meltwater. The freshwater inventory of the 0–100 m layer was equivalent to 7.4 m in western Davis Strait: 8 m from the Arctic outflow and −0.6 m from brine rejection. In eastern Davis Strait, the freshwater inventory was 4 m: 3 m from meteoric water and 1 m from sea ice meltwater. The Arctic outflow contributed 82–99 mSv to the southward freshwater transport about 67–81% of the total; glacial meltwater contributed the largest northward transport of 10–30 mSv.
 Freshwater dynamics in the northern North Atlantic Ocean have important implications for global climate. Modeling studies have indicated that the response of the Atlantic meridional overturning circulation (MOC) is sensitive to changes of the surface buoyancy fluxes [e.g., Cheng and Rhines, 2004; Gregory et al., 2005]. Perturbation experiments showed weakening of the thermohaline circulation by 30% after 100 years in response to an additional 0.1 Sv (1 Sv = 106 m3 s−1) freshwater input in the northern North Atlantic [Stouffer et al., 2006]. Freshwater input to in the Labrador and the Nordic Seas potentially influences deep convection regimes, therefore, the MOC. Observed changes in sea ice cover in the Arctic Ocean [Stroeve et al., 2008], hydrological cycles at high latitudes [Peterson et al., 2006] and decreasing volume of the Greenland Ice Sheet [Velicogna, 2009; Mernild et al., 2009] will potentially alter the buoyancy flux to the deep convection regions. Identifying the freshwater sources and their contributions to the high-latitude ocean is important for predictions of future climate.
 In the northern North Atlantic, freshwater sources include local precipitation, river runoff, sea ice, icebergs and glacial meltwater, and the Arctic Ocean outflow which carries less saline water into the North Atlantic. The influence of different freshwater sources varies in magnitude, time scales and the areas they affect. Changes in freshwater fluxes can influence not only circulation and mixing, but also air-sea CO2 uptake, nutrient cycles, biological responses [Else et al., 2008] and ocean acidification [Bates et al., 2009; Yamamoto-Kawai et al., 2009; Azetsu-Scott et al., 2010].
 Davis Strait is an ideal gateway to observe the propagation of changes from the Arctic into the Labrador Sea, the intrusion of the warm and saline Atlantic water into Baffin Bay and the freshwater contribution of the Greenland Ice Sheet to the northern North Atlantic. The volume transport and freshwater fluxes through Davis Strait are similar in magnitude to those through Fram Strait [Curry et al., 2011], the other major pathway of Arctic water to the subpolar North Atlantic. The purposes of this study are to quantify the freshwater sources and their distributions and to estimate the fluxes of each freshwater component through Davis Strait.
2. Sampling Area and Methods
 Water exchange between Baffin Bay (central depth 2300 m) and the Labrador Sea (central depth 3700 m) occurs in the ∼300 km wide Davis Strait, with a maximum sill depth of 640 m (Figure 1). The northward West Greenland Current has two components, a less saline branch called the West Greenland Shelf Current (WGSC) and more saline component from the Irminger Sea (West Greenland Irminger water, WGIW) concentrated on the continental slope. At Davis Strait, these currents can divide, with part of the flow entering Baffin Bay on the eastern side of Davis Strait and contributing to the cyclonic circulation in the Bay and part continuing westward as the Labrador Sea cyclonic circulation [Tang et al., 2004]. The cold and fresh Baffin Island Current (BIC) is the integrated Arctic outflow through the Canadian Arctic Archipelago (CAA) mainly from Nares Strait and Lancaster Sound. The BIC flows southward along Baffin Island coast and slope, and through western Davis Strait. Modified by Hudson Strait outflow, the BIC becomes a component of the Labrador Current, which continues flowing southward, mainly confined to the shelf and upper slope. The BIC and Labrador Current are modified in transit by local inflows, ice formation and melting, and mixing with offshore waters.
 The sections of temperature, salinity and potential density (σθ) and the T-S diagram just north of the sill in Davis Strait are shown inFigure 2. Four distinct water masses are identified following the definitions of Tang et al.  and Curry et al. . Arctic water (AW) occupies the surface 300 m on the western side of the section with the core T < 0°C and S < 33.7; AW represents the water that has flowed through the CAA. The northward flowing WGSC is found in the upper 100 m with T > 4°C and S < 33.5 on the Greenland Shelf. The WGIW (T > 2°C and S > 34.5) flows northward along the West Greenland Slope at the depths of 200–500 m and can recirculate in the western Strait or continue northward as part of the Baffin Bay cyclonic circulation. The densest (σθ > 27.65) Baffin Bay deep water (BBDW) occupies the deepest part of the section with T < 2°C and S > 34.5.
 Water samples were collected from 22 September to 4 October 2004 on board the R/V Knorrusing CTD-Rosette system (SeaBird 911+). Along the mooring line, water samples were collected at 13 of the 18 hydrographic stations (designated ML,Figure 1). The current measurements were made at 9 mooring sites using acoustic Doppler current profilers and single point current meters [Curry et al., 2011]. The deepest current meters were at 500 m. Nutrient samples were frozen and later analyzed at Bedford Institute of Oceanography following the World Ocean Circulation Experiment (WOCE) protocols using a Technicon Autoanalyzer with the precision of 0.19 μmol/kg for nitrate and nitrite (NO3− + NO2−), and 0.04 μmol/kg for phosphate (PO4). Oxygen isotope samples were collected in 30 ml HDPE Nalgene sample bottles, wrapped with a Parafilm and stored at room temperature. They were analyzed with a FISONS PRISM III with a Micromass multiprep automatic equilibration system at Lamont-Doherty Earth Observatory. Two milliliter subsamples were equilibrated with CO2 gas (8 h at 35°C). Data are reported with respect to standard mean ocean water (SMOW) with the δ18O notation. The external precision based on replicates and standards is ±0.033‰.
3. Freshwater Decomposition
 A conventional freshwater decomposition using salinity and δ18O measurements has three end-members, namely sea ice meltwater, meteoric water and seawater. When several types of seawater coexist, the first step is to determine the composition of seawater masses. In this study, we reference two base water masses, the Arctic and the Atlantic waters. After determining the ratio of these two water masses, seawater end-member values for salinity andδ18O are calculated for each sample. Then the fractions of sea ice meltwater, meteoric water and seawater are determined for each sample using measured salinity and δ18O with three end-member values. Arctic water is further decomposed into Pacific water, sea ice meltwater and meteoric water using the end-member values from the formation site in the Arctic Ocean.
3.1. Seawater End-Member Composition
 The first step in the determination of freshwater components is to estimate the Arctic and Atlantic water contributions using the nutrient relationships. Jones et al.  estimated Pacific water fraction using the sum of nitrate and nitrite (NO3− + NO2−) versus phosphate (PO4) relationships. This method is based on a nutrient characteristic of Pacific water entering the Arctic Ocean, specifically lower NO3− + NO2− concentrations due to denitrification in the Bering Sea [Cooper et al., 1997]. Yamamoto-Kawai et al.  took into account the possible overestimate of the Pacific water fraction caused by the further denitrification of Pacific water on the Arctic shelves by including the ammonium (NH4+) concentration. Based on the T-S properties (Figure 2), we have established an Arctic water line, instead of the Pacific water line, in order to evaluate the integrated freshwater transport by the Arctic outflow to the North Atlantic. Further decomposition of Arctic water into Pacific, sea ice meltwater and meteoric water allows us to differentiate contributions of meteoric and river runoff in the local Davis Strait area from those in the distant Arctic.
Figure 3 shows NO3− + NO2− and PO4 concentrations from the mooring and northern lines (ML and NL, Figure 1). The Atlantic water reference line was calculated from the Labrador Sea data (BioChem database, http://www.meds-sdmm.dfo-po.gc.ca/biochem/Biochem_e.htm) collected in 2001–2007 with S > 34.2 and T > 1.14°C (n = 795) defined by the T-S diagram. The Arctic water line was determined from samples from this study, with S < 33.7 and T < 2°C (Figure 2) and PO4 > 0.7 μmol/kg (n = 15). The Pacific lines used by Yamamoto-Kawai et al.  and Jones et al.  are also shown in Figure 3. We had NH4+ measurements only from the NL (Figure 1), consequently NH4+concentrations along ML were estimated using their relationship with T-S and depth. These ammonium adjustments caused the total inorganic nitrogen (TIN), the sum of NO3−, NO2− and NH4+ concentrations, on the Greenland Shelf to fall on the Atlantic line (Figure 3). The NH4+ concentrations were zero for the western and central parts of the NL, therefore no corrections were necessary for these observations. The δ18O and salinity values for seawater end-members were then determined for each sample using the ratio of Arctic and Atlantic water and given theδ18O and salinity values in Table 1. For example, if the seawater end-member were a mixture of 20% Arctic water and 80% Atlantic water for the given sample determined from the method above, seawater end-member values for salinity andδ18O would be S = 0.2 × 31.70 + 0.8 × 34.75 = 34.14 and δ18O = 0.2 × (−1.83) + 0.8 × 0.19 = −0.21‰, respectively.
Table 1. End-Member Values and Their Uncertainties Used in This Study
 In this method, the slopes of two source seawater lines (Figure 3) must be equal, otherwise the composition of two water masses cannot be determined uniquely and an infinite number of solutions will exist. This assumption, however, has often been ignored in earlier studies. The slopes for Arctic and Atlantic water lines derived in this study were tested using an Ftest and shown to be equal within the statistical uncertainty. This is one of the advantages of using the Arctic water as a part of the seawater end-member rather than the Pacific water references from other studies, whose slopes were not the same as that of Atlantic water (Figure 3). Uncertainties for the estimate of source seawater composition arise from the errors related to determination of the Arctic and Atlantic lines. The 95% confidence intervals for each line are shown in Figure 3. While the Atlantic line was robust and uncertainties due to the regression estimate are less than 1%, the Arctic line is less robust and includes large uncertainties at higher nutrient values. Uncertainties of seawater end-member estimates caused by the linear regression of the Arctic line are <10% at PO4 = 0.9 μmol/kg and 25% at PO4 = 1.2 μmol/kg. Higher nutrient concentrations (PO4 > 1.2 μmol/kg) were observed at depths >400 m.
3.2. Freshwater Composition
 Fractions of sea ice meltwater (FSIM), meteoric water including precipitation, river runoff and glacial meltwater (FMW) and seawater (FSW) are quantified using δ18O and salinity (S) with specified end-member values (Table 1) using the following relationships:
where the subscripts, sample, SIM, MW and SW, represent the sample measurement, sea ice meltwater, meteoric water and seawater (mixture of Arctic and Atlantic waters), respectively. End-member values ofδ18O and salinity and their uncertainties for meteoric water, sea ice meltwater and the two seawater types are summarized in Table 1. From FSWand the results from the Seawater End-Member Composition in the previous section, Arctic water fraction (FArc) and Atlantic water fraction (FAtl) were determined (Figure 4a).
 The sensitivities to the variability of end-member values for calculating FSIM, FMW and FSW were examined. The salinity and δ18O of the end-member values were varied one at a time by ±one standard deviation and compared with the mean values. For example, the sensitivity of FSW, FSIM and FMW for salinity of sea ice meltwater variation was evaluated using SSIM = 3 and 5 and obtained FSW, FSIM and FMW were compared with those calculated using the mean value, SSIM= 4, while other end-member values including salinity andδ18O of Arctic and Atlantic waters, δ18OSIM and δ18OMW, were fixed. The paired ttest was used to determine how the variations in end-member values influence FSW, FSIM and FMW estimates. The results are summarized in Table 2. Here tα equals 4.13, 3.44 and 2.65 for p = 0.0001, 0.001 and 0.1, respectively. In Table 2, when |t| ≥ tα (α is a significance level), the differences in FSIM, FMW or FSWcalculated from the varied end-member values are significant at p =αlevel. This provides an assessment of the relative importance of each end-member value for the estimates of freshwater components, FSW, FSIM and FMW. For example, the variation of the salinity end-member value for sea ice meltwater (SSIM) by ±1 does not influence the estimates of FSW, FSIM and FMW significantly, while the salinity variation in Atlantic water (SAtl) by one standard deviation, ±0.14, resulted in highly significant differences in the FSW and FSIM estimates. The differences caused by SAtl for FSIM and FMWestimates were 0.3 ± 0.2% and 0.01 ± 0.01%, respectively. Since most of the samples were near the mixing line of seawater and meteoric water, particularly close to the seawater end-member (Figure 4b), the FSIM and FMWcalculations are more sensitive to the seawater end-member values, less so to meteoric water end-member values, and least influenced by the sea ice meltwater end-member values. Therefore, it is important to define salinity andδ18O end-member values for Atlantic water carefully.
Table 2. Sensitivity Analyses of the End-Member Variation for Estimating the FW Componentsa
t Values for the Paired t Test
For p < 0.0001, t = 4.13, for p < 0.001, t = 3.44 and for p > 0.01, t = 2.65.
Sea ice meltwater
 Consequently, we used salinity and δ18O observations from the adjacent Labrador Sea for the Atlantic water end-members to avoid modification over extended advective pathways.
 Overall, uncertainties of the FSIM, FMW and FSWcalculations are due to a combination of the seven end-member variations above (SMWis 0 with 0 S.D.). They were evaluated using Monte Carlo simulations. For the seven end-member values, a thousand random numbers, which were normally distributed with the averages and standard deviations given inTable 1, were created. Using these randomly created end-member values, the propagation of uncertainties was calculated for each sample. Distributions of uncertainties for FArc, FAtl, FSIM and FMW are shown in Figure 4c. Uncertainties are not uniform either among freshwater fractions or in space for each freshwater fraction type. Here, absolute values of uncertainties are less important than the relative uncertainties, namely the coefficient of variation (CV = standard deviation/mean × 100%). The highest uncertainties for FArc are observed in the surface waters of Stations 2, 3, 5 and 7. Since FArc is high (>90%, Figure 4a) in this region, the CV is low (2.6%). Therefore, FArc estimates are considered robust. In contrast, the highest CV was observed for FSIM in the same region where it is close to zero (Figure 4a); thus the calculated uncertainties (Figure 4c) make the FSIM estimates less reliable.
3.3. Flow Fields
 The fluxes of total freshwater and the individual components, namely Arctic water, sea ice meltwater and meteoric water, were calculated using two current fields, one based on data from an extensive mooring array (MA) [Curry et al., 2011] and the second, a geostrophic field (GS) estimated using the CTD observations from 18 stations and assuming near bottom velocities of 0 (Figure 5). Both current fields exhibit bimodal southward flows between the Baffin Island coast and ∼150 km offshore. Currents are southward over the Baffin Island Shelf and Slope, and in the central part of the Strait. These two features are separated by a region of weak northward flow, centered on mooring C2 (MA) and CTD stations 4 and 5 (GS). Northward flow dominates on the Greenland Shelf and Slope (MA), but considerably less so in the geostrophic field. The spatial structure of both current fields is qualitatively similar. However, there are differences.
 The major differences between these two flow fields are a lack of northward flow over the Greenland Slope in the GS estimate compared to the MA field and the velocities of the northward flow on the Greenland Shelf (∼0.06 m s−1 in MA, compared to ∼0.02 m s−1 in GS). Less obvious differences are seen in the northward flow off the Baffin Island Slope (>0.02 m s−1 in MA and almost 0 m s−1 in GS), and in the southward flow at the central part of the Strait (0.08 m s−1 in GS and 0.04 m s−1 in MA). In addition, the MA field extrapolated to the Baffin Island coast is considerably stronger than the GS flow in the same area. While the spatial resolution of the hydrographic field (18 stations, observations from the surface to the bottom) is better than that of the mooring array (9 sites, deepest measurements at 500 m), the major difference in the two fields arises from the assumption of zero velocity near the bottom for the GS estimates. This is particularly true for the Greenland Slope regions where the mooring array indicates a significant barotropic current characterized by mean bottom flows of 0.06 m s−1 (averaged between moorings C5 and WG1), with an associated volume transport of 0.75 Sv, not realized in the geostrophic (baroclinic) estimate. In the deep area of the Strait between the 200 m isobaths, the RMS difference of the MA and GS bottom currents is 0.022 m s−1, giving rise to an increased northward transport of 0.58 Sv in the MA field. Moreover, the mooring data provide a synoptic picture of the flow, whereas the CTD section, the basis for the GS field, took several days to complete. Curry et al.  gave uncertainties of 30–40% for flux estimates from the mooring measurements. For the geostrophic current field, an error was estimated using temperature time series from moored instruments and vertical temperature gradient to determine the variance of isopycnal displacement, applying the displacement randomly at CTD sites for 10,000 trials. This yielded an uncertainty of the GS estimate of about 35%.
4. Results and Discussion
4.1. Freshwater Distribution and Sources
 Since salinity and temperature ranges were used to define the nutrient relationship for Arctic water, it is not surprising that Arctic water identified by multiple chemical tracers corresponds well with that determined by salinity and temperature alone. An advantage of using multiple chemical tracer method, however, is to provide additional constraints on the estimate of Arctic water in the surface and subsurface layers, where salinity and temperature can be strongly influenced by surface warming and by various freshwater sources. This is demonstrated by identifying a surface expression of Arctic water and a clear subsurface extension of Arctic water toward West Greenland Slope (Figure 4a), although the latter is also seen in temperature section (Figure 2). The Arctic water fraction (FArc) is centered at 50–100 m in western Davis Strait close to Baffin Island, and extends over 150 km toward the eastern side of Davis Strait (Figure 4a). The bottom topography along the western margin of Davis Strait and the tendency of currents to follow f/h contours (f is Coriolis parameter and h is a bottom depth) may divert the Arctic water eastward as it approaches the sill just south of the ML section [Tang et al., 2004; Cuny et al., 2005]. This subsurface spread of Arctic water in Davis Strait may facilitate mixing of Arctic water in the northern Labrador Sea and therefore the transport of Arctic water into its interior. On the other hand, the subsurface recirculation of saline and warm water close to Baffin Island originating from the North Atlantic observed in CTD measurements (Figures 2 and 4a) could not be resolved due to the sparser sampling of the chemical tracers.
 A significant fraction (≈ 20%) of Arctic water is also observed below the sill depth (>600 m). The origin of Baffin Bay deep water is still debated [Tang et al., 2004]. Our results indicate that a significant fraction of this deep water mass is composed of Arctic water.
 Sea ice meltwater is only a small fraction (<2.5%) of the surface waters in central and eastern Davis Strait (Figure 4a); its contribution is not significantly different from zero in the deeper waters of the Strait. Distribution of sea ice meltwater is highly seasonal and this may contribute to the low values seen in this study. Brine rejection, shown as FSIM < 0, was observed in the subsurface and the bottom layers of the section. Samples with high brine rejection (FSIM < −2) occur in water with high FArc, suggesting that ice formed in Arctic water is separated from the Arctic outflow and melted elsewhere or that the sea ice meltwater and Arctic water arrive at Davis Strait at different times.
 Meteoric water is largely found on the West Greenland Shelf, where the highest fraction (6%) is observed, and it extends over the slope as a thin surface layer. The runoff from the glaciers and direct precipitation contribute to this meteoric water. The summer runoff contribution from the glaciers in Disko Bay (Figure 1) was estimated as 800–1000 m3 s−1 [Ettema et al., 2009; Rignot et al., 2010]. The average monthly precipitation from July to October, when the highest values are observed, at Disko Island (69°15′N, 53°31′W) is 0.05–0.09 m with mean annual average of 0.44 m (based on 1991–2004 observations [Hansen et al., 2006]). The freshwater contribution from precipitation, which is estimated for the shelf region immediately outside Disko Bay (shelf width of 100 km and 75 km along the axis parallel to Greenland) during summer, is equivalent to 145 to 260 m3 s−1. Therefore, the contribution from glacier meltwater is the dominant component of the observed meteoric water fraction on the central Greenland Shelf. Although Disko Bay is north of our study site, glacier inputs south of the study area will influence the observed FW components similarly. No direct regional observations of δ18O values for glacial meltwater are available. The linear regression of salinity and δ18O observations from the Greenland Shelf shows that the intercept at zero salinity is −24‰ (δ18O = 0.70 × salinity − 23.9, r2 = 0.86), which agrees well with observations adjacent to the glacial terminus of an East Greenland fjord at a similar latitude [Azetsu-Scott and Tan, 1997]. Using δ18O = −24‰ as an end-member value for glacial meltwater, the meteoric fraction on the Greenland Shelf changes from 6% (usingδ18O = −20‰) to ∼5%, a change of less than 1%, which is balanced by less than 1% increase in sea ice meltwater fraction.
 In summary, the Arctic water contributed about 80% of the freshwater on the western side of Davis Strait close to Baffin Island in 2004. The largest contribution from meteoric water was over the Greenland Shelf with a maximum value of 6%. Sea ice meltwater contributions generally amounted to only 1–2%, with negative values corresponding to the thick cold intermediate layer off Baffin Island, indicative of brine rejection.
4.2. Inventories of Total Freshwater and Contributions From Sea Ice Meltwater, Meteoric Water and Arctic Water
 Freshwater inventories of each fraction were calculated. First, the total freshwater fraction (FWtotal) was calculated by FWtotal = 1 − Ssample/SAtl, where SAtl is the reference salinity of Atlantic water (34.75; Table 1). The contribution of Arctic water to FWtotal, (CFArc), was calculated as CFArc = FArc × (1 − SArc/SAtl). The FArc and SArc are the fraction and salinity of Arctic water estimated previously and defined in Table 1. In a similar manner, the contribution of sea ice meltwater (CFSIM) to the total freshwater was calculated by CFSIM = FSIM × (1 − SSIM/SAtl), while the contribution of meteoric water to the total freshwater CFMW is CFMW = FMW. By integrating over depth with respect to FWtotal, CFArc, CFMW and CFSIM, inventories of the total and individual components of freshwater were estimated for each station (Figures 6a and 6b).
 For the entire depth, more than 17 m of freshwater was calculated for stations 2–7, within 90 km of Baffin Island. This high freshwater content was due to the contribution from Arctic water of over 20 m, a much smaller contribution from meteoric water (<3 m), and offsetting brine rejection of 4–5 m. On the West Greenland Shelf and Slope, meteoric water from the Greenland Ice Sheet was the dominant component and contributed 3–5 m of freshwater, while less than 1 m was accounted for by sea ice meltwater; there was no freshwater contribution from Arctic water.
 Meteoric water on the Greenland side is considered to originate predominantly from glacial meltwater. The freshwater inventory was calculated using an glacial water end-member,δ18OMW = −24‰, and compared to the results using δ18OMW = −20‰ (Table 1). The difference was a reduction by 0.6 m on average for meteoric water using δ18OMW= −24‰ and an increase of sea ice meltwater contribution by the equivalent amount at stations 13–17. For the western side of the Strait, the differences derived by using the two end-members were not significant.Figure 6shows the freshwater inventories in the upper 100 m (using an end-member valueδ18OMW = −24‰ to estimate the glacial meltwater component) to eliminate biases due to the bottom topography. The relative importance of Arctic water at the western side of Davis Strait is evident with about 8 m of freshwater, while glacial meltwater on the Greenland Shelf and Slope had ∼3 m of freshwater.
4.3. Fluxes of Total Freshwater and Their Components
 The gridded flow fields were multiplied by the freshwater fractions interpolated to the same grid. Although there are differences in the two current fields, the freshwater fluxes are similar. The southward component of the total freshwater flux is 125 ± 44 mSv (MA) and 120 ± 42 mSv (GS), while the northward component is 52 ± 18 mSv (MA) and 11 ± 4 mSv (GS) (Figure 7). The difference of the northward flux estimates is attributed mainly to the flow over the Greenland Shelf and Slope (see section 3.3) (Figure 5). Arctic water is the dominant component of the southward freshwater flux and contributes around 99 mSv (MA) and 82 mSv (GS). Sea ice meltwater and meteoric water fluxes are small (both about 20 mSv) for the GS estimates. The MA estimates also have small southward flows for both components (12–14 mSv) but a much stronger meteoric water flux to the north (31 mSv) due to glacial meltwater on the Greenland Shelf. The net sea ice meltwater flux estimates (7 mSv for MA and 18 mSv for GS, Figure 7) are comparable to Ingram and Prinsenberg's value for the mean annual solid sea-ice export through Davis Strait of 35 mSv (liquid water equivalent) and the smaller values, 11 to 12.9 mSv, ofCurry et al.  and Cuny et al. . The northward Arctic water flux is around 17 mSv (MA), which we attribute to the recirculation off Baffin Island (centered on mooring C2), and which was very weak in the GS field.
 The Arctic water component is further deconvoluted to Pacific water, sea ice meltwater and meteoric water portions using end-member values reported previously [Ekwurzel et al., 2001; Yamamoto-Kawai et al., 2008] in the Arctic (Table 1). Pacific water is the largest contributor for the southward flux of freshwater (105 mSv in the MA and 88 mSv in the GS estimates; Figure 7 inserts). This Pacific water component in Arctic water is also responsible for the low pH and calcium carbonate saturation state in this region [Azetsu-Scott et al., 2010]. The meteoric flux is 77 mSv (MA) and 64 mSv (GS), most likely dominated by outflow from the Mackenzie River. No significant flux of sea ice meltwater was observed, instead the brine rejection component is large (84 mSv in MA and 70 mSv in GS) in the Arctic water flux, which almost compensates the freshwater component due to Pacific water flux. The total ice volume export to Baffin Bay through Lancaster Sound is estimated to be 112 km3 a−1 [Agnew et al., 2008]. If a 3 month melting period is assumed, this would produce a little over 10 mSv of sea ice meltwater. However, the high brine rejection observed in this study indicates that ice formed in Arctic water was separated from the main Arctic outflow before reaching Davis Strait and melted elsewhere or sea ice meltwater and brine rejection reach Davis Strait at a different time. Fram Strait and Canadian Arctic Archipelago are the major conduits of Arctic waters to the North Atlantic. Fresh and cold Arctic water flows southward to the North Atlantic on the western side and warm and salty Atlantic water flows northward on the eastern side of Fram Strait with a net freshwater flux estimated to be 119 ± 40 mSv to the south [Rudels et al., 2008]. Meredith et al.  estimated 3680 km3 yr−1 (116 mSv) for a meteoric water flux in the summer of 1997 and 2000 km3 yr−1(63 mSv) for 1998 and half that amount for the sea ice flux. They used different end-member values, including higher Atlantic water salinity as a saline end-member (SAtl = 34.92), which influences the fraction calculation as well as freshwater inventories and fluxes. Therefore direct comparison is difficult, however, the indication is that the compositions of the freshwater fluxes differ between the two straits: a lower flux of sea ice melt together with high brine rejection to the south through Davis Strait than through Fram Strait; a distinct glacial meltwater flux to the north through Davis Strait compared to none in Fram Strait.
 The rapidly changing polar environment such as melting sea ice in the Arctic and the Greenland ice sheet melt can affect the freshwater composition and fluxes in Davis Strait. This study is a part of an ongoing time series program and the analysis for interannual variability from 2004 to 2011 is underway.
 Freshwater composition and inventory including sea ice meltwater, meteoric water and Arctic water, across Davis Strait were estimated using nutrient relationships, δ18O and salinity observations collected from late September to early October of 2004. Further, fluxes of each component were calculated using current fields estimated from the mooring array and from a geostrophic calculation based on hydrographic measurements. The Arctic water, less saline than the receiving Atlantic water, dominates both in the freshwater inventory and the southward flux into the Labrador Sea. Total freshwater transport to the south is around 120 mSv with 70–80% transported by Arctic water. This Arctic outflow from the Canadian Arctic Archipelago extends eastward about 200 km from Baffin Island to central Davis Strait as far as the base of the Greenland Slope. Within the Arctic water component, Pacific water transports 88–105 mSv of freshwater followed by meteoric water (64–77 mSv); these are compensated by a flux of high brine rejection water (70–84 mSv). On the Greenland Shelf and Slope, meteoric water, considered to be largely glacial meltwater from the Greenland Ice Sheet, dominates the inventory and the total freshwater flux to the north which ranges from 11 to 52 mSv.
 The authors thank the captain and crew of R/V Knorrfor their seamanship and Richard Fairbanks and Rick Mortlock for oxygen isotope analyses. Beth Curry provided the current field derived from the mooring array. This study was funded by U.S. National Science Foundation Freshwater Initiative (2004–2007) and the International Polar Year and Arctic Observing Network (2007–2010) programs under grants OPP0230381 and OPP0632231. Also additional support was provided by the N-CAARE, Department of Fisheries and Oceans, Canada, and IPY-Canada.