Journal of Geophysical Research: Oceans

Pacific freshwater, river water and sea ice meltwater across Arctic Ocean basins: Results from the 2005 Beringia Expedition

Authors


Abstract

[1] Pacific water, sea ice meltwater, and river water are the primary sources of freshwater in the Arctic Ocean. We have determined their relative fractions on a transect across the Arctic Ocean Section 2005 Expedition onboard IB Oden, which took place from 21 August to 23 September 2005. The transect began north of Alaska, continued through the central Canada Basin to the Alpha Ridge and into the Makarov Basin, and ended in Amundsen Basin. Pacific freshwater and river water were the major sources of freshwater throughout the central Canada Basin and into Makarov Basin, with river water fractions sometimes considerably higher than Pacific water in the top ∼50 m. Pacific freshwater extended to depths of about 200 m. Pacific water found over the Alpha Ridge and in the Amundsen Basin is suggested to have been transported there in the Transpolar Drift. The inventories of Pacific freshwater and river water were roughly constant along the section through most of the Canada and Makarov basins. River water fractions were greater than those of Pacific freshwater in the Amundsen Basin. Sea ice meltwater fractions were negative (reflecting net ice formation) or near zero throughout most of the section. A comparison of freshwater inventories with those at stations occupied during expeditions in 1991, 1994, and 1996 indicated an increase in river water inventories in the Makarov and Amundsen basins on the Eurasian side of the Arctic Ocean.

1. Introduction

[2] Understanding climate and climate change is a main motive for determining the freshwater budget of the Arctic Ocean, i.e., freshwater sources, their distributions and their pathways. The Atlantic Ocean experiences more evaporation than precipitation. Much of this excess evaporated water falls as rain into the Pacific and Arctic oceans and into river drainage basins, all of which feed into the Arctic Ocean. This freshwater leaves the Arctic Ocean, entering the Greenland Sea and the Canadian Arctic Archipelago on the way back to the Atlantic Ocean. The freshwater is not immediately transported to where it was evaporated, however. The climate change concern is that, in returning to its evaporation sites, the freshwater passes through regions of deep convection in the Nordic and Labrador seas, the “headwaters” of thermohaline circulation [Aagaard and Carmack, 1989]. To understand the impact of freshwater on deep convection, we must know the sources and timing of the freshwater flows to regions where the deep convection occurs. How the freshwater sources are redistributed within the Arctic Ocean together with the place and timing of their exit from the Arctic Ocean are of direct relevance to the development of models giving scenarios of changes, possibly abrupt, in the Atlantic thermohaline circulation [e.g., Rahmstorf, 1996].

[3] Freshwater within the Arctic Ocean includes Pacific water (fresher than North Atlantic water), river water, and sea ice meltwater as well as precipitation. Serreze et al. [2006] estimated freshwater inputs to the Arctic Ocean to be large, ∼8500 km3, but an order of magnitude less that the oceanic freshwater storage within the Arctic Ocean. Changes in upper waters of the Arctic Ocean are well documented. Part of the redistribution of freshwater in the Arctic Ocean has been related to changing from one Arctic Oscillation mode to another [Proshutinsky et al., 2002; Hakkinen and Proshutinsky, 2004]. River water distribution within the Arctic Ocean has been observed to vary over a period of years [Anderson et al., 2004; Björk et al., 2002; Schlosser et al., 2002; Macdonald et al., 1999; Steele and Boyd, 1998]. River discharge into the Arctic Ocean from the six largest Eurasian rivers has increased [Peterson et al., 2002] but discharge from northern Canada has decreased [Déry and Wood, 2005].

[4] River water and sea ice meltwater mix with waters of both Pacific and Atlantic origin to form the Polar Mixed Layer. In the central Arctic Ocean in summer, sea ice meltwater can introduce freshwater (buoyancy) into the Polar Mixed Layer. In winter, brine, produced as sea ice forms, results in “negative” sea ice meltwater fractions that reflect sea ice growth and adds negative buoyancy to the Polar Mixed Layer. In shallower shelf waters the brine can mix to the bottom adding to waters below the Polar Mixed Layer, with some eventually reaching deep waters [e.g., Rudels et al., 2004]. Also, sea ice meltwater in the central Arctic Ocean originates as relatively warm Atlantic water entering the Arctic Ocean encounters and melts sea ice to form what evolves into the lower halocline of the Arctic Ocean [Rudels et al., 1996, 2004].

[5] During summer 2005, as part of the Beringia 2005 Expedition, the icebreaker Oden crossed the Arctic Ocean from north of Barrow, Alaska, to Svalbard, carrying out the Arctic Ocean Section (AOS-05). The expedition took place from August 21 through September 23, 2005. IB Oden was the first surface ship to cross the central and northern Canada Basin and the first to obtain measurements of Arctic Ocean freshwater components in this large, previously considered inaccessible region. The AOS-05 section began north of Alaska, proceeded across the Canada Basin, over the Alpha Ridge and into the Makarov Basin to the Lomonosov Ridge, and followed a path roughly parallel to the Lomonosov Ridge to a gap in the ridge previously thought to be deep enough for cold Eurasian Basin water to pass through [Jones et al., 1995]. The section continued across the ridge and proceeded through the Amundsen Basin to the Gakkel Ridge (Figure 1). Heavy ice in the Makarov and Amundsen basins slowed steaming speeds, leaving no time for the continuation of the section across the Nansen Basin.

Figure 1.

The AOS-05 section stations (solid circles). The different bathymetry colors correspond to depths of 500 m, 1000 m, 1500 m, 2000 m, 3000 m, 4000 m, and 4500 m, respectively. Geographical regions are identified for this figure and Figure 3e: Canada Basin (CB), Chukchi Cap (CC), Makarov Basin (MB), North Pole (NP), Amundsen Basin (AB), Nansen Basin (NB), Alpha Ridge (AR), Lomonosov Ridge (LR), and Gakkel Ridge (GR). Pacific freshwater lies on the North American side of a front denoted by the solid orange line (see sections in Figure 3).

[6] We report relative Pacific freshwater, river water and sea ice meltwater fractions and their inventories in the surface waters along the section.

2. Approach

[7] Nutrient concentration maxima have traditionally been invoked to trace Pacific water within the Arctic Ocean [e.g., Wheeler et al., 1997] and in waters exiting from it [e.g., Anderson and Dyrssen, 1981]. A difficulty with the nutrient maxima approach is that throughout the Canadian Basin silicate and nitrate are mostly depleted in the upper low salinity regions where the freshwater fractions are highest. Ekwurzel et al. [2001] employed the concept of PO* (PO* = 175PO4 + O2) [Broecker et al., 1985] to distinguish Pacific water. An analogous difficulty with the PO* approach in surface waters is that oxygen exchange with the atmosphere renders PO* no longer a conservative tracer.

[8] It is by now well-established that nutrient relationships can trace Pacific water in the Arctic Ocean and in waters flowing from it [Jones et al., 1998, 2003; Jones and Anderson, 2008; Carmack, 2000; Yamamoto-Kawai et al., 2006, 2008]. While nitrate and phosphate concentrations themselves are not conservative, relative changes resulting from biological processes, Redfield relationships, are conservative. Within the topmost layers, where nutrients are depleted by biological processes, nutrient relationships can distinguish Pacific water from Atlantic water, river water and sea ice meltwater.

[9] In the Arctic Ocean, alkalinity [e.g., Anderson and Dyrssen, 1981; Anderson et al., 1994, 2004; Yamamoto-Kawai and Tanaka, 2005] and oxygen isotopes, δ18O, [e.g., Ostlund and Hut, 1984; Macdonald et al., 1999; Ekwurzel et al., 2001] have been used to trace river or meteoric water. These are similar but not exactly equivalent tracers, a main difference being that alkalinity traces river runoff, whereas δ18O would include precipitation over land and within the Arctic Ocean. Here we designate all meteoric water as river water (see discussion below).

[10] Using alkalinity and salinity we can construct three equations to relate the relative fractions of Atlantic water, Pacific water, river source water, and sea ice meltwater:

equation image
equation image
equation image

where f is a water fraction, S is salinity, AT is alkalinity, and the superscripts, PW, AW, si, r, and m, designate Pacific water, Atlantic water, sea ice meltwater, river water, and measured values. In this approach, sea ice meltwater is the amount of freshwater otherwise not accounted for.

[11] Equations (1) to (3) with four unknowns do not admit to unique solutions. One more independent equation is required to determine the fractions, f. In the Arctic Ocean, waters of Pacific and Atlantic origin each have their own linear phosphate (PO4) vs. nitrate (NO3) relationship [Jones et al., 1998, 2003; Yamamoto-Kawai et al., 2006] (Figure 2). The fraction of Pacific water, fPW, in a sample with particular nitrate and phosphate concentrations can be determined using the nitrate phosphate relationships (Figure 2):

equation image
equation image
equation image

where AW* denotes Atlantic water together with river water and sea ice meltwater.

Figure 2.

Nutrient relationships based on AOS-05 data. The lines best represent data known to be of Pacific origin (red) and Atlantic origin (blue).

[12] The end-member slopes and intercepts of the nitrate-phosphate relationships were determined from regions and depths where the water type is present.

[13] The Pacific freshwater and total freshwater were calculated as follows:

equation image
equation image

[14] Within the Arctic Ocean, Pacific water has a range of salinity values from less than 30 to more than 33. We chose an intermediate value of 32, corresponding to the value suggested by Aagaard et al. [2006]. In contrast, the inflowing Atlantic water has a relatively well-defined salinity of 34.85 [Rudels and Friedrich, 2000].

[15] Pacific water entering the Arctic Ocean follows two paths [Jones et al., 1998; Shimada et al., 2001; Steele et al., 2004] with slightly different slopes and intercepts of the nitrate-phosphate relationship for each path [Jones et al., 2003]. The slightly different relationships give results within the expected precision of this approach. The Pacific and Atlantic alkalinity end-members each were chosen as averages of alkalinity values, 2250 μmol kg−1 and 2292 μmol kg−1, respectively, corresponding to their respective salinity end-members.

[16] A recently published representative average river alkalinity value for the Arctic Ocean of 831 ± 100 μmol kg−1 [Yamamoto-Kawai and Tanaka, 2005] is lower than that appearing in some publications [e.g., Anderson et al., 2004]. An overall average spring flow value of several rivers flowing into the Arctic Ocean is approximately 1100 μmol kg−1 [PARTNERShttp://ecosystems.mbl.edu/partners/data.html]. In an attempt to choose a single alkalinity value best representing river water and precipitation in the Arctic Ocean, we compared estimates of river water fractions from both the Eurasian and Canadian basins as well as the Beaufort Sea for which both alkalinity and δ18O data are available. We used the approach outlined above to determine Pacific and Atlantic water fractions and end-members for salinities for these waters and sea ice meltwater, but we used δ18O end-members in place of alkalinity values. By comparing data from three expeditions where both alkalinity and δ18O data were available [Ekwurzel et al., 2001], we found that river water alkalinity values of 1000 μmol kg−1 gave results reasonably consistent with δ18O results.

[17] We determined the end-member slopes and intercepts using data lying on straight lines most likely to include primarily Pacific water or Atlantic water. The Pacific water nutrient relationship was determined by a best fit to nutrient data from the Canada Basin (Stations 2–21) with salinities 32 < S < 33.1 and whose nitrate concentrations >1 μmol kg−1. The latter criterion is to ensure not including water that experienced denitrification. The Atlantic water nutrient relationship was determined by a best fit to nutrient data from the Atlantic layer (θ ≥ 0; 34.7 < S < 34.9). While slope and intercept end-members differ slightly from those used previously, the differences are well within the sensitivity estimates in Table 1.

Table 1. Sensitivity Estimates
FractionMean Standard DeviationEnd-Member Intervals
fr0.002ATr = (1100–900) μmol kg−1
fsi0.005Ssi = (6–2)
fAW0.009SAW = (34.93–34.8), ATAW changed accordingly
fPW0.009SPW = (32.5–31.6), ATPW changed accordingly
ffreshPW0.003nutrient relationships for Pacific and Atlantic waters were varied within a 10% change in slopes and intercepts
ffreshtotal0.001 

[18] River water and sea ice meltwater are taken to have nitrate-phosphate relationships similar to those of Atlantic source water [Jones et al., 1998, 2003]. We found that by including river and sea ice with Pacific water instead of Atlantic water made at most a 10% decrease in the calculation of river water fractions.

[19] We presume sea ice and sea ice meltwater to have bulk alkalinity values determined from their salinities relative to Atlantic water [Jones et al., 1998, 2003]. Sea ice meltwater salinity is taken to be that typical of old ice, S ∼ 2.5, and a corresponding alkalinity, ATsi ∼ 175 μmol kg−1.

[20] The sensitivity of the freshwater fraction calculations for different choices of end-member values was estimated by generating random numbers within chosen intervals for each of the end-members and then calculating the resulting freshwater fractions. This was done 100 times for all fractions and all samples. The mean standard deviation of each fraction as well as the intervals chosen is in Table 1. Note that in the calculation of the mean standard deviations of the freshwater fractions resulting from uncertainties in Pacific and Atlantic salinities, a linear salinity-alkalinity relationship was presumed for Pacific and Atlantic alkalinity end-members.

3. The Data

[21] Water was collected using a 36-bottle rosette sampler equipped with 10 liter Niskin type bottles and a CTD (Sea Bird 911+). Water samples for alkalinity and nutrient analyses were drawn soon after the rosette was secured in a room temperature container, and then analyzed onboard within hours of sampling. Alkalinity analyses were done using a potentiometric titration method, precision ∼±1 μmol kg−1 [Haraldsson et al., 1997]. Nutrients were determined with an AutoAnalyzer according to the WOCE protocols, accuracy near 1% full scale [Gordon et al., 1993]. Salinity samples collected from the rosette were analyzed onboard using a Guildline 8400 salinometer (accuracy ca. ±0.002).

4. Results

4.1. Distributions

[22] The freshwater components (relative to Atlantic inflowing water with a salinity of 34.85) are shown as sections of relative fractions (Figure 3). The total freshwater fraction gradually changes in both amount and depth across the AOS-05 section with the highest and deepest fractions in the southern Canada Basin, where near surface salinities were as low as 27 (Figure 3a).

Figure 3.

Section plots with freshwater fractions vs. distance along the cruise track: (a) total freshwater, (b) Pacific freshwater, (c) river water, (c) sea ice meltwater, and (e) bathymetry. Contour lines in Figure 3a show salinity distributions. Contour lines in Figures 3b–3d show the distributions of the freshwater fractions. The bathymetry (Figure 3e) reflects the approach to the Lomonosov Ridge, a change in course toward the north to deeper water and subsequently a return to the Lomonosov Ridge gap.

[23] Pacific freshwater and river water are the two main contributors to the total freshwater. Their distributions are somewhat different, however. Pacific freshwater fractions in near surface water were high in the central Canadian Basin, a region where prior to this work only speculations on their fractions existed [Jones et al., 1998]. Pacific freshwater is most abundant in the southern Canada Basin, where near surface fractions approach 0.07. The high fractions extend over the Mendeleyev Ridge into the Makarov Basin (Figure 3b). The presence of Pacific freshwater near the North Pole suggests the likely boundary of Pacific freshwater in the Makarov and Amundsen basins (Figure 1). Pacific water was not present in this region of the Nansen Basin.

[24] River water extends across the Canada Basin and was most abundant from about the middle of the southern Canada Basin to offshore from the Chukchi Cap (Figure 3c). The highest river water fractions, up to 0.13, are about double the highest Pacific freshwater fractions. River water, however, was generally confined to the top 50 m displacing Pacific water, whereas Pacific freshwater generally extends deeper, to depths of nearly 300 m in the southern Canada Basin. There was very little river water in the general vicinity of the Mendeleyev Ridge. River water fractions increase in the Makarov Basin, reaching a maximum in the central Makarov Basin, and decrease toward the Lomonosov Ridge. Along the slope and on top of the Lomonosov Ridge, Pacific water dominates as river fractions become low. River water fractions then increase to remain more or less constant across the Amundsen Basin up to near the Gakkel Ridge, where they decrease at a sharp front just inside the Nansen Basin.

[25] Sea ice meltwater, with fractions up to 0.08, was present in the upper 20 m with lower fractions extending to depths near 50 m (Figure 3d). It was found over much of the Canada Basin and generally coincides with Pacific freshwater. Sea ice growth extending to depths approaching 100 m was evident in the Makarov and Amundsen basins. Its distribution within the Amundsen Basin coincided with river water and higher salinities except near the North Pole, where there was Pacific freshwater and some sea ice meltwater as well as lower salinities.

[26] Inventories provide another way to describe freshwater distributions. Sections show vertical distributions. Inventories represent fractions of each freshwater component integrated from the surface to 250 m (Figure 4). In the southern Canada Basin total freshwater inventories reach 20 m, where Pacific freshwater and river water maximum inventories are ∼12 and ∼9 m, respectively. Sea ice meltwater maximum inventories reached up to ∼3 m in parts of the Canadian Basin, while sea ice growth inventories near −5 m were found in the Makarov and Amundsen basins. The inventories show that while upper freshwater fractions across the section decrease somewhat near the North Pole, the total amount was much reduced.

Figure 4.

Inventories (in m) along the cruise track of each freshwater source integrated to 250 m.

4.2. Variability

[27] We are most interested here in changes in Pacific freshwater and river water circulation patterns. Although the removal of brine to denser layers by sea ice formation is a process important to Arctic Ocean water mass formation and modification [e.g., Rudels et al., 1996], sea ice meltwater cycles between ice and surface water, with little effect upon the properties critical to our analysis. The distribution of river water found here adds to what was reported earlier showing the front varying across the Gakkel Ridge [Schlosser et al., 2002; Anderson et al., 2004]. The AOS-05 section offered the opportunity to compare freshwater components at the North Pole, near the Lomonosov Ridge in the Makarov and Amundsen basins, and in the vicinity of the Gakkel Ridge. To help illustrate changes, we show data at stations from expeditions in 1991, 1994, 1996, and 2005 that lie in close proximity to each other (Figure 5 and Table 2). (For a description of the previous expeditions, see Anderson et al. [1994], Swift et al. [1997], and Schauer et al. [2002].)

Figure 5.

Locations of stations for the freshwater inventory comparisons in Table 2. The bathymetry colors correspond to depths of 500 m, 1000 m, 2000 m, 3000 m, and 4000 m, respectively.

Table 2. Comparative Fresh Water Inventories (in m) in the Top 250 m
Year2005 PFW1996 PFW1994 PFW1991 PFW2005 RW1996 RW1994 RW1991 RW
North Pole (NP)3.8 2.23.03.7 5.69.5
Makarov Basin (MB) 86.3°N 151°E, 86.7°N 154°E∼01  104.5  
Lomonosov Ridge Amundsen Basin (LA) 89°N 149°E2.0 0.9 5.5 5.2 
Lomonosov Ridge Makarov Basin (LM) 88.5°N 159°E2.2  1.95.4  4.1
Lomonosov Ridge Slope (LS) 88°N 148°E  1.90.4  6.66.5
Gakkel Ridge Amundsen Basin (GA) 86.7°N 56°E0  08.5  4.8
Gakkel Ridge Nansen Basin (GN) 85.8°N 48.2°E0  03  ∼2

[28] At the North Pole (NP) comparisons can be made among results from 2005, 1991 and 1994. In 2005, Pacific freshwater and river water inventories were about the same, whereas in 1994 and 1991 the river water inventories were considerably greater than the Pacific freshwater inventories.

[29] In the Makarov Basin (MB) (86.3°N 151°E; 86.7°N 154°E) there was essentially no Pacific freshwater in either 1996 or 2005. There was clearly more river water in 2005, about twice as much as in 1996.

[30] In the vicinity of the Lomonosov Ridge, comparisons of 2005 results were made with those from 1991 in the Makarov Basin (LM) (88.5°N 159°E) and with those from 1994 in the Amundsen Basin (LA) (89°N 149°E). In addition, comparisons of 1991 results over the Lomonosov Ridge slope (LS) (88°N 148°E) were made with those from 1994. In the Makarov Basin (LM) freshwater inventories were quite similar in 1991 and 2005. The river water inventories were about twice that of Pacific freshwater (Table 2). In the Amundsen Basin (LA) the river water inventories in 2005 and 1994 were fairly similar to those in the Makarov Basin (LM) in 2005 and 1991, whereas the total freshwater (not shown) and Pacific freshwater inventories in 1994 were much lower than in 1991. River water inventories in 2005 and 1991 were similar, with their inventories about twice those of Pacific freshwater. Stations from 1994 and 1991 over the Lomonosov Ridge (LS) slope showed somewhat less total freshwater (not shown) and quite low values of Pacific freshwater (lower in 1991). River water was relatively abundant with inventories fairly similar to the other Lomonosov Ridge locations. The river water inventories here were much greater than those of Pacific freshwater and were the major component of the total freshwater.

[31] Near the Gakkel Ridge, comparisons of 2005 results were made with those from 1991. On the Amundsen Basin side of the Gakkel Ridge (GA) (86.7°N 56°E) inventories were not much different in 1991 and 2005. There was essentially no Pacific freshwater. River water fractions, up to 0.08, and inventories were greater in 2005. Sea ice growth (not shown in Table 2) was also greater in 2005, fractions up to −0.05. On the Nansen Basin side of the Gakkel Ridge (GN) (85.8°N 48.2°N) as on the Amundsen Basin side, there was essentially no Pacific freshwater. River water inventories were somewhat greater in 2005 than in 1991. There were, however, marked differences in freshwater inventories between the Amundsen and Nansen basins. The total freshwater (not shown) was slightly less in the Nansen Basin, but river water was much reduced in contrast to some other years when the river water front appeared to be less sharp [Anderson et al., 2004]. There was also only a very small amount of sea ice meltwater and no sea ice growth on the Nansen Basin side.

5. Discussion

[32] There were marked fronts in freshwater distributions along the section in the vicinity of the Alpha Ridge and into the Amundsen Basin. Pacific freshwater disappeared at a front near the Alpha Ridge (orange line in Figure 1) just inside the Makarov Basin, where fractions went from 0.05 to essentially 0 as the section deviated south toward the Eurasian side of the Arctic Ocean. Pacific freshwater then reappeared at a front over the Lomonosov Ridge toward the North Pole and disappeared at a front in the Amundsen Basin. River water essentially disappeared at a front over the Alpha Ridge, where Pacific freshwater fractions were still near 0.05, and reappeared with fractions greater than 0.1 at the front where Pacific freshwater was absent in the Makarov Basin. River water fractions became low at the Pacific freshwater front near the Lomonosov Ridge, then increased at the Pacific freshwater front in the Amundsen Basin. These same characteristics found over the Alpha Ridge and in the Amundsen Basin strongly suggest that these waters are the same and indicate that the section passed through the Pacific freshwater front just inside the Makarov Basin and crossed the front back into the same water in the Amundsen Basin.

[33] Across the Canada Basin the fraction of Pacific freshwater in the upper 50 m is nearly the same. River water is present, with the fractions highest in the center of the basin. The near absence of river water in Pacific freshwater over the Alpha Ridge and in the Amundsen Basin would seem to distinguish these waters from those found in the Canada Basin. Steele et al. [2004] described two components of summer Pacific halocline water (31 < S < 33), summer Bering Sea Water (sBSW) and Alaskan Coastal Water (ACW). They suggested that, when the Arctic Oscillation Index (AO) is strongly positive, sBSW was transported directly across the Arctic Ocean via the Transpolar Drift, while during years with a lower AO index, both ACW and sBSW are entrained in a larger Beaufort Gyre. The AO (December-March) abruptly became positive in 1988 with a 3 year running mean maximum in 1989–1990 that decreased but remained positive but decreasing to zero in 2005 [Hurrell, 1995]. The salinities of these waters over the Alpha ridge and in the Amundsen Basin are typically between 30 and 31 and undoubtedly contain Pacific water. Their salinities are in the lower range of sBSW characterized by Steele et al. [2004]; nevertheless it would appear that the Pacific freshwater described here is contained in sBSW carried by the Transpolar drift and that the Pacific freshwater in the Canada Basin in the Beaufort Gyre could be a mixture of ACW and sBSW.

[34] River water at the North Pole was least abundant in 2005 and most abundant in 1991. There was a decrease of river water at the North Pole from 1991 to 2005. While river water and the sum of river water and Pacific freshwater did change significantly from 1991 to 1994, this sum did not change significantly from 1994 to 2005. Björk et al. [2002] suggested that in 2000 river water from the Siberian shelves formed a strong freshwater front in the Amundsen Basin extending from the Gakkel Ridge to the Lomonosov Ridge. They argue “the low salinity layer is most likely caused by river input since the freshwater content is much too high in order to be formed by ice melting and it is far away from the inflow of low salinity Pacific water supplied through the Bering Strait.” Our results (Table 2) show Pacific freshwater to be a not insignificant component of freshwater at the North Pole, with river water being greater than Pacific freshwater in 1991 and 1994 but being nearly equal in 2005.

[35] Johnson and Polyakov [2001] (models) and Boyd et al. [2002] (measurements) discuss changes in locations of lower salinity shelf water in the Eurasian Basin associated with changing wind fields. Johnson and Polyakov [2001] ascribed changes in salinity to the eastward diversion of Russian rivers and increased brine formation due to enhanced ice production in numerous leads in the Laptev Sea ice cover. Similarly, Boyd et al. contended that the disappearance of the cold halocline layer in the early 1990s was due to a shift of low Siberian shelf waters from the Laptev Sea to the East Siberian Sea with a subsequent recovery to the earlier pattern during the 1998–2000 period. Our results indicate a doubling of river water in the Makarov Basin from 1996 to 2005, consistent with the earlier freshwater increases in the general vicinity of the North Pole reported by Boyd et al. [2002]

[36] In 1997 meteoric (mostly river water) and sea ice melt fractions were measured along the Surface Heat Budget of the Arctic Ocean (SHEBA) drift track [Macdonald et al., 2002]. In the southern Canada Basin, one of their stations (5 November 1997) at the eastern end of the drift track was close to our Station 8. In the near surface, above 40 m, they reported meteoric fractions of ∼0.15 and sea ice fractions of ∼0.025 in the top 10 m with negative values beginning at ∼20 m (their Figure 6). Their results at this location compare reasonably well with ours. We cannot make too much of these comparisons, though, because the sampling dates were separated by several years, and tracers (alkalinity cf. δ18O) were not the same. Also, Macdonald et al. [2002] used different end-members and followed a different procedure using three instead of four unknown fractions in their calculations to calculate freshwater fractions.

[37] At the North Pole, relatively high inventories of Pacific freshwater in 1991 were followed by lower inventories in 1994 and subsequently returned to higher inventories in 2005. We interpret these changes to suggest that the Pacific water front migrated from inside the Amundsen Basin in 1991 to the Makarov Basin in 1994 and back into the Amundsen Basin in 2005.

6. Conclusions

[38] We have evaluated freshwater distributions across three of the four major Arctic Ocean basins in a section extending from the southern Canada Basin to the Gakkel Ridge that includes the central Canadian Basin, a previously unsurveyed region for many oceanographic properties including a determination of freshwater fractions and inventories. We established limits of Pacific water and river water in the Makarov and Amundsen basins as they existed in summer 2005 and have proposed that the Pacific water found there is summer Bering Sea Water as described by Steele et al. [2004]. We have contributed to a time series spanning 1991 to 2005 documenting changes in freshwaters in the Makarov, Amundsen and Nansen basins north of 85°N. Anderson et al. [2004] noted a correlation with a 4–6 year delay between changes in the Arctic Oscillation Index and river runoff distributions in the Makarov-Amundsen-Nansen basins. With the few data reported here, no unambiguous relationship between the changing freshwater distributions and the Arctic Oscillation index was found.

[39] Pacific freshwater and river water shared the upper 50 m water column over much of the AOS-05 section. Inventories of Pacific freshwater diminish further north along the AOS-05 section up to the Mendeleyev Ridge where they were low. The distributions of Pacific freshwater and river water in 2005 suggest a near-surface Pacific water front was somewhere to the Siberian side in the Makarov Basin and roughly in the middle of the Amundsen Basin. Pacific water distributions were similar to those reported earlier [Jones et al., 1998], though fractions in the central Canada Basin were higher than they had speculated on the basis of much less data.

[40] River water fractions were greatest in the Canada Basin, where river water displaces some Pacific freshwater, and in the central Makarov Basin, where no Pacific freshwater was present. Pacific freshwater and river water fractions were roughly equal in the vicinity of the North Pole with Pacific freshwater disappearing at a front in the Amundsen Basin and river water disappearing at a strong front near the Gakkel Ridge. Changes in river water inventories in the Makarov and Amundsen basins indicate an increase of river water on the Eurasian side of the Arctic Ocean.

[41] We do not discuss sea ice, though freshwater export in the form of sea ice could be significant under present conditions as well as under future conditions [Vinje, 2001; Rudels et al., 2005]. At present the export of sea ice meltwater seems to be the least likely to influence thermohaline circulation, though this could play a large role with changing ice conditions [Jones et al., 2008].

[42] The inventories reported here would not reflect total inventories entering the Arctic Ocean. Pacific water has a single point of entry but follows different pathways to exit through the Canadian Arctic Archipelago and Fram Strait, likely mainly through the former [Rudels et al., 2004]. River water enters at many locations and follows different pathways to exit also through the Canadian Arctic Archipelago and, apparently, more through Fram Strait [Jones and Anderson, 2008].

[43] The distribution of freshwater source waters can determine which of them contribute to the Canadian Arctic Archipelago throughflow and which exit through Fram Strait to regions of deep convection in the Labrador and Nordic seas. For example, Falck et al. [2005] noted a dramatic decrease in the Pacific water north of Fram Strait, though there was not a large change in near surface salinities. River water and sea ice meltwater may be the components that change most in climate change scenarios because of changes in precipitation in the large river drainage basins feeding into the Arctic Ocean and because of the changes in sea ice meltwater arising from ice-free Arctic Ocean summers. This and other similar work make clear that climate change scenarios dealing with freshwater budgets of the Arctic Ocean should not use a single parameter representation of freshwater as freshwater has different sources and different distributions, all subject to different forcing.

Acknowledgments

[44] This work was supported by the Canada Panel on Energy Research and Development (EPJ), the Swedish Polar Research Secretariat, the Swedish Research Council, the EU project CARBOOCEAN, project 51176 (GOCE) (LGA, SJ, LM), and a National Science Foundation grant OPP-0425350 (JHS). Most figures were produced using Ocean Data View [Schlitzer, 2007].

Ancillary