Surface distributions within Hudson Bay (Fig. 2) reveal the dominant role that freshwater input plays in altering the carbonate system parameters of surface waters, particularly in the southern and eastern parts of the bay. Distributions of TA (Fig. 2b), DIC (Fig. 2c), and ΩAr (Fig. 2d) mimic that of salinity (Fig. 2a), with maxima in the high-salinity waters of the Hudson and Foxe straits and minima along the southern coast of Hudson Bay. One notable exception, however, is near the Nelson River (Fig. 1), where stations exhibit low salinities but relatively high DIC, TA, and ΩAr. The dilution of carbonate parameters by freshwater is more pronounced for sea-ice melt (DIC and TA 300–600 µmol kg-1; e.g., Miller et al., 2011) compared to river outflow (river DIC and TA ∼ 700–1800 µmol kg-1; Azetsu-Scott et al., 2014, and this study), and thus distributions of DIC, TA, and ΩAr in the southern Hudson Bay likely reflect the variable impacts of river runoff and sea-ice melt on the carbonate system.

The Nelson River has the highest freshwater discharge of any single river in the Hudson Bay system, and nearly half of the river runoff into Hudson Bay enters via James Bay. These hydrographical features are well illustrated by the distribution of meteoric water fractions (f_MW) throughout the region (Fig. 2f). In contrast, the fractions of sea-ice melt (f_SIM) are highest along the southern coast of Hudson Bay (Fig. 2e), near the comparatively small Severn and Winisk rivers (with respective discharges one-quarter and one-sixth that of the Nelson River, Table 1). This region of high f_SIM corresponds well to the observed sea-ice distributions shown in Fig. 1. It is important to note that the high positive f_SIM observed near the Nelson River is due to the use of a single flow-averaged δ18O end-member for meteoric water (-14 ‰, Granskog et al., 2011) near a river with a considerably higher δ18O (-10.8 ‰, Table 1). Overall, riverine δ18O signatures measured here agree well with those observed in previous studies, especially when comparing measurements made in/around the month of July (Table 1). Also note that, despite considerable freshwater input near James Bay, aragonite remains supersaturated (ΩAr > 1) at all stations in surface waters (Fig. 2d).

The surface distributions shown in Fig. 2 also align with the general circulation patterns of the Hudson Bay system. For example, differences can be seen between stations in the northern part of Hudson Strait, where Labrador Sea waters with characteristically high salinity, DIC, and TA flow into the Hudson Bay system, and stations at the southern side of the strait, where Hudson Bay outflow with lower salinity and TA, as well as higher f_MW, travels east towards the Labrador Sea. The high-salinity water in northwestern Hudson Bay represents water recently introduced from Foxe and Hudson straits, with salinities generally decreasing in a counterclockwise fashion due to freshwater inputs. At the mouth of James Bay, the salinity and carbonate parameters are higher on the western edge, where waters enter the bay, compared to the eastern edge, where waters altered by large James Bay rivers exit back into Hudson Bay. At the eastern edge of the mouth of James Bay, we recorded the minimum salinity (S= 21.71), TA (1705 µmol kg-1), and DIC (1624 µmol kg-1), as well as the second lowest ΩAr (1.04), among the “seawater” samples from our entire study.

The surface distributions of carbonate parameters throughout the Hudson Bay system are similar to those reported by Azetsu-Scott et al. (2014) for September 2005. Concentrations within Hudson Strait and Foxe Strait are comparable, and the north–south gradient within Hudson Strait is apparent in both years. Within Hudson Bay, concentrations are distinctively lower than in waters outside the bay, with values decreasing further from the northwest toward the southeast. Aragonite saturation (ΩAr) in surface waters does not show any significant differences between 2005 and 2010, with surface waters approaching saturation at the mouth of James Bay in both years. However, subtle differences do exist between these datasets. Stations in southern Hudson Bay have lower salinity, DIC, TA, and ΩAr in 2010, whereas stations located further downstream, near the eastern edge of the bay, have lower values in 2005. Rather than reflecting basin-wide changes over the 5-year period, these differences can be ascribed to the seasonal timing of sample collection and general circulation patterns. That is, stations along the southern coast are impacted by substantial SIM in July as the remaining ice is melting, and the cyclonic circulation is capable of transporting such low-salinity water along the eastern edge of Hudson Bay over a 2-month period (i.e., from July to September; cf. Granskog et al., 2009).

The main conduits for Pacific and Atlantic waters to the Hudson Bay system (HBS) are Fury and Hecla Strait (located north of Foxe Strait) and Hudson Strait, respectively. Yet it remains unclear whether Hudson Bay is composed primarily of Pacific or Atlantic waters. Jones et al. (2003) surmised that Pacific waters were more prevalent in Hudson Bay but were unable to provide conclusive evidence, due to limitations in the quantity and quality of their data. In regions like Hudson Bay, with multiple seawater and freshwater end-members, mass-balance calculations for unraveling freshwater components are particularly complicated, and in contrast to systems such as the coastal Beaufort Sea, where three rather distinct end-members (the Mackenzie River, sea-ice melt, and a primarily Pacific polar mixed layer) can be defined by two tracers (Macdonald et al., 1995), a similarly rigorous analysis in Hudson Bay would require additional end-members and tracers.

When examining the relationship between TA and salinity throughout the HBS (Fig. 3a), waters below 100 m depth in Hudson Bay are indistinguishable from those in Foxe Strait, and both are distinctly different from deep waters in Hudson Strait. Furthermore, data from Hudson Bay and Foxe Strait are well aligned with the Pacific water end-member, while data from Hudson Strait fall along a mixing line with the Atlantic water end-member (Shadwick et al., 2011; Fig. 3a). This indicates that deep water from Hudson Strait does not enter Foxe Strait and that the vast majority of water that flows over the 185 m sill into Hudson Bay is likely of Pacific origin. In addition, some samples in the Hudson Bay/Foxe Strait dataset have higher salinity and TA than the Pacific water end-member, yet still fall along the extension of the general mixing line (Fig. 3a), suggesting an additional source of both TA and salinity. One plausible source of excess TA and salinity in deep waters is from brine rejection. It is known that deep water formation occurs in Foxe Basin polynyas, and this water can flow into Foxe Strait and potentially further into Hudson Bay (Defossez et al., 2010). However, there is evidence in Hudson Bay that deep waters can also form locally due to brine rejection (Granskog et al., 2011). The same general patterns are shown in a plot of deep water DIC against salinity (Fig. 3b), except that distinctly higher DIC is observed in Hudson Bay compared to Foxe Strait due to the build-up of respiratory DIC in Hudson Bay, which is discussed in greater detail below.

Prior studies have revealed the importance of watershed geology in governing the highly variable compositions of Hudson Bay rivers (Mundy et al., 2010; Granskog et al., 2011; Azetsu-Scott et al., 2014; see also Thomas and Schneider, 1999), yet relatively little is known regarding the impact of varying river inputs on the inorganic carbon system of Hudson Bay. Variations in TA with salinity in the top 60 m of the water column of Hudson Bay (Fig. 4) reveal two distinct freshwater end-members: a lower TA end-member of ∼ 700 µmol kg-1, corresponding to the vast majority of stations in the bay, and a high-TA end-member of ∼ 1900 µmol kg-1 for samples collected within the Nelson River estuary (NRE, Figs. 1 and 5). Figure 4 also highlights the various impacts of sea-ice melt, as samples with high f_SIM (purple circles) show distinct positive deviations from the δ18O vs. salinity mixing line (Fig. 4, inset), as well as negative deviations from the TA vs. salinity mixing line, due to the enriched δ18O signature (δ18OSIM= 0 ‰) and low TA concentration characteristic of sea-ice melt. The negative TA deviation in sea-ice meltwaters also indicates that there is no significant source of TA from the dissolution of ikaite (CaCO3) during sea-ice melt (see Rysgaard et al., 2011) in this area.

Considering that samples with potential inputs of brine (i.e., deeper waters) and discernable sea-ice melt input (i.e., stations with f_SIM ≥ 0.05) are not included in the best-fit regressions, the lower alkalinity end-member (that is, TA = 689 µmol kg-1) likely represents the average riverine TA for the entire Hudson Bay. This end-member aligns with the prior estimate for “rivers” of 754 µmol kg-1, also derived from regression lines of samples collected within the bay and not from the rivers, themselves, along the eastern shore of Hudson Bay (Azetsu-Scott et al., 2014). Stations in southwestern Hudson Bay, nearest to the Churchill, Nelson, and Hayes rivers (Fig. 4, red squares, “705” stations and station 706; see Figs. 1 and 5), have a similar salinity–TA relationship to that observed throughout Hudson Bay but are vertically offset to a slightly elevated TA. This TA offset is not related to the merging of the two TA datasets, as the stations labeled as “SW Hudson Bay” (see Fig. 4) comprise a small subset of the TA data measured at IOS. Also, the magnitude of this vertical TA offset (∼ 80 µmol kg-1) is much larger than the average deviation observed between the datasets. Instead, this offset likely indicates input of carbonate ion from the dissolution of carbonate minerals. For salinity vs. DIC, the same offset is visible but is less pronounced, which also points towards benthic carbonate ion addition. Azetsu-Scott et al. (2014) attributed their observations of high TA in the deep waters of southwestern Hudson Bay to dissolution the carbonate-rich bedrock (defined as the Hudson Bay Lowland) that underlies this area of the bay, as well as much of the Nelson River drainage basin.

We attribute the high-TA end-member (1870 µmol kg-1) to the Nelson River. A similarly high TA end-member has been reported for the Mackenzie River (1540 µmol kg-1; Cooper et al., 2008) and time-series measurements collected from the Churchill River (Fig. 1) during the summer of 2007 reveal an average TA of 1394 ± 80 µmol kg-1 (Stainton, 2009). According to Mundy et al. (2010), the Churchill, Nelson, and Hayes rivers (Fig. 1) have dissolved organic carbon (DOC) concentrations that are 2–6 times greater than other Hudson Bay rivers (Table 1), and these authors attributed variability in riverine DOC to the vastly differing watershed geologies throughout the Hudson Bay region. The weathering of organic-rich soils throughout the extensive Nelson River drainage basin would contribute to higher riverine DIC and TA compared to rivers further north with drainage basins in the relatively organic-poor Arctic tundra. Furthermore, weathering processes become more active in subarctic, or even temperate, regions of the Hudson Bay drainage area, yielding higher TA inputs from those regions than from polar regions. This regional dependence on riverine composition is also observed in the δ18O signatures, both across the basin and within a single river catchment, with the southwestern rivers having distinctly less depleted δ18O signatures than the rivers further north (Table 1; see also Granskog et al., 2011, and Smith et al., 2015). Nevertheless, the TA end-member calculated here is substantially higher than the prior estimate for the Nelson River (1022 µmol kg-1; Azetsu-Scott et al., 2014), and we think this discrepancy is due to sampling depth within this highly stratified system, as well as to the fact that our samples extended further up the estuary, into fresher waters.

Data obtained via rosette sampling in southwestern Hudson Bay (Fig. 4, red squares) yield a riverine TA end-member of 710 µmol kg-1, and the uncertainty on this value places it in the same range as the end-member calculated for the rest of Hudson Bay. According to the CTD data, samples from southwestern Hudson Bay were collected between 1.4 and 10.3 m depth, with the former being considered “surface” samples. In contrast, samples collected from the surface in the inner NRE using the small barge yield an end-member of 1870 µmol kg-1. Given the close proximity of these inner NRE stations to others in southwestern Hudson Bay, the discrepancy in end-members indicates that surface samples collected via the rosette captured a different water mass than those collected via the small barge. This is illustrated further using a cross section of salinity in the NRE (Fig. 5).

Salinity across the NRE, as measured in situ by the CTD attached to the rosette system, and as measured from discrete surface bottle samples collected either from the barge or using a bucket dropped from the bow, is shown in Fig. 5. At stations 705a, b, and c, salinity samples collected from surface waters via bucket have considerably lower salinities than any waters measured by the CTD. Furthermore, according to the CTD pressure data, surface Niskin bottles from stations 705a, b, and c were closed at 1.6, 1.6 and 1.4 m respectively (Fig. 5, black X's), but the salinities measured in these Niskin bottles correspond to the CTD salinities observed at depths of 2.5, 3.6, and 7.1 m, respectively (Fig. 5, vertical black arrows), indicating that the Niskin bottles contained some deeper waters entrained into the surface by the upward-moving rosette. This highlights the importance of taking bottle salinity samples from Niskins in parallel with parameters such as DIC and TA, particularly in highly stratified waters. Here, all DIC and TA samples collected in southwestern Hudson Bay are accompanied by a high-precision salinity sample taken from the same Niskin bottle.

At stations B6, B8, and B12 (Fig. 5), salinities measured using the 3 L Niskin bottles (deployed to 1 m depth from the small barge) are very low and show an increasing salinity with distance away from the Nelson River mouth that is consistent with the bucket samples collected at stations 705a, b, and c. Clearly, the samples collected from the small boats (barge and Zodiac) captured low-salinity waters very near the surface, while the “surface” rosette bottles captured water from a deeper layer with considerably higher salinities. Assuming the low-salinity layer represents the Nelson River plume, this water is only accurately captured using the samples taken via small boats, which explains why samples taken via barge in the inner NRE have a markedly different TA vs. salinity relationship from samples taken via rosette throughout the remainder of Hudson Bay (Fig. 4). Additionally, at station 705c (Fig. 5), the bucket sample near the surface does not appear to have captured the low-salinity river plume, demonstrating the extent to which river inputs are constrained to the coastline during cyclonic transport around the bay (Fig. 1). Large CTD salinity gradients in the upper 2 m were observed at other stations along the southern coast, suggesting that data from the surface Niskin bottle at these stations may also not fully reflect the true “surface water” properties.

Hudson Bay waters undergo significant compositional changes during their cyclonic transit around the basin (Granskog et al., 2009; Mundy et al., 2010). Here, we focus specifically on changes in the carbonate system along a coastal transect that follows the general counterclockwise circulation pattern from northwestern Hudson Bay to the northeast (Fig. 6, transect A–B). For stations within this transect, TA, DIC, and δ13CDIC are plotted against in situ density anomaly (σt, observed density -1000 kg m-3) to illustrate deviations at constant density levels (Fig. 6a–c). For TA (Fig. 6a), there is little variability between stations along the transect and the vast majority of the data fall along the river mixing line. There are, however, two notable exceptions: the first being high-TA deviations at stations within the NRE that we attribute to dissolution of carbonate minerals, and the second being low-TA deviations in samples with high sea-ice melt fractions (f_SIM ≥ 0.05), all of which are measured at or near the surface along the southern coast.

For along-transect DIC (Fig. 6b), the impacts of both freshwater input and biological activity are apparent. Similar to TA, positive deviations from a general river mixing line are observed in the NRE, as are negative deviations along the southern coast, but a distinct increase in DIC is also visible along the transect due to a build-up of respiratory DIC in subsurface waters. The maximum offset between waters along the west coast (blue–purple points) and the east coast (red–yellow points) is between the 24 and 26 kg m-3σt horizons, indicating that organic matter respiration rates are highest at these density levels. The structures of these density horizons along the coastal transect are shown in Fig. 6d. The lack of any visible offset in bottom water TA suggests that any products of anaerobic respiration that may be produced in sediments are not transported into the water column.

Unlike TA and DIC, freshwater input appears to play a relatively minor role in governing variability in the stable carbon isotope signature of DIC (δ13CDIC) of Hudson Bay waters (Fig. 6c). This may be due to the fact that Hudson Bay rivers are very isotopically enriched (average δ13CDIC= -3.2 ‰, Table 1) compared to other coastal regions (North Sea δ13CDIC= -12 to -16 ‰; Burt et al., 2016) and thus are much closer to typical seawater values (∼ 0–2 ‰). Biogenic soils are highly depleted in δ13CDIC (∼ -25 ‰), while carbonate bedrock has a δ13CDIC near 0 ‰ (Spiker, 1980). Given that the majority of Hudson Bay rivers have rather low DOC (Mundy et al., 2010) and that much of the bedrock surrounding the bay is carbonate-rich (Ross et al., 2011), it is unsurprising that Hudson Bay rivers are isotopically enriched compared to other regions. However, δ13CDIC was not sampled along the southern coast; thus, the Churchill, Nelson, and Hayes rivers, which are likely more depleted in δ13CDIC, are not included in Table 1 and are not well represented in Fig. 6c and d.

Conversely, the δ13CDIC distributions in Fig. 6c do highlight the impact of biological activity along the coastal transect. Waters become isotopically lighter during their transit around the basin, because respiration of organic matter releases isotopically light DIC into the water column. In accordance with Fig. 6b, relative changes between waters along the western shore (blue–purple points) and eastern shore (red–yellow points) are greatest in intermediate waters. In the North Sea, Burt et al. (2016) showed that a respiratory DIC increase of 1 µmol kg-1 could be roughly equated to a 0.012 ‰ decrease in δ13CDIC. Taking this ratio, and assuming that waters are transported along lines of constant density, changes in δ13CDIC at intermediate depths (between ∼ 25 and 50 m, at σt= 25 kg m-3; see Fig. 6c and d) between western and eastern waters equate to an addition of approximately 110 µmol kg-1 DIC. Observed increases in DIC at the same density level (Fig. 6b) are approximately 90 µmol kg-1, which is comparable given the rough nature of this calculation. In deeper waters (between ∼ 40 and 100 m, at σt= 26 kg m-3; see Fig. 6d), both observed DIC increases (shown in Fig. 6b) and increases based on differences in δ13CDIC (shown in Fig. 6c) equate to ∼ 70 µmol kg-1. It is important to note that the ratio taken from Burt et al. (2016) assumes that all respired organic matter is of marine origin, and although the δ13CDIC derived from respired terrestrial organic matter may have a similar isotopic signature to that of marine-derived material, it is not clear how a significant fraction of terrestrial material in the organic matter pool would affect this first-order calculation. However, in the coastal marine system, the organic matter turnover (i.e., cycles of production and respiration) is generally rapid, and thus any terrestrial signature within the organic matter pool will quickly become difficult to detect.

The effects of biological activity, sea-ice melt, and variable riverine input on the carbonate system of Hudson Bay are well summarized in the coastal transect of aragonite saturation state (ΩAr) shown in Fig. 6e. In surface waters, ΩAr is high in the northwest but decreases along the southern coast due to the dilution of both TA and DIC by freshwater input. In the NRE, ΩAr remains high despite low salinities (see Fig. 2a, d) because of carbonate dissolution either in the drainage basin (as visible in the high alkalinity of the Nelson River) or in the sediments in the estuary (Fig. 4). Surface waters near James Bay show minima in ΩAr, likely due to the strong sea-ice melt signal in these waters, as well as large inflows of low-alkalinity river waters (see Fig. 2). Along the eastern shore, surface water ΩAr is comparable to (i.e., only slightly lower than) values along the west coast. This is likely due to the fact that the sea-ice meltwaters, which dilute TA and DIC to a much greater degree than river input, have yet to be transported to these stations. In September 2005, surface water ΩAr along the southern coast was higher than observed here, while further downstream along the eastern coast, ΩAr was lower than observed here (Azetsu-Scott et al., 2014). This may simply reflect the transport of low-ΩAr sea-ice meltwaters from the southern coast in July to the eastern shore by September. In accordance with this, Granskog et al. (2009) noted that in October 2005, virtually no sea-ice melt was observed along the southern shore, further suggesting that these waters had been transported northeast by the coastal current.

Focusing on subsurface waters along the transect, the saturation horizon shoals significantly from 60–70 m along the west coast to 30–40 m along the east coast. Furthermore, deep waters along the west coast are only slightly undersaturated, while deep waters along the east coast are heavily undersaturated. These patterns likely reflect the build-up of respiratory DIC in deep waters during transit around Hudson Bay, as aerobic respiration releases DIC while consuming TA, thus causing a relatively strong decrease in ΩAr. Deep waters along the eastern shore are likely older than waters along the western shore, allowing the products of aerobic respiration to build up over time. Exchange with open ocean waters is limited by the shallow sills at the mouth of Hudson Bay, and very low saturation states are observed at deeper stations in the central Bay, as well as in the coastal transect (Fig. 6e).

Subsurface ΩAr, both along the coastal transect (Fig. 6e) and in a cross-basin section (not shown), is very similar to the 2005 distributions shown by Azetsu-Scott et al. (2014), suggesting that the acidification state of Hudson Bay had not changed discernibly between 2005 and 2010. However, with only two such datasets taken at slightly different times of the year, more information is required to make conclusive statements regarding acidification rates in this region.