Observations of sea surface fCO2 distributions and estimated air-sea CO2 fluxes in the Hudson Bay region (Canada) during the open water season


  • Brent G. T. Else,

    1. Centre for Earth Observation Science, Department of Environment and Geography, University of Manitoba, Winnipeg, Manitoba, Canada
    Search for more papers by this author
  • Tim N. Papakyriakou,

    1. Centre for Earth Observation Science, Department of Environment and Geography, University of Manitoba, Winnipeg, Manitoba, Canada
    Search for more papers by this author
  • Mats A. Granskog,

    1. Centre for Earth Observation Science, Department of Environment and Geography, University of Manitoba, Winnipeg, Manitoba, Canada
    2. Arctic Centre, University of Lapland, Rovaniemi, Finland
    Search for more papers by this author
  • John J. Yackel

    1. Foothills Climate Analysis Facility, Centre for Alpine and Arctic Climate Research, Department of Geography, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author


[1] The lack of baseline estimates of air-sea CO2 exchange in Arctic and sub-Arctic regions represents a major shortfall in our ability to understand how climate change may affect CO2 fluxes at high latitudes. The 2005 ArcticNet cruise of Hudson Bay (Canada) provided a rare comprehensive oceanographic survey of one such region. Ship-based observations of sea-surface fugacity of CO2 (fCO2sw) were made at 56 locations between 15 September and 26 October and were found to range from 259 μatm in Hudson Strait to 425 μatm at the entrance to James Bay. Strong relationships between fCO2sw and river discharge were identified, with coastal waters observed to be supersaturated with respect to the atmosphere, while offshore waters were undersaturated. High correlations of fCO2sw with salinity, sea surface temperature, and colored dissolved organic matter suggest that thermodynamic effects and the oxidation of riverine carbon are driving supersaturation in the coastal zone. Calculated instantaneous fluxes of CO2 ranged from +16.5 mmol m−2 d−1 in James Bay to −19.6 mmol m−2 d−1 in Foxe Channel. Using National Centers for Environmental Prediction wind speed climatologies, a net sink in Hudson Bay of −0.73 (±0.4) mmol m−2 d−1 was estimated for study period, substantially lower compared to many other Arctic shelf environments. This initial study provides a preliminary examination of fCO2sw dynamics in Hudson Bay; future analyses and field measurements will be necessary to properly constrain CO2 fluxes in this season and over an annual cycle.

1. Introduction

[2] Continental shelf seas are very important in the global carbon cycle because of their role as an intermediary between the oceanic, terrestrial and atmospheric carbon reservoirs [Walsh et al., 1981; Thomas et al., 2004]. However, the precise budgeting of these waters as either sources or sinks of atmospheric CO2 lacks a firm consensus [Cai and Dai, 2004; Chen, 2004; Thomas et al., 2004]. This ambiguity is largely due to the variability of biogeochemical processes in coastal zones, and a dearth of baseline observations in many regions. Borges et al. [2005] conducted an extensive literature review of reported air-sea CO2 fluxes and found that high-latitude and temperate coastal oceans tend to be a sink of CO2 (integrated exchanges of −0.10 Pg C a−1 and −0.13 Pg C a−1 respectively) while subtropical and tropical coastal oceans tend to be sources of CO2 (0.18 Pg C a−1). Cai et al. [2006] conducted a similar review with similar results. These studies identified a great deal of spatial and temporal variability in air-sea CO2 flux both within and between latitudinal bands, which provides impetus for conducting studies in the large fraction of coastal ocean that remains unstudied.

[3] Two general patterns of spatial variability in air-sea CO2 fluxes in coastal waters have been observed. The first pattern relates to the different processes that occur in the proximal coastal regions compared to the distal open shelves [Chen, 2004]. In the proximal coastal regions a great deal of terrestrial carbon and nutrients can be delivered by rivers, and waters tend to be supersaturated in fCO2sw [see, e.g., Frankignoulle et al., 1998]. Most estuaries have been found to be sources of atmospheric CO2 [Borges et al., 2005] although seasonal undersaturations have been observed in some instances [e.g., Salisbury et al., 2008]. In the distal open shelves the influence of river input is less direct and the source/sink status becomes less clear. In these regions, there appears to be a latitudinal pattern in the variability of fluxes [Cai et al., 2006]. High-latitude and temperate shelves tend to be a net sink of CO2, largely because of the presence of strong spring phytoplankton blooms [Frankignoulle and Borges, 2001; Kaltin et al., 2002; Thomas et al., 2004]. Low-latitude open shelves can also be very productive (e.g., in coastal upwelling zones), but often experience high CO2 production relating to terrestrial carbon input [Borges, 2005]. As a result, they tend to act as sources of CO2 [Cai et al., 2003; Zhai et al., 2005].

[4] Temporal variability in atmosphere-ocean CO2 fluxes arises from the seasonal and interannual variability of fCO2sw. In the open ocean seasonal variations in biological productivity, sea surface temperature, and upwelling drive an annual amplitude of fCO2sw that typically reaches ∼40 μatm [Takahashi et al., 2002; Sarmiento and Gruber, 2006]. In the coastal and shelf regions this amplitude is greatly enhanced by riverine input and complex coastal processes, and can easily exceed 1000 μatm in estuary systems [Frankignoulle et al., 1998; Borges et al., 2006] and has been observed as high as 400 μatm on shelves [Thomas and Schneider, 1999; Borges et al., 2006].

[5] Understanding the total contribution of the Earth's coastal seas to the atmospheric CO2 budget is a key concern for climate studies. Ideally, this would be achieved by thoroughly studying annual and interannual variability in every coastal sea on the planet. Since this is obviously not logistically feasible, most investigators have focused on upscaling results from well-studied coastal seas [Borges, 2005; Borges et al., 2005; Cai et al., 2006]. However, given the temporal and spatial variability described above, it is still important to gather information in as many coastal zones as possible in order to assess the reliability of these upscaling studies. Of particular concern is the paucity of studies in Arctic and sub-Arctic regions, especially given the likelihood of these areas experiencing significant climatic change in the near future.

[6] The Hudson Bay system, composed of the Hudson, James and Ungava bays and Hudson Strait, is a sub-Arctic/Arctic continental shelf system that to date has not been studied with respect to air-sea CO2 exchange. In fact, the region has not been included in past coastal CO2 budgets at all, a significant omission considering that the total area (including Foxe Basin) is ∼1.2 × 106 km2, or 4.5% of the total continental shelf area cited in the paper by Cai et al. [2006]. Hudson Bay forms the outlet of a major drainage basin with a mean annual discharge of 30 900 m3 s−1; more than three times the discharge of the St. Lawrence or Mackenzie river systems (for a discussion of discharge characteristics of this region, refer to Déry et al. [2005]). The Hudson Bay system is covered by seasonal sea ice 8–9 months of the year, and the freeze/thaw cycle has an important influence on oceanography and biota [Anderson and Roff, 1980; Prinsenberg, 1986b]. Sea ice formation tends to be most rapid in northeastern Hudson Bay, which increases salinity through brine rejection. Currents and wind subsequently concentrate ice in southeastern Hudson Bay, which results in a stronger freshwater pulse in these areas during the melt season [Prinsenberg, 1986b; Granskog et al., 2007]. The freshwater flux associated with melt and runoff creates strong vertical stratification across the Hudson Bay region, which likely reduces nutrient supply to the surface ocean and may limit phytoplankton growth [Anderson and Roff, 1980].

[7] The primary objective of this study is to describe the spatial variability of fCO2sw observed in Hudson Bay during the 2005 open water season. We estimate instantaneous fluxes of CO2 from field data using the bulk aerodynamic approach. A rough estimate of the integrated air-sea CO2 exchange in Hudson Bay is computed for the study period in order to gain an approximation of the magnitude and direction of CO2 exchange. This study provides a preliminary look at the air-sea CO2 fluxes in Hudson Bay, allowing us to place this large and previously unstudied continental shelf system within the context of the latitudinal and provincial carbon-budgeting approaches taken by Borges et al. [2005] and Cai et al. [2006].

2. Methodology

2.1. Study Area and Sampling

[8] Field data for this study was collected in the Hudson Bay region from 15 September to 26 October during Leg 2 of the 2005 ArcticNet (http://www.arcticnet-ulaval.ca) expedition onboard the research icebreaker CCGS Amundsen (Figure 1). fCO2sw measurements were made at discrete stations in Hudson Bay where the vessel stopped to conduct sampling. At each station water was pumped through a hose (∼8 m long) using a submersible pump that delivered water from a depth of ∼2 m at a rate of ∼10 LPM to a custom-built equilibration system on the ship deck. The equilibration system uses a peristaltic pump to pass water at a rate of 0.4 LPM through a membrane contractor (LiquiCel® MiniModule model x50) which equilibrates the dissolved gases in the water with an air stream that is looped through the device [see Hales et al., 2004]. The equilibrated air stream is passed through a drying agent (silica gel) to a nondispersive infrared (NDIR) spectrometer (Li-Cor® model LI-820) which measures the dry air CO2 mole fraction (xCO2) and atmospheric pressure. The span and zero of the LI-820 was checked periodically (at least every two weeks) using a synthetic air blend (CO2 at 380.3ppm) and ultrahigh-purity N2 (99.999%). The span gas was calibrated against traceable WMO standards. We found the gas analyzer to be very stable (drift typically < 0.5ppm). While on station all variables were sampled every 2 s and recorded as 1 min averages on a data logger (Campbell Scientific Inc. model CR-10X). The LI-820 was turned on at least 30 min prior to operation in order to initialize the sensor, and the system was typically deployed for 15 to 30 min at each station to allow for a stable sample of surface water xCO2.

Figure 1.

Map of study area and sampling locations. The thin line denotes the ship track, and the numbers indicate the date (in day of year, 2005) when the ship arrived at that location. Stars indicate sites where fCO2sw measurements were made in conjunction with hydrocasts. Dots indicate sites where only hydrocasts were made. Arrows represent the general surface circulation as per Prinsenberg [1986a].

[9] A water temperature correction of the fCO2sw measurement to account for changes from in situ sea surface temperature (SST) [e.g., Takahashi et al., 1993] was not performed because of a malfunction in our inline water temperature thermistor. We expect the deviation from in situ SST to be very small because of the short travel time between the sea surface and equilibrator (<10s), and because the system was deployed outside where air temperatures were on average within ±1.3°C of SST. Nevertheless, even a small change in temperature can result in a large correction factor to fCO2sw. For example a change of ±0.25°C creates an average error of ±3.7 μatm at the fCO2sw levels measured in this study. Since we cannot completely rule out the possibility of a temperature change, we adopt ±3.7 μatm as an estimate of the error involved in omitting these computations. By including the typical calibration drift experienced (±0.5 μatm) we calculate an overall RMS error for the fCO2sw system of ±4 μatm.

[10] In addition to the fCO2sw sampling, traditional rosette hydrocasts were made at each station. The rosette was equipped with a conductivity-temperature-depth (CTD) sensor (Sea-Bird Electronics Inc. model SBE-911plus) and a chlorophyll fluorometer (Seapoint Sensors Inc.). The CTD, in combination with the ship's flow-through thermosalinograph (Sea-Bird Electronics Inc. model SBE-45) provided measurements of SST for this study. A surface water sample was also collected at each station prior to the ship disturbing the mixed layer. From this surface sample, colored dissolved organic material (CDOM) absorption was determined using a UV-visible spectrophotometer (Ocean Optics Inc. model S2000), and salinity was determined using a benchtop salinometer (Guildline Autosal model 8400). The CDOM and salinity sampling methodologies are outlined by Granskog et al. [2007].

2.2. Atmospheric Observations

[11] A meteorological tower was deployed on the foredeck of the Amundsen to monitor atmospheric variables relevant to air-sea gas exchange. Sensors on the tower measured horizontal wind speed and direction at a nominal height of 14.6 m above the sea surface using a conventional propeller anemometer (RM Young Co. model 15106MA). Temperature and relative humidity was measured using a RH/Temperature probe (Vaisala model HMP45C212) housed in a vented sunshield. Incoming shortwave and longwave radiation was measured using a pyranometer and a pyrgeometer (Eppley Laboratory, Inc. models PSP and PIR, respectively). Data from these instruments was measured every second and recorded as 1 min averages on a data logger (Campbell Scientific Inc. model CR-10X). The tower also employed an open path NDIR gas analyzer (Li-Cor® model LI-7500) which was installed 12.2 m above the sea surface and output atmospheric CO2 and H2O measurements (in units of molar concentration) at 10 Hz to a data logger (Campbell Scientific Inc. model CR-23X). The CO2 and H2O measurements were combined with temperature and pressure measurements from the other instruments to calculate the dry air xCO2.

[12] Wind speed was corrected to height of 10 m using the NOAA/COARE v3.0 algorithm [Fairall et al., 2003]. The algorithm was used to compute the stability parameters L (Monin-Obukhov length), u* (friction velocity) and zo (roughness length) on the basis of input of wind speed, air temperature, humidity and radiation (from the meteorological tower) and SST (from either CTD casts or the thermosalinograph). The output parameters were used to modify the log linear wind speed profile and estimate wind at 10 m following Stull [1988]. The NOAA/COARE algorithm has not been widely validated for application in coastal seas, but preliminary assessments have shown that it performs adequately [Sopkin et al., 2007].

2.3. The fCO2 Calculations

[13] For this study, fCO2 measurements are used. fCO2 is the partial pressure of CO2 (pCO2) corrected for the nonideal behavior of the gas [Weiss, 1974]. For practical applications, fCO2pCO2 (∼1% difference for the range of concentrations and temperatures typically observed), but we adopt this approach for the sake of scientific rigor [McGillis and Wanninkhof, 2006]. fCO2sw was derived from pCO2sw which was calculated using the xCO2 output from the gas analyzer following Department of Energy [1994]:

equation image

where pCO2eq is the partial pressure of CO2 in the equilibration chamber, P is the atmospheric pressure, pH2O is the saturation vapor pressure of air in the equilibration chamber (determined empirically from water temperature and salinity using the equations of Weiss and Price [1980]), and xCO2eq is the dry air mixing ratio in the equilibrator. The pCO2 value was then converted to fCO2 using the equations of Weiss [1974] [see also McGillis and Wanninkhof, 2006] and bulk SST measured from the CTD casts or the thermosalinograph. Atmospheric fCO2 (fCO2atm) values were also computed following the same approach, but replacing SST with air temperature.

2.4. Flux Calculations

[14] CO2 flux is commonly parameterized using the equation:

equation image

where image is the flux of carbon dioxide (negative values indicate a sink into the ocean while positive values indicate a source), k is the gas transfer velocity, s is the solubility of CO2 at in situ SST and salinity, and ΔfCO2 is the air-sea gradient of fCO2 calculated as ΔfCO2 = fCO2swfCO2atm. This approach can be further refined to account for the fact that fCO2sw is sampled in the bulk surface water, yet fCO2 at the ocean surface skin drives gas exchange. Since the skin layer is usually cooler than the bulk seawater temperature by about 0.1 to 0.4 degrees [Fairall et al., 1996], fCO2 will be slightly lower because of the thermodynamics of carbonate reactions [Takahashi et al., 1993]. We therefore follow the approach adopted by Hare et al. [2004] to calculate instantaneous fluxes of CO2:

equation image

where the subscripts s and sw represent variables calculated at the ocean skin and in the bulk surface respectively, and ΔT = Tsw − Ts.

[15] The largest uncertainty in fluxes calculated using this approach arises from the determination of k. Laboratory and field studies have identified a relationship between k and 10 m wind speed, but there is still controversy about the nature of this relationship at high (>10 m s−1) wind speed. For this study we adopted the Sweeney et al. [2007] (hereinafter referred to as S07) formula which is based on global inventories of bomb 14C. This parameterization coincides well with direct measurements of k at low to moderate wind speeds [e.g., McGillis et al., 2001] and at high wind speed [Ho et al., 2006], and is not as likely to be biased to location-dependant variables as direct measurement studies [i.e., Wanninkhof and McGillis, 1999; Nightingale et al., 2000]. However, since a reliable k parameterization still awaits more field experimentation, we find it prudent to use a range of formulations as an estimate of uncertainty as suggested by Nightingale et al. [2000]. To this end, the Wanninkhof [1992] (hereinafter referred to as W92) and the Nightingale et al. [2000] (hereinafter referred to as N00) parameterizations were used as upper and lower bounds (respectively) on our estimates of k.

[16] Solubility was calculated using SST and salinity following the equations of Weiss [1974]. Ts was estimated using the NOAA/COARE algorithm and Tsw was measured either by the thermosalinograph or CTD casts made concurrently with the fCO2sw measurements. By utilizing equation (3) with site-specific fCO2atm measurements, the instantaneous flux of CO2 was estimated for each site sampled in Hudson Bay.

3. Results

3.1. The fCO2sw Distributions

[17] Figure 2 shows the observed distribution of fCO2sw. A minimum value of 259 μatm was observed in Hudson Strait, while a maximum value of 425 μatm was observed near the mouth of James Bay. In general, higher fCO2sw values were observed near the coast with particularly high values along the southeast shoreline. The measurements made offshore and those along Foxe Channel and in northern Hudson Bay showed considerably lower fCO2sw.

Figure 2.

Observed distribution of fCO2sw in Hudson Bay.

[18] To understand the patterns shown in Figure 2, knowledge of the different water masses in the region is required. The review by Ingram and Prinsenberg [1998] shows that Arctic marine water is imported into Hudson Bay from Foxe Basin and via the passage between Coats and Mansel Islands (see Figure 1). Once in the bay, this marine water circulates counter clockwise and is diluted by river runoff. Freshwater tends to be concentrated along the southern coast, especially near the outlet of James Bay where river runoff and ice melt rates are highest [Granskog et al., 2007]. Water exits the bay primarily to the east of Mansel Island, but also in a narrow band just to the west.

[19] Many of these circulation features are reflected in the distribution of fCO2sw (Figure 2). The marine offshore waters of Hudson Bay and Hudson Strait have lower amounts of dissolved CO2, while the river-influenced waters have high amounts of dissolved CO2. The transect between Coats and Mansel Islands also seems to be influenced by circulation; the west side of the transect (coinciding with the inflow of marine water) has low fCO2sw, while the east side (coinciding with outflow from the bay) has higher fCO2sw. The relationship between fCO2 and river runoff is examined in detail in section 4.1.

3.2. Atmospheric Conditions

[20] Figure 3 shows the 10 m wind speeds observed during the experiment. The average wind speed during the cruise was 7.8 m s−1 and ranged from 1 to 16 m s−1. This corresponds well to scatterometer-derived wind climatologies for Hudson Bay [Young, 1999], and for comparison is lower than the North Atlantic in the fall season (mean of ∼11 m s−1). It should be noted that on several instances the ship took shelter in protected bays in order to avoid stormy weather conditions; therefore Figure 3 is likely biased toward low wind values, and should not be taken as representative of conditions throughout Hudson Bay for the study period. Nevertheless, Figure 3 shows that on several occasions winds in excess of 10 m s−1 were sustained for significant periods of time. This is an important threshold for two reasons; first because at high wind speeds gas transfer becomes increasingly significant, and second because at this point the various parameterizations of transfer velocity (k) begin to diverge [Wanninkhof and McGillis, 1999].

Figure 3.

Observed wind conditions from the micrometeorological tower: (a) the frequency distribution of wind speeds and (b) a time series of hourly averaged wind speed (use day of year labels on Figure 1 to determine ship location).

[21] fCO2atm observations for the cruise are illustrated in Figure 4. For most of the experiment observations varied within ∼5–10 μatm of the mean value (367 μatm) which corresponds well with seasonal CO2 observations taken in the Canadian Arctic at Alert, Nunavut (see http://www.cmdl.noaa.gov/). Occasional strong departures from the mean value were observed (a minimum value of ∼325 μatm near the mouth of James Bay and high value of ∼400 μatm off the southwest coast of Hudson Bay). These variations in atmospheric CO2 concentration likely reflect regional and synoptic controls on the characteristics of the prevailing air mass [Wang et al., 2007].

Figure 4.

Observed atmospheric fCO2 values during the cruise. The data is choppy because the LI-7500 does not operate well in moist conditions and because instances where wind is blowing from the stern of the ship must be removed to avoid contamination. The dashed line shows the average fCO2atm value.

3.3. Calculated Fluxes

[22] Figure 5 shows observations of the air-sea fCO2 gradient using the station average fCO2atm. The pattern follows the expected distribution from Figure 2, with undersaturation of surface waters for the offshore samples, and supersaturation along the coastal regions. The strongest undersaturation was located in Hudson Strait (−92 μatm) and was associated with a very low fCO2sw sample (286 μatm, Figure 2) and a fairly high fCO2atm observation (378 μatm, Figure 4). The surface waters of northern Hudson Bay were found to be significantly undersaturated as a result of low fCO2sw values (Figure 2) and fCO2atm near the mean value. The highest supersaturations were observed near the mouth of James Bay, and along the southeast coast of Hudson Bay. These observations were the result of high fCO2sw values and near-mean fCO2atm values. Observations on the southwest coast near the outflow of the Nelson and Churchill rivers also showed supersaturation, supporting the hypothesis that river discharge is strongly affecting the carbon regime in Hudson Bay [Granskog et al., 2007].

Figure 5.

Observed instantaneous air-sea fCO2 gradient (ΔfCO2). Dark columns indicate positive ΔfCO2, and white columns indicate negative ΔfCO2.

[23] The computed instantaneous fluxes of CO2 based on the ΔfCO2 values and calculated values of solubility and transfer velocity are shown in Figure 6. Outgassing of CO2 was observed in the river-influenced regions of southern Hudson Bay, while absorption of CO2 was observed toward the center of the Bay and along the northern channels. The importance of wind speed, and hence transfer velocity is very apparent in Figure 6. For example, the highest estimated evasions of CO2 on the James Bay transect occurred near the western coast despite the sites on the western shore being less strongly supersaturated (Figure 5). This pattern is a result of the relatively low wind speeds encountered at the east end of the transect in comparison to the west end. The highest observed absorptions of CO2 were in northern Hudson Bay, where strongly undersaturated waters were combined with high wind speeds. The range of solutions based on the different parameterizations of k are shown as error bars in Figure 6. This approach allows for the identification of the range within which the true flux of CO2 likely falls.

Figure 6.

Estimated instantaneous flux of CO2. Dark columns indicate positive fluxes (evasion) and white columns indicate negative fluxes (absorption) calculated using the Sweeney et al. [2007] (S07) parameterization. The error bars indicate the range of computed fluxes using Wanninkhof [1992] (W92) (high bar) and Nightingale et al. [2000] (N00) (low bar).

4. Discussion

4.1. Freshwater Influences on fCO2sw Distributions

[24] Understanding how river discharge may affect fCO2sw in Hudson Bay requires an understanding of CO2 dynamics in rivers and estuaries. Rivers are commonly supersaturated in fCO2 upstream of estuaries largely as a result of high CO2 in groundwater and net heterotrophy in the rivers themselves [Hope et al., 1994]. Peatland streams (which compose a major portion of the Hudson Bay watershed) have been found to be particularly supersaturated in CO2 [Hope et al., 2004; Worrall et al., 2005]. The discharge of highly saturated river water has the potential to increase fCO2sw in estuaries and even in adjacent coastal seas through simple mixing, however Borges et al. [2006] have calculated that this effect is usually small.

[25] Once river water enters an estuary, it undergoes intense biogeochemical modification. In nearly all inner estuaries studied to date, the net result of these modifications is supersaturation of fCO2sw [Frankignoulle et al., 1998; Raymond et al., 2000; Borges et al., 2005]. Generally, this is thought to be caused by estuaries acting as net heterotrophic systems (respiration exceeds primary production), although studies confirming this hypothesis are sparse [Raymond et al., 2000]. According to this theory, an abundance of labile organic carbon of allochthonous origin provides a constant source for bacterial respiration [Smith and Hollibaugh, 1993; Raymond et al., 2000], which in turn outstrips production that may be light limited because of high turbidity [Cloern, 1987].

[26] After water is modified in the inner-estuarine zone, it enters the outer estuary (or river plume) where the CO2 source/sink status becomes more ambiguous (see, for example, the review by Abril and Borges [2004]). The CO2 balance of outer estuaries depends on two main processes: (1) the balance between heterotrophic and autotrophic biology (as in inner estuaries) and (2) the advection of fCO2sw rich waters from inner estuaries. In outer estuaries that are stratified and characterized by river outflow with high nutrient concentrations a strong autotrophic drawdown of fCO2sw can occur leading to a sink of CO2 (see for example the study by Körtzinger [2003] on the Amazon River plume). Conversely, the Scheldt River plume has been found to be consistently supersaturated in fCO2sw (except during a spring phytoplankton bloom) as a result of thorough tidal mixing, net heterotrophy and advection from the inner estuary [Borges and Frankignoulle, 2002].

[27] Other possible mechanisms that modify fCO2sw in inner and outer estuaries include photomineralization of CDOM [Johannessen and Miller, 2001] and the mixing of river and seawater with different carbonate chemistries. Photomineralization of CDOM is most effective where light is not limited, so it tends to occur most rapidly offshore (i.e., in outer estuaries and beyond). The role of mixing and buffering between river and oceanic end-members can be significant, but is also complicated; for example Körtzinger [2003] found that the mixing of Amazon river water with high-alkalinity ocean water creates a significant drawdown of CO2, while Abril et al. [2003] found the role of calcite buffering to be an important source of CO2 in the Loire River, and an important sink of CO2 in the Loire River estuary.

[28] To examine how river and estuarine input affects fCO2sw in Hudson Bay, it is instructive to look at where freshwater influences are most extreme. Figure 7a shows the distribution of salinity in Hudson Bay, which is a reasonable proxy for river runoff. Freshwater outflow from the Nelson, Churchill, Severn and Winisk rivers produces a band of low salinity (<28 practical salinity units) that closely follows the southern coast [Prinsenberg, 1986b; Granskog et al., 2007]. The lowest salinities were observed near James Bay, where the influence of freshwater has a particularly large extent; in this case extending well out into the bay along the southeast coast. River discharge also has an influence on SST by stratifying the surface water, allowing for heating by solar radiation without mixing with cooler waters at depth [Prinsenberg, 1986b]. As a result, the pattern of SST in Hudson Bay largely follows the salinity pattern with higher temperatures along the southern coasts, particularly near James Bay (Figure 7b).

Figure 7.

(a) Distribution of surface salinity in practical salinity units as determined from conductivity-temperature-depth (CTD) sensor casts and (b) distribution of sea surface temperature in °C as determined from CTD casts. The open dots are sampling locations (see Figure 1).

[29] To examine the relationships between fCO2sw and freshwater signals in the Hudson Bay system, a correlation matrix was created (Table 1). fCO2sw and salinity exhibit a strong negative correlation (−0.79), while fCO2sw and SST exhibit a strong positive correlation (0.89), confirming the fCO2sw/freshwater relationships apparent in Figures 2 and 7. Table 1 also shows that fCO2sw is strongly positively correlated with CDOM (0.72). Granskog et al. [2007] found that CDOM in Hudson Bay is largely related to terrestrial input, and can therefore be used as an indicator of terrigenous carbon input. Rivers entering Hudson Bay have been shown to be significant sources of dissolved organic carbon (DOC) and particulate organic carbon (POC), which would create a source of organic material for respiration [Meybeck, 1982; Hudon et al., 1996; Granskog et al., 2006]. Primary production in the nearshore zone is likely limited by low nutrient concentrations in the rivers [Hudon et al., 1996] and low light availability due to an abundance CDOM [Granskog et al., 2007]. Despite these limitations, Anderson and Roff [1980] found the coastal waters of Hudson Bay to be higher in Chla than offshore waters (likely a result of nutrient supply from coastal resuspension), which explains the weak positive correlation (0.29) between Chla and fCO2sw.

Table 1. Pearson's r Correlation Coefficients for Hydrocast/fCO2 Variables (n = 56)a
 fCO2SWSSTChlaSalinityCDOM (A355)
  • a

    SST, sea surface temperature; CDOM, colored dissolved organic matter.

  • b

    Values are significant to the 99% confidence interval.

  • c

    Values are significant at the 95% confidence interval.

CDOM (a355)0.72b0.73b0.23c−0.92b1b

[30] Without a more comprehensive time series or distinct end-member samples of fCO2sw from river water, it is difficult to draw any firm conclusions about what may be driving supersaturations of fCO2sw in the nearshore environments of Hudson Bay. On the basis of observed salinities (Figure 7a) a large portion of Hudson Bay can be considered to be an extended outer estuary. As discussed earlier, the source/sink status of these regions depends on the balance between production and respiration, and the advection of CO2 rich estuarine water. Given the largely oligotrophic nature of Hudson Bay during this time period, it is likely that advection of river/estuarine water is playing a key role in driving supersaturations near the coast.

[31] In the more marine areas away from the coast, there is an oceanic uptake of atmospheric CO2 (Figures 2 and 6). These undersaturations might be indicative of a lesser influence of advected estuarine water, or may be a relict of oceanic processes that have depleted fCO2sw in the source water regions to the north (refer to the general circulation of the area depicted in Figure 1). The low fCO2sw values in Foxe Channel and Hudson Strait (Figure 2) suggest that water flowing into Hudson Bay is undersaturated, but unfortunately the sampling program did not survey rivers along the northwest coast to determine possible contribution of fCO2sw by those rivers. Another possibility is that effects of melting sea ice on fCO2sw are more pronounced in these regions, where meltwater contributes a more significant fraction of the freshwater budget [Granskog et al., 2007]. Ice melt should dilute the concentration of dissolved inorganic carbon (DIC) and alkalinity which would in turn lower fCO2sw. Furthermore, these offshore areas of the bay are cooler than the onshore areas (Figure 7b), which should keep fCO2sw low because of thermodynamic effects [Takahashi et al., 1993].

4.2. Estimation of Air-Sea Carbon Flux Budget for the Open Water Season

[32] The calculation and interpretation of instantaneous CO2 fluxes (Figure 6) from short-term cruises can provide a process-level understanding of hourly trends and patterns, but reveals little about the overall source-sink status of the region. Atmospheric conditions (wind, fCO2atm) change rapidly (on timescales of minutes to days), but the oceanic conditions (SST, salinity, fCO2sw) change much more slowly (on timescales of weeks to months). In this section, we exploit the conditions of slow oceanic change to extend our observations temporally and spatially to compute a rough estimate of the integrated flux of CO2 across the bay during the study period.

[33] The obvious problem with deriving a flux estimate for an extended time period is that the rate of processes perturbing fCO2sw may not be constant, thus biasing any estimate to the time of sampling. In the case of Hudson Bay, it is possible that biological activity and the rate of river discharge may have changed through the sampling period, creating biases in the estimate of CO2 flux. Although these potential pitfalls should be kept in mind, we feel that the resulting errors will be acceptable given that Hudson Bay has been found to be oligotrophic during this time period [Anderson and Roff, 1980] and since river discharge is fairly consistent [Déry et al., 2005].

[34] We have spatially limited our estimate to Hudson Bay proper, because of a lack of data from Hudson Strait, Ungava Bay, Foxe Basin and James Bay. To create the estimate, surface observations of salinity and SST from CTD casts were extrapolated onto a 5 km grid (see Figure 7) and used to calculate solubility and Schmidt number. National Centers for Environmental Prediction reanalysis long-term mean wind observations [see Kalnay et al., 1996] averaged for the August–October period and scaled to 10 m assuming neutral conditions (Figure 8) were also interpolated onto the grid and were used to calculate gas transfer velocity using long-term formulations of the S07, W92 and N00 parameterizations. fCO2sw observations were extrapolated spatially using the salinity grid and a regression relationship between fCO2sw and salinity (fCO2sw = −14.3(SAL) + 767.4, r2 = 0.78 (Figure 9a); fCO2sw was not extrapolated independently because of a lack of observations in some regions). Combining these products, fluxes of CO2 were calculated for Hudson Bay for the open water season (Figure 9b). Spatially averaging the fluxes yields an integrated flux of −0.73 (±0.4) mmol m−2 d−1 using the S07 parameterization (−0.92 (±0.4) mmol m−2 d−1 and −0.56 (±0.3) mmol m−2 d−1 following W92 and N00, respectively), indicating that the region acted as a modest sink of CO2 during this time period.

Figure 8.

Long-term mean 10 m wind speed for August–October in m s−1. Calculated from the National Centers for Environmental Prediction/National Center for Atmospheric Research long-term monthly mean data derived from the years 1968–1996.

Figure 9.

(a) Calculated open water fCO2sw distribution in μatm and (b) estimated flux of CO2 in mmol m−2 d−1. The open dots are sampling locations (see Figure 1).

[35] The open water season in Hudson Bay typically lasts 3 months (August–October) with some interannual variability [Markham, 1986]. Since the field data for this study was collected during a good portion of this season, it is tempting to use this estimate to calculate an overall flux for the season. For example, by multiplying the S07 value by the spatial extent of Hudson Bay (7.32 × 105 km2) and accounting for the 3 month open water season, an absorption of −0.58 (±0.3) TgC can be estimated for the season. The problem with this estimate, however, is that it assumes that fCO2sw modifying processes remain stationary beyond the study period. While this stationarity might be argued for biological processes and river runoff, satellite SST climatologies (see for example, http://oceancolor.gsfc.nasa.gov/) show that temperatures cool significantly between the beginning of August and the end of October. Since water temperature exerts a strong thermodynamic influence on fCO2, the lack of fCO2sw observations in August makes such an estimate somewhat questionable.

[36] An estimate of open water season CO2 flux is desirable since sea ice is generally regarded as a barrier to gas transfer, making it likely that the majority of air-sea CO2 flux occurs during these 3 months. In theory, this flux could then be added on to the Cai et al. [2006] coastal ocean carbon budget, since they did not include Hudson Bay. However, recent observations from the Arctic and Antarctic have called into question the traditional paradigm of ice as a barrier to gas transport, suggesting that fluxes during periods of the ice-covered season cannot be ignored [Papakyriakou et al., 2003; Semiletov et al., 2004; Delille et al., 2007]. Furthermore, it could be argued that ice-out may be an important time for air-sea CO2 fluxes in this region. Recent studies have shown that winter fCO2 values measured under ice in northern rivers can be very high [Striegl et al., 2007] and that the spring freshet carries a high DOC signal [Finlay et al., 2006] that may be mineralized to CO2 in the ocean. This raises the possibility of a buildup of under-ice fCO2sw during the winter followed by a burst of outgassing similar to that predicted for frozen lakes at ice-out [e.g., Striegl et al., 2001]. Finally, the nature of our scaling exercise implicitly ignores the potential of a spring phytoplankton bloom that may cause a drawdown of CO2. Studies of these seasonal contributions of CO2 flux will be necessary to properly constrain the annual exchange of CO2 in the Hudson Bay region.

4.3. Comparisons to Other Arctic Coastal Seas

[37] The results of this study indicate that the offshore continental shelf regions of Hudson Bay act as a sink of CO2, while the nearshore outer-estuary systems act as a source. These observations are consistent with other high-latitude continental shelves, and seem to indicate that Hudson Bay fits well within the coastal CO2 flux budgeting approach for high latitudes suggested by Borges et al. [2005]. However, the results in the present study underline the fact that caution should be taken when assessing analogs for coastal seas (as in the papers by Borges et al. [2005] and Cai et al. [2006]). Table 2 shows that Hudson Bay does not compare well with much higher open water season sinks observed in the Chukchi Sea (−30 to −90 mmol m−2 d−1) [Bates, 2006], the Beaufort Sea (−12 mmol m−2 d−1) [Murata and Takizawa, 2003] and the Barents Sea (−9.5 mmol m−2 d−1) [Kaltin et al., 2002]. Strong absorptions in these regions are primarily driven by a highly productive spring bloom, followed by cooling of the sea surface in the autumn that sustains low fCO2sw. Better analogs for Hudson Bay appear to be the Laptev (−2.1 mmol m−2 d−1) and East Siberian Seas (+0.3 mmol m−2 d−1) [Nitishinsky et al., 2007]. This is not surprising, because the Laptev and East Siberian seas are similar to Hudson Bay in the sense that they both receive a great deal of river outflow and terrestrial carbon. The Gulf of Bothnia is an extreme example of similar processes; high river outflow and associated terrigenous carbon creates net heteorotrophy throughout most of the seasonal cycle, driving strong annual outgassing of CO2.

Table 2. Comparison of Computed Fluxes for Hudson Bay With Fluxes Computed for Other Arctic Shelves
RegionOpen Water Flux (mmol m−2 d−1)Annual Flux (mmol m−2 d−1)Reference
  • a

    Annual flux calculated by assuming a 100 day open water season and no contribution of fluxes from other seasons (as per Cai et al. [2006]).

Chukchi Sea−30 to −90−14.8Bates [2006]
 −12.0−3.3aMurata and Takizawa [2003]
Chukchi Sea/Bering Shelf-−7.1Kaltin and Anderson [2005]
Barents Sea-−2.9Fransson et al. [2001]
 −9.5−2.6aKaltin et al. [2002]
 -−4.3Omar et al. [2007]
Laptev Sea−2.1−0.6aNitishinsky et al. [2007]
East Siberian Sea+0.3+0.1aNitishinsky et al. [2007]
Beaufort Sea−12.0−3.3aMurata and Takizawa [2003]
Northeast Water Polynya-−5.5Yager et al. [1995]
North Water Polynya-−1.0Miller et al. [2002]
Greenland Sea-−2.9Anderson et al. [2000]
 -−9.6Hood et al. [1999]
Gulf of Bothnia-+7.1 to +9.7Algesten et al. [2004]
Hudson Bay−0.73−0.2apresent study

5. Conclusions

[38] This study has shown a strong dichotomy between undersaturation of fCO2sw in the marine-influenced waters on the continental shelf of Hudson Bay and supersaturation of fCO2sw in the coastal, riverine-influenced waters of southern Hudson Bay and James Bay. This difference suggests that a combination of thermodynamic effects and the oxidation of riverine carbon are driving outgassing of CO2 along the coast. In the offshore regions absorption is most likely driven by different source water or cooler temperatures and sea ice melt. The broad patterns match well with past studies of high-latitude coastal systems, but the extent of freshwater influence (and hence the extent of ocean acting as a source of CO2) is anomalously large. Furthermore, the biological drawdown of fCO2sw appears to be lower than expected for Arctic shelves.

[39] The contribution of the large CO2 source area and depressed photosynthetic activity make the region a weaker sink than expected. A rough calculation revealed a net absorption of −0.73 mmol m−2 d−1 during the study period, which is significantly lower than other Arctic/sub-Arctic shelves [Borges et al., 2005; Cai et al., 2006]. However, a solid estimate of CO2 flux for the open water season requires observations closer to the ice-out period. The ultimate goal of calculating an annual budget of CO2 in Hudson Bay will also require observations from other seasons and a better understanding of ocean-sea ice-atmosphere CO2 flux involving a landfast or mobile sea ice cover. Future field work will focus on revisiting Hudson Bay in the spring, summer and winter seasons to better understand intraannual and interannual CO2 flux variation. Attempts will also be made to improve the spatial and temporal scaling of CO2 flux by exploiting strong fCO2sw/CDOM/SST relationships (Table 1) to develop remotely sensed products of regional air-sea exchange [e.g., Lohrenz and Cai, 2006].

[40] The main contribution of this paper is a preliminary look at air-sea CO2 exchange in the Hudson Bay region. Since this area is expected to be subject to significant climate change [Gagnon and Gough, 2005], and changes to river outflow from hydroelectric development [Déry et al., 2005] such baseline observations will be crucial in establishing how the air-sea CO2 flux regime is affected in the future.


[41] We wish to thank C.-J. Mundy, J. Ehn, M.-E. Rail, S. Blondeau and the captain and crew of the CCGS Amundsen for logistical and technical support in the field. We are grateful to two anonymous reviewers whose suggestions improved the quality and content of this paper. The authors of this paper are members of ArcticNet, funded in part by the Networks of Centres of Excellence (NCE) Canada, the Natural Sciences and Engineering Research Council (NSERC), the Canadian Institute of Health Research and the Social Sciences and Humanities Research Council. This work was supported in part by NSERC CGS-M and Alberta Ingenuity Student Scholarship grants to B. Else and by research grants 107708 and 108150 from the Academy of Finland to M. Granskog. Additional funding for logistics was provided by an NSERC Discovery grant to T. Papakyriakou and by the Canada Foundation for Innovation. We gratefully acknowledge the continued logistical and analytical support from the Foothills Climate Analysis Facility of the Center for Alpine and Arctic Research at the University of Calgary and the Centre for Earth Observation Science (CEOS) at the University of Manitoba.