Journal of Geophysical Research: Oceans

Coupling primary production and terrestrial runoff to ocean acidification and carbonate mineral suppression in the eastern Bering Sea

Authors


Abstract

[1] Water column pH and carbonate mineral saturation states were calculated from dissolved inorganic carbon (DIC) and total alkalinity data collected over the eastern Bering Sea shelf in the spring and summer of 2008. The saturation states (Ω) of the two most important carbonate minerals, calcite (Ωcalcite) and aragonite (Ωaragonite) were strongly coupled to terrestrial runoff from the Yukon and Kuskokwim rivers, primary production in the surface waters, and remineralization of organic matter at depth over the shelf. In spring, before ice melt occurred, pH over the shelf was largely confined to a range of 7.9–8.1 and Ωcalcite and Ωaragonite ranged from 1.5 to 3.0 and 0.8 to 2.0, respectively. At the stations closest to river outflows, aragonite was undersaturated in the water column from the surface to the bottom. During the summer sea ice retreat, high rates of primary production consumed DIC in the mixed layer, which increased pH and Ωcalcite and Ωaragonite. However, Ωcalcite and Ωaragonite decreased by ∼0.3 in the bottom waters over the middle and outer shelf. Over the northern shelf, where export production is highest, Ωaragonite decreased by ∼0.35 and became highly undersaturated. The observed suppression and undersaturation of Ωcalcite and Ωaragonite in the eastern Bering Sea are correlated with anthropogenic carbon dioxide uptake into the ocean and will likely be exacerbated under business-as-usual emission scenarios. Therefore, ocean acidification could threaten some benthic and pelagic calcifying organisms across the Bering Sea shelf in the coming decades.

1. Introduction

[2] Since preindustrial times, the oceans have absorbed approximately 127 Pg (Pg = 1015 g C) of anthropogenically produced carbon dioxide (CO2) from the atmosphere [Sabine and Feely, 2007]. While this has mitigated the increase in atmospheric CO2 concentrations by ∼55% [Sabine et al., 2004; Sabine and Feely, 2007], it has changed the carbonate chemistry of seawater chemical speciation [e.g., Caldeira and Wickett, 2003; Andersson and Mackenzie, 2004; Feely et al., 2004; Orr et al., 2005; Millero, 2007] with potentially significant impacts to current and future ocean biota [Andersson et al., 2007; Fabry et al., 2008, 2009; Cooley and Doney, 2009; Feely et al., 2009]. Most notably, the absorption of atmospheric CO2 by the ocean has resulted in a lowering of pH, especially over the last few decades [e.g., Bates, 2007; Byrne et al., 2010], with a subsequent decrease in the availability of carbonate ions ([CO32−]) and a suppression of the saturation states (Ω) of calcium carbonate minerals (CaCO3) which could result in a reduction of suitable habitat for marine calcifiers. These processes, collectively termed ocean acidification (OA), have occurred naturally over geologic time scales [e.g., Zhuravlev and Wood, 2009; Zachos et al., 2005] but have been accelerated due to anthropogenic emissions from industrial processes and changes in land use [Feely et al., 2004; Sabine et al., 2004; Orr et al., 2005; Caldeira and Wickett, 2005].

[3] As CO2 levels rise in the atmosphere, the increased partial pressure of carbon dioxide (pCO2) in seawater contributes to OA and the suppression of biologically important carbonate mineral concentrations, such as calcite and aragonite, through a series of well-known reactions

equation image
equation image
equation image
equation image
equation image

Following dissolution (equation (1)), dissolved CO2 undergoes hydration reactions to form carbonic acid (equation (2)), which rapidly dissociates to form carbonate and releases hydrogen ions (equations (3) and (4)). Almost all of the produced carbonate ions react with calcium to form mineral solids (equation (5)), preventing this reaction from contributing to dissolved alkalinity. Further, most of the free hydrogen ions produced react with the naturally dissolved alkalinity in seawater, reducing carbonate ion concentrations. The remaining hydrogen ions contribute to the lowering of pH. Carbonate mineral saturation states are dependent on the concentration of free carbonate ions according to the following equations, such that a reduction in available CO32− (equation (5)) decreases the saturation states of both aragonite and calcite:

equation image
equation image

[4] Cold ocean temperatures increase the solubility of CO2 and precondition the seawater to have lower calcium carbonate concentrations and saturation states compared to more temperate ocean environments, leaving polar and subpolar shelves particularly vulnerable to OA [Orr et al., 2005; Bates and Mathis, 2009; Fabry et al., 2009; Steinacher et al., 2009]. In addition to this temperature effect, several other processes affect the carbonate system and can contribute to the intensification of OA in polar and subpolar regions, including seasonally high rates of primary production, river runoff, and sea ice formation and melt processes [e.g., Bates and Mathis, 2009; Bates et al., 2009]. For example, seasonally intense periods of primary production are uncoupled from grazing in most polar environments [e.g., Springer et al., 1996; Macdonald et al., 2009] leading to high rates of organic matter export from the surface layer [e.g., Mathis et al., 2007]. While this export production supports the biologically diverse benthic communities in these regions it leads to elevated rates of remineralization in bottom waters and sediments [Grebmeier and McRoy, 1989; Devol and Christensen, 1993; Christensen, 2008; Alonso-Sáez et al., 2008; Garneau et al., 2009]. Thus, ocean biology tends to drive seasonally divergent trajectories for seawater chemistry, with primary production in the euphotic zone increasing Ω in the mixed layer while an accumulation of dissolved inorganic carbon (DIC) in subsurface waters through remineralization suppresses Ω [Bates et al., 2009].

[5] The reduction and undersaturation of carbonate minerals, particularly in bottom waters of polar and subpolar seas could have profound implications for benthic ecosystems. The subpolar continental shelf of the eastern Bering Sea (Figure 1) sustains a vast and commercially valuable benthic fishery [Cooley and Doney, 2009; Cooley et al., 2009] that produces approximately 47% of the U.S. fish catch by weight. This fishery is critical to both the regional and national economy and subsistence communities in Alaska with some species already potentially at risk (e.g., walleye pollock, pink salmon, king crab, tanner crab, ribbon seals) [Fabry et al., 2008, 2009; Chilton et al., 2010; Boveng et al., 2008]. Further decreases in pH and Ω could have significant consequences for the benthic and pelagic ecosystems in a region where organisms are already struggling to adapt to changing environmental conditions [Løvvorn et al., 2003; Moore et al., 2003; Overland and Stabeno, 2004; Grebmeier et al., 2006]. Given the importance of the Bering Sea fishery, we must determine the controls and extent of OA and carbonate mineral saturation states in this region.

Figure 1.

Map of the eastern shelf of the Bering Sea shown with generalized circulation in solid arrows, the location of the Yukon and Kuskokwim rivers in dashed lines, and the four transect lines (SL, MN, NP, 70M) occupied in spring and summer of 2008 in dotted lines. The coastal, middle, and outer domains are labeled.

[6] Here, we describe the seasonal variability of the seawater carbonate system over the eastern Bering Sea shelf in spring and summer of 2008 and investigate the impacts that primary production, sea ice processes, and terrestrial inputs have on carbonate mineral saturation states.

2. Background

[7] The eastern Bering Sea contains a wide, shallow shelf covering over 500,000 km2 [Askren, 1972; Coachman, 1986] from the Aleutian Islands to Bering Strait (Figure 1). Semipermanent frontal structures associated with wind, tidal mixing and bottom topography naturally divide the shelf into three along-shelf domains [Askren, 1972; Muench, 1976; Coachman, 1986; Kachel et al., 2002; Stabeno et al., 1999, 2006] with differing vertical and horizontal structure largely controlled by the penetration of atmospheric forcing and tidal mixing. The coastal domain extends from the western shores of Alaska to the 50 m isobath. Within this region, wind and tidal currents vertically mix the water column to the bottom, although some stratification occurs in spring as the result of freshwater input from river runoff (Figure 1) and sea ice melt. The Inner Front constitutes the boundary between the coastal domain and middle domain and approximately follows the 50 m isobath [Kachel et al., 2002]. A well stratified, two-layer system exists in the middle domain, where wind mixes the surface waters over a denser, tidally mixed bottom layer. The Central Front generally follows along the 100 m isobath, marking a gradual transition from the middle domain to the outer domain. A two-layer system is also present in the outer domain, although the transition between the surface and bottom layers is more gradual than in the middle domain. The outer domain is divided from the Bering Sea Basin at the shelf break.

[8] Large-scale circulation on the Bering Sea shelf is dominated by the advection of Pacific Ocean water from the Alaskan Stream and tidal energy dominates most of the shelf circulation, although some along shelf flow following the bathymetry to the northwest is evident in the coastal and outer domains [Coachman, 1986, 1993; Overland and Roach, 1987, Stabeno et al., 1999]. In addition, there is some seasonal cross-shelf flow directed onshore in the outer domain during spring as a result of eddies and the funneling of water through submarine canyons [Coachman, 1982, 1986; Stabeno and van Meurs, 1999; Schumacher and Stabeno, 1998; Mizobata and Saitoh, 2004]. Upwelling of deep Bering Seawater can occur over the northern shelf as a result of shoaling topography [Nihoul et al., 1993]. On-shelf flow contributes nutrients to the shelf [Stabeno et al., 1999, 2006; Nihoul et al., 1993], while tidal mixing transports coastally derived iron offshore toward the deep Bering Sea [e.g., Aguilar-Islas et al., 2007]. The highest concentrations of iron and macronutrients tend to coincide at the Central Front, producing a region of elevated phytoplankton production known as the “Green Belt” [e.g., Springer et al., 1996; Aguilar-Islas et al., 2007; Mathis et al., 2010].

[9] The physical environment of the Bering Sea shelf is seasonally dominated by sea ice advance and retreat [Luchin et al., 2002; Walsh and Johnson, 1979]. Sea ice is produced in the northern Bering Sea and advected southward by winds [Stabeno et al., 2007]. Large-scale atmospheric forcing manifested in winter storm tracks generally dictate the extent of sea ice cover and the timing of sea ice retreat for a particular year, causing large interannual variations [Niebauer, 1998; Wyllie-Escheverria, 1995]. Over the long-term, maximum sea ice extent coincides with the negative phase of the Pacific Decadal Oscillation (PDO), although changes in the Arctic Oscillation (AO) and annual oscillations of the Aleutian Low are also causally related to recent changes in sea ice extent [Stabeno et al., 1998, 2001; Stabeno and Overland, 2001]. Because of the relatively long flushing time (>3 months) [Coachman, 1986] of the middle domain, the formation and melting of sea ice usually results in the formation of a cold pool of bottom water isolated to the middle domain [Wyllie-Escheverria and Wooster, 1998; Kachel et al., 2002].

[10] Sea ice melt is the primary source of freshwater that influences the central and outer domains [Aguilar-Islas et al., 2008]. However, like most other arctic and subarctic shelves, the eastern Bering Sea receives a disproportionately large volume of freshwater input from rivers along the coast [Opsahl and Benner, 1997; Opsahl et al., 1999; Peterson et al., 2002; Hansell et al., 2004; Wheeler et al., 1997]. The principle sources of terrestrial runoff onto the Bering Sea shelf are the Yukon and Kuskokwim rivers [Lisitsysn, 1969] (Figure 1). The Yukon River Basin spans 853,300 km2 and is the fifth largest drainage basin in North America in terms of average annual discharge [Brabets et al., 2000; Schumm and Winkley, 1994], delivering ∼200 km3 water into the northern Bering Sea annually [Brabets et al., 2000; Striegl et al., 2005; Stabeno et al., 2006]. The Kuskokwim River Basin is smaller (130,000 km2) and contributes 34 km3 annually, but has a greater impact over the southern and central parts of the eastern Bering Sea [Feely et al., 1981]. Seasonal discharge peaks during May and June for both rivers, in conjunction with peak snowmelt [Brabets et al., 2000; Dornblaser and Striegl, 2007]. A secondary pulse of increased discharge occurs in August due to peak glacial melt in the Yukon River Basin [Dornblaser and Striegl, 2007].

[11] Shelf circulation of river discharge is primarily restricted to the coast by a series of persistent nearshore fronts. The runoff from the Kuskokwim River enters Kuskokwim Bay and flows along the shelf to the north, directed by tidal and wind driven currents [Feely and Cline, 1976] (Figure 1). A semipermanent front at the mouth of Kuskokwim Bay restricts cross shelf flow south of Nunivak Island [Belkin and Cornillon, 2005; Danielson et al., 2006]. At Nunivak Island, a portion of the flow breaks away from the coast and passes along the western side of the island as it travels north. In contrast, flow from the Yukon River is less restricted to the coast. Although most Yukon River discharge is directed into Norton Sound, a front at the mouth of the sound directs some flow directly toward Bering Strait (Figure 1).

[12] Overall, river runoff is mostly isolated to the coastal domain by the Inner Front. Its influence is extensive enough to significantly impact the vertical structure in the coastal domain [Coachman, 1986; Kachel et al., 2002], especially during seasonal periods of increased discharge (May–June) [Brabets et al., 2000]. Mixing of this water with Bering shelf water produces a unique, low-salinity water mass known as Alaskan Coastal Water (ACW) [Coachman, 1986] and limited westward penetration of discharge from both rivers can occur under certain wind forcings [Danielson and Kowalik, 2005; Amos and Coachman, 1992; Coachman and Shigaev, 1992].

3. Methods

3.1. Cruise Information and Water Column Sampling

[13] Physical, chemical and biological measurements were made from the USCGC Healy during spring (April/May) and summer (July) cruises to the eastern Bering Sea in 2008 as part of the Bering Ecosystem Study (BEST) project. Stations were occupied on three east–west transect (SL, MN, and NP lines) lines and one north–south transect along the 70 m isobath (Figure 1). The SL line was the northernmost line extending from nearshore across the broad northern part of the shelf to a depth of ∼90 m. The central line (MN) extended roughly from the southern tip of Nunivak Island, across the shelf south of St. Matthew Island and terminated at the shelf break (2000 m deep). The southern line (NP) extended from the southern tip of Nunivak Island southwest past the 150 m isobath. The north–south line followed the 70 m isobath (70M) down the length of the shelf from the SL line and ended southeast of the NP line. At the beginning of the spring cruise, sea ice cover was near 100% at all stations with the exception of stations at the southern end of the 70M line and some minor leads, particularly around the islands. Toward the end of the spring cruise, sea ice had diminished and the southern stations of the 70M line were sea ice free when sampled. During summer, the entire Bering Sea shelf was sea ice free.

[14] At each CTD/hydrocast station, water samples were collected for salinity, inorganic nutrients (ammonium, nitrate, nitrite, phosphate, reactive silicon, and urea), DIC, total alkalinity (TA) and dissolved oxygen (DO). Seawater samples for DIC/TA were drawn from Niskin bottles into precleaned ∼300 mL borosilicate bottles. These samples were subsequently poisoned with mercuric chloride (HgCl2) to halt biological activity, sealed, and returned to the laboratory for analysis. Sea ice cores were collected at seven locations across the Bering Sea shelf during the spring cruise. Cores were partitioned into 10 cm sections and kept frozen until analysis. Cores were allowed to thaw and the meltwater was transferred into borosilicate bottles, poisoned with HgCl2 and analyzed for DIC and TA. All sampling and analysis was performed in compliance with the guide to best practices for ocean acidification research and reporting [Riebesell et al., 2010].

3.2. Laboratory Analysis and Calculation of Carbonate Parameters

[15] DIC and TA samples were analyzed using a highly precise and accurate gas extraction/coulometric detection system [Bates, 2001]. The analytical system consists of a VINDTA 3C (Versatile Instrument for the Detection of Total Alkalinity; http://www.marianda.com) coupled to a CO2 coulometer (model 5012; UIC Coulometrics). TA samples were also determined by potentiometric titration using the VINDTA 3C. Routine analyses of Certified Reference Materials (CRMs, provided by A.G. Dickson, Scripps Institution of Oceanography) ensured that the accuracy of the DIC and TA measurements were within 0.05% (∼1 μmoles kg−1) and stable over time.

[16] Seawater pH and CaCO3 saturation states for calcite (Ωcalcite) and aragonite (Ωaragonite) were calculated from DIC, TA, temperature, salinity, phosphate, and silicate data using the thermodynamic model of Lewis and Wallace [1995]. The carbonic acid dissociation constants of Mehrbach et al. [1973] (as refit by Dickson and Millero [1987], i.e., pK1 and pK2) were used to determine the carbonate parameters. The CO2 solubility equations of Weiss [1974], and dissociation constants for borate [Dickson, 1990], silicate and phosphate [Dickson and Goyet, 1994] were used as part of the calculations. Uncertainty in the calculation of Ωcalcite and Ωaragonite were ∼0.02.

4. Results

4.1. Seasonal Variability in Seawater Carbonate Parameters

[17] TA in spring across the shelf and throughout the water column ranged from ∼2150 to ∼2440 μmoles kg−1 over a salinity range of 31 to 34.5 (Figure 2). In summer, TA was reduced by as much as 100 μmoles kg−1 in the surface layer (salinity 29.5–31; Figure 2). In general, TA was higher over the northern regions of the shelf (>60°N) and offshore in spring (Figure 3a). TA decreased the most between spring and summer over the northern part of the shelf (Figure 3b) and was higher over the southern shelf in summer. A similar trend was observed in DIC, where the highest drawdown between spring and summer occurred over the northern shelf [Mathis et al., 2010].

Figure 2.

Distribution of TA (μmoles kg−1) plotted against salinity for spring and summer of 2008. The black arrow shows the impact that ice melt and freshwater inputs have on TA, particularly in the upper water column.

Figure 3a.

Surface (upper 20 m) distribution of TA (μmoles kg−1) across the shelf and in the Yukon (Pan-Arctic River Transport of Nutrients, Organic Matter, and Suspended Sediments Project (PARTNERS) [2010] project data) and Kuskokwim [Wang, 1999] rivers (see Tables 1 and 2) for spring 2008.

Figure 3b.

Surface (upper 20 m) distribution of TA (μmoles kg−1) across the shelf and in the Yukon (PARTNERS [2010] project data) and Kuskokwim [Wang, 1999] rivers (see Tables 1 and 2) for summer 2008.

[18] In spring, pH ranged from ∼7.87 to 8.30 over the shelf, with most values falling between 8.0 and 8.1. In summer, pH increased in the surface waters by as much as 0.2. However, pH decreased in bottom waters by as much as 0.3. In surface waters, there was a gradient with pH increasing along all three transect lines moving offshore in spring (Figure 4a). Between spring and summer, pH increased in surface waters at all stations (Figure 4b).

Figure 4a.

Surface (upper 20 m) distribution of pH across the shelf and in the Yukon (PARTNERS [2010] project data) and Kuskokwim [Wang, 1999] rivers (see Tables 1 and 2) for spring 2008.

Figure 4b.

Surface (upper 20 m) distribution of pH across the shelf and in the Yukon (PARTNERS [2010] project data) and Kuskokwim [Wang, 1999] rivers (see Tables 1 and 2) for summer 2008.

[19] Increasing pH in surface waters over the shelf between spring and summer caused an average increase in Ω (Figures 5a and 5b) for both Ωcalcite and Ωaragonite. In spring, Ωcalcite and Ωaragonite ranged from 1.5 to 3.0 and 0.8 to 2.0, respectively. For Ωaragonite, nearly the entire water column exhibited Ω near the saturation horizon (Ωaragonite = 1). Average Ωaragonite was 1.25, although undersaturation was evident at the surface in some areas (Figure 5a). In summer, Ωcalcite and Ωaragonite ranged more widely, from 1.1 to 4.5 and 0.65–3.2 (Figure 5b). Increasing saturation states were particularly pronounced in the surface waters. In the upper 30 m of the water column, Ωcalcite and Ωaragonite increased by ∼1.5 and ∼1, respectively, between spring and summer. Bottom water saturation states decreased by ∼0.20 in comparison.

Figure 5a.

Calcite and aragonite saturation states (Ωcalcite and Ωaragonite) plotted against depth (m) in the upper 300 m over the shelf of the Bering Sea in spring 2008.

Figure 5b.

Calcite and aragonite saturation states (Ωcalcite and Ωaragonite) plotted against depth (m) in the upper 300 m over the shelf of the Bering Sea in summer 2008. The dashed line at 1.0 indicates the saturation horizon.

[20] In spring, along the northernmost transect line (SL, Figure 1), bottom waters were undersaturated (Ω < 1) with respect to aragonite from the coastal domain to the Inner Front (Figure 6a). At the central front (SL 10 and 11), the water column was undersaturated with respect to aragonite from the surface to the bottom (Figure 6a). During this time, the water column was vertically stratified at the Central Front, which likely lead to mixing of remineralized DIC from the bottom waters and the sediment. High silicate concentrations throughout the water column at the Central Front compared to the surrounding water masses indicated a strong benthic remineralization signature. Along the SL Line in summer, aragonite saturation states increased at the inshore stations with the exception of SL 1, where the water was undersaturated from the surface to the bottom. On either side of the central front (SL 7–14) the bottom waters (40–85 m) were undersaturated (Ωaragonite < 0.7) in aragonite. However, in the surface waters above this feature, Ωaragonite values had increased (Ωaragonite < 2.5) relative to spring values (Figure 6b).

Figure 6a.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the SL transect line in spring. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

Figure 6b.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the SL transect line in summer. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

[21] In spring, along the central transect line (MN), the water column was also vertically stratified. At the inshore stations (MN1 and 2) the water column was undersaturated with respect to aragonite from the surface to the bottom. Saturation states increased offshore with the highest values (Ωaragonite > 1.5) present in the surface waters of the outer domain (Figure 6c). In summer, the entire water column of the coastal and middle domains were saturated with respect to aragonite likely due to the drawdown of DIC throughout the water column (nitrate was depleted from the surface to the bottom in this region). Ωaragonite in the surface waters of the outer domain had also increased (Ωaragonite > 2.5). However, the saturation states in the bottom waters at the Central Front (MN 13) and in the outer domain decreased (Ω = 1 at 80 m; Figure 6d).

Figure 6c.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the MN transect line in spring. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

Figure 6d.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the MN transect line in summer. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

[22] The water column along the NP line in spring was supersaturated with respect to aragonite (Figure 6e). These saturations states increased during summer throughout the entire water column from the coast to the Central Front. However, the saturation horizon for aragonite outcropped to within 40 m of the surface (NP14–16), likely due to upwelling of deep Bering Seawater along this part of the shelf (Figure 6f).

Figure 6e.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the NP transect line in spring. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

Figure 6f.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the NP transect line in summer. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

[23] An analysis of saturation states from north to south along the 70 m isobath (70M) showed that there was a gradient between the northern and southern parts of the shelf. Saturation states for aragonite were lower north for 60°N, with undersaturated conditions present in the bottom waters at stations 50–58. Ω undersaturations were also observed in the southern part of the shelf in the bottom waters at stations 4 and 8, but these conditions were less prevalent than those in the north (Figure 6g). In summer, saturations states increased, with the exception of persistent undersaturations at station 58. The increase was greatest (Ωaragonite > 2.5) in the surface waters of the northern shelf (Figure 6h).

Figure 6g.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the 70M transect line in spring. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

Figure 6h.

Contoured sectional plots (depth in m) of aragonite saturation states (Ωaragonite) along the 70M transect line in summer. The dashed contour lines represent locations where the observed saturation states were below 1.0. The major domains, fronts, and station numbers are identified at the top.

4.2. Carbonate Parameters in Sea Ice

[24] Sea ice cores were collected across the shelf in spring of 2008 and analyzed for DIC and TA. Both DIC and TA increased with increasing salinity in these cores (Figure 7a) with the lowest concentrations found closest to the sea ice-atmosphere interface and the highest concentrations observed near the ice-water interface, where brine concentration is highest. DIC and TA were tightly correlated at all locations, exhibiting consistent TA:DIC ratios of ∼1.14, similar to the ratios obtained by Rysgaard et al. [2007] (Figure 7b). This indicates that in-ice productivity was at a minimum, while brine rejection had a substantial effect on the DIC/TA concentrations in the ice. These DIC and TA data were used to calculate the saturation states in the meltwater. The Ω values were low for both aragonite and calcite, ranging from 0.05 to 1.35 and 0.08 to 2.25, respectively (Figure 7c). All of the cores exhibited both calcite and aragonite undersaturation, except near the ice-water interface. The degree of undersaturation was not correlated to sampling location.

Figure 7a.

Geographically and vertically integrated carbonate parameters of seven ice cores collected across the Bering Sea shelf in spring of 2008 showing concentrations of DIC and TA (μmoles kg−1) plotted against salinity. Both parameters consistently decrease with decreasing salinity.

Figure 7b.

Geographically and vertically integrated carbonate parameters of seven ice cores collected across the Bering Sea shelf in spring of 2008 showing concentrations of DIC plotted against TA, showing a correlation between these two parameters (R2 = 0.97).

Figure 7c.

Geographically and vertically integrated carbonate parameters of seven ice cores collected across the Bering Sea shelf in spring of 2008 showing saturation states for calcite (Ωcalcite) and aragonite (Ωaragonite) from ice cores plotted against salinity. The dotted line indicates the saturation horizon (Ω = 1).

5. Discussion

[25] Using the data collected in 2008 and historical observations from the Yukon and Kuskokwim rivers, we can describe the influence of phytoplankton primary production, river runoff, and sea ice processes on pH and CaCO3 mineral saturation states across the shelf. The controls and impacts on the carbonate system in surface waters are described in section 5.1, and the processes driving saturation states of subsurface waters are discussed in section 5.2. The unique processes controlling carbonate chemistry in the nearshore environment are discussed in section 5.3. Finally, we discuss how the conditioning of waters over the Bering Sea shelf might influence seawater carbonate chemistry in the western Arctic Ocean in section 5.4.

5.1. Carbonate Chemistry of the Surface Waters

[26] In early summer, the combination of nutrient-rich slope waters, winter-renewed iron concentrations, nearly continuous solar irradiance, and consistent stratification over the eastern shelf of the Bering Sea creates one of the world's most productive environments [Walsh et al., 1989]. In spring of 2008, DIC concentrations ranged from 1900 to 2400 μmoles kg−1 [Mathis et al., 2010] but were drawn down in the mixed layer by as much as 150 μmoles kg−1 in summer. Meanwhile, DIC concentrations increased in bottom waters likely due to the remineralization of exported organic matter [Mathis et al., 2010]. From this seasonal change in DIC concentrations, shelf-wide average net community production (NCP) was estimated at 28 ± 10 mmoles C m−2 d−1 in 2008 with the highest rates observed in the “Green Belt” at the Central Front (40–47 mmoles C m−2 d−1), where micro- and macronutrient inputs as well as stratification are usually at their peak during late spring and summer [Springer et al., 1996; Mathis et al., 2010]. In contrast to these highly productive regions, phytoplankton blooms in the coastal domain tend to rapidly deplete all available nitrate shortly after ice retreat [Walsh et al., 1989], thereby limiting total production. NCP [Mathis et al., 2010] in the outer domain is also limited, due to low micronutrient concentrations (i.e., iron [Aguilar-Islas et al., 2007]). In 2008, there was a northwest gradient in productivity, with higher NCP trending toward northern regions in the outer and middle domains, and toward southern regions in the coastal domain [Mathis et al., 2010]. A summary of domain-specific rates of productivity is given in Table 3.

[27] Large phytoplankton blooms consume DIC in the surface layer, thereby raising pH and increasing Ω. Figures 8a8d show the changes in aragonite saturation states between spring and summer along the four transect lines. Increases in Ω can be seen in the surface waters along each line, and were particularly pronounced above the Central Front and in the northern regions of the 70M line, where our previous work indicated the highest NCP [Mathis et al., 2010]. The greatest increases in Ω between spring and summer corresponded to regions where dissolved oxygen (DO) concentrations were highest in summer, further indicating the coupling between productivity and increased Ω.

Figure 8a.

Contoured sectional plot (depth in m) of the difference in aragonite saturation states (Ωaragonite) between spring and summer along the SL transect line. The dashed contour lines represent locations where the observed saturation states were negative indicating a suppression of Ω. The major domains, fronts, and station numbers are identified at the top.

Figure 8b.

Contoured sectional plots (depth in m) of the difference in aragonite saturation states (Ωaragonite) between spring and summer along the MN transect line. The dashed contour lines represent locations where the observed saturation states were negative indicating a suppression of Ω. The major domains, fronts, and station numbers are identified at the top.

Figure 8c.

Contoured sectional plots (depth in m) of the difference in aragonite saturation states (Ωaragonite) between spring and summer along the NP transect line. The dashed contour lines represent locations where the observed saturation states were negative indicating a suppression of Ω. The major domains, fronts, and station numbers are identified at the top.

Figure 8d.

Contoured sectional plots (depth in m) of the difference in aragonite saturation states (Ωaragonite) between spring and summer along the 70M transect line. The dashed contour lines represent locations where the observed saturation states were negative indicating a suppression of Ω. The major domains, fronts, and station numbers are identified at the top.

[28] Table 3 shows a comparison between the rates of net community production and the change in saturation states in the upper 30 m. A loose trend between the rate of productivity and the increase in surface layer saturation states is obvious: The lowest change in saturation states (∼0.1) coincides with the lowest rate of productivity on the shelf, in the northern outer domain; and the greatest change in mixed layer saturation states occurred in the southern outer domain, in conjunction with the second highest rate of production, in the southern outer domain. However, the drastically low change in saturation state in the northern coastal domain seems too great to be completely due to low rates of production, and the inconsistent relationship between higher rates of productivity and greater increases in saturation states indicate that other factors must be influencing saturation states in the surface waters.

[29] Warming sea surface temperatures between spring and summer may also be contributing to the increases in saturation states. Increased temperatures raise the partial pressure of carbon dioxide, promoting outgassing events which decrease the concentration of carbon dioxide in the surface layer, in turn increasing saturation states. These effects are particularly dominant in the southern outer and middle domains [Bates et al., 2010], which may contribute to the particularly high increases in surface layer saturation states in these regions.

[30] These effects may be mitigated in some regions by the influence of ice melt [e.g., Bates et al., 2009; Yamamoto-Kawai et al., 2009]. Both DIC and TA are rejected with brine during the formation of sea ice [Glud et al., 2002; Papadimitriou et al., 2004; Delille et al., 2007; Rysgaard et al., 2007] and contribute substantially to the high-latitude carbon pump [Kelley, 1968; Gibson and Trull, 1999; Anderson et al., 2004; Semiletov et al., 2004; Omar et al., 2005; Rysgaard et al., 2007]. As the ice ages throughout the winter, in-ice productivity [Glud et al., 2002; Gleitz et al., 1995] and brine rejection can substantially alter the carbonate parameters in sea ice [Gleitz et al., 1995]. Together, nutrient exhaustion and brine rejection precondition meltwaters to have particularly low DIC and TA concentrations (Table 2), which leads to suppression and undersaturation of Ωaragonite and Ωcalcite in the ice (Figure 7). During the melt period, the mixing of meltwaters with low Ωaragonite and Ωcalcite with the surface layer likely created a divergent trajectory for Ω in the surface waters as NCP increased Ωaragonite and Ωcalcite. The increases in Ωaragonite observed in the surface layer in 2008 (Figures 8a8d) were likely moderated by meltwater. Unlike the Arctic, there is no perennial sea ice in the Bering Sea so there will not be an expansion of the influence of the low-Ωaragonite and -Ωcalcite meltwater in response to decreases in seasonal sea ice cover. However, as surface waters continue to absorb CO2 from the atmosphere the seasonal levels of Ωaragonite and Ωcalcite prior to water column productivity will continue to decrease. Because this is a macro- and micronutrient-limited system, the removal of DIC through NCP each year will not compensate for the anthropogenically induced increases in seawater pCO2. Therefore, as OA continues to decrease Ωaragonite and Ωcalcite, the onset of ice melt each year could cause Ωaragonite to become undersaturated. This effect may be particularly apparent over the inner and outer shelf where NCP is reduced, much like the undersaturations observed in the oligotrophic Canada Basin in the Arctic Ocean [Yamamoto-Kawai et al., 2009].

5.2. Carbonate System of the Subsurface Waters

[31] In response to high export production, the remineralization of organic matter increases the concentration of DIC and pCO2 in bottom waters and suppresses carbonate mineral saturation states to a varying degree across the shelf. Over the northern part of the shelf and through the Central Front, where bottom temperatures are lowest and export production is highest, we observed the strongest seasonal suppression of aragonite (∼−0.35; Figures 8a and 8b) in subsurface water. This suppression of Ωaragonite corresponded to high apparent oxygen utilization (AOU) values and elevated silicate in the bottom waters indicating both pelagic and benthic remineralization. The subsurface effects of remineralization can be especially significant during periods of intense production when Ω increases at the surface. These biologically driven, seasonally divergent trajectories of Ω, or the “Phytoplankton-Carbonate Saturation State” (PhyCaSS) Interaction, have been observed in the Chukchi Sea [Bates et al., 2009; Bates and Mathis, 2009], and are likely typical of highly productive polar and subpolar shelves.

[32] The PhyCaSS Interaction could be particularly influential on benthic calcifiers (i.e., crabs) in the Bering Sea because the lowest Ω coincide with areas of highest export production and the bottom water cold pool. It appears that the export production which provides the food source at the bottom is causing the undersaturation that could inhibit shell and test growth in calcifying organisms. However, because the Bering Sea has been highly productive since well before industrial times, we must quantify whether the observed undersaturations are a natural phenomenon or due to the absorption of anthropogenic CO2 emissions.

[33] Ideally, the amount of anthropogenic CO2 in a given system can be estimated by directly calculating the age of the water mass, but a paucity of data in this region prevents this approach. However, based on the origin of the water on the Bering Sea shelf and the observed density constraints, we can approximate anthropogenic CO2 inventories to evaluate the preindustrial state of the carbon cycle in the Bering Sea. Sabine et al. [2004] estimated that ∼35 μmoles kg−1 anthropogenic CO2 has penetrated into waters of the North Pacific Ocean to the 26 kg m−3 isopycnal surface. Because the source waters for the Bering Sea shelf are derived from the North Pacific Ocean and the density of waters we sampled ranged from 23.6 to 27.71 kg m−3, and averaged ∼25.5 kg m−3 for both spring and summer, we assume that the concentrations of anthropogenic CO2 in this region is ∼35 μmoles kg−1.

[34] To determine the impact of OA due to the uptake of anthropogenic CO2, we subtracted 35 μmoles kg−1 from our DIC observations while keeping the remaining parameters constant (TA, salinity, etc.) and recalculated the seawater Ωaragonite and Ωcalcite using the thermodynamic model of Lewis and Wallace [1995]. In this scenario, the entire water column over the shelf was supersaturated with respect to aragonite in both spring and summer. The only aragonite undersaturations present were below 100 m at the shelf break of the NP line. While there are a number of weaknesses associated with this first-order approximation, the calculation suggests that OA has resulted in persistent aragonite undersaturation in northern domain bottom waters and within the coastal domain and a suppression of Ωcalcite across the shelf. As atmospheric CO2 concentrations increase, it is likely that these undersaturations will spread across the bottom waters of the shelf for at least parts of the year.

[35] The timing of sea ice retreat may also have a substantial effect on Ω in subsurface waters. Ice retreat exerts a significant control on the fate of the organic matter produced during the phytoplankton blooms [Hunt and Stabeno, 2002; Hunt et al., 2002]. Zooplankton grazing of seasonal production is minimal during blooms associated with colder surface water temperatures favoring the benthic ecosystem [Coyle and Pinchuk, 2002]. In contrast, warmer years increase zooplankton production by up to 50% [Coyle and Pinchuk, 2002]. Thus, colder waters are expected to be associated with higher export production to the benthos, and large remineralization signals will be generated at depth, corresponding to increases in pCO2 and decreases in Ω. Warmer water blooms will retain carbon in the mixed layer and contribute to increased pelagic production and reduced bottom water remineralization.

[36] Variation in the timing of sea ice retreat could change the mode of production over the shelf. The earlier retreat of sea ice in recent years [Overland and Stabeno, 2004; Grebmeier et al., 2006; Moore and Laidre, 2006] indicates that the blooms have been occurring in colder water, favoring export production. The biological effects of this retreat have been documented in the benthos of the southern shelf, although the effect may be impacting the northern shelf as well [Grebmeier et al., 2006]. If ice continues to retreat earlier in the spring it could lead to a dichotomy for benthic scavengers such as crabs. On the one hand, higher rates of export production should lead to increased food supply and an expansion of biomass. However, if high rates of export production coupled to increasing anthropogenic CO2 inventories over the shelf cause expanded aragonite undersaturations it could lead to a reduction in habitat.

5.3. Carbonate System of the Nearshore Waters of the Bering Sea Shelf

[37] The nearshore waters of the Bering Sea shelf are seasonally dominated by terrestrial runoff from both the Yukon and the Kuskokwim rivers (Figure 1), and it is likely that the complete aragonite undersaturation of the inner stations along the MN and SL lines is the result of freshwater influence. Chemical processes occurring in both rivers precondition runoff waters to have low pH and Ω. Furthermore, the influence of productivity indicated by the divergent trajectories of Ω and AOU in the waters of the middle and outer shelf are absent in the inshore region (Table 3 and Figures 8a8d) leading us to conclude that biology is not the dominant control on Ω in the nearshore environment.

[38] In 2008, pH values in the coastal domain closely matched the pH values of river discharge (Figures 4a and 4b) during both spring and summer, indicating that even minimum rates of discharge (December–April) [Brabets et al., 2000] have a significant impact on the pH in the coastal domain. Seasonal variability in the rates of discharge exert some control over the carbonate parameters within the river (Table 1) [Striegl et al., 2005]. During spring, seasonal peaks in glacial melt and precipitation significantly dilute TA and DIC values. This dilution of DIC is minimally balanced through summer and autumn through the remineralization of peak dissolved organic carbon (DOC) concentrations resulting from increased soil drainage [Striegl et al., 2005, 2007]. Increased soil drainage in summer does not increase TA significantly as the drainage basin for both rivers is relatively carbonate poor [Cai et al., 2008; Brabets et al., 2000; Cooper et al., 2008; Dai and Trenberth, 2002], with the exception of White River, which is carbonate rich but only accounts for ∼10% of the total Yukon River discharge [Eberl, 2004; Brabets et al., 2000]. DIC concentrations reach a maximum during winter, when ice cover over the river reduces air-sea exchange causing high concentrations of pCO2 and low pH (Table 1).

Table 1. Seasonal Variation of the Carbonate Parameters at the Mouth of the Yukon Rivera
 SpringSummer-AutumnWinter
  • a

    DIC, DOC, and pCO2 data are taken from Striegl et al. [2007]. The pH and TA values are taken from the PARTNERS [2010] provisional online data set. DIC and pCO2 concentrations peak in winter due to ice cover and near continuous remineralization, creating a pH minimum.

Striegl et al. [2007]   
 DOC (μmoles kg−1)900 ± 7.0430 ± 10.0220 ± 2.7
 DIC (μmoles kg−1)1480 ± 4.21890 ± 3.04100 ± 1.9
 pCO2 (μatm)1530 ± 9.51650 ± 24.08280 ± 3.6
PARTNERS [2010]   
 pH7.87.97.0
 TA (μmoles kg−1)619.4898.61300.8

[39] Although there is a paucity of carbonate data in the Kuskokwim River, basin lithology links the Yukon and the Kuskokwim drainage areas [Gallant et al., 1995] and processes occurring during downstream transport should be similar. However, the shorter length of the Kuskokwim limits the amount of preconditioning that waters can undergo. For example, summer pH, TA and organic carbon are higher in the Kuskokwim River (Table 2), suggesting less remineralization occurs during downstream flow. In the Yukon, nearly all organic carbon that reaches the coastal margin is nonlabile, as is typical in other high-latitude shelves [Dittmar and Kattner, 2003; Fernandes and Sicre, 2000; Krishnamurthy et al., 2001; Neumann, 1999; Boucsein and Stein, 2000; Fahl and Stein, 1997]. However, because of the shorter length of the Kuskokwim River, some labile organic carbon may reach the coastal margin, where it is likely remineralized in nearshore estuaries rather than during downstream transport. While the mouth of the Kuskokwim River may exhibit higher pH and lower DIC values, we expect that coastal modification processes, such as primary productivity and remineralization, will balance these differences, and the net effect of Kuskokwim waters discharged to the inner shelf should be comparable to that of the Yukon River.

Table 2. A Comparison of the Carbonate Parameters for the Yukon and Kuskokwim Rivers and Sea Icea
 YukonbKuskokwimcSea Ice
  • a

    DOC and pH in the Kuskokwim River are higher than in the Yukon while DIC concentrations are lower, indicating a shorter time for remineralization to occur. Due to nearshore estuarine modification, the net effect of the discharge of both rivers should be similar despite these differences. Comparatively, the freshwater input from ice melt to the shelf has a much lower DIC and TA.

  • b

    Striegl et al. [2007].

  • c

    Wang [1999].

DOC (μmoles kg−1)900 ± 7.0199.8
DIC (μmoles kg−1)1480 ± 4.2384.8
pCO2 (μatm)1530 ± 9.5
pH7.88.1
TA (μmoles kg−1)619.4639.4424.6

[40] Overall, these processes precondition the freshwater discharge of the Yukon and Kuskokwim rivers to have particularly low Ω for two different reasons: (1) When river discharge rates are relatively low, DIC concentrations are highest and the shelf is covered with sea ice [Makkaveev, 1994] and outgassing of high-pCO2 water is prevented at the air-sea interface, maintaining supersaturation with respect to carbon dioxide. (2) When river discharge rates are highest, the low alkalinity of river waters effectively causes the greatest dilution of shelf alkalinity [Salisbury et al., 2008].

[41] Complete undersaturation of MN 1 and 2 (Figure 6c) occurs in spring, and may indicate the influence of peak concentrations of high-pCO2 water in Kuskokwim River outflow beneath the sea ice over the shelf. Primary productivity occurring in the coastal domain draws down DIC values in summer, mitigating the springtime undersaturation caused by river discharge. In contrast, undersaturation at SL 1 (Figure 6b) occurs during summer, indicating that primary productivity cannot balance the carbonate mineral suppression caused by the dilution of alkalinity through river discharge (Table 3). Therefore, we can assume that peak river discharge exerts the greatest influence over shelf carbonate parameters in summer, and that the dilution of alkalinity is the primary driver of riverine carbonate mineral suppression on the inner Bering Sea shelf. Based on these observations, it is likely that the coastal waters from Kuskokwim Bay to Bering Strait, including Norton Sound are undersaturated in aragonite at least part of the year due to the direct influence of river discharge.

Table 3. A Comparison of the Rate of Average Net Community Production in 2008 in the Different Domains of the Eastern Bering Sea Shelf and the Average Changes in Aragonite and Calcite Saturation States Between Spring and Summer in the Upper 30 m of the Water Columna
DomainNCP Rate (mmol C m−2 d−1)Summer-Spring ΔΩaragoniteSummer-Spring ΔΩcalcite
  • a

    The different domains of the eastern Bering Sea shelf are as described by Mathis et al. [2010]. The greatest changes in saturation states occur in conjunction with high rates of net community production and warm sea surface temperatures. The lowest changes in saturation state are seen in the northern coastal domain, near the outlet of the Yukon River.

Northern coastal19 ± 250.08 ± 0.820.12 ± 1.31
Southern coastal23 ± 150.74 ± 0.471.17 ± 0.75
Northern middle37 ± 130.97 ± 0.921.55 ± 1.46
Southern middle26 ± 110.94 ± 0.941.49 ± 1.51
Southern outer34 ± 101.38 ± 0.952.2 ± 1.03

5.4. Preconditioning Surface Waters of the Western Arctic Ocean

[42] Model predictions indicate that the Arctic Ocean will experience the impacts of OA before other areas due to colder water temperatures, the loss of sea ice, and high volumes of terrestrial runoff [Orr et al., 2005; Salisbury et al., 2008]. Our observations in the Bering Sea show that biogeochemical preconditioning of Pacific Ocean inflow that occurs during transport to the Arctic could also contribute significantly to OA. The seasonal drawdown of DIC in the mixed layer, coupled with terrestrial runoff and remineralization processes, has a major impact on the carbonate chemistry of the Bering Sea shelf and may strongly influence the Arctic Ocean. Bering Sea shelf waters flow through Bering Strait, where they are then transported northward over the shallow Chukchi Sea shelf [Coachman and Barnes, 1961; Overland and Roach, 1987; Roach et al., 1995; Woodgate and Aagaard, 2005; Woodgate et al., 2005a, 2005b] where the PhyCaSS Interaction further suppresses bottom water Ω by as much as 0.34 on a seasonal basis [Bates et al., 2009; Bates and Mathis, 2009].

[43] Pacific Ocean waters and freshwater inputs biogeochemically modified over these shelves make up the majority of the upper water masses in the western Arctic Ocean [Macdonald et al., 2002; Kadko and Swart, 2004; Cooper et al., 2008]. Furthermore, the suppression of Ω in bottom waters is expected to increase, especially due to the increasing phytoplankton primary production observed in the western Arctic Ocean resulting from reduced sea ice cover and a lengthened production season [Arrigo et al., 2008]. Additionally, carbonate-poor ice melt and river discharge typical of the Arctic Ocean is also increasing [McClelland et al., 2006; White et al., 2007], further diluting TA. These mounting drivers of OA may have negative consequences for both pelagic and benthic calcifying biota in the region [Buddemeier et al., 2004; Kuffner et al., 2008; Fabry et al., 2008, 2009; Bates et al., 2009; Bates and Mathis, 2009; Doney et al., 2009].

6. Conclusions

[44] Our observations have shown that freshwater inputs from the Kuskokwim and Yukon rivers seasonally suppress Ω in the coastal waters of the eastern Bering Sea while remineralization of organic matter in bottom waters across the shelf reduces Ωaragonite and Ωcalcite by ∼0.20 and pH by ∼0.25 causing areas of aragonite undersaturation. In spring, when the water column under the ice was vertically stratified aragonite undersaturations reached all the way to the surface in an area where NCP is the elevated. In summer, primary production in surface waters significantly raises Ωaragonite and Ωcalcite and pH, particularly in the highly productive “Green Belt.” Our analysis shows that both persistent and seasonal undersaturations are likely a result of the uptake of anthropogenic CO2 in the region. Under business-as-usual projections for CO2 emissions [Intergovernmental Panel on Climate Change, 2007], it has been estimated that the surface ocean will absorb an additional 50–100 μmoles kg−1 of anthropogenic CO2 over the next 100 years resulting in further suppression of carbonate mineral saturation states. As this happens, seasonal processes like the suppression of Ω from sea ice melt could have a more substantial impact on the surface waters of the Bering Sea.

[45] Current and future reductions in pH and Ω could have significant effects on the Bering Sea and its associated economy and subsistence inhabitants. In addition to impacting the ability of calcifying organisms to maintain and form shells and tests, reductions in pH could elicit physiological responses from noncalcifying organisms through less obvious pathways. Of great concern is that climate forcing and the uptake of anthropogenic CO2 could result in an ecosystem wide shift in speciation which could be less economically viable for current fisheries. Although several current estimates of the effects of ocean acidification on economic systems have been published [Cesar et al., 2002; Burke et al., 2004; Stern, 2006; Cooley and Doney, 2009], predictions of the future effects of changing climate and biogeochemistry on the Bering Sea shelf ecosystem and economy remain unclear. However, we should not wait until a major shift occurs to prioritize this region for expanded study.

Acknowledgments

[46] The authors thank the officers and crew of the USCGC Healy for their tireless efforts in supporting our work. Without their commitment, none of the science would have been possible. We also thank the hydrographic team from PMEL, NOAA for providing data and helping in sample collection. Finally, we thank our colleagues in the BEST-BSIERP project, supported by NSF and NPRB. The work presented in this paper was supported by the Alaska OCS Region, U.S. Minerals Management Service and the Coastal Marine Institute, University of Alaska Fairbanks under agreement M08AC12645.