The Chukchi Sea, a shallow sea-ice covered coastal sea adjacent to the Arctic Ocean, exhibits an intense bloom of phytoplankton each year due to the exposure of nutrient-laden surface waters during the brief summertime retreat and melting of sea-ice. The impact of phytoplankton production and other factors on the seasonal dynamics of carbon and air-sea CO2 fluxes were investigated during two survey cruises (5 May–15 June 2002, and 17 July–26 August 2002), as part of the Western Arctic Shelf-Basins-Interactions (SBI) project. In springtime, most of the Chukchi Sea was sea-ice covered (>95%) and remnant winter water was present across the shelf. Surface layer seawater partial pressure of CO2 (pCO2) ranged from ∼200–320 μatm, indicative of undersaturation with respect to atmospheric pCO2, although sea-ice cover kept rates of air-to-sea CO2 flux generally low (<1 mmoles CO2 m2 d−1). By summertime, after sea-ice retreat, seawater pCO2 contents had decreased to very low values (<80–220 μatm) in response to high rates of localized primary and net community production (NCP) and biological uptake of dissolved inorganic carbon (DIC). In the seasonally sea-ice free regions of the Chukchi Sea shelf, rates of air-to-sea CO2 fluxes, determined using the quadratic wind speed-transfer velocity relationships of Wanninkhof (1992), were high, ranging from ∼30–90 mmoles CO2 m−2 d−1. In regions of the Chukchi Sea slope (and western Beaufort Sea shelf and Arctic Ocean basin) where sea-ice cover remained high (>80%), air-to-sea CO2 fluxes remained generally low (<2 mmoles CO2 m−2 d−1). Seasonal (i.e., May to September) and annual net air-to-sea CO2 fluxes from the Chukchi Sea shelf were estimated at ∼27 ± 7 Tg C yr−1, and 38 ± 7 Tg C yr−1, respectively. The Chukchi Sea represents the largest oceanic CO2 sink in the marginal coastal seas adjacent to the Arctic Ocean. An active continental shelf pump of carbon, driven by the northward transport of nutrient-rich water of Pacific Ocean origin, high rates of primary and net community production during the sea-ice free period, and lateral export of organic carbon, maintains the Chukchi Sea shelf and slope as a perennial ocean CO2 sink.
 The coastal ocean is the active interface where the terrestrial, ocean and atmosphere components of the Earth's biogeochemical system interact. However, CO2 dynamics, and CO2 sink or source status of the coastal ocean domain remains poorly understood due to past undersampling and limited historical studies. The CO2 source or sink status of the coastal ocean continues to be highly debated. In early papers, Walsh first suggested that the continental margins sequestered significant amounts of CO2 and that the global coastal ocean was a significant sink of atmospheric CO2[Walsh et al., 1985; Walsh, 1991]. In contrast, Smith and Hollibaugh  argued that the coastal ocean (and open ocean) was net heterotrophic and thus a potential source of CO2 to the atmosphere. More recently, other studies have suggested that the CO2 source or sink status [Borges, 2005; Borges et al., 2005], and the metabolic status of continental shelves is highly variable in time and space [e.g., Wollast, 1998; Gattuso et al., 1998]. High-latitude and temperate coastal seas tend to be net annual sinks of atmospheric CO2, while nearshore, upwelling dominated, and subtropical to tropical coastal environments tend to release CO2 to the atmosphere [Borges et al., 2005] (see references in Table A1 of Borges et al. ). In a recent review, Ducklow and McAllister  concluded that the global coastal ocean is presently net autotrophic, and a potential sink for atmospheric CO2.
 There are multifold and complex factors that influence the CO2 sink or source status of the coastal ocean. The net metabolism (either net autotrophy or heterotrophy) of a continental shelf region is influenced by a balance of transports and exchanges, such as: the input of organic matter (OM) from terrestrial sources, OM production and consumption in the coastal ocean, retention or export of OM from the shelf and, the net air-sea exchange of CO2. The balance of these processes dictates the metabolic status (i.e., autotrophic or heterotrophic) of the continental shelf, but can also influences the potential CO2 sink or source status. The spatiotemporal variability of CO2 on the coastal ocean is driven by a complex interplay of connected and quasi-independent physical and biological factors. For example, the physicochemical process of air-sea CO2 exchange can proceed independently of the biological processes (e.g., production versus consumption of OM) that dictate the metabolism of the coastal ocean. Thus, as Ducklow and McAllister  point out, the variability of pCO2 (or ΔpCO2, difference between seawater and atmospheric pCO2) and separation of biological and physical influences can lead, for example, to a net heterotrophic open (or coastal) ocean still being a sink for atmospheric CO2.
 In high-latitude polar and subpolar coastal ocean environments, Yager et al.  put forward the “seasonal rectification hypothesis” to explain seasonal seawater pCO2 variability. For example, in the North East Water (NEW) polynya [Yager et al., 1995], Ross Sea [Bates et al., 1998; Sweeney et al., 2000; Sweeney, 2003], and east of Baffin Island [Miller et al., 2002], there are brief but large drawdowns of seawater pCO2 (and DIC) due to intense primary production during the ice-free periods of summertime (austral summer in Antarctica). During the remainder of the year, Yager et al.  suggested that the sea-ice cover prevented gas exchange in the NEW polynya, thereby allowing wintertime rectification of seawater pCO2 to near equilibrium values through remineralization of OM produced earlier. The seasonal asymmetry in the air-sea CO2 fluxes, driven by primary production and wintertime sea-ice cover, is thought to cause these coastal regions to be net sinks for atmospheric CO2 [Yager et al., 1995; Miller et al., 2002]. However, it is not clear whether these shelves are weak or strong sinks, since this depends on whether OM is preferentially retained and remineralized to CO2 on the shelf (weaker CO2 sink) or OM (and DIC) is preferentially exported to the open ocean (stronger CO2 sink) during the sea-ice covered period.
 Despite the importance and highly dynamic physical and biological nature of the Chukchi, Beaufort and Bering Seas, knowledge about the distribution of CO2, air-sea CO2 fluxes and the net metabolism of the region is limited due to a lack of sufficient observations. For example, in the Bering Sea (an important preconditioner for the Chukchi Sea and Arctic Ocean basin), early studies suggest that surface waters of the highly productive “Green belt” of the shelf [e.g., Springer et al., 1996; Springer and McRoy, 1993] are potential sinks for atmospheric CO2 [Codispoti et al., 1982, 1986]. More recent studies have focused on the southwestern periphery of the Bering Sea, for example, seawater pCO2 or dissolved inorganic carbon (DIC) data has only been collected close to the western Aleutian Islands [e.g., Murphy et al., 2001; Nedashkovskii and Sapozhnikov, 2001] or outside the Bering Sea in the subarctic gyre of the North Pacific [Midorikawa et al., 2002; Andreev and Watanabe, 2002].
 On the Chukchi and Beaufort Sea continental shelves, studies on CO2 distributions or air-sea CO2 fluxes are also extremely limited. For example, inorganic carbon data were collected at a few conductivity-temperature-pressure (CTD) stations in the Chukchi Sea during the Arctic Ocean Section in 1994 [Anderson et al., 2003]. In September 1996, seawater pCO2 and DIC were calculated from measurements of pH and total alkalinity (TA) in the Laptev, East Siberian and Chukchi Seas [Semiletov, 1999; Pipko et al., 2002]. On several cruises between 1998 and 2001, Murata and Takizawa  measured surface seawater pCO2 and DIC. In this paper, we report the surface distributions of seawater DIC, TA, and calculated seawater pCO2 data in the Chukchi and western Beaufort Sea shelves, collected as part of the 2002 field surveys of the Western Arctic Shelf-Basin Interactions (SBI) project. The seasonal distributions of seawater pCO2, air-sea CO2 fluxes and the net metabolism of the Chukchi Sea shelf are examined in context of a comprehensive suite of supporting water column physical and biological carbon measurements [e.g., Codispoti et al., 2005; Bates et al., 2005a] obtained during the SBI survey cruises. The spatiotemporal CO2 data collected in the Chukchi Sea will be used to evaluate the attributes and drivers of carbon dynamics in the region, and comparisons to the “continental shelf pump” and “seasonal rectification” hypotheses. In the Chukchi Sea, it is hypothesized that the region is a perennial oceanic sink of atmospheric CO2 and that the CO2 sink-source status does not oscillate seasonally compared to the Bering Sea, for example.
2. Overview of Physical Circulation and Water Masses in the Chukchi Sea and Adjacent Arctic Ocean Basin
 Physical transformations and seasonal sea-ice cover play a major role in shaping the Chukchi Seawater masses and ecosystem. During the winter and much of the year, sea-ice covers much of the Chukchi and Beaufort Sea shelf areas. Furthermore, the water column on the Chukchi Sea shelf is confined to a small range of temperature-salinity space [Woodgate et al., 2005a], indicating mixing and homogenization by ventilation, brine rejection and mixing. During the summer, the sea-ice briefly retreats northward from the Bering Sea, through Bering Strait into the Chukchi and Beaufort Seas to the shelf break and beyond into the Canada Basin of the Arctic Ocean. On the Chukchi Sea shelf, local sea-ice melt transforms water of Pacific Ocean origin to relatively warm, fresher Polar Mixed Layer (PML) water (upper 0–∼30 m, salinity typically <31; temperature >−1.5). In the later part of the sea-ice free season, the temperature and salinity properties of the PML on the Chukchi Sea shelf widens from wintertime T-S space. The surface water of the Chukchi Shelf, outflows through the major shelf-slope conduits (i.e., Barrow Canyon, Herald Valley), merges and interjects with the Arctic Ocean basin PML waters off the slope of the Chukchi and western Beaufort Seas. On the continental shelf of the Chukchi Sea, beneath the mixed layer, are waters of the Upper Halocline Layer (UHL). The UHL layer is typically >∼30 m deep, and has a core layer with a salinity of 33.1, nitrate concentration of ∼14 ± 2 μmoles kg−1 and phosphate concentration of ∼1.8 ± 0.2 μmoles kg−1 [Codispoti et al., 2005]. The UHL is also transported off the shelf into the Canada Basin, and can be traced to the Eurasian Basin exits from the Arctic Ocean region [e.g., Jones et al., 2003].
3.1. Sampling and Analyses
 Comprehensive sampling of seawater physical, biogeochemical and biological properties on the Chukchi and Beaufort Sea shelves and Canada Basin of the central Arctic Ocean were conducted from the icebreaker USCGC Healy during spring (HLY 02-01) and summer (HLY 02-03) cruises, as part of the 2002 field phase of the SBI project [Grebmeier and Harvey, 2005]. Forty and forty-five CTD/rosette stations were occupied during spring (5 May–15 June 2002) and summer (17 July–26 August 2002), respectively. Both cruises sampled at the Bering Strait, over the Chukchi and Beaufort Sea shelves, the shelf-slope region and into the central Arctic Ocean (Figure 1). Three sections, extending across the Chukchi and Beaufort Sea shelf and slope into the Arctic Ocean, were repeated each cruise, including: (1) West Hanna Shoal (WHS); (2) East Hanna Shoal (EHS) transect, and; (3) Barrow Canyon (BC) transect. In addition, during summer field activities, a fourth section East of Pt. Barrow (EBC), and a section extending from the Alaskan side of the Bering Strait to the Diomede Islands were also taken.
 Seawater samples for DIC and total alkalinity were drawn from the Niskin samplers into precleaned ∼300 mL borosilicate bottles, poisoned with HgCl2 to halt biological activity, sealed, and returned to the Bermuda Biological Station For Research (BBSR) for analysis. DIC samples were analyzed using a highly precise (∼0.025%; <0.5 μmoles kg−1) gas extraction/coulometric detection system [Bates et al., 1996, 1998; Bates, 2002; Bates et al., 2005a]. Total alkalinity samples were analyzed using an automated potentiometric method with a precision of <∼1 μmoles kg−1 and potential inaccuracy of ∼0.1% (∼2 μmoles kg−1). Routine analyses of Certified Reference Materials (provided by A. G. Dickson, Scripps Institution of Oceanography) ensured that the accuracy of the DIC and TA measurements was 0.05% (∼0.5 μmoles kg−1) and 0.1% (∼2 μmoles kg−1) respectively.
 Seawater pCO2 concentrations were calculated from DIC, TA, temperature and salinity data using the algebraic relationships given in Peng et al. , the CO2 solubility equations of Weiss , and dissociation constants for carbonic acid (i.e., pK1 and pK2), borate [Dickson, 1990], phosphate[Dickson and Goyet, 1994]. The carbonic acid dissociation constants of Mehrbach et al.  (as refit by Dickson and Millero ) were used to determine seawater pCO2, although other dissociation constants were also used [Goyet and Poisson, 1989; Roy et al., 1993] (F. J. Millero et al., Dissociation constants of carbonic acid in seawater as a function of salinity and temperature, submitted to Marine Chemistry, 2006) (hereinafter referred to as Millero et al., submitted manuscript, 2006) to examine the potential inaccuracy of the seawater pCO2 calculation. It should be noted that since the experimental determinations of carbonic acid dissociation constants [i.e., Mehrbach et al., 1973; Goyet and Poisson, 1989; Roy et al., 1993; Millero et al., submitted manuscript, 2006] had lower limits of −1°C or 0°C, equations governing pK1 and pK2 were extrapolated to temperatures colder than −1°C. The difference in calculated seawater pCO2 between different pK1 and pK2 values was relatively small (<5 μatm) at temperatures less than 0°C. For example, seawater pCO2 data were calculated using the carbonic acid dissociation constants of Mehrbach et al.  (as refit by Dickson and Millero ) and Goyet and Poisson . The majority of samples from the Chukchi Sea were collected from waters with a temperature range of ∼−1.8°C to 0°C, and computed seawater pCO2 was slightly higher for the carbonic acid dissociation constant of Mehrbach et al.  compared to Goyet and Poisson , with an average difference of ∼2.5 μatm. The difference in calculated seawater pCO2 was very small relative to the large ΔpCO2 values observed in the region (−100 to −250 μatm). This, in turn, has minor impact on air-sea CO2 fluxes. The difference between seawater pCO2 computed from Mehrbach et al.  and Goyet and Poisson  pK1 and pK2 constants was higher (i.e., ∼5–10 μatm) for the few samples collected in warmer waters (i.e., ∼6–10°C in water at Bering Strait or in southernmost stations of the Chukchi Sea). Overall, the inaccuracy of seawater pCO2 calculations was less than 10 μatm.
3.2. Air-Sea CO2 Gas Exchange Considerations
 The net air-sea CO2 flux (F) was determined by the following formula:
where k is the transfer velocity, s is the solubility of CO2 and, ΔpCO2 is the difference between atmospheric and oceanic partial pressures of CO2. The ΔpCO2, or air-sea CO2 disequilibrium, sets the direction of CO2 gas exchange while k determines the rate of air-sea CO2 transfer. Here, gas transfer velocity-wind speed relationships for short-term and long-term wind conditions based on a quadratic (U2) dependency between wind speed and k [i.e., Wanninkhof, 1992] were used to determine air-sea CO2 fluxes:
where U10 is wind speed corrected to 10 metres, and Sc is the Schmidt number for CO2. The Schmidt number was calculated using the equations of Wanninkhof  and s (solubility of CO2 per unit volume of seawater) was calculated from the observed temperature and salinity using the equations of Weiss .
 The ΔpCO2 data sets used here to estimate air-sea CO2 fluxes were computed from seawater pCO2 data across the Chukchi and western Beaufort Sea shelves, and atmospheric pCO2 data from Point Barrow in Alaska. The atmospheric data was downloaded from http://www.cmdl.noaa.gov (National Oceanographic and Atmospheric Administration, NOAA, Climate and Meteorological Diagnostics Laboratory, CMDL), and corrected for water vapor pressure.
 Daily averaged 6-hourly wind speed data from the NCEP (National Centers for Environmental Prediction)/NCAR (National Center for Atmospheric Research) reanalysis 2 data assimilation model was used to calculate k values (http://www.cdc.noaa.gov/cdc/data.ncep.html). NCEP/NCAR Reanalysis 2 (i.e., NNR) data were used rather than shipboard meteorological data reports in order to allow regional estimates of air-sea CO2 fluxes to be made. The spatial resolution of the NNR data assimilation model is 2.5° by 2.5° (for example, 167.5°W–170°W, 70°N–72.5°N).
 In the Chukchi Sea, western Beaufort Sea and Arctic Ocean (i.e., Canada Basin), rates of air-sea CO2 fluxes were determined at each CTD/rosette station. Net air-sea CO2 fluxes were calculated using observed ΔpCO2 data and k values (using equation (2) for short-term winds) calculated from mean daily NNR wind speed in the nearest 2.5° by 2.5° box for the duration of each relevant cruise (i.e., 5 May–15 June 2002 and 17 July–26 August 2002). Rates of net air-sea CO2 flux were initially calculated assuming sea-ice free conditions (i.e., 0% sea-ice). It is assumed here that sea-ice provides an effective barrier to air-sea CO2 gas exchange, and that air-sea CO2 fluxes are a linear function of sea-ice coverage. Thus, air-sea CO2 fluxes were corrected to observed sea-ice coverage (shipboard reports and satellite data; http://www.joss.ucar.edu/sbi/). For example, in sea-ice coverage conditions of 95%, 50%, and 0%, multipliers of 0.05, 0.5 and 1.00, respectively were applied to the air-sea CO2 fluxes. In those regions with 100% sea-ice coverage, it is possible that air-sea CO2 gas exchange can occur through leads and fractures in the ice, and also directly through sea-ice [Semiletov et al., 2004]. In these regions, a multiplier of 0.01 (equivalent to 99% sea-ice cover) was used to allow for minor rates of air-sea CO2 exchange through leads, fractures and brine channels [Semiletov et al., 2004].
 Net air-sea CO2 fluxes were also calculated for the Chukchi Sea bounded by the geographic range of 65–75°N, and 150–170°W (an area of 595000 km2). Although the SBI cruises only spanned a four-month period (May to August), physical and biogeochemical observations captured the full seasonal range of wintertime (complete sea-ice coverage) to summertime (sea-ice minima) conditions. On the spring SBI cruise, mixed layer water had a very narrow range in temperature, salinity, and density across most of the Chukchi Sea. The physical data collected for the spring SBI cruise was also similar to wintertime data (i.e., December to May) collected from mooring locations at Bering Strait and across the Chukchi Sea slope [Woodgate et al., 2005a, 2005b]. Mixed-layer DIC concentrations observed over most of the Chukchi Sea shelf and slope during the spring SBI cruise also had a small range, with the average salinity normalized DIC in the 0–30 m layer of 2357 ± 4 μmol kg−1 [Bates et al., 2005a]. As such, spring DIC (and TA) data were chosen as representative of sea-ice covered, wintertime conditions and, seawater pCO2 values were calculated for each 2.5° by 2.5° area of the Chukchi Sea region for the December to May period (Table 1). Seawater pCO2 values observed in summertime (i.e., July and August; HLY 02-03 cruise) during the seasonal minima in sea-ice coverage (and after inorganic nutrient exhaustion) were chosen as representative of the seasonal minima in seawater pCO2. Mean seawater pCO2 values for each 2.5° by 2.5° area of the Chukchi Sea region was determined from station data or the nearest representative stations. At other times, during seasonal transitions (i.e., June to July, and September to November), seawater pCO2 values were estimated using simple linear interpolation between wintertime and summertime conditions. Net air-sea CO2 fluxes were thus calculated for each 2.5° by 2.5° area of the Chukchi Sea region in 2002 using monthly averaged seawater and atmospheric pCO2 data (i.e., monthly values of ΔpCO2) and NNR wind speed data. Average sea-ice coverage each month (from shipboard reports and satellite data composites; http://www.joss.ucar.edu/sbi/) for each 2.5° by 2.5° area of the Chukchi Sea region was used as multiplier to correct air-sea CO2 fluxes for sea-ice cover. If an error of seawater pCO2 data (e.g., 20 μatm higher) and wind speed (e.g., 0.2 m s−1 higher) is assumed, net annual air-sea CO2 fluxes for the entire Chukchi Sea have an error of 7 Tg C yr−1 and 2 Tg C yr−1, respectively. In addition, air-sea CO2 gas exchange is allowed for during 100% sea-ice coverage by using a multiplier of 0.01 (equivalent to 99% sea-ice coverage). If no air-sea CO2 gas exchange through 100% sea-ice coverage occurs, net annual air-sea CO2 fluxes would be lower by ∼0.04 Tg C yr−1.
Table 1. Seasonal, Annual, and Geographic Net Air-to-Sea CO2 Fluxes (Tg C) for the Chukchi Sea Bounded by the Areal Range of 65–75°N and 150–170°W (an Area of 595000 km2)a
Net Air-to-Sea CO2 Fluxes, Tg C or 1012 g C
Net air-to-sea CO2 fluxes were calculated using NNR wind speed and ΔpCO2 data sets for each 2.5° by 2.5° area of the coastal sea. ΔpCO2 were determined from seawater pCO2 data across the Chukchi and western Beaufort Sea shelves, and atmospheric pCO2 data from Point Barrow Alaska (data from http://www.cmdl.noaa.gov). Steady/long-term wind speed k formulation (equation (3)) of Wanninkhof ) was used in the calculation of CO2 flux. Negative values denote CO2 flux into the ocean.
Details of the error calculation are given in the methods section.
4.1. Springtime Physical and Biogeochemical Distributions
 The seasonal physical and biological transformations on the Chukchi Sea and western part of Beaufort Sea shelves during the 2002 SBI field program are described in detail elsewhere [Codispoti et al., 2005; Bates et al., 2005a, 2005b]. During the springtime, sea-ice cover was high at >80–100% over much of the Chukchi Sea shelf (Figure 2a), except in the region of Bering Strait. Mixed layer and UHL temperatures (−1.8°C to −1.5°C; Figure 2c), and salinities (>32.5–33.2; Figure 3a) were confined to a small range in temperature and salinity space. TA concentrations ranged from ∼2150–2350 μmoles kg−1, with similar spatial gradients to surface salinity (Figure 3b). Nitrate (∼10–15 μmoles kg−1) and phosphate (∼1.8 μmoles kg−1) concentrations in the PML and UHL were fairly uniform across the Chukchi Sea shelf [Codispoti et al., 2005; Bates et al., 2005b], and similar to nutrient concentrations in the UHL measured at Bering Strait. Chlorophyll a concentration was also generally low (<0.8 μg L−1 [Hill and Cota, 2005]). In summary, across much of the Chukchi Sea shelf and slope regions, the physical properties of water present on the shelf did not appear to have been significantly transformed from wintertime conditions.
 During springtime, surface DIC concentrations across the Chukchi Sea shelf ranged from ∼2100–2200 μmoles kg−1, with concentrations decreasing northward into the Arctic Ocean basin (Figure 4a [Bates et al., 2005a]). However, over the Chukchi Sea shelf, DIC corrected to a constant salinity (i.e., nDIC) had a very small range (∼2200–2230 μmoles kg−1; Figure 4c) at the surface. Furthermore, the mean nDIC concentration in the surface layer (0–30 m deep) also had a very small range of 2357 ± 4 μmoles kg−1. The nDIC data is corrected to the mean salinity of the UHL (i.e., 33.1) from which the PML is modified, and it is assumed that normalizing DIC and TA data to a constant salinity does not introduce potential biases [Friis et al., 2003]. The homogeneity of the nDIC data across the Chukchi Sea shelf, slope and Arctic Ocean basin also suggests that remnant winter waters were still present on the shelf during the springtime field survey.
 The springtime distributions of surface seawater pCO2 showed distinct spatial variability. On the central Chukchi Sea shelf, seawater pCO2 (∼300–350 μatm; Figure 5a) values were slightly undersaturated with respect to atmospheric pCO2 conditions (∼370 ppm). Surface seawater pCO2 values, however, decreased northward from the Chukchi Sea shelf and slope stations to the Arctic Ocean stations. At the Arctic Ocean stations under near perennial sea-ice cover, seawater pCO2 values were highly undersaturated (∼200–250 μatm; Figure 5a) with respect to atmospheric pCO2 conditions.
4.2. Summertime Physical and Biogeochemical Distributions
 Two months later in August 2002, the physical and biological properties of the Chukchi Sea were transformed from springtime conditions. Seasonal sea-ice retreat had occurred, with most of the Chukchi Sea shelf sea-ice free. In the northern areas of the Chukchi Sea shelf, ∼30–50% sea-ice coverage remained, with a higher sea-ice coverage (>80%) in the region of the Chukchi Sea slope, Beaufort Sea shelf and Arctic Ocean (Figure 2b). Seasonal warming of the surface layer was evident from the Bering Strait across the Chukchi Sea shelf to ∼71°N (Figure 2d), with surface layer salinities highly variable (<10 to 32) due to sea-ice melt (Figure 3b). Surface total alkalinities were more variable compared to spring (Figure 3d), with low values (<1600 μmoles kg−1) in areas of significant freshening due to sea-ice melt. The biological transformation was also evident, with inorganic nutrient concentrations exhausted, reduced to less than 0.2 μmoles kg−1 at all stations [Codispoti et al., 2005; Bates et al., 2005b]. Across much of the Chukchi Sea shelf (and western Beaufort Sea shelf), large concentrations of chlorophyll a were present, and high rates of in situ 14C primary production and net community production (>0.3–2.8 g C m2 d−1 [Hill and Cota, 2005; Bates et al., 2005a]), high rates of vertical organic matter export [Moran et al., 2005] and horizontal export of suspended particulate organic matter from the Chukchi Sea shelf into the Arctic Ocean [Bates et al., 2005b] were observed.
 The summertime DIC distributions in the mixed layer of the Chukchi Sea were also transformed from springtime conditions. Surface DIC concentrations had a very wide range of concentrations (∼526–1900 μmoles kg−1) with the lowest values occurring at stations where large amounts of sea-ice melt have diluted DIC and salinity. The changes in nDIC between springtime and summer were used to estimate rates of net community production [Bates et al., 2005a]. The highest rates of NCP (∼1–2.8 g C m2 d−1) and in situ primary production [Hill and Cota, 2005] occurred on the shelf in the Barrow Canyon region of the Chukchi Sea and east of Point Barrow in the western Beaufort Sea. A total NCP rate of 20 x 1012 g C for the growing season (∼120 d) in 2002 was estimated for the eastern Chukchi Sea shelf and slope region (area of 13.9 x 104 km2 [Bates et al., 2005a]). At the Arctic Ocean stations, DIC and nDIC concentrations remained almost constant between spring and summer, with NCP rates estimated to be very low ∼0.015 +0.030 g C m−2 d−1[Bates et al., 2005a], similar to previous studies [e.g., English, 1961; Moran et al., 1997; Anderson and Kaltin, 2001].
 There were also large changes in seawater pCO2 conditions on the Chukchi Sea shelf. By summer (July–August), low pCO2 conditions (<150 μatm) were observed in the surface of the PML in the Barrow Canyon region of the Chukchi Sea shelf and western Beaufort Sea shelf (Figure 5b). The low pCO2 conditions occurred in regions where there was, for example: (1) intense pelagic and benthic production [Hill and Cota, 2005; Bates et al., 2005a]; (2) complete depletion of inorganic nutrients and large drawdown of DIC (∼100–200 μmoles kg−1 [Bates et al., 2005a]); (3) production and export of organic matter [Bates et al., 2005b; Moran et al., 2005; Mathis et al., 2006] and, large increases in phytoplankton [Hill and Cota, 2005] and zooplankton biomass [Ashjian et al., 2005]. Extremely low pCO2 conditions (<100 μatm) were observed in the upper 5 m of the PML in the Beaufort Sea shelf in a region of mixing between Mackenzie River runoff, PML waters and sea-ice melt. Vertically, these regions were highly stratified with seawater pCO2 increasing with depth.
 The data reported here represent the first seawater pCO2 observations during the early growing season (i.e., May and June) and summer season (July and August). However, the SBI field survey in 2002 did not sample during the month of September before surface freezing and sea-ice advance in early October. In previous studies, Pipko et al.  and Murata and Takizawa  surveyed the Chukchi Sea in September, and a composite view of seasonality on the Chukchi Sea emerges. In 1996, Pipko et al.  found seawater pCO2 values (calculated from pH and total alkalinity) of ∼280–320 μatm from the Central Channel west of Cape Lisburne to Barrow Canyon. Along the same track, Murata and Takizawa  found similar seawater pCO2 values (∼290–350 μatm) in 1998. Subsequently, Murata and Takizawa  observed low seawater pCO2 conditions in 1999 (∼240–280 μatm) and 2000 (∼180–220 μatm), similar to the 2002 SBI observations. Thus, during period before surface freezing and return of sea-ice cover, seawater pCO2 conditions remained highly undersaturated with respect to the atmosphere.
 The late season increase in seawater pCO2 conditions from summertime minima likely reflects the net balance of organic matter remineralization (including benthic OM respiration), air to sea CO2 gas exchange and temperature changes. The first two processes add CO2 to the surface layer, increasing seawater pCO2, while temperature changes act to decrease seawater pCO2. In the fall period (during sea-ice advance), surface layer waters in the Chukchi Sea shelf, slope and Arctic Ocean region would have cooled from summertime values (∼−0.5°C to +0.5°C) to wintertime values of ∼−1.5–1.7°C. Assuming that the thermodynamic influence of temperature on seawater pCO2 is ∼4.2% °C−1 [i.e., Takahashi et al., 1993], the average cooling of surface water by ∼2°C would have decreased seawater pCO2 by ∼15 μatm, opposite in direction to the additions from OM remineralization and air to sea CO2 gas exchange.
 In the adjacent slope and Arctic Ocean basin, surface seawater pCO2 had a small seasonal range (Figure 3a; May to September; ∼200–275 μatm). Pipko et al.  and Murata and Takizawa  also surveyed along the sea-ice edge of the Beaufort Sea during the month of September (∼150–160°W, 72–74°N), finding seawater pCO2 values of ∼160–280 μatm, in the same range as observed on the SBI 2002 field survey. Combined, these observations suggest that the sea-ice covered surface waters of the Arctic Ocean (specifically the Canada Basin) remain perennially undersaturated with respect to CO2 in the atmosphere.
4.3. Air-to-Sea CO2 Fluxes on the Chukchi Sea and Western Beaufort Sea Shelves
 Across the Chukchi Sea and western Beaufort Sea, the ΔpCO2 values in both seasons were negative (i.e., seawater pCO2 conditions undersaturated compared to atmospheric pCO2). Thus, the direction of CO2 gas exchange in the region was uniformly from air to sea. Distinct spatiotemporal distributions of air-to-sea CO2 flux were observed on the Chukchi Sea and western Beaufort Sea shelves.
 In springtime, sea-ice coverage at each CTD/rosette station was typically 100% and an effective barrier to air-sea CO2 gas exchange. If minor air-sea gas exchange is allowed for through leads and fractures in the sea-ice, rates of air to sea CO2 gas exchange in most areas were very low (<0.1 to 1 mmol CO2 m−2 d−1; Figure 5c) despite seawater pCO2 levels being undersaturated (ΔpCO2 values of −50 to −150 μatm). Only in the region of Bering Strait were air-to-sea CO2 fluxes >4 mmol CO2 m−2 d−1 due to sea-ice retreat and early break up of sea-ice.
 By summertime, much of the Chukchi Sea shelf was sea-ice free and seawater pCO2 was highly undersaturated with respect to the atmosphere (ΔpCO2 values of −50 to −150 μatm), particularly in those regions with very low (∼80–150 μatm) seawater pCO2 levels due to intense primary production [e.g., Hill and Cota, 2005; Bates et al., 2005a]. In the Central Channel (east of Herald Shoals), Hanna Valley and Barrow Canyon regions, air-to-sea CO2 fluxes were very high, ranging from ∼30–90 mmol CO2 m−2 d−1(Figure 5d). In contrast, air-to-sea CO2 fluxes remained low (<0.1–2 mmoles CO2 m−2 d−1) in regions of the Chukchi Sea slope, Arctic Ocean basin, and western Beaufort Sea stations where sea-ice coverage remained high (>80%).
5.1. Comparison of Air-Sea CO2 Fluxes for the Chukchi Sea and Other Polar Shelves
 Previous estimates of the net oceanic CO2 sink in the Arctic Ocean and adjacent shelves (excluding the Bering Sea) have relied on indirect mass balance considerations rather than direct air-to-sea CO2 flux considerations. Early estimates of the net oceanic CO2 sink for the Arctic Ocean (and adjacent shelves) ranged from 70 to 120 Tg C yr−1 [Anderson et al., 1990, 1994; Lundberg and Haugen, 1996] (note that the latter paper included the Norwegian Sea). More recently, Anderson et al. [1998b] revised the flux estimate downward to 24 ± 18 Tg C yr−1. In the Barents Sea, the annual estimate of atmospheric CO2 uptake into the ocean has been estimated at ∼9 Tg C year−1 [Fransson et al., 2001; Kaltin et al., 2002], who revised the Arctic Ocean air-to-sea CO2 flux upwards slightly to ∼31 Tg C year−1. All these annual rates of net air-to-sea CO2 flux have considerable uncertainty due to the mass balance approaches used.
 In this study, the mean air-to-sea CO2 flux at stations on the Chukchi Sea shelf was 40 ± 10 mmol CO2 m−2 d−1. If this flux rate is extrapolated to the entire Chukchi Sea (i.e., 595000 km2), net air-sea CO2 flux during the sea-ice free period (∼100 d) was 29 ± 8 Tg C. Since sea-ice provides an effective barrier to gas exchange during the rest of the year, this rate is also equivalent to an annual flux. Alternatively, if air-sea CO2 fluxes are estimated for each 2.5° by 2.5° using monthly averaged seawater pCO2 and daily wind speed data, the net air-to-sea CO2 flux for the Chukchi Sea (the region bounded by 150°W to 170°W, and 65°N to 75°N) for the May to September period was 27 ± 6 Tg C (Table 1 and Figure 6). The highest net air-to-sea CO2 flux occurred during August and September during the sea-ice coverage minima and lowest seawater pCO2 conditions. For example, for the month of September, a net air-to-sea CO2 flux of 8.6 Tg C is calculated here compared to the Pipko et al.  estimate of ∼2 Tg C during September 1996. Since then, late season surface pCO2 conditions appear to have decreased (1998–2000 [Murata and Takizawa, 2003]), with areas of lowest undersaturated waters (seawater pCO2 values of 120–250 μatm) observed during the summertime 2002 SBI field survey.
 The annual flux of CO2 (for 2002) can be calculated for the Chukchi Sea. If wintertime months are included in the estimate of CO2 flux (i.e., January to April; October–December), the annual net air-to-sea CO2 flux for the Chukchi Sea was 38 ± 7 Tg C yr−1 (Table 2). This compares to the mass balance estimate of net air-to-sea CO2 flux of 22 and 31 ± 18 Tg C yr−1 calculated for the Chukchi Sea [Kaltin and Anderson, 2005] and Arctic Ocean basin [Fransson et al., 2001] and adjacent seas [Anderson et al., 1998a]. More recently, mass balance calculations indicate a 9 Tg C yr−1 oceanic sink for atmospheric CO2 in the Barents Sea [Fransson et al., 2001]. Scaling these mass balance estimates to the Arctic Ocean and adjacent seas yields an annual net air-to-sea CO2 flux estimate of ∼31 Tg C yr−1 (Table 2). This mass balance approach yields a small oceanic sink for CO2 in the Chukchi Sea (∼3.8 Tg C yr−1) compared to estimates (i.e., 38 ± 7 Tg C yr−1) reported by direct means here. If rates of air-to-sea CO2 flux for the Chukchi Sea (this study) are combined with mass balance estimates [Anderson et al., 1998a, 1998b; Fransson et al., 2001] for other adjacent Arctic Ocean seas, the annual oceanic sink for CO2 is ∼66 Tg C yr−1(Table 2). Additionally, if minor rates of CO2 gas exchange (e.g., 0.2 mmoles CO2 d−1) occur within the perennially sea-ice covered Arctic Ocean basin, the annual rate increases by 5 Tg C yr−1.
Table 2. Annual Net Air-to-Sea CO2 Fluxes (Tg C) for the Arctic Ocean (Including Adjacent Continental Shelves), Chukchi Sea, and Other Adjacent Continental Shelves (i.e., Barents, Laptev, Kara, East Siberian, and Beaufort Seas)a
Net Air-to-Sea CO2 Fluxes, Tg C or 1012 g C
A, Annual Arctic Ocean and adjacent sea CO2 flux estimates from previous studies are given. B, Annual Arctic Ocean and adjacent sea CO2 flux estimates are scaled to the Arctic Ocean and adjacent seas using the areas for each coastal sea and fluxes from the previous studies of Anderson et al. [1998a, 1998b] and Fransson et al. . C, Annual Arctic Ocean CO2 flux estimates compiled from direct (this study) and mass balance considerations.
Indirect estimate of net air-to-sea CO2 fluxes from mass balance considerations.
Direct estimate of CO2 flux from seawater pCO2 observations with a spatial resolution of 2.5° by 2.5°.
Direct estimate of CO2 flux extrapolated from a mean net air-to-sea CO2 fluxes of 40 ± 10 mmol CO2 m−2 d−1 to the entire Chukchi Sea assuming a sea-ice free period of 120 d. If this flux rate is extrapolated to the entire Chukchi Sea (i.e., 595000 km2), net air-sea CO2 flux during the sea-ice free period (∼120 d) was 29 ± 8 Tg C.
 Knowledge about the spatiotemporal variability of CO2 species and air-sea CO2 fluxes in the Bering Sea (particularly the continental shelf region) are poorly known and contradictory. The high levels of primary production observed in the southeastern Bering Sea result in drawdown of inorganic nutrients and DIC [Codispoti et al., 1982, 1986]. From a very limited pCO2 data set, the extrapolated seawater pCO2 climatology (4° by 4° resolution) compiled by Takahashi et al.  also shows supersaturated pCO2 conditions in February in the open-ocean and continental shelf of the Bering Sea. Subsequently, undersaturated pCO2 conditions (due to biological drawdown) are shown for the summer period and the Bering Sea (open-ocean and coastal ocean) is inferred to be a net sink for atmospheric CO2. There appears to be a seasonal oscillation in the sink and source conditions in the Bering Sea. Chen et al.  suggested that the open ocean of the Bering Sea is a net oceanic sink of CO2. Walsh and Dieterle  calculated a mean invasion of CO2 of 4.3 mol m−2 yr−1. Scaling this value to the Bering Sea yields a net annual oceanic sink of ∼57 Tg C yr−1, similar in magnitude to the Chukchi Sea. A further complication to understanding the source-sink status of the Bering Sea is the recent observations by Murata and Takizawa  of elevated pCO2 conditions (>400 μatm) and total alkalinity drawdown in coccolithophore blooms (E. Huxleyi) of the Bering Sea in 2000. These widespread coccolithophore blooms have only occurred since 1997 [e.g., Overland et al., 2001; Napp and Hunt, 2001] with the implication that the spatiotemporal distributions of CO2 and possibly the CO2 sink-source status of the Bering Sea is changing in concert with the observed ecosystem regime shifts.
5.2. A Continental Shelf Pump for CO2 in the Chukchi Sea
 In upwelling regions, the CO2 sink or source term can be highly variable. For example, vertical transport of remineralized inorganic nutrients (and CO2) to the surface can support high levels of primary production and drawdown of pCO2 significantly below equilibrium with the atmosphere [e.g., Pérez et al., 1999; van Geen et al., 2000; Borges and Frankignoulle, 2001; Friederich et al., 2002]. However, in the Arabian Sea, pCO2 conditions remain above or close to atmospheric values and portions of the coastal and central Arabian Sea are thought to be a perennial source of CO2 to the atmosphere [Körtzinger et al., 1997; Goyet et al., 1998; Sarma, 2003]. In this region, Ducklow and McAllister  have suggested that upwelling of remineralized CO2 is not fully compensated for by primary production (or net community production) thereby driving the system to a CO2 source. The variability in CO2 sink or source conditions of upwelling regions may also relate to differences in the carbon, nitrogen and phosphorus stoichiometry of upwelled waters.
 In the East China Sea [Tsunogai et al., 1999; Wang et al., 2000; Liu et al., 2000] and parts of west European shelves (e.g., northern North Sea [Thomas et al., 2004; Bozec et al., 2005]), seawater pCO2 concentrations remain below atmospheric CO2 values for most of the year. These continental shelves are net sinks for atmospheric CO2 [e.g., Tsunogai et al., 1999; Wang et al., 2000; Borges and Frankignoulle, 1999; Thomas et al., 2004; Bozec et al., 2005]. Tsunogai et al.  suggested that a “continental shelf pump” of carbon from the shelf maintained oceanic CO2 sink conditions. In this scenario, winter cooling on the shelf depressed seawater pCO2 below saturation. During the subsequent summertime, inorganic nutrient supply, continued CO2 drawdown in the summer due to primary production, and the subsequent vertical export of organic matter and lateral export of DIC maintain the CO2 sink status of the shelf throughout the year. Not all continental shelves act in this manner, for example, in the Middle Atlantic Bight, thermally driven increase in pCO2 during summertime appears not to be compensated for by sufficient nutrient input and primary production [DeGrandpre et al., 2002]. This continental shelf region acts as a weak sink for CO2 due to the seasonal oscillation driven by winter cooling (undersaturated pCO2 conditions) and summertime heating (oversaturated pCO2 conditions) rather than primary production and remineralization.
 In the polar seas, Yager et al.  put forward the “seasonal rectification hypothesis” to explain seasonal pCO2 variability in the North East Water (NEW) polynya. There is a brief but large pCO2 drawdown in the NEW polynya in response to local primary production during the ice-free periods of summertime. During the remainder of the year, Yager et al.  suggested that the sea-ice cover prevented gas exchange, thereby allowing wintertime rectification of seawater pCO2 to near equilibrium values through remineralization of OM produced earlier in the season. Air-sea CO2 gas exchange may, however, contribute to rectification since there is the potential for CO2 gas exchange through leads, fractures, and brine channels within sea-ice [Semiletov, 1999].
5.2.2. Carbon Dynamics of the Chukchi Sea
 In the Chukchi Sea, surface waters are undersaturated with respect to CO2 in the atmosphere during sea-ice cover and the brief seasonal sea-ice free period (Figure 5). The dominant processes maintaining the Chukchi Sea shelf as a perennial oceanic CO2 sink relate to factors such as the northward transport of water, seasonal cooling, warming and sea-ice cover, and very high rates of seasonal primary and net community production during the brief, summertime exposure of nutrient-laden surface waters of Pacific Ocean origin.
 The northward transport (∼0.8 Sv) of Pacific Ocean water from the Bering Sea through Bering Strait clearly acts to precondition the Chukchi Sea by providing inorganic nutrients that the pelagic plankton communities utilize during seasonally suitable light and sea-ice free conditions. Water mass transit times from Bering Strait to the northern margins of the Chukchi Sea are approximately 3–6 months [Woodgate et al., 2005a, 2005b], with waters of the Chukchi Sea shelf renewed and replenished from the Bering Sea each year (note that there are minor exchanges with the East Siberian Sea and Canada Basin). Limited knowledge about CO2 dynamics in the Bering Sea makes it difficult to assess how the Bering Sea preconditions the Chukchi Sea. The SBI data from Bering Strait suggests that surface waters transiting into the Chukchi Sea are undersaturated with respect to the atmosphere. The seawater pCO2 climatology maps of Takahashi et al.  also suggest that surface waters in the northern Bering Sea are undersaturated with respect to atmospheric pCO2 in summertime (August), and neutral in the wintertime (February).
 Surface water cooling during the continual northward transport of surface water across the Chukchi Sea from Bering Strait partly contributes to seawater pCO2 undersaturation on the shelf. During the summertime seasonal sea-ice minima (e.g., June to September), relatively warm waters from the Bering Sea (particularly Alaskan Coastal Current waters) can cool by ∼5–7°C as they transit northward across the shelf. Given the thermodynamic seawater pCO2 change of ∼4.2% per °C [Takahashi et al., 1993], seawater pCO2 could potentially decrease by 8–12 μatm per °C (or ∼40–80 μatm) during the transit northward. Previously, the northward shelf to slope decrease in pCO2 observed in September was primarily attributed to the effect of temperature rather than production or other effects [Pipko et al., 2002; Murata and Takizawa, 2003]. Murata and Takizawa  suggested that the net air-to-sea CO2 flux on the Chukchi Sea shelf was driven primarily by an enhancement of the solubility pump (i.e., temperature decrease northward) rather than primary production. At other times (e.g., ∼October to May), during sea-ice advance and wintertime sea-ice cover, the latitudinal gradients in temperature across the Chukchi Sea shelf are small (<2°C), and cooling has less impact on seawater pCO2.
 The major determinant of seawater pCO2 undersaturation and strong air-to-sea CO2 flux on the Chukchi Sea shelf is summertime primary (and net community) production. The strong oceanic sink for atmospheric CO2 on the Chukchi Sea results from the biogeochemical modification of seawater CO2 conditions (rather than temperature) and enhancement of the capacity of surface waters to absorb CO2.
 Observations from the seasonal SBI cruises indicate that temperature changes have a smaller impact on seawater pCO2 compared to the impact of primary or net community production. Seawater pCO2 can be corrected for the thermodynamic effect of temperature (e.g., ∼4.2% per °C [Takahashi et al., 1993]). On the Chukchi Sea shelf, the largest changes occur between the spring and summertime, when temperature corrected seawater pCO2 decreased by ∼100–220 μatm (Figure 7), particularly in those regions with high rates of primary and net community production [e.g., Hill and Cota, 2005; Bates et al., 2005a]. In the regions of high net community production, summertime surface water temperatures were only slightly warmer (<1°C) than wintertime conditions, and the thermodynamic impact of cooling had minor influence on the large seawater pCO2 (and DIC) drawdown.
 In the region of highest primary (and net community) production, the removal of inorganic and organic carbon from surface waters lead to very low seawater pCO2 conditions (<100–150 μatm) (Figure 8). As a consequence, primary or net community production drives the Chukchi Sea shelf to be a very strong oceanic sink for atmospheric CO2 (Figure 8). In the surface layer, the summertime shelf outflow of water from the Chukchi Sea shelf exports warm, primary production-modified surface water with a DIC deficit, a suspended POM load, and low seawater pCO2 (<200 μatm) content that merges with polar mixed layer waters of the Arctic Ocean (Figure 9). Furthermore, much of the summertime primary production on the Chukchi Sea shelf is vertical exported as organic matter from the surface layer to the underlying waters of the Upper Halocline Layer and to the shelf floor (e.g., export production rates were ∼20–40 mmol C m−2d−1 [Moran et al., 2005]). Subsequently, lateral water transport advects suspended POM off the Chukchi Sea shelf into the Arctic Ocean (Figure 9). At Barrow Canyon, high concentrations of suspended POM (∼60 μM) were found in the upper part of the Upper Halocline Layer extending northward from the shelf into the Arctic Ocean basin [Bates et al., 2005b]. Similar features were observed in the outflow of Hanna Valley [Bates et al., 2005b] and the Herald Valley [Weingartner et al., 2006]. The lateral export of Upper Halocline Layer water from the Chukchi Sea shelf also exports DIC to the Arctic Ocean (in contrast to the surface layer).
 Summertime primary (and net community) production also enhances the capacity of Chukchi Sea surface layer water to absorb CO2 and act as a strong oceanic sink for atmospheric CO2. The capacity of seawater to absorb CO2 depends primarily on the buffer capacity or Revelle Factor of seawater [Takahashi et al., 1993; Sabine et al., 2004]. The Revelle Factor quantifies the rate of change of pCO2 to dissolved inorganic carbon (DIC) in water masses and reflects the underlying seawater charge balance and ratios of DIC to total alkalinity. Tropical and subtropical waters typically have low Revelle Factors (e.g., 8–10) [Takahashi et al., 1993; Sabine et al., 2004], with a greater potential capacity to absorb atmospheric CO2 than temperate and sub-polar waters with high Revelle Factors (e.g., 11–14). Here, the Revelle Factor (i.e., LNtpCO2/LNnDIC) was calculated for surface waters of the Chukchi Sea shelf and slope waters. In accordance with other studies, seawater pCO2 was normalized to a constant temperature (in this study 0°C) using the thermodynamic rate of ∼4.2% change in pCO2 per °C change. DIC data were also normalized to a salinity of 33.1 (the core salinity of the upper halocline layer). Previous observations of polar surface waters from other polar and sub-polar seas, suggest that surface waters of the Chukchi Sea should have high Revelle Factors and low capacity to absorb CO2. However, SBI data from 2002 indicate that the surface waters of the Chukchi Sea shelf and slope have unusually low Revelle Factors (∼3.5–6.5; Figure 10) and a high capacity to absorb CO2. This results primarily from a decrease in the DIC to total alkalinity ratios of surface waters of the Chukchi Sea due to the removal of DIC by primary production while total alkalinity remains constant (due to the absence of pelagic calcification in the Chukchi Sea). This in turn, reduces the Revelle Factor and consequently enhances the capacity of Chukchi Sea surface waters to absorb CO2 during summertime.
 Do the carbon dynamics on the Chukchi Sea shelf have the attributes of either the “continental shelf pump” [Tsunogai et al., 1999] or the “seasonal rectification hypothesis” [Yager et al., 1995]. The “seasonal rectification hypothesis” of Yager et al.  requires that all of the CO2 fixed into organic matter during a period of brief primary production is retained on the shelf and subsequently remineralized back to CO2 during the wintertime. The strength of the continental shelf pump as a weak or strong CO2 sinks is thus dependent on whether organic matter is preferentially retained and remineralized to CO2 (weaker sink) or organic matter is preferentially export to the open ocean (stronger CO2 sink) during the sea-ice covered period. On the Chukchi Sea shelf, carbon (either as suspended POM or remineralized CO2) is lost from the shelf to the Arctic Ocean basin during summertime. As such, during the subsequent sea-ice covered period, the depression of seawater pCO2 conditions cannot be fully rectified or recharged by CO2 through organic matter remineralization, and shelf seawater pCO2 conditions remain undersaturated.
 The “continental shelf carbon pump” mechanism hypothesized for the East China Sea [Tsunogai et al., 1999] and other coastal systems requires wintertime cooling to depress seawater pCO2 conditions, summertime production, vertical export of OM, and lateral export of remineralized CO2 as DIC from the shelf. Although there are similarities to this general process, Chukchi Sea carbon dynamics have some unique attributes. In the Chukchi Sea, it is the depression of seawater pCO2 by primary production, facilitated by seasonal conditions of light availability and sea-ice retreat, combined with cooling and uncertain preconditioning of waters in the Bering Sea that drives surface waters to be a strong oceanic sink for atmospheric CO2 during their one-way transit northward across the Chukchi Sea. The removal of DIC from brief and intense primary production occurring during the summertime increases the capacity of surface waters to absorb CO2, and subsequent lateral export of DIC and suspended particulate organic carbon from the shelf to the Arctic Ocean basin maintains the Chukchi Sea as a perennial ocean CO2 sink. Furthermore, during sea-ice cover in the wintertime, water transport into the Chukchi Sea from the Bering Sea, and release of CO2 from the benthic community [Grebmeier and Harvey, 2005] and shallow sediments of the Chukchi Sea shelf (∼50–100 m deep) must partially rectify seawater pCO2 conditions toward atmospheric pCO2 values. Finally, the Chukchi Sea shelf is also strongly net autotrophic on annual time-scales since the output of OM from the shelf to the Arctic Ocean is greater than the OM input through Bering Strait.
 How potentially important is the Chukchi Sea sink for atmospheric CO2 compared to the global coastal ocean. Overall, global marginal coastal seas are thought to be a sink of CO2 of ∼450 Tg C yr−1 [Borges et al., 2005]. This may be compensated for by nearshore coastal systems acting as sources of CO2 of ∼400 Tg C yr−1[Borges et al., 2005]. The Chukchi Sea and Arctic Ocean CO2 sinks of 38 ± 7 and 66 ± 7 Tg C yr−1 estimated here are ∼8% and 15%, respectively, of the global marginal coastal sea sink for CO2 estimated by Borges et al. . In the Arctic Ocean, sea-ice extent and volume has decreased over the last few decades[e.g., Cavalieri et al., 2003; Rothrock and Zhang, 2005; Stroeve et al., 2005], with complex interactions and feedbacks between the atmosphere, oceans and sea-ice that influence heat and freshwater budgets and exchanges within the Arctic Ocean region. The inputs of freshwater and organic carbon inputs to the Arctic Ocean [e.g., Peterson et al., 2002; Hansell et al., 2004; Frey and Smith, 2005] are likely to change in the future with hard to predict changes in the Arctic Ocean carbon cycle. Thus, the relatively large contribution of the Chukchi Sea and Arctic Ocean to the global marginal coastal CO2 sink may change significantly, with implications for the global carbon cycle.
 In springtime (May–June) of 2002, surface water properties on the Chukchi Sea shelf were characteristic of remnant winter water. Surface layer DIC (and nDIC) had a small concentration range and seawater pCO2 contents on the Chukchi Sea shelf and slope regions were undersaturated with respect to atmospheric pCO2. Two months later, in response to high rates of localized primary production, low seawater pCO2 conditions (<150 μatm) were observed in the surface layer in the Central Channel and Barrow Canyon region of the Chukchi Sea shelf and western Beaufort Sea shelf. At some stations, very low pCO2 conditions (<150 μatm) were observed. In the Arctic Ocean basin, spring and summertime surface layer DIC (and nDIC) had a small concentration range and seawater pCO2 contents (∼200–250 μatm) were highly undersaturated with respect to atmospheric pCO2.
 In the seasonally sea-ice free regions of the Chukchi Sea shelf, rates of air-to-sea CO2 fluxes were high, ranging from ∼50–90 mmoles CO2 m−2 d−1. Air-to-sea CO2 fluxes decreased to <5 mmoles CO2 m−2 d−1, in slope, Arctic Ocean basin, and western Beaufort Sea stations where sea-ice cover remained very high (>80%). The Arctic Ocean basin region appears to act as a potential oceanic sink for atmospheric CO2 through this is suppressed by perennial sea-ice conditions. Annually and during the May to September period, the net air-to-sea CO2 flux from the Chukchi Sea shelf was estimated at ∼38 ± 7 and ∼27 ± 7 Tg C, respectively. An active continental shelf pump of carbon, driven by the northward transport of nutrient-rich water of Pacific Ocean origin, high rates of primary and net community production during the sea-ice free period, and lateral export of organic carbon from the shelf maintains the Chukchi Sea shelf and slope as a perennial ocean CO2 sink.
 Jackie Grebmeier, Lou Codispoti, Terry Whitledge and Lee Cooper are thanked for their dedicated contributions to the SBI program. I am indebted to the captain, crew and all scientific participants on the two 2002 USCGC Healy cruises. In particular, my thanks to Charlie Farmer (RSMAS), Cindy Miller (RSMAS), Christine Pequignet (BBSR) and Paul Lethaby (BBSR) for their participation in the field activities of SBI program. Christine Pequignet and Julian Mitchell (BBSR) are thanked for their organization of precruise and postcruise logistics. Margaret H. P. Best and Julian Mitchell (BBSR) are thanked for the analyses of DIC and total alkalinity. Two anonymous reviewers provided very detailed and thorough comments that improved the paper. I thank Wei-Jun Cai, Alberto Borges, Lou Codispoti, Dennis Hansell, Jeremy Mathis and Brad Moran for helpful discussion. NSF grant OPP-0124868 supported this research.