Rapid changes in surface water carbonate chemistry during Antarctic sea ice melt

The effect of sea ice melt on the carbonate chemistry of surface waters in the Weddell–Scotia Confluence, Southern Ocean, was investigated during January 2008. Contrasting concentrations of dissolved inorganic carbon (DIC), total alkalinity (TA) and the fugacity of carbon dioxide (fCO2) were observed in and around the receding sea ice edge. The precipitation of carbonate minerals such as ikaite (CaCO3·6H2O) in sea ice brine has the net effect of decreasing DIC and TA and increasing the fCO2 in the brine. Deficits in DIC up to 12 ± 3 ¼mol kg-1 in the marginal ice zone (MIZ) were consistent with the release of DIC-poor brines to surface waters during sea ice melt. Biological utilization of carbon was the dominant processes and accounted for 41 ± 1 ¼mol kg-1 of the summer DIC deficit. The data suggest that the combined effects of biological carbon uptake and the precipitation of carbonates created substantial undersaturation in fCO2 of 95 ìatm in the MIZ during summer sea ice melt. Further work is required to improve the understanding of ikaite chemistry in Antarctic sea ice and its importance for the sea ice carbon pump.


Introduction
The Southern Ocean greatly influences the climate system through the biological and physical pumps that facilitate the uptake of atmospheric carbon dioxide (CO 2 ) and transport of carbon to the deep ocean (Heinze et al., 1991). Marginal ice zones (MIZ) form the boundary from dense sea ice cover to the open ocean and show large spatial and temporal variability in sea surface CO 2 (Bakker et al., 1997;Gibson and Trull, 1999;Hoppema et al., 1999;Hoppema et al., 2000;Alvarez et al., 2002;Bellerby et al., 2004;Shim et al., 2006). The significance of sea ice regions in the oceanic carbon cycle has been demonstrated by a dramatic revision of the Southern Ocean (50-62 • S) CO 2 sink estimate, from −0.34 to −0.06 Pg C yr −1 in the reference years 1995 and 2000, respectively (Takahashi et al., 2009). This trend has been governed by increased data coverage that highlights the dominance of winter CO 2 release, as a result of upwelling and respiration, over CO 2 uptake upon seasonal sea ice melt.
The eastward-flowing Antarctic Circumpolar Current (ACC) transports heat and dissolved constituents around the Antarctic continent (Rintoul and Sokolov, 2001). The dominant water mass in the ACC is Circumpolar Deep Water (CDW), a broad layer of warm, saline and nutrient-rich water, which shoals to the south with the uplifting isopycnals at the Southern Boundary (SB; Sievers and Nowlin, 1984;Orsi et al., 1995;Pollard et al., 2002).
The Weddell Sea is dominated by a cyclonic gyre, bounded to the west and south by the Antarctic continent but, open to the north and east for interaction with waters of the ACC. CDW is advected into the eastern side of the gyre and can be identified by maxima in temperature, salinity and nutrients within the gyre (Deacon, 1979;Whitworth and Nowlin, 1987;Gouretski and Danilov, 1993;Schroder and Fahrbach, 1999). A substantial part of the ocean is ventilated in the Weddell Gyre as a result of the formation of deep and bottom waters due to strong atmosphere-ice-ocean interactions (Gill, 1973;Carmack and Foster, 1975). Early studies largely regarded this region as a source of CO 2 due to the upwelling of the CO 2 -rich deep water (Deacon, 1979;Weiss et al., 1979;Takahashi et al., 1993). Since then, several investigations into the carbonate chemistry of the Weddell Gyre have shown this area to be a net sink for atmospheric CO 2 (Hoppema et al., 1995;Hoppema et al., 1999;Stoll et al., 1999;Hoppema et al., 2000;Stoll et al., 2002;Bakker et al., 2008).
In the western Atlantic sector of the Southern Ocean, the SB lies just north of the Weddell-Scotia Confluence, which extends from the Antarctic Peninsula to about 20 • W (Patterson and Sievers, 1980;Whitworth et al., 1994), separating the Weddell Sea to the south from the ACC in the Scotia Sea to the north (Orsi et al., 1995). The Weddell-Scotia Confluence provides a transition zone where the recently ventilated waters flow out of the Weddell Sea and into the Scotia Sea (Patterson and Sievers, 1980;Orsi et al., 1999;Naveira Garabato et al., 2002). Therefore, it is important to improve the understanding of the ocean carbonate chemistry of the Weddell-Scotia Confluence and its contribution to the Southern Ocean carbon cycle.
Field and laboratory studies have shown that ikaite, a hydrous calcium carbonate mineral (CaCO 3 ·6H 2 O), precipitates out within brines during sea ice formation (Marion, 2001;Papadimitriou et al., 2004;Delille et al., 2007;Papadimitriou et al., 2007;Rysgaard et al., 2007;Dieckmann et al., 2008): The net effect of ikaite precipitation in brine is to reduce the concentration of dissolved inorganic carbon (DIC) and total alkalinity (TA), whilst increasing the fugacity of CO 2 (f CO 2 , eq. 1). Any brine rejected during winter or released during ice melt transfers these inorganic carbon characteristics to the underlying water. One implication of these processes is that during sea ice growth, the winter mixed layer becomes isolated from the atmosphere (Klatt et al., 2002), allowing levels of f CO 2 to reach supersaturation beneath the sea ice. During sea ice melt in spring and summer, any remaining carbonate minerals within the ice are thought to be released into the water column where they re-enter the carbon cycle through dissolution. According to eq. (1), this would increase TA and DIC and reduce f CO 2 of the water.
The activity of ice algae also leads to changes in the carbonate system of the brine and the water at the ice-sea interface (Gleitz et al., 1995;Gleitz et al., 1996;Gibson and Trull, 1999;Krembs and Engel, 2001;Meiners et al., 2009). During summer, increased light levels promote photosynthesis, reducing DIC and the f CO 2 of the brine. As the ice melts, these carbonate characteristics are transferred to the surrounding water, becoming superimposed on effects from carbonate mineral chemistry.
The sea ice carbonate chemistry processes have been described as a 'sea ice CO 2 pump', assuming dissolution of ikaite in the summer promotes atmospheric CO 2 uptake and that DICand TA-poor, but CO 2 -rich brines are transported out of the surface waters during the winter (Rysgaard et al., 2007). The MIZs of the Southern Ocean are of great interest due to the effects of enhanced biological activity and, more recently, calcium carbonate chemistry on the oceanic carbon cycle. However, few data exist to investigate and quantify the carbonate chemistry processes in melting polar sea ice and to infer the implications of the sea ice carbon pump for the global carbon cycle. In this paper, we present direct measurements of f CO 2 , DIC and TA from waters at and around the receding ice edge. The distribution of upper ocean carbonate chemistry in the Weddell-Scotia Confluence during austral summer is evaluated within the context of physical and biological controls, including (1) the influence of sea ice cover; (2) the effect of recent sea ice melt and an ice edge bloom and (3) the influence of the SB.

Underway and station sampling
Data were collected in the Weddell-Scotia Confluence during cruise JR177 on the RRS James Clark Ross in January 2008 (Fig. 1). The ship reached the South Orkney Islands on 8 January 2008 and made two transects in the Weddell-Scotia Confluence, southward and northward, passing the ice edge overlying the South Orkney shelf at station 5 (Table 1, Fig. 2). The southward transect ended in the sea ice at station 1 (62.61 • S 43.24 • W), over the break of the South Orkney shelf on 11 January. Between stations 1 and 6 the ship sailed through waters with partial sea ice cover, melting sea ice and areas of open water. The ship left the MIZ on 13 January and continued northward to the SB at station 13 (59.14 • S 43.69 • W) on 17 January.
Continuous measurements of temperature and salinity in surface water were made using seawater from the non-toxic underway supply (bow intake 6.5 m below the surface). High resolution, vertical profiles of potential temperature and salinity were obtained during the downcast using a conductivity, temperature and depth (CTD) sensor (Seabird SBE9+). An offset in sea surface temperature (0.37 • C) was detected between the ship's oceanlogger temperature sensor and the CTD surface temperature, which has been applied to the underway temperature dataset. All salinity values are reported on the practical salinity scale.
The summer mixed layer depth (MLD) is defined here as the depth where the potential density exceeds that measured at 10 m by 0.05 kg m −3 (Brainerd and Gregg, 1995). This definition   was selected to provide a depth that is located between the active mixing layer and the seasonal mixed layer. MLDs were calculated from 2 dbar profiles of the potential temperature and salinity from the CTD deployment at each station. The winter mixed layer, marking the remnant of the Winter Water (WW), is defined here as the depth of the potential temperature minimum (θ min ) (Brainerd and Gregg, 1995). For all stations, the WW layer was identified by a well-defined θ min at depths in the range of 60-100 m.

Fugacity of CO 2 in surface seawater and the atmosphere
Quasi-continuous measurements of the f CO 2 in surface water were made using an underway f CO 2 system and seawater from the underway supply (Schuster and Watson, 2007). Atmospheric samples were taken from an air inlet located forward at 15 m height on the ship's bridge. Mixing ratios of CO 2 and moisture in the equilibrator headspace and the outside air were determined by circulation through a non-dispersive infrared analyser (Li-COR 7000). The Li-COR was calibrated using three secondary gas standards of known CO 2 concentrations of 249.1 ± 0.2, 356.5 ± 0.2 and 457.4 ± 0.2 μmol mol −1 in an air mixture (21% O 2 , 79% N 2 ). All gases underwent pre-and post-cruise calibration against certified, high precision, primary standards from the National Oceanic and Atmospheric Administration (NOAA). Samples from the equilibrator headspace and marine air were partially dried by being passing through an electric cool box at about 2 • C, prior to analysis in the Li-COR. The f CO 2 for equilibrated water samples and the air was computed from the partly dried mixing ratios and ambient barometric pressure and then corrected for seawater vapour pressure (assuming saturation at the sea surface). Two platinum resistant thermometers positioned in the upper and lower part of the seawater stream determined the temperature of the water in the equilibrator. Warming of the seawater between the intake and the equilibrator was on average 0.8 • C (σ = 0.1 • C; n = 250). Sea surface f CO 2 data were corrected to sea surface temperature to account for this warming (Takahashi et al., 1993). The difference in f CO 2 between the seawater and overlying air [ f CO 2 (wa)] was calculated continuously during the cruise. The time lag between sampling and the seawater reaching the equilibrator was 116 seconds, which has been corrected for. Sea surface f CO 2 from the underway supply was co-located to the CTD casts by interpolation to the time of the 5 m sample from the upcast of the CTD. Gaps in the data are due to the seawater supply being turned off in thick ice cover or contamination from the ship exhaust gases. The accuracy of the marine air f CO 2 data was determined as 0.6 μatm from comparisons with atmospheric data from Jubany Station, South Shetland Islands (62.23 • S 58.67 • W) (Ciattaglia et al., 1999). The precision of the f CO 2 data is 3.0 μatm.

Carbonate system
Underway samples for sea surface DIC and TA analysis were collected from the underway seawater supply. Vertical profile samples were taken from the 24 × 10 L Niskin bottles mounted onto the CTD rosette at 5, 10, 30, 50, 80, 120, 160 and 200 m depth on the upcast of the CTD. Seawater (250 or 500 ml) was collected in borosilicate glass bottles and a small volume was extracted to create a head space. Saturated aqueous mercuric chloride was immediately added (0.02% vol/vol) and the bottle was sealed, shaken and stored in a dark location at ambient temperature for shipment to the UK.
The DIC concentration was determined by coulometric analysis (Johnson et al., 1987) and TA analyses were carried out by potentiometric titration with hydrochloric acid (Dickson, 1981) using a VINDTA instrument from April to May 2008. Duplicate analyses were made for each 500 ml bottle. Two bottles of certified reference material (CRM) from batches 76 or 81 (DOE, 2007) were analysed in duplicate per CTD cast and per 10 samples from the underway supply. DIC and TA values were corrected to account for the dilution of the seawater by the addition of mercuric chloride solution. The accuracy for the DIC and alkalinity measurements is determined as 2 μmol kg −1 (n = 96) from the average difference between certified and calculated values for each CRM analysis. The precision for the DIC and alkalinity measurements is estimated as better than 2 μmol kg −1 (n = 96) based on the average difference between duplicate CRM bottle analyses. DIC and TA from all sea surface samples were used to calculate f CO 2 , f (DIC,TA), using the CO2Sys programme (Lewis and Wallace, 1998) with thermodynamic dissociation constants for K 1 and K 2 by Mehrbach et al. (1973) and the re-fit by Dickson and Millero (1987).
All surface alkalinity samples showed strong salinity dependence with a positive intercept of 269 μmol kg −1 (TA 0 ) at zero salinity. To correct for dilution or concentration effects, TA data were normalized to salinity 34.3, the average salinity of the WW for all stations in the Weddell-Scotia Confluence, using eq. (2) to account for the non-zero end-member (Friis et al., 2003) This was repeated for the DIC data, where a value of 966 μmol kg −1 was determined for DIC 0 . A series of de-ionized water samples were analysed to determine a 'blank' for DIC and TA at zero salinity. Average values for TA 0 de-ionzed and DIC 0 de-ionzed of 74 ± 11 and 36 ± 3 μmol kg −1 (n = 6), respectively, indicate that the TA 0 and DIC 0 determined for the Weddell-Scotia Confluence samples do not result from the analytical technique but from natural sources, which could include dissolution of carbonate minerals during the analysis and bacterial degradation. DIC and TA measurements reported in the text are not salinity normalized unless stated.

Macronutrients
Underway samples from the underway seawater supply and CTD water bottle samples were filtered through a mixed ester membrane (Whatman, pore size 0.45 μm), and the filtrate was analysed colorimetrically for dissolved silicate (SiO 4 4− ), phosphate (PO 4 3− ) and nitrate (NO 3 − ) with a segmented-flow analyser (Technicon, Whitehouse, 1997). Analyses for nitrate included nitrite (NO 2 ), which is not considered separately as its concentration had little variation and was typically <1% of total NO 3 + NO 2 . Nutrient data were normalized to salinity 34.3 using a direct normalization procedure by multiplication with 34.3 and division by the in situ salinity. Nutrient concentrations reported in the text are not salinity normalized unless stated.

Chlorophyll-a
Underway samples for sea surface chlorophyll-a were collected from the non-toxic seawater supply approximately every hour as the ship was in transit. Water samples were filtered through glass fibre filters (Whatman GF/F) under low (<70 mm Hg) vacuum pressure and immediately frozen and stored at −20 • C until analysis onboard. The samples on the filters were then extracted in acetone (10 ml, 90%) in the dark for 24 h (Parsons et al., 1984). Fluorescence of the extract was measured before and after acidification with 1.2 M HCl on a TD-700 Turner fluorometer. The instrument was calibrated against commercially prepared chlorophyll-a standards (Sigma).

Sea ice
Visual observations of sea ice coverage were made from the vessel's bridge as part of the daily position report filed by the Captain. The observations included primary ice types present, the sea ice coverage and meteorological conditions. Daily sea ice concentration data at 4 km resolution were obtained through the Operational SST and Sea Ice Analysis (OSTIA) service, a component of the Group for High-Resolution Sea Surface Temperature (GHRSST) at http://ghrsstpp.metoffice.com/pages/latest_analysis/ostia.html (Stark et al., 2007). The transition area between open ice-free waters to icecovered waters is referred to as the MIZ, encompassing stations 1, 3, 5 and 6 during the present study. The ice edge is defined here as the position where consolidated sea ice was first encountered (station 5) on the southern transect.

Seasonal biogeochemical deficits
The seasonal depletion in DIC, alkalinity and nutrients was determined for each station from the difference between the average concentration in the summer mixed layer and the concentration at the depth of the θ min , in the WW (Jennings et al., 1984). The biogeochemical properties of the WW have previously been used as proxies for conditions present in Southern Ocean surface waters during the preceding winter (e.g. Minas and Minas, 1992;Ishii et al., 1998;Rubin et al., 1998;Pondaven et al., 2000;Ishii et al., 2002). The validity of this method for the Weddell-Scotia Confluence data is discussed below. For the sea ice stations (1-6), values for the θ min were between −1.60 and −1.70 • C (Table 2). Satellite derived sea ice cover indicated that the break-up of the sea ice (last day when ice cover was ≥90%) occurred in early December, 4 weeks prior to sampling. Therefore, top-down warming was very recent and any erosion of the winter mixed layer was attributed to vertical mixing of warmer waters below the thermocline. To estimate any bottom-up influences, the value of the θ min was compared to the temperature at the base of the thermocline, assuming an end member winter mixed layer temperature of −1.70 • C and that the specific heat capacity of seawater was constant. For the range of θ min values, mixing from deeper waters from winter to summer is estimated at 4-7%. This could lead to a slight over estimation in winter DIC and TA as highlighted during a comparable seasonal Table 2. Winter mixed layer depth (WMLD, m), temperature ( • C), salinity, silicate (SiO 4 , μmol kg −1 ), phosphate (PO 4 , μmol kg −1 ), nitrate (NO 3 , μmol kg −1 ), dissolved inorganic carbon (DIC, μmol kg −1 ) and total alkalinity (TA, μmol kg −1 ) of the winter mixed layer (representing the WW) for all stations in the Weddell-Scotia Confluence from 62.61 • S (station 1) to 59.14 • S (station 13) Notes: The absolute value for each station is shown measured from 1 bottle at the depth of the potential temperature minimum (θ min ). Stations 6 and 13 have interpolated nutrient concentrations from proximal stations 8 and 11, respectively. study in the Indian sector of the Southern Ocean (Jouandet et al., 2008). The total DIC deficit ( DIC deficit ) can be expressed as a sum of contributing processes, represented by the following: Deficits in DIC due to salinity changes are determined from the difference between the measured and salinity normalized DIC deficit. A composite error of 2 μmol kg −1 is associated with the DIC salinity term. Modifications in DIC due to abundance of organic matter through photosynthesis and respiration ( DIC Corg ) have been estimated from normalized nitrate deficits (Sweeney et al., 2000), assuming a carbon to nitrogen uptake ratio of 117:16 moles (Anderson and Sarmiento, 1994). Comparisons to phosphate utilization by Redfield stoichiometry (Redfield et al., 1963) indicate uncertainties in DIC Corg of 1-9 μmol kg −1 . The deficit in alkalinity, corrected for salinity and organic matter effects, indicates carbonate mineral (ikaite) precipitation or dissolution based on 2:1 changes in TA and DIC (Zeebe and Wolf-Gladrow, 2001). A composite error of 3 μmol kg −1 accompanies the carbonate term. The residual term accounts for the remaining seasonal deficit as a result of processes including CO 2 air-sea exchange, surface advection and slight vertical mixing (see above). The associated error is the sum of that for each term in eq. (3).

Hydrography
Surface waters (0-20 dbar) showed a strong northward gradient in potential temperature from about −1.4 • C beneath the sea ice to greater than 1.0 • C at about 59 • S (Fig. 3b). Surface salinity showed a similar trend, increasing from values close to 33.5 beneath the sea ice to values above 33.8 at 59 • S (Fig. 3a). A notable exception to this trend was the distinct reduction in salinity in the region of substantial sea ice melt to the south of the South Orkney Islands, at and around station 6 (Table 3, Fig. 2b).
A distinct WW layer could be observed for all stations, with an average potential temperature minimum (θ min ) of −1.55 ± 0.12 • C and a winter MLD of 79 ± 15 m (n = 9, Table 2). Potential temperature of the WW predominantly followed the increasing northward trend of summer mixed layer potential temperature. In comparison, WW salinity was much more variable, from the most saline WW beneath the sea ice decreasing to the freshest water in the region of substantial ice melt. North of the MIZ, WW salinity increased from values close to 34.2 over the South Orkney shelf up to about 34.3 at 59 • S.

Chlorophyll-a and phytoplankton species distribution
Chlorophyll-a concentrations ([chl-a]) varied between 1.85 and 2.84 mg m −3 (Fig. 3f) in the open waters of the Weddell-Scotia Confluence, south of 59 • S to about 60 • S (stations 10-13). A summer maximum in chlorophyll-a of 3.51 mg m −3 was observed in the melt waters at and around station 6, signifying the presence of a phytoplankton bloom (defined here as [chla] ≥ 2.0 mg m −3 ). Chlorophyll-a concentrations sharply decreased to a minimum of 0.21 mg m −3 beneath the sea ice at station 1. For the MIZ (stations 1-6), the phytoplankton community was dominated by naked, heterotrophic dinoflagellates with diatoms accounting for between 21 and 46% of the total cell abundance (Korb et al., 2010). North of the South Orkney Islands to 59 • S (stations 8-13) the community consisted of a mixture of dinoflagellates and cryptophytes, with less than 20% diatoms (Korb et al., 2010).

Carbonate chemistry
Contrasting concentrations of f CO 2 , DIC and TA were observed in and around the sea ice edge and within the frontal waters of the SB (Figs 3c-e). The f CO 2 varied from supersaturation with f CO 2 of 14 μatm in waters beneath the sea ice to strong undersaturation with f CO 2 of −95 μatm in the region of substantial ice melt (Fig. 2). Calculated f CO 2 [f (DIC,TA)] generally showed a close agreement to measured f CO 2 , apart from a positive offset due to higher calculated f CO 2 values from the ice edge to the meltwater region (Fig. 3c). The concentration of sea surface DIC showed similar patterns and predominantly followed the trend of the f CO 2 observations (Fig. 3d). The highest concentration of DIC, TA and the highest value of f CO 2 were measured beneath the sea ice close to station 1 (Table 4). Relative to all other MIZ stations, DIC, TA and f CO 2 reached distinct summer minima in the region of extensive ice melt (Fig. 3). Sea surface normalized DIC (DIC 34.3 ) predominantly followed the trend in DIC concentrations, decreasing from values greater than 2190 μmol kg −1 beneath the sea ice to values around 2180 μmol kg −1 at the SB (Table 4). An exception to this gradual decreasing trend was at station 6, where a substantial reduction in DIC 34.3 of up to 40 μmol kg −1 was observed, relative to the average DIC 34.3 in the summer mixed layer at all other stations. Sea surface normalized TA (TA 34.3 ) showed a similar variation, with the highest values of about 2332 μmol kg −1 in the area of greater sea ice coverage and the lowest values in the region of extensive sea ice melt (stations 5 and 6).
Across the Weddell-Scotia Confluence, WW DIC and DIC 34.3 were relatively homogenous with average concentrations of about 2220 μmol kg −1 (Table 2). Alkalinity showed more variation, with distinct low values at MIZ stations in and near the area of greatest sea ice melt. There was an overall northward increase in TA 34.3 of the WW.

Sea surface carbonate chemistry during summer ice melt
The processes affecting carbonate chemistry in waters at and around the ice edge will be investigated by comparing surface  water f CO 2 , DIC and TA measured at the contrasting sites in the MIZ. To the south of the South Orkney Islands (stations 1-3), the sea surface was characterized by large plates of consolidated sea ice as well as regions where the ice pack had begun to break up. Macronutrient concentrations were the highest (Table 3) and chlorophyll-a reached the lowest concentration of 0.21 mg m −3 (Table 4) for whole Weddell-Scotia Confluence. Surface temperatures in excess of the freezing point of seawater and reduced salinity, compared to station 5 at the ice edge, indicate that slight sea ice melt had occurred (Fig. 4). Effects of sea ice melt were most pronounced at station 6 ( Fig. 2b), where the site was largely open water with sporadic ice floes. Substantial sea ice melt had stabilized the upper water column by forming a warm and shallow meltwater lens at the surface (Fig. 4), which supported elevated chlorophyll-a (Fig. 3f). This corresponded to a rapid reduction in sea surface DIC by 53 μmol kg −1 and f CO 2 by 55 μatm, compared to all other stations, driving a strong CO 2 undersaturation of 74 μatm at station 6 ( Fig. 2a). Melting sea ice forms shallow mixed layers with favourable light conditions that promote the development of phytoplankton blooms in the meltwaters (Smith and Nelson, 1986;Lancelot et al., 1993). Blooms associated with the receding ice edge have been previously observed in the MIZ of the Weddell-Scotia Confluence region (Buma et al., 1992;Lancelot et al., 1993;Kang et al., 2001;Korb et al., 2005). The bloom in the meltwaters at station 6 could be partially sustained by natural iron enrichment from melting sea ice (Löscher et al., 1997;Lannuzel et al., 2007;Lannuzel et al., 2008), vertical mixing over shallow topography (Table 1), upwelling in the ACC (Hoppema et al., 2003;Blain et al., 2007) or advection from the Antarctic Peninsula region (Nolting et al., 1991;Sanudo-Wilhelmy et al., 2002;Dulaiova et al., 2009). Recent evidence strongly suggests that iron is advected into the Scotia Sea from the South Shetland Islands and the Antarctic Peninsula, supporting phytoplankton growth in this downstream region (Ardelan et al., 2010). It has been inferred that it is the interaction between iron and light that limits the initiation and maintenance of blooms in the MIZ (Korb et al., 2005, and references cited therein).
In addition, diatoms in sea ice can act as an inoculum and seed the formation of phytoplankton blooms when released upon sea ice melt (Smith and Nelson, 1985;Ackley and Sullivan, 1994). Diatom activity reduces the concentrations of macronutrients and DIC in the sea ice brine. These chemical signatures are transferred to the surrounding water as the sea ice melts. It is likely that a combination of these processes stimulated the formation and growth of the bloom in the meltwaters of the Weddell-Scotia Confluence and contributed to the reductions in DIC and macronutrient concentrations within the shallow mixed layer (Figs 3 and 4).
Changes in concentrations of macronutrients were compared to DIC 34.3 to investigate whether biological utilization of nutrients had occurred (Figs 5a-c). Nitrate and phosphate showed similar decreasing concentrations with decreasing DIC 34.3 , which suggests nutrient depletion as a result of phytoplankton growth. The ratio of the change in DIC 34.3 with respect to nitrate (7) is very similar to that determined from phytoplankton decomposition analyses, C:N:P of 106:16:1 (Redfield et al., 1963), and to that calculated by examination of spatial changes in nutrient concentrations, C:N:P of 117:16:1 (Anderson and Sarmiento, 1994). The gradient of the phosphate regression line (48) is much smaller than that predicted by either nutrient utilization ratio. The decreasing concentration of silicate indicates the activity of the diatoms observed in this region (Korb et al., 2010). Despite the bloom being dominated by naked dinoflagellates, with little photosynthetic ability, it is predicted that the few diatoms present would account for the majority of the photosynthetic activity and carbon utilization. Based on the DIC 34.3 , TA 34.3 and nutrient deficits, photosynthesis had reduced nutrients and DIC by approximate Redfield ratios. Deviations from predicted nutrient depletion ratios may be due other processes, Fig. 4. Vertical profiles of upper water column (a, d, g) potential temperature (θ , solid line, • C) and salinity (dashed line), (b, e, h) dissolved inorganic carbon (DIC, black solid line, μmol kg −1 ), total alkalinity (TA, black dashed line, μmol kg −1 ), salinity normalized DIC (grey solid line, μmol kg −1 ), salinity normalized TA (grey dashed line, μmol kg −1 ) and (c, f, i) nitrate (NO 3 , solid line, μmol kg −1 ), phosphate (PO 4 , dashed line, μmol kg −1 ) and silicate (SiO 4 , dot-dashed line, μmol kg −1 ), in meltwaters (station 6), at the ice edge (station 5) and beneath the sea ice (station 1).
including sea ice carbonate chemistry, which will now be investigated.
A relationship between DIC 34.3 and TA 34.3 of 1:2 is indicative of ikaite (CaCO 3 ·6H 2 O) precipitation, following eq. (1). Changes in DIC 34.3 and TA 34.3 for the MIZ stations (1-6) group together in a decreasing trend close to 1:2, with the exception of station 6 (Fig. 5d). During sea ice formation and growth, precipitation of ikaite creates DIC-and TA-poor and f CO 2 -rich brines. Any brine remaining in channels and pockets within the sea ice matrix is released into the underlying water during sea ice melt, which would have the net effect of reducing TA 34.3 and DIC 34.3 (2:1) at the surface. Using salinity as a proxy for ice melt (Fig. 2b), in conjunction with in situ and satellite observations, sea ice melt was evident to some degree at all stations in the MIZ during summer 2008. We put forward the hypothesis that ikaite precipitation had occurred in the sea ice brines of the Weddell-Scotia Confluence during the preceding winter and the resultant DIC-and TA-poor brines were released into the underlying water as the ice melted, transferring the inorganic carbon characteristics into the summer mixed layer. The absence of calcifying phytoplankton in this region (Korb et al., 2010) supports the proposed mechanism of the observed depletion in DIC and TA at the sea surface.
An outlier to the principle trend is station 6 in the meltwaters supporting a phytoplankton bloom. Although not numerically dominant within the bloom, diatoms had rapidly reduced DIC 34.3 by 35 μmol kg −1 at the sea surface, compared to ice edge station 5. Based on the DIC 34.3 and nutrient distributions for the MIZ stations (Fig. 5), biological carbon uptake had further reduced the DIC signature from the released brine, driving a strong undersaturation in f CO 2 at the sea surface.
A notable feature in sea surface f CO 2 was the higher calculated f CO 2 values along the strong f CO 2 gradient that existed from the ice edge to the meltwater region (Fig. 3c). This suggests that samples from this location had either enriched DIC or depleted TA, or both, upon analysis. DIC increases occur from dissolution of carbonates, CO 2 uptake and remineralization of organic matter (Fig. 5d). Alkalinity deficits result from carbonate mineral precipitation and remineralization of organic matter. With respect to the high DIC 0 (Section 2.3), it is proposed that there is an additional source of DIC in the sea ice, such as bacterial degradation or the presence of labile organic matter. In the region of very recent ice melt, prior to biological assimilation, the 'extra' DIC would contribute to the observed f CO 2 offset.

Seasonal changes in carbonate chemistry
The depth of the winter mixed layer and temperature of the WW increased northwards from waters beneath sea ice to waters proximal to the ACC at 59 • S (Table 2). Away from the region of greatest sea ice melt, WW salinity and concentrations of macronutrients, DIC and TA had little variation. North of the MIZ, the phytoplankton community was made up of cryptophytes and naked dinoflagellates (Korb et al., 2010), with the cryptophytes accounting for the elevated concentrations of chlorophyll-a north of the South Orkney Islands (Fig. 3f). This corresponded to the undersaturation in sea surface f CO 2 observed approaching the SB (north of station 10) and supports the idea of the frontal waters as an area of high productivity. In contrast, WW values of DIC and DIC 34.3 are relatively constant for nearly all stations and hence processes that affect DIC concentrations, either at the surface during the winter or through advection and slight vertical mixing in the winter mixed layer during the summer (Section 2.7), act uniformly across the Weddell-Scotia Confluence.
The seasonal change in carbonate chemistry can be explored by comparing DIC 34.3 and TA 34.3 at depth intervals from the surface to below the winter mixed layer (Fig. 6). Trends in DIC 34.3 and TA 34.3 in the summer mixed layer for stations located in the MIZ were explained previously (Section 4.1). Stations north of the MIZ, showed comparatively reduced DIC in the summer mixed layer at a narrow range of TA 34.3 of less than 10 μmol kg −1 . With the exception of station 11, surface TA 34.3 and DIC 34.3 for stations north of the MIZ tended to follow a 2:1 trend, but from a different starting point compared to the MIZ stations (Fig. 6a). As WW DIC 34.3 and TA 34.3 for all stations are quite similar (Fig. 6c), the greater DIC 34.3 deficit observed north of the MIZ is attributed to the greater photosynthetic activity that had occurred in the northern Weddell-Scotia Confluence, compared to the MIZ, by the summer.
Below the summer mixed layer, there was a gradual increase in DIC 34.3 and TA 34.3 throughout the water column for all stations (Fig. 6b). Values of TA 34.3 remained relatively constant compared to DIC 34.3 at all depths. This pattern indicates the influence of biological carbon uptake on summer Similarly to surface biogeochemical distributions, the principal exception to the WW trends was station 6, where salinity and the concentrations of macronutrients and DIC were the lowest observed for the whole Weddell-Scotia Confluence (Table 2). Vertical profiles at station 6 show a deflection below the winter mixed layer, at about 130 m, of relatively cold, fresh water (Fig. 4a). This corresponds to reduced DIC and macronutrient concentrations compared to station 5 (Figs 4d-f). This could be a result of the deep penetration of brines where biological activity has amplified the reduction in DIC. As station 6 is located over shallow topography, the interaction of water masses and bathymetry (Section 4.1) could enhance vertical mixing and potentially erode the winter mixed layer from below (Section 2.7). These results highlight the potential for overestimation of hydrographic and biogeochemical parameters in the WW, however this effect is estimated to be small (Section 2.7).
To further investigate the processes that affect the sea surface carbonate chemistry across the Weddell-Scotia Confluence during the summer, the seasonal depletion of inorganic carbon was determined for all stations, as detailed in Section 2.7 (Table 5).
Changes in salinity as a result of freshwater input accounted for between 14 and 32 ± 2 μmol kg −1 of the seasonal change for all stations (Table 5), highlighting the influence of sea ice melt on the carbonate chemistry of the Southern Ocean. The seasonal depletion in DIC due to the photosynthetic production of organic carbon showed a general northward increase, with prominent utilization of DIC in and around the receding ice edge and at the SB. This is in agreement with previous accounts of enhanced biological activity at these locations (Comiso et al., 1993;Helbling et al., 1993;Holm-Hansen and Hewes, 2004;Hewes et al., 2008). Surface waters for the whole region showed TA 34.3 and DIC 34.3 deficits characteristic of ikaite precipitation (Figs 5d and 6). For most stations this was strongly correlated to salinity, which suggests that continual sea ice melt and release of DIC-and TA-poor brines into the surface acts to rapidly change the carbonate chemistry of the Weddell-Scotia Confluence during summer sea ice melt. This was exemplified where the greatest extent of sea ice melt had occurred and an ice edge bloom had developed. Upper ocean distributions of DIC and TA showed a general decrease from waters beneath sea ice to the region of substantial sea ice melt and enhanced biological activity (Fig. 3). This culminated with the largest DIC deficits due to photosynthetic production of organic carbon and carbonate processes of 41 ± 1 and 12 ± 3 μmol kg −1 , respectively, and was accompanied by a strong drawdown of CO 2 (Fig. 2a). Following the same hypothesis, carbonate minerals in melting Nordic sea ice have been proposed to enhance the uptake of CO 2 during summer sea ice melt (Rysgaard et al., 2009).

Comparison to a surface water CO 2 climatology
The CO 2 sink of −0.06 Pg C yr −1 , between 62 and 52 • S in the reference year 2000, is a result of an average f CO 2 of −5 μatm, ranging from −15 μatm in January to 12 μatm in September (Takahashi et al., 2009). Data from the Weddell-Scotia Confluence (62.6-59.1 • S) show that the region was dominated by CO 2 undersaturation at the sea surface, with an average f CO 2 of −25 μatm (range from −95 to 14 μatm, Fig. 2a) during summer 2008. The potential CO 2 source of 14 μatm observed beneath the sea ice cover near station 1 is very similar to the f CO 2 maximum from the climatology (12 μatm). The strong undersaturation of f CO 2 in the region of substantial sea ice melt far exceeded that of the climatology, by up to 80 μatm. In reference to the climatological f CO 2 maximum of about 12 μatm, it is hypothesized that the Weddell-Scotia Confluence is a region of CO 2 uptake of greater than 0.06 Pg C yr −1 . However, investigations into the function of Antarctic sea ice as a permeable 'barrier' to CO 2 exchange may require these estimates to be re-evaluated (Nomura et al., 2006).

Conclusion
The precipitation of carbonate minerals such as ikaite in sea ice has the net effect of decreasing DIC and alkalinity and increasing the f CO 2 in the sea ice brine. Deficits in salinity normalized DIC and alkalinity observed in the summer mixed layer of the Weddell-Scotia Confluence are consistent with the release of brines from melting sea ice, where ikaite precipitation has taken place. Across the MIZ, ikaite precipitation accounted for up 12 ± 3 μmol kg −1 , or 13%, of the summer DIC deficit and was strongly correlated to the amount of sea ice melt that had taken place. Photosynthetic production of organic carbon had the largest effect on summer DIC concentrations, resulting in deficits of 41 ± 1 μmol kg −1 at the receding ice edge and 66 ± 9 μmol kg −1 at the SB. The effects of biological carbon uptake became superimposed on those of the sea ice carbonate chemistry, rapidly reducing f CO 2 compared to waters beneath the sea ice. These processes created substantial undersaturation in f CO 2 of 95 μatm in the MIZ during the summer thaw. The 'sea ice CO 2 pump' hypothesis suggests that the dissolution of calcium carbonate minerals in meltwater during the summer promotes CO 2 uptake from the atmosphere. The carbonate chemistry data for the MIZ of the Weddell-Scotia Confluence suggest that it is the transfer of DIC-and TA-poor brines into the surface water during ice melt, in the presence of phytoplankton blooms, that drive a strong sink for CO 2 during the summer. Further work should be directed at improving the understanding of sea ice carbonate chemistry and its role in the sea ice carbon pump of the polar oceans.

Acknowledgments
The authors would like to thank the Captain, officers, crew and scientists involved in cruise JR177 and the British Antarctic Survey for making the RRS James Clark Ross available to us. The cruise was a component of the British Antarctic Survey's DISCOVERY-2010 programme. This work is part of PhD research by EM Jones at the University of East Anglia funded by the Natural Environment Research Council (NERC) under the auspices of the Centre for observation of Air-Sea Interactions and fluXes (CASIX) (NER/F14/G6/115). Participation in the cruise was supported by the Antarctic Funding Initiative (AFI) through the Collaborative Gearing Scheme (CGS8/28). The authors acknowledge the constructive comments by the editors and an anonymous reviewer.