Dense shelf water formed in the Mertz Polynya supplies the lower limb of the global overturning circulation, ventilating the abyssal Indian and Pacific Oceans. Calving of the Mertz Glacier Tongue (MGT) in February 2010 altered the regional distribution of ice and reduced the size and activity of the polynya. The salinity and density of dense shelf water declined abruptly after calving, consistent with a reduction of sea ice formation in the polynya. Breakout and melt of thick multiyear sea ice released by the movement of iceberg B9B and the MGT freshened near-surface waters. The input of meltwater likely enhanced the availability of light and iron, supporting a diatom bloom that doubled carbon uptake relative to precalving conditions. The enhanced biological carbon drawdown increased the carbonate saturation state, outweighing dilution by meltwater input. These observations highlight the sensitivity of dense water formation, biological productivity, and carbon export to changes in the Antarctic icescape.
 Antarctic coastal polynyas, areas of open water within the sea ice pack, are sites of intense air-sea interaction, sea ice production, dense water formation, and biological activity [Zwally et al., 1985; Arrigo and van Dijken, 2003; Williams and Bindoff, 2003; Tamura et al., 2008]. Coastal polynyas form where strong katabatic or synoptic winds drive newly formed sea ice away from the coastline or other fixed boundary such as a glacier tongue, which can also act as a barrier preventing the movement of ice into the region [Massom et al., 1998]. Polynyas produce and export large amounts of sea ice; brine released as sea ice forms increases the salinity and density of the underlying water. Coastal polynyas can therefore produce large volumes of dense shelf water, the precursor to Antarctic Bottom Water (AABW). The primary sources of dense water in the Southern Ocean are the Weddell Sea, the Ross Sea, and the polynyas near the Mertz Glacier and neighboring embayments (hereafter referred to as the Mertz Polynya) [Orsi et al., 1999]. Prior to the calving event in February 2010, the Mertz Polynya was one of the largest in East Antarctica and was the third largest Antarctic polynya in terms of sea ice production [Tamura et al., 2008]. AABW formed in the Mertz Polynya feeds the deep branch of the overturning circulation that supplies water and oxygen to the lower layers of the Indian and Pacific Oceans [Rintoul, 1998; Orsi et al., 1999]. Polynyas are often also regions of enhanced biological production because the presence of open water in early polar spring exposes the surface ocean to sunlight and stimulates primary production [Arrigo and van Dijken, 2003].
 Iceberg calving and grounding can alter polynya dynamics, dense water production, and regional circulation, as found by Nøst and Østerhus  and Grosfeld et al.  following calving of icebergs from the Filchner Ice Shelf in the Weddell Sea. The ungrounding of iceberg B9B and the subsequent calving of the Mertz Glacier Tongue (MGT) resulted in substantial changes to the regional icescape (Figures 1a and 1b). Tamura et al.  used remote sensing data and atmospheric reanalyses to infer that these changes reduced the volume of sea ice formed in the Mertz Polynya in the winter of 2010 (2011) by 14% (20%) relative to the 2000–2009 mean. The numerical simulations of Kusahara et al. [2011a] suggested that calving of the MGT changed the regional circulation, reduced sea ice formation, and decreased the export of dense shelf water by 23%. However, no in situ observations were available to test these inferences. Here we use observations of water mass properties from before and after the calving event to examine the consequences of changes in the icescape for dense water formation, water column stratification, and biological carbon uptake.
 We use data from four voyages to the Mertz Polynya region: a winter voyage in 1999 [Williams and Bindoff, 2003] and summer voyages in 2001 [Sambrotto et al., 2003], 2008 [Lacarra et al., 2011], and 2011. A small number of colocated stations from other cruises were used to extend the time series (Figure 1). Station locations are shown in Figure 1c, and cruise details are given in the caption. Measurements of temperature, salinity, and nutrients were available for each voyage (see references above for sampling details). Inorganic carbon system measurements were made on the 2008 and 2011 voyages. Dissolved inorganic carbon (DIC) and total alkalinity (TA) were determined by coulometric and potentiometric titration, respectively, following standard procedures [Dickson et al., 2007]. Regular analysis of Certified Reference Materials [Dickson et al., 2007] ensured the accuracy of the DIC and TA measurements, and reproducibility was better than 1 and 2 µmol kg−1, respectively. The computation of aragonite saturation state (Ω) was made using the CO2Sys program of Lewis and Wallace  and the equilibrium constants of Dickson and Millero . Surface pCO2 measurements were made by underway continuous flow equilibration [Dickson et al., 2007] and have an uncertainty of less than 3 µatm. Air-sea CO2 fluxes were computed with (short term) in situ winds measured on board at 10 m, following Wanninkhof .
 Net community production (NCP) was calculated from the depth integrated (0–100 m) deficit of salinity-normalized DIC (nDIC) in 2008 and 2011. Profiles of nDIC from both 2008 and 2011 indicate nearly constant concentrations below a depth of approximately 200 m. We therefore chose to use the 250 m bottle measurement of nDIC (2263 µmol kg−1) as representative of the winter water, since the winter profiles of salinity (and density) indicate that the depth of winter mixing extends well below this depth [e.g., Williams and Bindoff, 2003]. The standard deviation in nDIC, nitrate, and silicate at this depth was on the order of 2, 1, and 3 µmol kg−1, respectively. The air-sea exchange of CO2 was not accounted for in the computation of NCP. If we assume that the largest uptake of atmospheric CO2 in 2011 (80 mmol C m−2 day−1) persisted for 30 days, this would change the nDIC concentration in the upper 100 m by roughly 20 µmol kg−1, and correspond to an underestimate of less than 0.5 mol C m−2 in the computation of NCP. A similar computation was made to estimate the (depth-integrated) NO3 and Si(OH)4 deficits using winter water (250 m) values of 31 and 78 µmol kg−1, respectively. DIC measurements were not made in 2001; therefore, NCP was computed from the NO3 deficit as described above, scaled by the Redfield ratio of C:N = 6.6.3.
3 Results and Discussion
 The densest shelf water formed in winter in the Mertz Polynya, called high salinity shelf water (HSSW), accumulates in the Adélie Depression on the continental shelf (see Figure 1c), allowing the HSSW to be sampled during summer voyages. Near-repeat stations occupied at three sites in the depression allow the time history of dense water formation to be assessed (Figure 2). As no direct velocity observations are available to calculate changes in transports of dense water, we focus on changes in water mass properties. The most saline and dense HSSW is found in the deepest water at the eastern side of the depression, near the former location of the MGT, and the largest changes in salinity are observed there (Figure 2a). From 2008 to 2012, the salinity of the bottom waters decreased by more than 0.15. For perspective, the most rapid freshening so far observed in the Southern Ocean is in the Ross Sea, where shelf waters freshened at a rate of 0.03/decade between 1958 and 2008 [Jacobs and Giulivi, 2010]. The observed reduction in HSSW salinity after the MGT calving is equivalent to 50 years of freshening at the long-term rate observed in the Ross Sea.
 Interannual variability in HSSW salinity is observed at all three sites prior to calving, consistent with modeling studies, suggesting that dense water formation varies from year to year [Marsland et al., 2004; Kusahara et al., 2011b]. The 2001 profiles, in particular, tend to be fresher than those from 2008 or from earlier cruises. Tamura et al.  show that the amount of sea ice formed in the winter of 2000 was the second lowest observed in the 2000–2009 period and almost as low as observed following calving of the MGT. At all three locations, however, the abrupt decrease in salinity after the calving event exceeds the range of interannual variability (Figure 2d). Whilst the definition of a single “critical density” for AABW formation is problematic [Williams et al., 2010], the densest shelf waters observed in summer 2012 are substantially fresher (S = 34.589) and lighter (σ0 = 27.8499 kg m−3) than water with the criteria of Bindoff et al.  for AABW formation (S = 34.63, σ0 = 27.88 kg m−3) and approach the minimum density of shelf water (27.85 kg m−3) observed to be exported from the Adélie Depression prior to calving [Williams et al., 2010].
 Substantial changes were also observed in the upper ocean after calving. The mixed layer was fresher by 0.755 and 33 m shallower in summer 2011 than in summer 2008 (fresher by 0.500 and 9 m shallower than 2001) based on median values for each voyage (Figure 3a and Table 1; the distributions of mixed layer depth and salinity for the 2001, 2008, and 2011 voyages are shown in Figures A1 and A2). For comparison, the change in surface salinity driven by the annual formation and melting of sea ice is 0.209 (Figure 3a, winter 1999 to summer 2008, neglecting interannual variability). The freshening of the upper 200 m in 2011 relative to 2008 requires an input of 1.16 m of freshwater per m2 of ocean surface. Melting of the thick multiyear sea ice released by the movement of iceberg B9B and the MGT (Figure 1) likely supplied this additional freshwater. After calving (January 2011), the area of multiyear fast ice between 145°E and 154°E (5504 km2) was less than 27% of the mean annual minimum extent (20,425 km2) observed between 2000 and 2009. The multiyear ice east of the MGT was unusually thick (up to 10–55 m thick just east of the MGT [Massom et al., 2010] and 2–3 m thick east of B9B [Massom et al., 2001]) and covered by more than 3 m of snow (measured on floes adjacent to the ship on the 2011 voyage).
Table 1. Properties of the Mixed Layer in the Mertz Polynya System in 2001, 2008, and 2011 with E and W Referring to East and West of the Former Position of the MGT (see Figure 1).a
aMixed-layer depth is in m, temperature is in °C, nutrient and (salinity-normalized) inorganic carbon concentrations are in µmol kg−1, and ΔpCO2 are surface values in units of µatm. Carbon deficits (NCP, mol C m−2) and nutrient deficits (mol m−2) relative to winter values are given along with the ratio between silicate and nitrate utilization (def(Si):def(NO3)) and between silicate and carbon utilization (def(Si:NCP). Decreases in def(Si):NCP may indicate a relief of iron deficiency.
 The abrupt and large changes in the salinity of shelf waters can therefore be attributed to two distinct, but related, phenomena. First, the ungrounding and drift of the large iceberg B9B precipitated the calving of the MGT [Young et al., 2010]. These events reduced the area of the Mertz Polynya and sea ice production declined by 24 km3 (14%), relative to the 2000–2009 mean [Tamura et al., 2012]. The freshening of HSSW reflects this reduction in polynya activity and brine rejection. Second, the movement of B9B and the MGT released a large volume of sea ice, which drifted into the polynya area and melted (Figures 1a and 1b). The dramatic freshening of surface waters between 2008 and 2011/2012 was caused by this meltwater input. The continued decline in the salinity of HSSW in 2012 reflects both a further reduction in ice formation in the winter of 2011 [Tamura et al., 2012] and the preconditioning of surface waters by the large input of sea ice meltwater.
 Substantial increases in the seasonal drawdown of inorganic carbon and nutrients in 2011 (relative to 2008) indicate an increase in regional primary production and export (Table 1 and Figure A2). The largest drawdown was observed east of the former position of the MGT, coincident with the lowest surface salinities and shallowest mixed layers (Figure A1), and where the largest input of freshwater from melting sea ice is expected (Figure 1, Figure A2, and Table 1). This response is in contrast to that reported for the Ross Sea Shelf after the movement of iceberg B-15, which resulted in heavier spring ice cover and decreased phytoplankton production [Arrigo et al., 2002]. NCP in the Mertz Polynya in 2011 (2.4 mol C m−2) exceeded NCP in 2008 (1.1 mol C m−2) and 2001 (1.3 mol C m−2) by about a factor of 2. NCP in January 2011 (~0.98 g C m−2 day−1) was nearly double the 5 year mean January (gross) primary production estimated by remote sensing (0.52 ± 0.29 g C m−2 day−1) [Arrigo and van Dijken, 2003] and was similar to that observed on the Ross Sea Shelf [Bates et al., 1998; Sweeney et al., 2000], considered among the most productive regions of the Southern Ocean [Arrigo and van Dijken, 2003].
 Carbon utilization enhanced the surface ocean sink for atmospheric CO2 from 15 mmol C m−2 day−1 in 2008 to 30–80 mmol C m−2 day−1 in 2011 (Table 1). The uptake of CO2 by the ocean lowers pH and reduces the concentration of carbonate ions, reducing the calcium carbonate saturation state (Ω). Sea ice melt reduces Ω through dilution, as recently observed in the Arctic Ocean [Yamamoto-Kawai et al., 2009]. In contrast, our observations indicate that Ω increased in the Mertz Polynya in 2011 relative to 2008, despite the large input of meltwater (Figure 3b, inset). The enhanced biological carbon drawdown was sufficient to increase the saturation state by 10 times the expected decrease due to dilution. Blooms induced by sea ice melt can therefore buffer the decrease in Ω due to dilution.
 Iron was not measured during the 2011 voyage, but several lines of evidence suggest the bloom was supported at least in part by the input of iron. First, productivity of the “high-nutrient, low-chlorophyll” waters of the Southern Ocean is limited by the availability of iron [Martin et al., 1990; Boyd et al., 2000] (though to a lesser extent in coastal regions which have a sedimentary supply [Boyd et al., 2012]). Sambrotto et al.  linked seasonal patterns of nutrient uptake in the region (before calving) to iron supply, implying that the enhanced 2011 bloom may have been sustained with additional iron. Second, filtered seawater samples indicate the 2011 bloom was dominated by large diatoms Corethron and Chaetocerous. In contrast, Phaeocystis dominated the bloom in the eastern sector in 2001 [Sambrotto et al., 2003; Vaillancourt et al., 2003], and a mixed community of small diatoms was observed in 2004 [Beans et al., 2008]. Since large diatoms tend to outcompete Phaeocystis in iron-rich, well-stratified waters [de Baar et al., 2008], their presence in 2011 supports the hypothesis that the availability of both light and iron was increased. Third, the lowered ratio of silicate utilization to NCP in the eastern Mertz Polynya region in 2011, relative to 2008, is consistent with diatoms blooming in iron-replete conditions [Takeda, 1998], though changes in species composition may also impact these ratios (Table 1). Finally, the decaying multiyear fast ice that dominated the sea ice distribution in 2011 represents a potential source of iron to the surface waters [Sedwick and DiTullio, 1997]. Iron concentrations in sea ice are several orders of magnitude higher than those measured in ice-free waters [Lannuzel et al., 2007]. During the melting period, sea ice provides more iron to Antarctic surface waters than other sources such as atmospheric deposition, extraterrestrial iron, vertical diffusion, and upwelling [Lannuzel et al., 2007]. Melting of the thick multiyear fast ice and pack ice released by the movement of B9B and the MGT would therefore supply substantial iron to surface waters. Particulate matter released into the water column by the ungrounding of iceberg B9B may have also contributed iron. Large diatom blooms are often associated with enhanced carbon export as particulate organic matter [Romero and Armand, 2010]. Assuming that all the estimated inorganic nutrient deficits result from biological drawdown and subsequent export, the larger nitrate drawdown in 2011 relative to previous years indicates an increase in carbon export (Table 1).
 Additionally, photographs of the seafloor revealed a thick mat of fresh organic matter (Figure A3) and microscopic examination of filtered samples showed that the suspended material was largely organic; in particular, Corethron and Chaetoceros cells were identified at depths greater than 500 m, most with cytoplasm present but disrupted, consistent with rapid sinking of organic matter. Thick siliceous sediments deposited in Antarctic troughs like the Adélie Depression during glacial retreat (ca. 10,000 y. B.P.) have been associated with intense blooms of large diatoms like the Chaetocerous and Corethron species that dominated in 2011 [Leventer et al., 2006]. These blooms were likely supported by the input of iron-rich glacial meltwater [Leventer et al., 2006]. Our observations of a substantial bloom of large diatoms driven by input of meltwater, in this case from thick multiyear sea ice, may provide a modern analogue of the hypothesized deglacial conditions, though in this case the perturbation in local freshwater input and biological production is likely to be transient.
 Changes in the distribution of ice associated with the calving of the MGT, including the movement of iceberg B9B and the subsequent breakout and melt of thick sea ice, had substantial impacts on the water mass properties and biogeochemistry of the region. The change in the regional distribution of ice reduced the area and activity of the Mertz Polynya, decreased sea ice formation and brine rejection, and lowered the salinity of HSSW close to the threshold needed to form AABW. Input of meltwater freshened the upper ocean, providing light and micronutrient conditions favorable for phytoplankton growth. The resulting bloom of large diatoms doubled biological carbon uptake and export, and increased the carbonate saturation state (Ω) despite the effects of dilution. While calving of glacier tongues is a natural process, thinning and decay of ice shelves and glacial tongues have been linked to increased basal melting by warm ocean waters [Jacobs et al., 2011; Pritchard et al., 2012]. Future warming of the ocean surrounding Antarctica is expected to drive further thinning and decay of floating glacial ice, with implications for ice sheet mass balance [Pritchard et al., 2012] and the regional distribution of ice. Our study suggests that changes in the Antarctic icescape can also have substantial consequences for dense water formation, carbon uptake, and biological productivity of Antarctic shelf waters.
 This work was supported by the Australian Government's Cooperative Research Centres Program through the Antarctic Climate and Ecosystems Cooperative Research Centre (ACE CRC) and by the Department of Climate Change and Energy Efficiency through the Australian Climate Change Science Program. We acknowledge the use of data products from the Land Atmosphere Near-real time Capability for EOS (LANCE) system operated by the GSFC/Earth Science Data and Information System (ESDIS) with funding provided by NASA/HQ. We acknowledge the use of the MERIS data provided through ENVISAT project AO-745 (CI N. Young). J. Smith publishes with permission of the CEO of Geoscience Australia.