Nutrient Distribution in East Antarctic Summer Sea Ice: A Potential Iron Contribution From Glacial Basal Melt

Antarctic sea ice can incorporate high levels of iron (Fe) during its formation and has been suggested as an important source of this essential micronutrient to Southern Ocean surface waters during the melt season. Over the last decade, a limited number of studies have quantified the Fe pool in Antarctic sea ice, with a focus on late winter and spring. Here we study the distribution of operationally defined dissolved and particulate Fe from nine sites sampled between Wilkes Land and King George V Land during austral summer 2016/2017. Results point toward a net heterotrophic sea-ice community, consistent with the observed nitrate limitation (<1 μM). We postulate that the recycling of the high particulate Fe pool in summer sea ice supplies sufficient ( ∼ 3 nM) levels of dissolved Fe to sustain ice algal growth. for bacterial activity. results indicate potential relationship between glacial meltwater discharged from the Moscow University Ice Shelf the occurrence of Fe-rich ( 4.3 μM) platelet ice in its vicinity. As climate change is expected to result in enhanced Fe-rich glacial discharge and changes in summer sea-ice extent and quality, the processes influencing Fe distribution in sea ice that persists into summer need to be better constrained.

A high percentage of sea-ice dissolved Fe (DFe, filtered through a 0.2 µm pore size membrane), our best empirical measurement of potentially bioavailable Fe (Morel et al., 2008), is lost to the seawater in a rapid pulse over a few days during spring (Duprat et al., 2019;Lannuzel et al., 2008). This early release is likely rapidly consumed by phytoplankton with very low DFe concentrations reported for seawater after the spring bloom (Bertrand et al., 2011;Sedwick et al., 2000Sedwick et al., , 2011. Sea-ice melt may provide an additional later pulse of Fe to surface waters in the particulate form (PFe, > 0.2 µm), as the flushing of this pool from the sea ice is initially limited by brine channel transport (Lannuzel et al., 2013;van der Merwe et al., 2011). Studies conducted during the last two decades indicate that sea-ice DFe concentrations are relatively unaffected by the proximity to Fe sources (Lannuzel et al., 2016, and references therein). This apparent contradiction has been attributed to the fact that in the presence of saturated organic ligands, the dissolution equilibrium should impose a maximum threshold for DFe concentrations in sea ice. When all organic ligands are saturated, excess DFe precipitates into PFe. Much of East Antarctica's free-floating pack ice is formed in leads and polynyas (Allison, 1989), and due to an increased distance to coastal sources, it generally carries a lower concentration of lithogenic PFe than land-fast sea ice (sea ice fastened to coastal features; . The accumulation of PFe in pack ice can still occur, either by uptake and conversion of DFe (e.g., from the continental shelf or icebergs) into PFe, or via direct incorporation of biogenic PFe from highly productive shelf waters where it originated .
In contrast to pack ice, fast ice can be affected by a greater input of Fe from lithogenic sources, such as coastal sediment (de Jong et al., 2012) and dust from ice-free continental areas Duprat et al., 2019). Less clear is the influence of subglacial meltwater on the fast-ice Fe pool via incorporation of ice platelets at the bottom of sea ice. The growth of ice platelets has been observed below sea ice under the influence of ice-shelf meltwater (Langhorne et al., 2015). There are several observations of platelet ice in the Ross and Weddell seas; otherwise, there are very few reports of platelet ice layers in East Antarctica (Langhorne et al., 2015). It is believed they are seeded by frazil crystals formed in supercooled ice shelf waters, although the processes behind the formation of these unique crystallographic structures are still poorly understood (Tison et al., 1998). The ice platelets can eventually be incorporated onto the base of the ice shelf (marine ice) via buoyant accumulation (Oerter et al., 1992), or be carried out from the ice shelf cavity and deposited beneath neighboring sea ice (subice platelet layers; Langhorne et al., 2015;Tison et al., 1998). Under sea ice, in-situ growth can continue, creating a porous and friable subice platelet layer from which the highest concentrations of sea-ice algae (up to 374 mg C L −1 and 6.5 mg Chla L −1 ) have been reported (Arrigo et al., 2010;Bombosch, 2013;Smetacek et al., 1992;van der Linden et al., 2020). The origin of marine ice Fe is often associated with highly reactive glacial flour produced from subglacial physical and chemical weathering (Wadham et al., 2010). The presence of buoyant frazil and ice platelets could facilitate the incorporation of sediment and biogenic material from the ocean into the marine ice layer (Treverrow et al., 2010). While platelet ice incorporation under ice shelves has been associated with higher concentrations of Fe DUPRAT ET AL.
2 of 19 10.1029/2020JC016130 (Herraiz-Borreguero et al., 2016), no link has yet been made for platelet ice incorporation under sea ice due to the paucity of observations and measurements.
To date, most sea-ice surveys focused on investigating Fe distribution and biogeochemistry in late winter and spring . In this study, we report Fe concentrations obtained from between Wilkes Land and King George V Land coast (East Antarctica) and sampled during a marine science voyage carried out from December 2016 to January 2017. We aim to answer our central question: is sea-ice primary production Fe-limited in summer due to the advanced stage of brine and nutrient drainage? A further objective was to evaluate how the Fe biogeochemistry relates to macronutrients and other biological variables in summer sea ice, such as algae-produced exopolysaccharides (EPS). The final aim of this study was to investigate a potential link between glacial basal melt from ice shelves encountered during our expedition (e.g., Moscow University, Mertz and Ninnis) and the formation of Fe-rich platelet ice, as an important spatial determinant in sea-ice PFe distribution. Results from this study could guide studies on the potential impacts of future ice mass balance changes on local primary production, and carbon export of highly productive polynyas found downstream of the Dalton, Mertz and Ninnis ice shelves (Paolo et al., 2015;Rignot et al., 2019).

Sampling Sites
Sea ice was sampled during an icebreaker Research Survey Vessel Aurora Australis voyage at nine different stations off the East Antarctic coast (Figure 1) during austral summer (December 2016 to January 2017). The mobility of the ice stations (e.g., fastened vs. free-floating) was used as criteria to categorize the stations into fast and pack ice (see Section 3.1

Cleaning Procedures
Plasticware used for sample collection (polypropylene [PP; Superlift®] melting containers) and storage (low-density polyethylene [LDPE; Nalgene®] sampling bottles) were kept in sealed plastic bags after being acid-cleaned according to GEOTRACES recommendations. First, new LDPE bottles were soaked in 2% (v:v) Decon90 baths for 1 week. Bottles were then thoroughly rinsed with ultra-high purity water (UHP water, Barnstead), filled with 6 M hydrochloric acid (HCl, 50% v:v reagent grade, Merck) and immersed in 2 M HCl (∼20% v:v reagent grade, Merck) for one month. Bottles were then rinsed with UHP water, filled with 1 M distilled HCl (∼10% v:v instrument quality, Seastar) and kept for 1 week on a hot plate (60°C). Finally, bottles were rinsed three times with UHP water in a class-100 laminar flow hood and stored triple bagged in plastic press seal bags until use. Between sampling days, plastic equipment (Teflon® filtration sets, plastic tubing, and plastic scoops) used for collection and processing were rinsed with UHP water and soaked in an acid bath (10%-20% v:v HCl, according to Nalgene® recommendations for each material) and thoroughly rinsed again with UHP water before use. Finally, glassware used for organic carbon and nitrogen filtrations were soaked in a 2% (v:v) HCl solution (reagent grade, Merck), rinsed with UHP water, wrapped in aluminum foil, and combusted at 450°C for 4 h.

Sampling Procedures
During field sampling, trace metal clean protocols were followed, as described by Lannuzel et al. (2006) and van der Merwe et al. (2009). The snow was collected using an acid-cleaned polyethylene (PE) hand shovel and stored in a 3.8 L wide-mouth Nalgene® high-density polyethylene (HDPE) bottle. Separate cores from the same vicinity were collected for temperature/salinity, particulate organic matter (POM), chlorophyll-a (Chla), particulate EPS (PEPS), macronutrient and Fe determinations. Ice cores were collected with an electric-powered, electro-polished stainless-steel corer of 0.14 m internal diameter (Lichtert Industrie, Belgium), previously shown appropriated for trace metal sampling (Lannuzel et al., 2006;van der Merwe et al., 2009). Ice cores were sectioned in the field using a medical-grade bone saw (Richards Analytical). Six or seven sections were obtained out of the entire core (from top, intermediate, and bottom layers), 10-15 cm thick, in order to represent different depth horizons of the ice cores. Snow and sea-ice samples were melted at room temperature immediately after collection, filtered (for the dissolved fraction, see details in paragraph 2.4.2) and poured into 125 mL Superlift® PP bottles, double bagged and stored at −20°C until analysis. Brine was collected from ∼1 m deep sack holes, while underlying seawater was collected through full-icedepth core holes, just below the ice (hereafter referred to as SW0m). Both brine and seawater were collected using an acid-cleaned 0.5 L perfluoroalkoxy (PFA) bottle (Nalgene®) attached to an improvised 2 m long bamboo pole and transferred into acid-cleaned 1L LDPE bottles (Nalgene®).

Physical Variables
Sea-ice temperatures were measured with a Testotherm720 thermometer (accuracy ± 0.1°C) immediately after ice core retrieval from small holes drilled in the core at 0.1 m intervals. Bulk salinities of brine, seawater, melted snow, and sea-ice sections were measured using a Guildline Autosal salinometer (Ocean Scientific International Ltd, UK; ± 0.001 practical salinity unit [PSU]). The texture of each ice core was analyzed by photographing vertical thin sections between cross-polarizing filters, following the method of Langway (1958). The physical ice properties are discussed in more detail in a companion paper. Finally, a light sensor was deployed in the core hole to measure the luminous ice emittance (per square meter) at 0.1 m increments as an indication of the light intensity that passed through different depth horizons of the sea ice.

Fe Physical Fractionation
Dissolved and particulate Fe from snow, seawater, brine, and melted sea-ice sections were obtained by filtration through 0.2 µm pore size 47 mm diameter polycarbonate (PC) membrane filters (Sterlitech) using Teflon® PFA filtration apparatus (Savillex, USA cell rupture during filtration. The dissolved fraction (<0.2 µm, which includes the soluble (<0.02 µm) and colloidal (0.02-0.2 µm) fractions) was collected in LDPE bottles, acidified to pH 1.8 by adding ∼1% (v:v) of 12 M ultrapure HCl (Seastar Baseline, analytical grade), triple-bagged and stored at room temperature until analysis. PC filters with the retained particulate fraction (>0.2 µm) were placed into acid-clean polystyrene Petri dishes, triple-bagged, and stored at −20°C in the dark until further processing and analysis at the University of Tasmania (Australia).

DFe Analysis
An automated off-line sample system (seaFAST-pico™, Elemental Scientific; Nobias Chelate-PA1 resin) was used to preconcentrate trace metals and remove sea-ice and seawater filtrate matrices from our samples according to the method described by Wuttig et al. (2019). Comprehensive method testing using a variety of internationally recognized community reference samples and oceanographic standards  as well as NASS-6 verified method accuracy for Fe. A mean Fe value of 9.24 ± 0.24 nmol kg −1 (n = 6) was measured for the NASS-6, in good agreement with the reported indicative value of 8.

Particulate Fe
Particulate Fe was quantified after complete filter digestion using a strong acid treatment . First, a mixture of 250 µL 12 M HCl, 250 µL 16 M nitric acid (HNO 3, SeaStar Baseline), and 500 µL 29 M hydrofluoric acid (Seastar Baseline, Choice Analytical) was added to each sample and blank filter and placed into a Teflon® PFA vial (Savillex, USA). Vials were heated to 120°C for 12 h over a Teflon® coated hotplate (SCP Science) housed within a class-100 fume hood. Vials were then left open until complete evaporation of the acids. Next, 9.9 mL of ultrapure 10% (v:v) HNO 3 was added to re-suspend the dried residues before the addition of 100 µL Indium solution (final concentration 87.1 nM) as an internal standard. Finally, an aliquot of 5 mL was transferred to PP tubes for Fe quantification using SF-ICP-MS following similar procedures as outlined in Section 2.4.2.1. The detection limit for Fe was calculated as 3× the standard deviation of the digested acid filter blank, determined as 8 × 10 −4 ppb (0.14 nM). The theoretical biogenic fraction (PFe bio ) was calculated as the total PFe-lithogenic PFe (inferred by the Fe:Al ratio of 0.33; Taylor & McLennan, 1985). It is worth noting that Fe enrichment relatively to Al can occur via flocculation and increasing sorting of lithogenic material with distance from the coast (Markussen et al., 2016), therefore results should be interpreted with care. For the purpose of this work, we refer to biogenic Fe as any PFe form associated to biogenic pools (living, detrital, and authigenic) that are not refractory particles originated from terrestrial sources of rock, such as silicate minerals (lithogenic origin).

Particulate Organic Matter
Melted sea-ice and seawater samples for particulate organic matter determination were gently homogenized before filtration (0.14-0.6 L) on precombusted (450°C, for 8 h) glass microfiber (GF/F) filters (0.7 µm nominal porosity, MGF Sartorius). Samples were dried overnight at 60°C. Inorganic carbon contamination was removed from the filters with the addition of 10% (v:v) distilled HCl (80 µL) for 12 h inside a closed PCR well tray (Martin, 1993). Samples were then packed in silver cups (Elemental Microanalysis, UK) and stored in a desiccator until analysis for elemental composition at the CSL. Particulate organic carbon (POC) and DUPRAT ET AL. 5 of 19 10.1029/2020JC016130 particulate organic nitrogen (PON) concentrations were determined by flash combustion and oxygen pyrolysis using a Thermo Finnigan EA 1112 Series Flash Elemental Analyzer (estimated precision ∼1%). POC and PON detection limits were calculated as 3× the standard deviation of the procedure blanks, determined as 2.5 and 1.8 µM, respectively (n = 6).

Particulate EPS
In aquatic systems, dissolved EPS are prone to coalesce into particulate form, also referred to PEPS. Sea-ice, brine, and seawater PEPS concentrations were measured using the semi-quantitative colorimetric Alcian Blue assay (Passow & Alldredge, 1994) following van der Merwe et al. (2009). Calibration was achieved by measuring five different extracted solutions from filters previously treated with 0, 2.5, 5, 7.5, and 10 mL of a 0.75 μg L −1 Xanthan gum solution, following the same staining protocol used for the sample filters. A linear regression (R 2 = 0.92; n = 16) was calculated between absorbance and the respective concentration of Xanthan gum. The detection limit (3× the standard deviation of the filter blank) was 28.6 μg Xanthan equivalent (xeq.) L −1 (n = 20).

Chla and Macronutrients
Samples for Chla analysis were melted at room temperature in the dark (Rintala et al., 2014). Chlorophyll a was concentrated by filtering samples (0.09-0.7 L) through GF/F filters under low vacuum (<0.13 bar) followed by extraction with 90% acetone (High Performance Liquid Chromatography grade) for 12-24 h. Chlorophyll a concentrations were determined fluorometrically with a Turner Designs 10AU fluorometer (in vitro detection limit 0.02 µg L −1 ). Fluorescence was measured at 750 nm before and after acidification with 0.1 M HCl, and phaeopigment-corrected Chla was calculated according to Arar and Collins (1997). Macronutrient samples were collected via filtration over 0.2 µm pore size filters (Acrodiscs) placed inside a filter holder and stored frozen (−20°C) in 60 mL HDPE bottles until analysis in the laboratory (CSIRO, Hobart, Australia ), and (ortho)silicic acid (Si(OH) 4 ) were determined using a segmented flow analyzer (SEAL Analytical Ltd AA3 HR Autoanalyzer) following standard colorimetric methods from Rees et al. (2018), and Kérouel and Aminot (1997) for NH 4 + . Salinity-normalized macronutrient concentrations (C*) were calculated to tease apart physical effects following the equation applied by Fripiat et al. (2017) where C is the measured nutrient concentration in the bulk sea ice, S w is the salinity of seawater and S the corresponding measured salinity of in the bulk ice sample.

Statistical Analysis
Distributions of variables were tested from all stations and within both the pack and fast ice groups. First, the normality and the homogeneity of variance of each variable was tested using the Shapiro-Wilk test and Levene's test, respectively. When the normality and homogeneity of variance could be verified, an analysis of variance test was run. For non-normal distributed variables, the Kruskall-Wallis test was used. Multiple (paired) comparisons corrected by the Holm test were performed. Differences between fast and pack ice were also tested using either a Wilcoxon test or a t-test, depending on the distribution. Potential correlations between all physical and biogeochemical parameters were investigated using a Spearman correlation coefficient test due to the non-normality observed in most variables. For this test, ice texture proportions were calculated for each section using rankings 0 for noncolumnar (granular and platelet) and 1 for columnar. All analyses were performed using the R programming language (version 3.3.1; R Core Team, 2013).

Ice Physics
All collected ice cores were relatively warm (>−2°C). Porosities were almost always above the percolation threshold of 5% for columnar ice and often greater than 10%, which likely allowed brine movement through the more random brine inclusions of granular ice found in the ice cores. Brines were stratified, indicating exchange with the underlying seawater was limited to diffusive processes. The overall low brine salinities (0.9-32.2 PSU) were indicative of surface-melt input which was likely driving desalination. Of the textures sampled near ice shelves, columnar ice was often interspersed with layers of granular ice or platelet ice. Granular ice was found near the surface, especially as snow ice, while platelet ice was found at greater depths, beyond ∼0.5 m (Totten and Mertz) and beyond ∼1 m in one instance in front of the Mertz Glacier Tongue (Table 1). Measured illuminance reached optimum diatom growth values (1,000 lx; Azam & Singh, 2013) at the ice ∼0.4 m depth. Average depth-adjusted illuminance varied from 2,730 lx at the ice surface to only 17 lx in basal ice.

Distribution of Fe
Dissolved and particulate Fe concentrations measured in the snow, sea ice, brine and seawater are summarized in Table 2. A detailed description of values for each station and ice section is available via the link in Section 6. The vertical DFe and PFe profiles for fast and pack ice show no clear pattern for most stations (Figure 2). DFe was enriched in sea ice relative to concurrent seawater (SW0m) and to values found in winter seawater (Figure 3a; van der Merwe et al., 2009). Concentration of PFe were up to an order of magnitude higher than the seawater below. The average concentrations of DFe and PFe for all pack and fast ice stations were 2.9 ± 1.9 nM and 323 ± 606 nM, respectively. The particulate fraction comprised 98% of the total Fe pool with average PFe:PAl ratios of 0.6 for fast ice and 1.0 for pack ice. Lithogenic PFe inferred from the ice PFe:PAl ratios represented approximately 70% in fast ice and 40% in pack ice. Fast ice showed significantly higher concentrations of DFe and PFe (4.7 ± 2.3 nM; 665 ± 980 nM) compared to pack ice (2.0 ± 0.7; 81.7 ± 48.2; p < 0.01). Totten 1 showed significantly higher DFe and PFe (p < 0.01) than the other fast ice stations. Higher DFe and PFe were also found in fast ice (3.5 ± 3.4 nM; ± 99.6 ± 64.3 nM) compared to pack ice (2.0 ± 1.3 nM; 83.2 ± 124 nM) when Totten 1 was treated as an outlier, albeit not statistically significant. 2.9 ± 1.9 2.9 ± 1.9 12.4 ± 8.6 0.4 ± 0.3 0.9 ± 0.5 31.0 ± 18.0 2.9 ± 2.8 20.6 ± 12.5 2.8 ± 1.7 7.9 ± 4.9 454 ± 223 Abbreviations: DFe, dissolved Fe; PEPS, particulate exopolysaccharides; PON, particulate organic nitrogen; POC, particulate organic carbon.   Table 2. Overall, lower concentrations of all macronutrients were found in our summer ice samples compared to other seasons (Fripiat et al., 2017). Nitrate was depleted relative to the Redfield N:P ratio (Figure 3b; Redfield, 1963). With exception of station SR3-5 and bottom ice data, Si(OH) 4 followed the Redfield-Brzezinski ratio for PO 4 3− relatively closely (Figure 3c; Brzezinski, 1985). Station SR3-5 presented significantly higher concentrations of PO 4 3− , NO 2 − and NO 3 − (p < 0.01). An average 6.4 ± 3.0 μM NO 3 − was found at this station, an order of magnitude higher than the other stations. SR3-5 station displayed a seawater-filled gap halfway through the ice cover, which could have supplied macronutrients to the ice.

Correlations Between Variables
Correlations between measured parameters were tested for fast and pack ice stations separately, and for all stations together ( Figure 5). Results highlight a strong and significant correlation between PFe and PAl for DUPRAT ET AL. 10 of 19 10.1029/2020JC016130 both pack (0.8; p < 0.01) and fast ice (0.9; p < 0.01). PEPS was correlated with POC, PON, and Chla (ρ > 0.7; p < 0.01) in fast, pack, and both ice types. There is also evidence of a moderate correlation of PEPS with salinity, ice porosity (0.5; 0.6; p < 0.05), macronutrients (0.5-0.8; p < 0.05) and NH 4 + (0.5; p < 0.01) in fast ice, while PEPS was moderately correlated with PFe (0.5; p < 0.01) in pack ice. Chlorophyll a, POC, and PON were all related to one or more types of macronutrient at different degrees (0.4-0.8; p < 0.05) in fast ice. When all stations were analyzed together, salinity was moderately to strongly correlated (0.5-0.8; p < 0.05) with macronutrients but not with (dissolved and particulate) Fe. Finally, there is also a tendency of inverse correlation (not significant) of columnar ice with salinity, macro-nutrients and Fe for all ice stations.

Summer Sea Ice Supports Heterotrophic Activity
Biomass and detritus are physically incorporated into newly forming sea ice and often show a uniform distribution over the ice thickness (Meiners et al., 2012;. Throughout spring and summer though, other processes, for example, in-situ production and grazing, start shaping the vertical distribution of biomass. The relatively higher POC concentrations (both in fast and pack stations) observed in the top and intermediate ice layers compared to other seasons Duprat et al., 2019;Janssens et al., 2016;Meiners et al., 2011;van der Merwe et al., 2011) suggest the upper ice cover becomes a more suitable habitat for sea-ice algae as the brine network connectivity increases during spring and summer. Heterotrophic processes are favored by the abundant presence of algal cell detritus and, potentially, EPS build up within these layers (Meiners, 2004(Meiners, , 2008. In the lowermost sections of the ice, we found C:N ratios (∼12) deviating from the Redfield ratio of 6.6 for phytoplankton (Redfield et al., 1963) and approaching those usually associated with the presence of anaerobic micro-organisms in the rest of the ice (Rysgaard & Glud, 2004). A shift from an autotrophic bottom ice in spring toward a hetero-, and mixotrophic dominant environment could be driven by a potential decline in primary productivity associated with reduced nutrient availability and/or increased internal self-shading by ice algae. A higher C:N ratio could also indicate a relative increase of carbon associated with PEPS, which is known to be highly enriched in carbon relative to nitrogen (26 mol:mol; Engel & Passow, 2001).
Sea-ice concentrations of PEPS obtained here were consistently higher (average 946 µg xeq L −1 ; 113-16,290 µg xeq L −1 ) than those reported by van de Merwe et al. (2009) in East Antarctic spring sea ice (average 493 µg xeq L −1 ; 2.84-2,690 µg xeq L −1 ) indicating an ongoing microbial productivity from spring through summer. PEPS production could have also been intensified as a response to nutrient stress or changes in temperature and salinity conditions (Ewert & Deming, 2013). The relationship between PEPS and Chla (ρ = 0.7; p < 0.01) suggests autotrophic protists are involved in PEPS production. To date, it is still unknown to what extent sea-ice bacteria hydrolyze and degrade PEPS as a source of nutrition. Nevertheless, observations of sea-ice EPS-particles colonized by bacteria indicate this substrate is available for bacterial degradation (Meiners et al., 2004(Meiners et al., , 2008Underwood et al., 2010). Therefore, PEPS could be used as a carbon source for bacteria, also facilitating grazing by heterotrophic and mixotrophic protists. Tightly coupled bacteria-algae interactions have been reported in sea-ice systems (Stewart & Fritsen, 2004) and the microbial loop is considered an important mechanism for recycling of organic matter in sea ice (Martin et al., 2012;Meiners and Michel, 2016).

Nitrate rather than Fe Limits Primary Production in Late Summer Sea Ice
Results from previous modeling (Saenz & Arrigo, 2014) and field studies (Lim et al., 2019;McMinn et al., 2000;Petrou et al., 2010) suggest that nutrients rather than light limit Antarctic sea-ice algae growth for the vast majority of the sunlit season. Sea-ice algae are also adapted to survive in a broad range of salinities, with particular tolerance to the low summer-ice salinity levels (Arrigo et al., 1993;Bates & Cota, 1986;Kirst & Wiencke, 1995;Palmisano et al. 1987;Ralph et al., 2007;Ryan et al., 2004;Torstensson et al., 2018). Based on these findings, macro-or micronutrients rather than light or salinity controlled autotrophic activity during our summer study.
Iron is considered the most limiting nutrient for phytoplankton growth in HNLC waters (Sunda, 2012). Seaice DFe levels in our study (∼3 nM) were above the limitation threshold (∼0.2 nM) for SO phytoplankton growth (Timmermans et al., 2004). We use Fe* to evaluate the potential Fe deficiency relative to PO 4 3− at this time of year. Scavenging of Fe commonly results in the decoupling of these elements (lower Fe relative to PO 4 3− ) in most of the ocean surface waters (Parekh et al., 2005), especially in the SO where input from aeolian deposition is low (Gao et al., 2001). In our study, Fe* was always positive, indicating DFe was present in sufficient concentrations for complete utilization of the available PO 4 3− . It is worth considering that Fe:PO 4 3− was estimated based on oceanic phytoplankton ratios and that it is plausible that sea-ice diatoms could have evolved to less efficiently use Fe compared to pelagic assemblages, due to the abundance of this nutrient in sea ice relative to seawater (Lim et al., 2019). However, a Fe:PO 4 3− uptake ratio an order of magnitude higher would be necessary to observe some degree of Fe depletion relative to PO 4 3− in summer sea ice. A positive Fe* is also observed for the majority of the sea-ice samples if Fe* is calculated based on Fe to total DIN (dissolved inorganic nitrogen) uptake ratios for oceanic diatoms under both Fe limited and replete conditions (0.1-2.1 mmol:mol; Price, 2005). Our positive Fe* values suggest organic ligands should be present in sea ice in excess relative to DFe concentrations, therefore preventing Fe precipitation. Fe does not seem to limit primary production in summer sea ice, although we cannot rule out a potential co-limitation imposed by other essential micronutrients such as manganese and cobalt (Corami et al., 2005;Koch & Trimborn, 2019;Pausch et al., 2019;Schoffman et al., 2016).
A concomitant drawdown of PO 4 3− and Si(OH) 4 following the Redfield-Brzezinski nutrient ratio (Figure 3c; Brzezinski, 1985) was observed for most data points. This suggests a large contribution of diatoms to sea-ice primary production. Median (and interquartile) salinity normalized PO 4 3− (2.0 µM (4.3 µM)) and Si(OH) 4 (16.1 µM (24.1 µM)) concentrations show that the studied sea ice was enriched and partially depleted respectively compared to seawater (0.9 µM (0.6 µM) and 44.5 µM (34.3 µM)) at this time of the year. However, following the "Liebig's Law" (von Liebig, 1940) which states that growth should be limited by the most deficient nutrient, the fact that PO 4 3− is above the Redfield N:P ratio (16; Figure 3b) indicates that DIN is the primary limiting nutrient in summer sea ice. Nitrate:PO 4 3− depletion was previously observed under Fe replete conditions during mesoscale Fe-enrichment experiments in HNLC waters (Boyd et al., 2000). In the present study, bulk nitrate and nitrite (NO x ) concentrations are close to exhaustion (<0.2 µM) at Casey 1, Totten 2, Mertz 1, 2, and 3. Brine concentrations < 0.2 µM were also observed at Mertz 2 and Mertz 3. In the absence of other sources of nitrogen NH 4 + would be readily consumed shortly after production. This could explain why our bulk NH 4 + concentrations are amongst the lowest ever reported in Antarctic sea ice (Fripiat et al., 2017). The low concentrations of N-sources found in most ice floes could therefore impose severe limitations for sympagic algal growth. Enhanced ice melting during summer exacerbates stratification within the ice, preventing convection-driven brine exchanges and replenishment of nutrients from seawater. This scenario would render sea ice particularly dependent on recycling and/or seawater intrusion to alleviate macronutrient limitation at this time of the year.

Sea-Ice PFe as a Capacitor for Summer Blooms
Warmer air temperatures lead to increased connectivity of the brine network forming new pathways for brine, meltwater and associated nutrient drainage toward the seawater underneath. As a consequence, DFe concentrations in sea ice collected at this time of the year are expected to be low. However, the sea-ice DFe concentrations up to 13.3 nM (Figure 3a) were comparable to those reported from other seasons (Lannuzel, et al., 2016, and references therein). DFe is therefore likely continuously supplied to sea ice either as truly soluble Fe or in the colloidal form, despite brine drainage and ice decay. Many studies have shown the importance of the PFe in replenishing the DFe pool via thermal, photochemical, biological and ligand-mediated dissolution (Borer et al., 2005;Ellwood et al., 2012;Kraemer, 2004;Sulzberger et al., 1989). These biogeochemical processes could have potentially maintained background levels of DFe (average 4.7 ± 3.8 nM and 2.0 ± 1.3 nM for fast and pack ice, respectively) at our sea ice stations. Particulate Fe is retained within sea ice longer than DFe because of the adsorption of particles to the walls of the brine channels, a process possibly facilitated by EPS van der Merwe et al., 2011). EPS are particularly abundant in waters characterized by high chlorophyll concentrations and in the sea ice (Gledhill & Buck, 2012). Although they are predominantly composed of neutral sugars, EPS contain a significant fraction of uronic acids which are known to complex Fe (Mancuso Nichols et al., 2004).
Relatively high concentrations of calculated PFe bio (∼50 nM) were observed for all pack ice stations over the entire ice thickness. Since Al is not appreciably accumulated into biogenic material, this result indicates the occurrence of biological assimilation and build-up of biogenic PFe at all ice depths. Based on the measurements of Fe concentration and sea-ice thickness from the sampled cores, if all of the sea ice represented by our samples melts by the end of austral summer (∼60 days between the sampling period and the end of February), an average of 4.6 ± 7.1 µmol m −2 d −1 PFe would be released from sea ice into East-Antarctic surface waters. This input could be particularly important during this time of the year due to the stronger ocean stratification driven by sea-ice melt and ocean surface warming, which hamper the mixing with Ferich deep waters. Pack ice formed in coastal polynyas is exposed to coastal Fe inputs and could therefore play a crucial role in delivering biologically processed (and potentially bioavailable) PFe to open waters. In this context, the contribution of summer sea ice as a source of (both biogenic and lithogenic) PFe cannot be neglected. The direct link between PFe solubility and phytoplankton growth however needs to be established, as well as the possible forward trajectories of ice floes into Fe-limited waters, before we will be able to adequately assess the effect of sea ice on the carbon cycle.

Can Sea Ice Enhance the Fertilization Potential From Glacial Fe?
Iron distribution in East Antarctic sea ice did not show significant spatial variation in this study, except for Totten 1, with DFe and PFe concentrations significantly higher than those found elsewhere (average 7.0 and 1,800 nM, respectively). The level of PFe measured in bottom sea ice at Totten 1 (∼5 μM) was only comparable to concentrations found in bottom sea ice collected in the Ross Sea (Grotti et al., 2001;Noble et al., 2013) and at a study site of very shallow (∼20 m depth) bathymetry near Casey station (van der Merwe et al., 2011). The average molar PFe:PAl (0.3 ± 0.1) obtained from Totten 1 ice is very close to values previously reported for the continental crust (0.3; Taylor & McLennan, 1985), suggesting a predominantly lithogenic source. Interestingly, this elemental ratio differs significantly (p < 0.01) from the ones measured at Casey 1 and 2 (0.5 ± 0.2), which were very close to the ratio reported for ocean floor sediments (0.48; Ravanelli, 1997). Therefore, while stations near Casey seem to be highly dependent on sediment resuspension as the main source of Fe, another source may prevail at Totten 1. The low PFe content in the snow at Totten 1 indicates atmospheric sources are unlikely. Iron must instead come from below, either by vertical or lateral transport. The most substantial geographical features in the proximity of Totten 1 are the Totten Ice Shelf and the Moscow University Ice Shelf (MUIS). We therefore hypothesize the presence of basal meltwater from the adjacent MUIS and potentially meltwater runoff from the dominant Totten Ice Shelf could explain the high Fe content observed at this station.
Seawater Fe levels have been previously reported to be significantly influenced by glacial melting around Antarctica (Death et al., 2014;Herraiz-Borreguero et al., 2016;I. Kim et al., 2015;Person et al., 2019). This effect is attributed to the fact that glacial meltwater can supply intensively physically and chemically weathered Fe-rich labile sediments (Raiswell et al., 2018 and references therein;van der Merwe et al., 2019). Glaciogenic minerals are also enriched in labile Fe 2+ nanoparticulates compared to other lithogenic sources which are mostly comprised of less bioavailable Fe 3+ oxyhydroxide and other Fe 3+ chemical-weathered oxide products Shoenfelt et al., 2017). While some of this glacial material can be incorporated into marine ice along with upwelling transport, a fraction can also be carried within the ice shelf meltwater plume and reach surface waters (Herraiz-Borreguero et al., 2016). Layers of granular and platelet ice were evident at Totten 1. Supercooled glacial meltwaters from the nearby MUIS could promote the formation of these ice structures. The abundant particles contained in these meltwaters could act as crystallization nuclei and therefore facilitate the formation of frazil ice. These crystals can then rise through the water column scavenging Fe-rich suspended sediments in shallow waters near the coast or at the grounding line of ice shelves to be incorporated into the sea ice, analogous to marine ice under an ice shelf Herraiz-Borreguero et al., 2016). The high proportion of platelet ice within the noncolumnar ice at Totten 1 strengthens the link between Fe-rich glacial meltwater and the high concentration of PFe found at the bottom of Totten 1.
Within the sea ice, Fe can be converted into more readily available forms via abiotic (photoreduction and chemical complexation) and biotic (biological and enzymatic reduction at the cell surface) processes (Genovese et al., 2018;Kaplan & Ward, 2013;Tagliabue et al., 2009). Fe dissolution reactions could be augmented by freezing processes. Repetitive freeze/ thaw cycles have been shown to continuously add DFe to the ice. Freezing concentrates water and solutes into microenvironments at grain boundaries (Jeong et al., 2020(Jeong et al., , 2012K. Kim et al., 2010). However, Fe 3+ addition to an enclosed system is limited by saturation, therefore Fe 3+ is prone to incorporation into the colloidal phase or adsorption onto ice crystals . When sea ice melts, the undersaturated meltwater can stimulate further Fe dissolution to occur. Layers of superimposed ice found in our ice-core samples suggest that melting and refreezing is common in summer sea ice in this area. Therefore, the interaction between glacially derived reactive sediments and sea ice could enhance productivity by releasing elevated concentrations of highly recycled and potentially bioavailable Fe during the melting season, when this stock becomes exhausted in surface waters (Sedwick et al., 2000). Additional offshore transport of sea ice formed in coastal areas could further enhance the extent of Fe fertilization.
Findings from Totten 1 raise the question why similar Fe enrichment in sea ice was not observed at the other stations near ice shelves that we sampled along the East Antarctic coast. Mertz 2 station was relatively close (∼1 km) to the Mertz glacier. Results from past oceanographic surveys could explain this difference. Waters on the continental shelf in East Antarctica are relatively cool, leading to low rates of basal melt from the ice shelves. The Totten and MUIS are exceptions to this general pattern as relatively warm modified Circumpolar Deep Water (mCDW) enters the cavity beneath the shelf through a deep trough, delivering sufficient heat to drive rapid basal melt (Rintoul et al., 2016). Silvano et al. (2017) showed that this mCDW is widespread over the continental shelf of the Sabrina Coast. Based on this oceanographic particularity, the Totten/MUIS glacial basin could offer the required hydrographic setting for the recurrent formation of Fe-rich platelet ice to potentially occur in large areas of the Sabrina Coast. While Fe supplied from these glaciers is not essential for the sea-ice communities, it might play an important role for phytoplankton communities when this sea ice melts along the Sabrina Coast.
Satellite Chla imagery from 2002 to 2016 supports the link between the presence of platelet ice and surface water fertilization in the surroundings of Totten 1 ( Figure 6). The summer climatology suggests that large phytoplankton biomass accumulates when the sea ice in front of the Totten Glacier breaks out. This contrasts with the rather low productivity of the neighboring Dalton Polynya that is in front of the MUIS . The outflow of glacial meltwater from the Totten Ice Shelf is located on the western side of the calving front (Silvano et al., 2018). The outflow of glacial meltwater from the MUIS is widespread along the coast and south of the Dalton polynya, with productivity highest in the latter region . To our knowledge, these observations could be the first evidence of a potential fertilization impact from the combined presence of platelet ice and Fe-rich glacial meltwaters. The Totten glacier mass loss has increased through time from 5.7 Gt y −1 in 1979-2003 to 7.3 Gt y −1 in 2003-2017; glaciers draining into MUIS show a slight net thinning in the last decades (Rignot et al., 2019). As climate change could drive the loss of up to 23% of the ice shelf volume in Antarctica by 2070 in a business-as-usual scenario (DeConto & Pollard, 2016;Rintoul et al., 2018), a better understanding of how marine and glacial sediment transport will be affected by increasing basal melting and grounding line retreat is needed.

Conclusion
The scarcity of inorganic nitrogen sources along with abundant particulate matter suggest the dominance of heterotrophic and potentially mixotrophic activity in sea ice during summer. At this time of the year, the sea ice can still retain a significant amount of Fe, both in the dissolved and, particularly, in the particulate fraction. Biotic and abiotic processes help sustain Fe supply during late summer. Projected changes in seaice thickness and snowfall could increase fluxes of macronutrients via surface flooding and brine percolation, alleviating nutrient depletion in summer, and potentially extending the period of habitability for seaice algae. Climate change might also impose indirect impacts on the sea-ice Fe distribution via reduction of glacial coverage in Antarctica. In East Antarctica, the Totten Ice Shelf is already experiencing rapid basal melt. A further increase in the supply of glacial meltwater to the East Antarctic shelf could potentially stimulate the formation of platelet ice layers in the surrounding sea ice. How phytoplankton will be affected by this reservoir of Fe will depend both on the timing of its release to surface waters and on the bioavailability of this source which is a function of the biological and physicochemical changes that Fe undergoes within the ice.

Data Availability Statement
Data supporting the analysis and conclusions here presented can be accessed through the Australian Antarctic Data Centre website (https://data.aad.gov.au/metadata/records/AAV2_seaice_BGC_2016).