A proposed seasonal cycle of dissolved iron-binding ligands in Antarctic sea ice

The Southern Ocean is anaemic (Martin et al., 1990), with dissolved iron (DFe) concentrations in surface waters rarely exceeding 0.2 nM (Boyd and Ellwood, 2010). More than 99% of this DFe pool is complexed organically (Gerringa et al., 2019; Rivaro et al., 2019; Ardiningsih et al., 2020). Organic Fe-binding ligands prevent Fe from precipitating as less soluble oxide forms (Millero et al., 1995; Johnson et al., 1997; Hassler and Schoemann, 2009; Hassler et al., 2011) or being adsorbed onto settling particles (Boyd and Ellwood, 2010; Boyd et al., 2010). Advection from the continental margin (e.g., sediments, ice shelves), aeolian deposition from ice-free continental areas (Duprat et al., 2019), upwelling, icebergs, hydrothermal vents (Tagliabue et al., 2017), as well as in-situ recycling or excretion of Fe-rich compounds by organisms (Ratnarajah et al., 2018) are among the main sources of Fe in the Southern Ocean and associated sea ice.

Sea ice starts forming in autumn, when atmospheric temperatures decrease, and seawater temperatures approach −2°C (Petrich and Eicken, 2010). Sea ice is a highly dynamic environment, where the fraction of solid ice versus liquid brine can change drastically with seasonal temperature changes (Golden et al., 1998). As a general rule, sea ice is considered permeable when the brine volume fraction (Vb/V, a proxy for sea-ice porosity) is higher than 5% (Golden et al., 1998). From autumn onwards, a thermal gradient forms within sea ice due to the low temperature of the overlying atmosphere relative to the underlying seawater (Fritsen et al., 1994). The continual exchange between bottom ice sections and underlying seawater depends on the temperature, density, and permeability of sea ice. Exchanges can occur via three main mechanisms, namely convection, brine drainage, and diffusion (Tison et al., 2008; Vancoppenolle et al., 2013). Convection may occur when there is a density imbalance between brine solution and the underlying seawater, resulting in a downward movement of denser brine counterbalanced by an upward flux of lighter seawater. Similarly, during brine drainage, brine moves downward but, in this case, it is displaced by melting and/or infiltration in the upper layers. When the density gradient between upper and lower ice is small, stratification occurs and diffusion becomes the dominant process of exchange between sea ice and the ocean.

Sea ice is enriched in DFe compared to seawater (Grotti et al., 2005; Lannuzel et al., 2011; Lannuzel et al., 2016b), with DFe being almost entirely complexed by ligands (Lannuzel et al., 2015; Genovese et al., 2018). The incorporation of Fe into sea ice starts within the first hour of growth (Janssens et al., 2016). The degree of enrichment depends on the location of the ice (land-fast versus mobile pack ice; hereafter, fast and pack ice), sea-ice porosity, and organic matter concentration (Janssens et al., 2018). Concentration and distribution of DFe fluctuate throughout the year, driven by changes in physical conditions (e.g., porosity and exchange mechanisms as above) and/or biological conditions (e.g., abundance and composition of microbial species; Lannuzel et al., 2016b and references within). Observations have shown that the development of ice-associated organisms occurs in parallel with the production of exopolymeric substances (EPS; Aslam et al., 2012; Underwood et al., 2013; Aslam et al., 2016). In the Arctic, high concentrations (4.7 x 10⁶ particles L⁻¹) of EPS have been observed at the ice-water interface, acting as hotspots for bacterial activity (Meiners et al., 2008). Likewise, copious amounts of EPS were observed in Antarctic sea ice (van der Merwe et al., 2009; Ugalde et al., 2016). EPS may complex iron, or favour bacterial remineralisation of particulate Fe into dissolved Fe (van der Merwe et al., 2009; van der Merwe et al., 2011). There is currently little information on other possible types of organic ligands in the sea-ice environment. However, observations in seawater have included humic substances and siderophores, associated with heterotrophic or photochemical degradation (Powell and Wilson-Finelli, 2003; Laglera et al., 2019; Hassler et al., 2020) and bacterial activity (Velasquez et al., 2011), respectively. In addition, inorganic types of ligands can also play a role, being supplied through hydrothermal plumes, aeolian deposition and iceberg melting (Gerringa et al., 2012; Fitzsimmons et al., 2015; Tagliabue et al., 2017; Hassler et al., 2020). Regardless of their nature, recent work by Cabanes et al. (2020) highlighted the importance of Fe-complexes in sustaining primary productivity during phytoplankton blooms. However, still open to debate is whether the uptake of Fe by marine microorganisms relies on complexed or free Fe.

Investigations focusing on the role of organic Fe-binding ligands in the ephemeral sea-ice environment are currently limited. This situation limits the current understanding of the seasonal distribution of ligands and DFe, from their incorporation into growing sea ice until their release into seawater during melting. The aim of this work was to evaluate the distribution of Fe-binding organic ligands in sea ice, and identify potential drivers (e.g., physical and/or biological). Concentrations and distributions specifically of dissolved Fe-binding ligands were compared in sea ice, collected across three different seasons (winter, spring, and summer) and from different locations around Antarctica (west and east), also considering different types of sea ice (fast and pack ice). Data from two previous studies (Lannuzel et al., 2015; Genovese et al., 2018) have also been considered when interpreting the current results.

Sampling sites are shown in Figure 1. The AWECS expedition took place in the Weddell Sea (West Antarctica, 60–70°S, 0–50°W) in winter (June 17 to July 31, 2013). In addition to collecting sea-ice cores (Stations 486, 488, 489, 496, 500), a sea-ice growth experiment was conducted during which newly formed ice was sampled at 6, 12, 24, and 48-hour time-steps from drilled holes (Stations 506 and 517) to quantify the incorporation rates of organic matter and Fe into growing sea ice (Janssens et al., 2016). Fieldwork near Davis (East Antarctica, 68°S, 78°E) took place during late spring (November 16 to December 2, 2015), where a suite of biogeochemical variables was obtained for fast ice at a single location over a 16-day period (Duprat et al., 2019). The V2 voyage took place in summer (December 8, 2016, to January 23, 2017) along the East Antarctic coast, where sea ice was collected near Casey station (C1, C2, C3), the Totten Glacier (T1, T2), Dumont d’Urville Sea (SR3-5), and Mertz Glacier (M1, M2, M3). The aim of this third voyage was to investigate the spatial distribution of biogeochemical parameters in sea ice (Duprat et al., 2020) across three distinct coastal polynyas (Moreau et al., 2019).

Cleaning, sampling and processing of the samples followed procedures reported previously in Lannuzel et al. (2006) and van der Merwe et al. (2009). Briefly, ice cores were extracted and divided into sections 10–15 cm thick (namely top, interior, and bottom ice) and were stored in acid-clean polyethylene containers. Samples were left to melt under a class-100 laminar flow hood and filtered onto 0.2-µm polycarbonate filters mounted on a PFA^© Teflon Savillex filtration system. Filters and filtration apparatus were soaked in 10% HCl (Seastar Baseline) for 1 week and rinsed with ultra-high purity (UHP) water before use. Filtrates for DFe were acidified and stored at ambient temperature, whereas filtrates for ligand analysis were transferred into acid-cleaned 250 mL LDPE bottles (Cutter et al., 2017), double-bagged and stored in the dark at −18°C until analysis. The filters from V2 were transferred into acid-clean Petri dishes, double-bagged, and stored in the dark at −18°C for the quantification of particulate EPS (PEPS) concentrations using the Alcian Blue assay (Passow and Alldredge, 1995). Companion cores were used for the determination of physical parameters (temperature, salinity, porosity) and ancillary biogeochemical parameters such as macro-nutrients (nitrate, ammonium, phosphate, and silicic acid), organic carbon and chlorophyll-a (Chl-a).

Dissolved Fe concentrations from AWECS samples were determined via isotopic dilution Inductively Coupled Plasma-Mass Spectrometry at the Université Libre of Bruxelles, adapting protocols from Sohrin et al. (2008) and Milne et al. (2010). Dissolved Fe concentrations from Davis and V2 samples were determined using an off-line SeaFAST preconcentration system followed by Sector Field Inductively Coupled Plasma-Mass Spectrometry (Wuttig et al., 2019) at the University of Tasmania. Blanks, detection limits, and quality control values are reported elsewhere (Duprat et al., 2019; Duprat et al., 2020).

Competitive Ligand Equilibration-Adsorptive Cathodic Stripping Voltammetry (CLE-AdCSV) was used for the determination of DFe speciation (concentration of natural ligands, conditional stability constant of Fe complexes K’_Fe’L, and related parameters; see Section 2.3). The approach considers the competition between natural and added artificial ligands (e.g., Gledhill and van den Berg, 1994; Monticelli et al., 2010). Samples from AWECS and Davis were titrated using 1-nitroso-2-naphthol (NN) as artificial ligand (adapted from Gledhill and van den Berg, 1994), whereas samples from V2 were titrated using both NN and salicylaldoxime (SA; Abualhaija and van den Berg, 2014). The procedures are reported and discussed in Genovese et al. (2022). Briefly, the melted samples were buffered with borate (1 M H₃BO₃/0.3 M NH₃, final concentration of 10 mM) and left to equilibrate for 10 (NN protocol) or 50 (SA protocol) minutes. A set of 10–12 acid-rinsed polypropylene vials were spiked with increasing DFe concentrations (0–20 nM for SA, 0–37 nM for NN, with at least four increments in the competition region and four increments where ligands are saturated), except for the first three which were left DFe-free. Then, the sample/buffer mixture was split among the vials and the artificial ligand was added at the final concentration of 5 µM. Vials were left to equilibrate overnight at room temperature, before analysis was performed using the instrumental conditions reported in Table S1. In the NN protocol, the cell was conditioned overnight using the same sample/buffer/ligand mixture described above. The SA protocol required pre-conditioning of the vials overnight, carried out three times (Abualhaija et al., 2015), using Antarctic seawater (DFe concentration = 0.2 nM). This conditioning was performed before analysis of the first sample, and vials analysed thereafter were considered equilibrated.

Sea-ice porosity was calculated according to Vancoppenolle et al. (2019), using sea-ice temperature and bulk salinity measurements. Conditional stability constants for titration with NN (logK’_Fe’NN3) for the Davis dataset were extrapolated from the empirical relation (based on Fe³⁺): logK’_Fe3+NN = 30.1 ± 0.09 − 1.04 ± 0.08 logS, by Gledhill and van den Berg (1994). The validity of this relation under the current experimental conditions had been previously confirmed by titration with EDTA in sea-ice samples of salinity 5.0 and 12.5, obtaining values in good agreement with those predicted (Genovese et al., 2018). A similar relationship was also found in a later work (logβ’_Fe’NN3 = 20.07 ± 0.30 − 1.01 ± 0.20 logS; Genovese et al., 2022). Data were processed according to Gerringa et al. (2014), using a non-linear fit of the Langmuir model written in R, which provided logK’_Fe’L values and concentrations of natural L. These parameters were further used to calculate the other DFe organic speciation parameters. In particular, the excess ligand concentration (L’) was calculated as the difference between C_L and DFe. The free Fe fraction (Fe’) was obtained from the equation:

The percentage of ligand-complexed Fe (FeL) was estimated as 100 × (DFe − Fe’)/DFe. The side reaction coefficient for the complexation of Fe with the natural ligands (logα’_Fe’L) was calculated as log (K’_Fe’L × C_L). IBM SPSS Statistics 24 was used for statistical analysis of the datasets.

Winter-time AWECS samples were collected as part of a study on Fe incorporation processes into newly formed pack ice (Janssens et al., 2016). From AWECS, only four stations (young sea ice at Stations 486, 500A and 500B, and in-situ ice growth experiment at 517) could be considered for the study of Fe-binding organic ligands in this work, as no samples were collected for determination of C_L for the rest of the expedition. Table 1 summarises the main physical and Dfe organic speciation data. Briefly, the sea ice sampled was unconsolidated, and its thinness (average ± standard deviation of 0.08 ± 0.04 m) was indicative of the first stages of ice growth. During the survey (June 17 to July 31) the general decrease in sea-ice temperature from −5.0 to −9.9°C was associated with an increase in brine salinity from 58 to 155 (average of 102 ± 36) and a decrease in sea-ice porosity from 28 to 6% (average of 15.8 ± 9.3%). Macronutrients generally followed the theoretical dilution line, indicating low biological activity. Sea ice was enriched in organic carbon, with a mean particulate organic carbon to particulate organic nitrogen (POC:PON) ratio of 12.5 ± 5.1 (n = 9), compared to seawater, where the ratio was below the typical Redfield value of 6.6 (Redfield, 1963). Concentrations of EPS in sea ice (0.54–0.98 µg Xeq L⁻¹) were up to 1.5-fold higher than in the underlying seawater (0.66 µg Xeq L⁻¹). Chl-a concentrations in sea ice varied from 0.04 to 0.52 µg L⁻¹, and were lower than 0.02 µg L⁻¹ in the underlying seawater. Bacterial biomass was comparable between sea ice (0.95–2.22 x 10⁵ cells mL⁻¹) and seawater (1.96 x 10⁵ cells mL⁻¹). Further information is reported in Janssens et al. (2016).

Concentrations of Dfe and C_L in sea ice were 1.89 ± 0.87 nM (0.66–3.47 nM) and 6.07 ± 1.84 nM (3.27–7.98 nM), respectively. Ligands were always in excess with respect to Dfe, with an average ratio of ligands to Dfe (L/Dfe) of 4.01 ± 3.01 and no significant correlation between the two variables (p > 0.05, n = 9). Average logK’_Fe’L and logα_Fe’L were 11.48 ± 0.31 and 3.25 ± 0.29, respectively. Overall, more than 98% of Dfe was complexed. These values were obtained using logα’_Fe’NN3 = 3.6 ± 0.3.

All physical and biogeochemical datasets for Davis samples are available in Table S2. The Dfe speciation data for the Davis time series are reported in Table S3 and summarized in Table 2, along with the main physical data. Spring fast ice at Davis showed an average thickness of 1.71 ± 0.89 m. Ice temperatures increased with depth and over time (overall average of −2.89 ± 1.2ºC, ranging from −5.6ºC to −1.2ºC). Bulk salinity was lower in the top and interior layers, and increased towards the bottom sea ice (Lim et al., 2019). Overall, Davis sea ice was porous (average Vb/V = 12.5 ± 9.7%; Figure 2a). Maximum Chl-a concentrations were recorded in bottom ice, with the highest value measured on the first day of the time series (Station 1, 267 µg L⁻¹). Macronutrients were most concentrated in bottom ice. Seawater macronutrient concentrations were lower than bottom ice values, with the only exception being silica (Lim et al., 2019). Dissolved Fe was generally higher in bottom ice, except for Station 1 (sub-surface peak at 0.3 m). Detailed biogeochemical data and discussion can be found in Duprat et al. (2019). Dissolved Fe concentration and C_L showed very similar distributions (Figure 2b), with generally higher values in bottom ice than in the other ice sections, except for Station 1 (maximum concentration at 0.3 m). Average C_L was 13.1 ± 15.3 nM (ranging from 1.7 nM to 74.6 nM), generally in excess relative to Dfe (10.7 ± 15.8 nM, ranging from 1.0 nM to 75.5 nM). A significant correlation was observed between Dfe and C_L (Pearson’s r = 0.99, p < 0.01, n = 42), with an average L/Dfe of 1.7 ± 0.8. More than 99% of Dfe was complexed, with an average logK’_Fe’L of 12.2 ± 0.3. These values were obtained using logα’_Fe’NN3 = 3.6 ± 0.3.

All physical and biogeochemical datasets for V2 samples are reported in Table S4. The Dfe speciation data for V2 samples are reported in Table S5 and summarized in Table 2, along with the main physical data. Average sea-ice thickness for V2 was 1.44 ± 0.36 m. Ice temperatures at the snow-atmosphere interface were ≥−1ºC, and decreased with depth. Bulk salinity increased with depth, while porosity was well above the 5% threshold for some percolation onset, and even exceeded 40% in a few sea-ice surface sections (average of 15.0 ± 8.6%; Figure 2c). Maximum Chl-a concentrations were recorded at different depths along the ice cores, with the highest concentration (130.6 µg L⁻¹) observed at Casey 2 (at 1.29 m, bottom ice; Table S4). Concentrations of macronutrients in sea ice were generally low (≤10 µM for nitrate, ≤5 µM for phosphate, ≤3 µM for ammonia) in summer sea ice. On the other hand, high concentrations of PEPS (up to 16,290 µg Xeq L⁻¹) were found in most samples (Duprat et al., 2020).

Dissolved Fe concentrations were higher in coastal ice (i.e., Totten 1 and Casey 2 stations, with values up to 10.7 nM and 13.3 nM, respectively), compared to the off-shore stations (ranging from 0.6 to 4.7 nM; Table2). Iron speciation for V2 was obtained over the course of two separate analytical sessions, using NN for samples with high Dfe concentration (Table S5). To evaluate the effect of the artificial ligand on the speciation results, eight samples were analysed with both ligands (Table S6). Ligand concentrations and logK’_Fe’L values differed significantly (p = 0.019 for C_L, and p = 0.002 for logK’_Fe’L), with higher values noted for NN, with a shift towards SA of 2.7 ± 5.4 nM and 0.65 ± 1.7 unit, respectively. Therefore, data summarized in Table 2 account for the analytical variability due to the different titration methods. Overall, the average C_L for all sea-ice samples was 5.1 ± 4.7 nM (range of 1.0−21.9 nM), with vertical profiles characterized by a patchy distribution (Figure 2d). A significant correlation was found between Dfe and C_L (Pearson’s r = 0.83, p < 0.01), and the average logK’_Fe’L value was 11.8 ± 0.7. The 56 sea-ice samples showed an average FeL of 98.8 ± 2.8% and an average L/Dfe of 3.3 ± 0.9.

This study offers both a spatial (Section 4.2.1) and temporal (Section 4.2.2) analysis of ligand concentrations and distribution in Antarctic sea ice. This analysis comes with the important caveats that the samples were collected across different years and from distinct regions around the continent (Figure 1), where the sampling sites differed with respect to oceanographic features (e.g., proximity to the coastline, bathymetry, ice type, presence of ice shelves). As there is a paucity of published data with respect to DFe complexation in sea ice, this work provides basic information to move in the direction of more robust spatial and temporal analyses in future.

From an analytical chemistry point of view, two methods were used, based on an initial assumption that the ligand results could be compared directly. Other studies, however, have reported a disagreement between results when using different artificial ligands (Slagter et al., 2019), as well as due to experimental conditions and data interpretation (Buck et al., 2012). The two ligand classification methods currently available lack accuracy for establishing a clear threshold between ligand classes. In the water column, Gledhill and Buck (2012) differentiated a stronger ligand pool in the upper ocean, from a weaker ligand pool in deeper waters. Hassler et al. (2017) defined the possible classes of ligands based on logK’_Fe’L. The latter method, although more specific to the nature of the compounds, can lead to data misinterpretation, as differences in logK’_Fe’L are very small and results can overlap across different classes. Recent work by Gerringa et al. (2021) questions the use of logK’_Fe’L as an indicator for the identification of a specific class of ligands, as the application of two different artificial ligands resulted in large differences in the logK’_Fe’L obtained. In this study, poor agreement was also observed between logK’_Fe’L and L values for the eight V2 samples titrated with both SA and NN (Table S6). Ligand concentrations and logK’_Fe’L values differed significantly (p = 0.019 for C_L, and p = 0.002 for logK’_Fe’L), with higher values noted for NN than SA. This difference can be ascribed to uncertainties arising from the CLE-AdCSV approach with different ALs, as discussed by Gerringa et al. (2021). Lastly, computations of C_L and logK’_Fe’L in a few samples containing very high DFe concentrations collected at Davis (e.g., Station 4, ice sections 5 and 7, with DFe up to 75.5 nM) was not achieved. Interpretation of CLE-AdCSV data is based on the assumption that all of the DFe in a given sample is exchangeable (Gledhill and Buck, 2012). However, natural environments offer inorganic or refractory types of ligands, which are less prone to exchange DFe with the added artificial ligand (Thuróczy et al., 2010), interfering with data processing and further interpretation.

Ligands are ubiquitous in the marine environment as they fulfill different eco-physiological functions (Gledhill and Buck, 2012). A study by Caprara et al. (2016) compiled seawater DFe organic speciation datasets over the period 1994–2015, showing that the ligand pool (mostly <5 nM) generally exceeds the DFe pool (mostly <1 nM). Data presented in this sea-ice work shows that both DFe and C_L in sea ice are 1–2 orders of magnitude higher than in seawater from open oceanic locations including polar oceans (Figure 3). In addition, the almost complete complexation of DFe is a persistent feature of sea ice, regardless of the DFe concentration range (0.6–80 nM). Most of the sea-ice samples (61%, n = 171) displayed C_L lower than 10 nM (Figure 4). When these C_L values were divided into four ranges (0–10 nM, 10–20 nM, 20–30 nM, and >30 nM), each range contributed differently to the proposed year-round evolution of ligands in Antarctic sea ice: winter and summer ice values fell within the lowest concentration range, whereas spring was best represented in the upper three ranges. In the next sections, we address potential drivers of this apparent ligand distribution, considering changes in the physical features of sea ice as well as the phenology of microorganisms.

To reduce any regional and seasonal effects, the examination of spatial variability in Fe-binding ligands was applied to the Davis spring dataset, alongside two published springtime datasets, namely SIPEX-2 (Genovese et al., 2018) and Casey (Lannuzel et al., 2015). Generally, a wider range of C_L was observed in fast ice (1.7–74.6 nM, median of 9.3 nM, n = 76), compared to pack ice (4.9–41.1 nM, median of 12.6 nM, n = 34), in accordance with the latitudinal gradient in DFe concentrations (1.0–81.0 nM for fast ice, and 1.5–17.4 nM for pack ice). Higher concentrations of compounds acting as Fe-binding ligands in sea ice in coastal areas can be attributed to sediment resuspension from shallow waters and advection from the continental margin (Gerringa et al., 2015), as well as to post-bloom photodegradation of organic matter (Barbeau, 2006; Hassler et al., 2020). The proximity of the coast can also facilitate the incorporation of lithogenic Fe while ice is growing, or the subsequent addition of Fe through aeolian fluxes from ice-free areas on the Antarctic continent, as observed at Davis (Duprat et al., 2019). This aeolian Fe would have entered the snow layer and percolated through the brine network, as sea-ice permeability rose above the 5% threshold (Golden et al., 1998; Golden et al., 2007), reaching interior and bottom sea-ice sections where higher concentrations of DFe and ligands were observed. Nevertheless, the increasing DFe concentration when approaching the coastline (Lannuzel et al., 2016b) is believed to saturate the ligands, as noted in the significantly lower L/DFe ratio observed in fast ice stations (paired t-test, significance level of 0.05) compared to pack ice. Therefore, even if coastal areas are enriched in organic compounds, potentially acting as DFe ligands, the L/DFe sits close to the 1:1 ratio. Conversely, the significantly higher L/DFe ratio noted in pack ice stations could result from the lower DFe concentration in offshore waters, external input of Fe-free ligands, or in-situ production/recycle of ligands. Focusing on the latter, remineralization within interior sea-ice sections would provide fresh inorganic and organic compounds (Fripiat et al., 2014; Roukaerts et al., 2016). In fact, the distribution of biogeochemical parameters is highly variable in pack ice, far from being strongly localized in bottom sea ice, as is the case for fast ice. Therefore, bottom-ice-associated organisms have limited access to the inorganic and organic material temporarily trapped in the interior ice sections; only when full-depth permeability is reached, which occurs approximately during the spring–summer transition (van der Merwe et al., 2011), does the L/DFe ratio become comparable to fast ice values. We therefore suggest that the spatial difference in the L/DFe ratios in spring datasets was mostly related to the permeability of the cores at the time of sampling and to the ice type and the microbial dynamics occurring within the ice.

The highest C_L across all samples and stations considered in this study were recorded between early and late spring, which generally corresponds to the bloom of ice algae (Meiners et al., 2012). The only time L correlated with Chl-a (Pearson’s r = 0.632, p = 0.0024) was during springtime (Davis dataset). Based on the seasonal concentrations of L and of Chl-a (Figure 5; values in Table S7), we suggest that the L inventory follows four sequential events: 1) physical incorporation during sea-ice formation (winter, AWECS); 2) production of L by heterotrophs via remineralization of the seawater-sourced organic material (early spring, SIPEX-2); 3) large inventory of L due to autotrophic production (spring/late spring, Casey and Davis) and aeolian Fe input (as in the case of Davis); and 4) release of L into the seawater when sea-ice melting is advanced (late spring/summer, Davis and V2).

Despite the short period between sampling days, young sea ice collected during AWECS was already enriched in both particulate and dissolved organic matter (Janssens et al., 2016). Ligands followed this enrichment, with sea-ice concentrations above those generally found in seawater (<5 nM, according to Caprara et al., 2016). Janssens et al. (2016) found EPS in young winter ice collected at AWECS, with concentrations slightly higher than those measured in seawater. In wintertime, EPS are excreted by sea-ice microbes (Meiners et al., 2004; Aslam et al., 2012; Underwood et al., 2013; Aslam et al., 2016), to overcome the large gradients of temperature, pH, and salinity experienced in growing sea ice (Krembs et al., 2002). As the ice was quite young (i.e., up to 48 hours old), these EPS were likely directly sourced from seawater rather than produced in situ. Thicker Antarctic winter ice (i.e., 0.5–0.8 m; Meiners et al., 2004) showed EPS concentrations an order of magnitude higher than under-ice seawater, which is consistent with the 5 nM reported as the initial concentration in Figure 5. Ligand inventory can build with time, either through continuous incorporation into growing sea ice, or derived from biological activity. During the pre-bloom stage, the sea-ice microbial food web is sustained by the recycling of detritus within the interior ice layers (Roukaerts et al., 2016). Heterotrophic organism products are likely to be the first seasonal source of L, in the form of in-situ produced/recycled cryoprotectants and osmotic regulation agents (Krembs et al., 2002; Meiners et al., 2004).

Ligand concentrations can increase rapidly, as biological growth accelerates in early springtime (Meiners et al., 2018). Early spring ice already contains EPS (Rivaro et al., 2021), and an increase in EPS concentrations is typically associated with prolonged algal growth (Aslam et al., 2016). This parallel growth is observed in Figure 5, with ligands peaking (Julian day 314) a few days before Chl-a (Julian day 318). In many marine settings EPS can accumulate and organize into three-dimensional sticky polysaccharidic matrices, known as biofilms (Decho and Gutierrez, 2017), which are also proposed to occur in sea ice (Roukaerts et al., 2021). Biofilms present good habitats for microorganisms and may explain the co-existence of high algal standing stocks and high nutrient concentrations observed at Davis (Roukaerts et al., 2021). Biofilms include EPS physically attached to the microbial cell and unbound EPS of low molecular weight, mainly in colloidal form (Underwood and Paterson, 2003). EPS retain nutrients within their chemical structure, via hydrophilic and hydrophobic bonds (Flemming and Wingender, 2001), and the negative charges of the EPS structure could help bind Fe (Mancuso Nichols et al., 2005). Because 75% of DFe at Davis was in the colloidal phase (Duprat et al., 2019), these L may act concomitantly as DFe-retaining agents and components of the biofilm matrix. Ligands were most concentrated in bottom ice (Figure 2b) and significantly correlated with Chl-a (Pearson’s r = 0.632, p = 0.0024), suggesting in-situ production by ice algae and/or by algal-associated bacteria. The occurrence of an active microbial network could also suggest the presence of siderophores as Fe-binding ligands (Boiteau et al., 2016; Baars et al., 2017). Siderophores have been reported to occur at picomolar levels in Pacific seawater (Boiteau et al., 2016; Repeta et al., 2016), which is a thousand times lower than the minimum L concentration in sea-ice samples collected at Davis. We thus suggest that siderophores constituted only a small fraction of the ligand pool in this study, while also acknowledging the (untested) possibilities that siderophores might be present in higher amounts in Antarctic waters and become concentrated in sea ice during ice formation. The potential balancing effect of their high affinity for DFe also remains an open question. The highest DFe concentrations in sea ice at Davis were recorded after a strong aeolian event (up to 75.5 nM), suggesting that a plausible majority of refractory and/or inorganic ligands could have directly or indirectly contributed to the spring ligand inventory. Although the number of sea-ice sections showing C_L < DFe concentration increased slightly after the blizzard, an overall excess of ligands relative to DFe was still observed. This excess could be explained by in-situ production of ligands, as consumption or loss of DFe from the sea ice would be associated with loss of the related complex (FeL > 99%).

The post-bloom condition and increasing temperature resulted in a general decrease of Chl-a concentration and C_L (Figure 5). Ligands can be released into seawater, as components of the organic matter that fluxes from melting sea ice (Lannuzel et al., 2015; Genovese et al., 2018), and potentially bind to Fe present in the water column or aggregate and sink towards the seafloor, thereby contributing to the biological carbon pump. The advanced stage of melting noted in early summer also decreases opportunity for sea-ice organisms to adhere to the ice while increasing osmotic challenges within the changing brine network (Ewert and Deming, 2013). Exceptionally high PEPS concentrations have been detected in summer sea ice (113–16,290 µg Xeq L⁻¹), at much higher levels than those found in previous seasons (van der Merwe et al., 2009), as well as in seawater (83–727 µg Xeq L⁻¹; Duprat et al., 2020). These PEPS could be the response of sea-ice algae and bacteria to the physical changes inherent to melting sea ice. If first depolymerized into dissolved EPS (Lelchat et al., 2019), PEPS might also constitute a substrate for heterotrophic microbial activity. Overall, however, the concentration of ligands evidently decreases at this time of the year (Figure 5). We conclude that full-depth permeability and porosity likely determine the loss of most of the dissolved and particulate matter (whether produced in the ice or sourced externally), and thus that thermodynamic and physical processes regulate the ending stage of the ligand cycle in sea ice, returning C_L to background concentrations (5 nM).

Brine convection operates between sea ice and seawater, as long as the sea ice is permeable enough and brine salinity remains higher than seawater salinity (Tison et al., 2008). Fluxes have been estimated previously using one month as the timeframe and assuming that material is released from melting sea ice gradually. These studies have shown that sea ice is a large source of DFe and ligands during spring (Lannuzel et al., 2015; Genovese et al., 2018). However, field and laboratory studies have shown that dissolved material is released together with salts before the release of particles and low salinity meltwaters (Lannuzel et al., 2013). This loss via convection occurs relatively quickly: around 70% of the initial DFe inventory was released within 10 days in fast ice (Duprat et al., 2019) and pack ice (Lannuzel et al., 2008). Convection was partially underway when samples were collected during SIPEX-2, as inferred from sea-ice porosity. Similarly, convection had also probably started at Casey, given the enrichment of seawater with respect to DFe (2.5 ± 0.6 nM, n = 24) and L (6.9 ± 1.1 nM, n = 24) concentrations (Lannuzel et al., 2015). Convection had likely concluded when Davis cores were sampled, as brine salinity in bottom sections (20 cm) was lower (average of 29.5 ± 4.4, n = 6) than seawater salinity. Brine salinity during V2 was much lower than seawater salinity, indicating that convection had ceased before the sampling. These surveys can thus depict the different stages of vertical transport of brine solution year-round. The convection-driven release of sea-ice material is likely to have drained a considerable amount of organically complexed DFe from the ice (Lannuzel et al., 2008). The time covered by the voyages considered in this work was about 106 days (September 29 to January 13), which sits within the range of 20–200 days for residence time of released DFe in seawater (Ussher et al., 2004). We thus infer that the DFe released from sea ice melting, from the beginning of the convection to the almost complete melting of the floe, could remain at the surface and support the growth of phytoplankton in surface seawater. In order to understand the contribution of melting sea ice to the underlying seawater and phytoplankton communities, the timing and residence time of this release must be considered as equally important as the Fe inventory and Fe bioavailability. This flux is especially important because the magnitude of phytoplankton growth depends on the time elapsed between Fe fertilization and uptake (Hannon et al., 2001).

This work provides a first attempt at an inter-seasonal study, proposing potential inventories, drivers, and distributions of Fe-binding ligands in Antarctic sea ice. A strong relationship was found between C_L and DFe concentration in Antarctic sea ice throughout the year, although the sources of ligands and DFe likely differed between seasons. In the first phase of sea-ice formation and growth, during the transition from autumn to winter, physical incorporation would drive the enrichment of ligands in the sea ice. During spring, Fe-ligands were likely associated with the development of sympagic organisms, particularly in the bottom ice, and potentially supplemented later with dust-derived ligands. The inaccuracy of the reported logK’_Fe’L values suggests that either the analytical technique adopted is not able to discriminate between the inorganic and organic nature of the complexes, or that the equilibrium was influenced by more refractory compounds, less prone to Fe exchange. The summer dataset was characterized by a higher degree of algal decay, with the subsistence of sympagic organisms possibly aided by production of a biofilm, as proposed in other sea-ice studies. An EPS-based biofilm would have retained DFe within the brine system in support of both heterotrophs and autotrophs. Given the importance of microorganisms for the trophic regime of sea ice, future fieldwork should consider the sampling of ancillary parameters (i.e., dissolved and particulate inorganic and organic matter; bacteria and viruses), beyond the standard physico-chemical variables, allowing for a better interpretation of the DFe speciation data.

Data are available as online supporting information and through the University of Tasmania Open Access Repository website (http://data.utas.edu.au/metadata/ed349d22-e126-4c48-b956-c769a09080e0). Any other details of interest will be made available from the authors.

The supplemental files for this article can be found as follows:

• Table S1. Instrumental conditions for the titration of natural samples.

• Table S2. Physical and biogeochemical data of sea-ice samples collected at Davis (2015).

• Table S3. Iron speciation data for sea-ice, seawater, and snow samples collected at Davis (2015).

• Table S4. Physical and biogeochemical data of sea-ice samples collected during V2 (2016/2017).

• Table S5. Iron speciation data for sea-ice samples collected during V2 (2016/2017).

• Table S6. Iron speciation data for eight V2 sea-ice samples analysed with both NN and SA.

• Table S7. Compilation of ligand and chlorophyll-a (Chl-a) concentration data.

Authors would like to thank officers and crew of RV Polarstern for their logistical support and sampling of AWECS samples in 2013; officers and crew of RV Aurora Australis for their logistic support during Davis voyage in 2015 and V2 voyage in 2016/17; and Dr. J. De Jong for the analysis of iron in sea-ice samples from the AWECS expedition. This manuscript is a contribution to Biogeochemical Exchange Processes at Sea Ice Interfaces (BEPSII 2).

This work was co-funded by the Australian Government Cooperative Research Centre Program through the Antarctic Climate and Ecosystems (ACE CRC), the Australian Antarctic Science (AAS) project no. 4291. Access to ICP-MS instrumentation at the University of Tasmania was supported through the Australian Research Council LIEF program (LE0989539). CG was supported by the Australian Research Council's Special Research Initiative for Antarctic Gateway Partnership (Project ID SR140300001).

Authors declare the submitted work was carried out with no personal, professional, or financial relationships that could potentially be construed as a conflict of interest.

• Contributed to conception and design: CG, DL.

• Contributed to acquisition of data: CG.

• Contributed to analysis and interpretation of data: CG, MG, FA.

• Drafted and revised the article: CG, MG, FA, MJC, LPD, KW, ATT, DL.

• Approved the submitted version for publication: CG, MG, FA, MJC, LPD, KW, ATT, DL.

How to cite this article: Genovese, C, Grotti, M, Ardini, F, Corkill, MJ, Duprat, LP, Wuttig, K, Townsend, AT, Lannuzel, D. 2022. A proposed seasonal cycle of dissolved iron-binding ligands in Antarctic sea ice. Elementa: Science of the Anthropocene 10(1). DOI: https://doi.org/10.1525/elementa.2021.00030

Domain Editor-in-Chief: Jody W. Deming, University of Washington, Seattle, WA, USA

Associate Editor: Jeff S. Bowman, Integrative Oceanography Division, Scripps Institution of Oceanography, La Jolla, CA, USA

Knowledge Domain: Ocean Science

Part of an Elementa Special Feature: New Insights into Biogeochemical Exchange Processes at Sea Ice Interfaces (BEPSII-2)