Sea ice samples were collected from 24 stations during nine voyages conducted between January 2016 and March 2020 (Fig. 1; Table 1). The collected sea ice was brash ice (drift ice <2 m in diameter) originating from first-year ice/older ice (FYI/OI) and newly formed ice (pancake ice and nilas). On the training vessel Umitaka-maru (Tokyo University of Marine Science and Technology, UM15-08, UM16-08, UM18-08, UM19-08), drift ice was scooped up by the crane using a plankton net frame covered with cloth (Takahashi et al., 2022). On the icebreaker Shirase (JARE57, JARE61) and the research vessel Hakuho-maru (KH19-01, KH20-1), samples were collected manually and randomly using steel baskets and fishing nets (Takahashi et al., 2023, 2025), respectively. For UM18-08, UM19-09, JARE61, and KH19-01 sea-ice salinity, nutrients and Chl a concentration of some samples are published in the previous studies (Takahashi et al., 2022, 2023, 2025; Nomura et al., 2023). Along with these datasets, we compiled unpublished data (e.g., POC, PN, and stable isotope ratios etc.) and archived in the Arctic and Antarctic Data Archive System (ADS). The collected sea ice was immediately packed into polyethylene bags and stored in the dark at -20 °C (Miller et al., 2015) until analysis for not longer than 12 months. Seawater was sampled almost twice daily in January 2019 (UM18-08 cruise) using a continuous underway pump (5 m depth) aboard Umitaka-maru. For POC/PN concentrations and δ13C and δ15N, a total of 1.0 to 2.3 L of seawater was filtered through the pre-combusted Whatman GF/F filters (450 °C, 4 h). The filters were packed into cryovials and transported to the laboratory in liquid nitrogen.

After measuring the total weight of sea ice, the ice blocks were roughly crushed with an ice pick to homogenize them, then randomly divided into 200–800 g portions (excluding UM15-08 and UM16-08). This ice was melted in 9 times the volume of 0.2 µm-filtered seawater at 4 °C in the dark, and the meltwater was filtered through Whatman GF/F filters to measure Chl a concentration. The filters were soaked in N,N-Dimethylformamide (Suzuki and Ishimaru, 1990) and stored at -20 °C in the dark. The remaining sea ice was then melted in polyethylene bags at room temperature in the dark (Roukaerts et al., 2019). Meltwater was filtered through pre-combusted Whatman GF/F filters and stored in a deep freezer. The subsequent drying and fumigation followed the procedure for seawater POM filters. Salinity was measured from a portion of the meltwater collected for nutrient analysis, which was then stored in a deep freezer (-80 °C). Sea ice volume was calculated from salinity, meltwater temperature (+4 °C), and weight of sea ice.

The pigments on filters were extracted with N,N-Dimethylformamide for at least 24 h and Chl a concentrations were measured using a pre-calibrated fluorometer (Welschmeyer, 1994). For sea-ice samples from Sts. UM15-08 and UM16-08, all Chl a samples were collected from meltwater without adding filtered seawater. However, the effect of osmotic changes on bulk Chl a concentration is considered insignificant (Dieckmann et al., 1998; Meiners et al., 2012).

The POM filters were dried at 60 °C for 24 h, fumed with 1 mol L⁻¹ hydrochloric acid (HCl) to remove particulate inorganic carbon, and stored in a desiccator. The filter samples were packed into tin capsules and analyzed for POC and PN concentrations and stable isotope ratios (δ13C, δ15N) using the isotope ratio mass spectrometry (Flash 2000-ConfoIV-DeltaV advantage, Thermo Scientific, USA; Europa PPZ GEO 20-20). δ13C and δ15N were calculated using the following equations based on the differential ratios relative to standard materials. 1δx=RsampleRstandard-1×103 where x is either ¹³C or ¹⁵N and Rsample and Rstandard are the isotope ratios (13C/12C and 15N/14N) of the samples and standard materials, respectively.

Nutrient concentrations (NO3-, PO43- Si(OH)₄) in sea ice collected by Umitaka-maru from 2016 to 2019 (UM15-08, UM16-08, UM18-08) and sea ice collected by Shirase in 2016–2017 and 2018–2019 (JARE57, JARE59) were measured using the methods described in Shimada et al. (2020). Sea ice from the Hakuho-maru voyages (KH19-01 and KH20-01) and the 2019–2020 Shirase voyage (JARE61) were measured in accordance with JGOFS (1994). To visually analyze the consumption and remineralization of each nutrient (NO3-, PO43-, Si(OH)₄) since sea ice formation, comparisons with the Theoretical Dilution Line (TDL) were conducted. The mean nutrient concentrations and salinity (34.3) from the summer minimum temperature layer, measured in the eastern Antarctic SIZ, were cited from Nomura et al. (2023) as initial (autumn to winter) seawater values. Bulk sea ice nutrient concentrations depend on brine discharge, sea ice freezing and melting, and snow accumulation, which could dilute bulk concentration. To isolate the effects of biological accumulation and consumption processes on the other components (e.g., POC concentration), the salinity-normalized nutrient concentrations (*NO3-, *PO43-, *Si(OH)₄) were calculated (Fripiat et al., 2017), which were expressed as: 2*C=C×34.3S Here, C represents the bulk nutrient concentration, and S represents the salinity of sea ice. It should be noted that this method focuses on the seasonal change in nutrient concentration (i.e., changes since sea-ice formation) and does not completely rule out the effect of brine convection or exchange with seawater which could dilute/supply nutrient in sea ice.

The conversion of algae in sea ice to carbon was estimated from the POC : Chl a ratio reported in previous studies (Becquevort et al., 2009; Meiners et al., 2011; Takahashi et al., 2023). The POC : Chl a ratio for isolated cultured strains has been reported to be 25–96 for the Antarctic diatom Fragilariopsis cylindrus and 41–284 for the haptophyte Phaeocystis antarctica (Arrigo et al., 2010). However, these values are highly dependent on culture light conditions and are unlikely to reflect sea ice in natural environments inhabited by diverse communities. Therefore, this study used the ratio of ice-algal carbon (IAC) to chlorophyll a concentration, estimated from field observations. Previously, the IAC : Chl a ratios for Antarctic ice algae (based on linear regression) were reported to range from 32.5 to 91.7 (Becquevort et al., 2009; Meiners et al., 2011; Takahashi et al., 2023). We calculated ice-algal carbon content (POC_algae, µgCL-1) by multiplying sea-ice Chl a concentration by the mean IAC : Chl a ratio (54.4±26.5). 3POCalgae=Chla×F Here, F is the mean IAC : Chl a ratio. Additionally, by dividing POC_algae by the POC concentration, we estimated the proportion of algae-derived carbon (%). This ratio still has uncertainty which affects our estimation of POC_algae because of the variabilities among studies (Fig. S1 in the Supplement). Natural ice-algal assemblages change their cellular carbon or Chl a content under the alleviated light condition (Gosselin et al., 1990).

To reconstruct the trajectories of sea ice after its formation, we estimated sea ice motion based on satellite observations (Kimura, 2004). Specifically, backward trajectories of virtual particles were calculated using daily gridded sea ice drift velocity data with a spatial resolution of 60 km. Each particle was advected backward in time following the ice motion field. When a particle reached a grid cell where no sea ice was present, that location was regarded as the point of initial sea ice formation for the particle (Corkill et al., 2023, etc.). However, it remains unclear whether sea ice was formed at this specific point. In this study, we estimated daily sea-ice production along the sea-ice trajectory to better identify its origin. First, the thermal ice thickness was estimated following Nakata et al. (2019) using microwave data from AMSR2 on board GCOM-W. The daily sea-ice production (m d⁻¹, 6.25 km grid resolution) was calculated by the heat budget analyses (Nakata et al., 2021) using ECMWF ERA5 reanalysis products (Hersbach et al., 2018, 2020) and the estimated thermal ice thickness. We extracted sea ice production only along the trajectory. The first date of sea-ice production is defined as the first time sea ice was produced (>0.00 m d⁻¹) across all available trajectory data, and the elapsed days from sea-ice formation to ice sampling (DSF) are calculated. Because sea-ice motion was estimated using cross-correlations of pixels in a 60 km × 60 km window, the trajectories and DSF data are missing in Sts. KH19-1 and ROV which lacked valid sea-ice data around the sampling site.

To estimate POC and algae-derived POC (Sect. 2.6) supply from sea ice, we determined the average POC and Chl a concentrations within sea ice during the melt season. Because this study randomly sampled broken ice, it included sea ice of various sizes. To estimate the POC concentration in meltwater when these sea-ice samples melt, it is necessary to account for differences in ice piece size (melt volume). Using the POC concentration of all sea ice (excluding newly formed ice) and the volume calculated from sea ice weight, we calculated the volume-weighted mean (VWM) POC and Chl a concentrations, following Takahashi et al. (2022). 4POCVWM=∑i=1nai×bi∑i=1nbi5ChlVWM=∑i=1nci×bi∑i=1nbi Here, n is the total number of ice samples, and i indexes each ice sample. ai and ci represent the concentrations of POC or Chl a (µgL-1), respectively, in the sea-ice sample number i, and bi is the volume (L) of the sample. Chl_VWM was converted to algal-derived POC by multiplying it by an IAC : Chl a ratio of 54.4±26.5 (Sect. 2.6).

We adopted the VWM method to determine POC content in meltwater in order to avoid biases when a mathematical mean was used. In extreme case, release of POC into seawater from quite small sea ice (sample volume: 0.4 L) with high POC concentration (5,000 µgL-1) is equivalent to that from larger ice (10 L) with low POC concentration (200 µgL-1) if both of them were melted (POC_VWM: 2000 µgL-1). In this situation, using a simple mean value (2600 µgL-1) would make the contribution of POC concentration in larger-sized sea ice minor and the other way around. Although the VWM method cannot calculate the variance SD, it can correct for such bias when dealing with the samples with different size or concentrations.

Next, we estimated the freshwater flux from sea ice melt (unit: sea ice melt thickness [m]). The freshwater flux was calculated using climate values from Komatsu et al. (2025) and averaged over the study area (20–160° E). The sea ice meltwater flux was multiplied by the POC_VWM concentration in sea ice to estimate the POC flux from sea ice to the water column per unit area (mg C m⁻² yr⁻¹). 6POCInput=Hi×ρi×POCVWM Here, Hi is ice melt thickness and ρi is the density of sea ice (0.90) used in Komatsu et al. (2025). To estimate the contribution of the POC flux from sea ice to primary production in the water column, we cited the means (± standard deviation: SD) of SIZ area (difference of ice extent in September and February) during 1979–2019 (Parkinson, 2019) and the primary production for the summer MIZ and MIZ shelves during 1997–2006 (Arrigo et al., 2008) in the Indian sector of the Southern Ocean. The SD for the contribution of sea-ice POC to the production in seawater was calculated by propagation of error using SD of POC input and primary production in the MIZ.

To identify factors influencing variations in sea-ice POC, PN concentrations, and POM stable isotope ratios (δ13C, δ15N), we investigated correlations between these data and salinity, Chl a, salinity-normalized nutrient concentrations, the Si(OH)₄ : NO3- ratio, the NO3-:PO43- ratio, and DSF. For each dataset, Q-Q plots were first used to assess whether parametric tests were appropriate. Since linear regression could not be applied to some parameter sets, a nonparametric Spearman's rank correlation test was used for all correlations, except the POC : PN ratio, to assess the presence of significant relationships and to calculate correlation coefficients. For POC, PN concentrations, δ13C, and δ15N in the four areas (LB, CD, VB, SC; Fig. 1), the datasets were non-normal according to the Kolmogorov–Smirnov test. Therefore, the Kruskal-Wallis test and Steel-Dwass tests were used to confirm differences in medians across areas. All statistical analyses were performed using the R software (https://www.r-project.org/, last access: 4 January 2026).

The POC concentrations in sea ice ranged 246.1–6365.3 µgL-1 and showed a linear relationship with PN (14.4–725.4 µgL-1) (Fig. 2a). The POC : PN ratio (g : g) ranged from 4.5 to 41.9, with an average of 13.4, and the ratio obtained from linear regression was 7.6. Sea-ice POM δ13C and δ15N ranged from -35.1 ‰ to -11.7 ‰ and -3.8 ‰ to +11.1 ‰, respectively, showing a significant positive correlation with each other (Spearman's rank test, r=0.34, p < 0.05), although the correlation coefficient was low (Fig. 2b).

The POC and PN concentrations in surface seawater showed lower values and smaller variations compared to sea ice ranging 79.7–302.5 and 13.9–64.3 µgL-1, respectively. Chl a concentration was also consistently low (0.2–1.8 µgL-1) except the maximum found at St. S35 (1.8 µgL-1). Seawater POM δ13C and δ15N were -28.8±1.0 ‰ (-30.9 ‰ to -27.0 ‰) and -0.5±1.4 ‰ (-3.1 ‰ to +1.5 ‰), respectively and no relationships were observed between them.

Sea-ice salinity ranged from 0.3 to 10.6 in FYI/OI-derived sea ice and from 9.4 to 18.4 in newly formed ice. NO3-, PO43-, and Si(OH)₄ concentrations in sea ice were 0.5–15.9, 0.0–8.0, and 0.0–21.3 µmolL-1, respectively (Fig. 2d–f). All nutrients showed a positive correlation with sea ice salinity, with the highest concentrations observed in the new ice from the SC. Compared to seawater (winter TDL), plots for NO3-, PO43-, and Si(OH)₄ fell below the TDL in 43 %, 7 %, and 93 % of samples, respectively. The proportion of sea ice samples falling below the TDL showed no significant regional variation for NO3-.

The tracked trajectory of the sea ice revealed that much of it drifted from the east (Figs. S1–S3). The number of days since the sea ice formation (DSF) ranged from 203 to 581 d (Table 1). Even within the same sea area (LB, CD, VB, and SC), the DSF varied significantly, ranging from sea ice less than 1-year old to ice likely to be second-year ice (DSF > 500 d).

To examine the regional variations and differences between sea ice types (FYI/OI and newly formed ice) for each parameter, the values were plotted for the four regions (Fig. 3). The sea ice salinity of the FYI/OI from all stations was significantly lower than that of the newly formed ice. Significant salinity differences were observed between the VB (mean 3.0) and SC (mean 5.4) (Steel-Dwass test, p<0.05). Chl a concentrations in FYI/OI ranged 0.01–134 µgL-1, showing higher average concentrations compared to newly formed ice (0.2–4.0 µgL-1). Except for CD, FYI/OI had significantly higher Chl a concentrations than the newly formed ice (Steel-Dwass test, p<0.05).

Similar to Chl a, POC concentrations of FYI/OI were higher than the average value for newly formed ice (738.3 µgL-1), but significant differences in POC concentration were only observed between those of CD and VB, and between CD and newly formed ice. Unlike POC, PN concentrations showed no significant differences between FYI/OI, whereas newly formed ice was significantly lower than VB, CD, and SC. The POC : PN ratio for FYI/OI ranged from 4.5 to 12.6, showing lower values and less variation compared to new ice (11.9–41.9). However, significant differences were observed only between the SC and newly formed ice (Steel-Dwass test, p<0.05).

The mean (± SD) δ13C and δ15N values for sea ice excluding newly formed ice were -24.4±3.2 ‰ and +1.6±2.2 ‰, respectively (-24.6±3.2 ‰ and +1.5±2.2 ‰ for all sea ice). The δ13C of newly formed ice was significantly lower than those of LB and SC, but no significant difference was observed among FYI/OI samples. For δ15N, no significant differences were found between regions or newly formed ice, except that SC was significantly higher than LB. The POC : Chl a ratio showed significant differences among the VB, SC, and newly formed ice, with the highest average value observed in the newly formed ice.

To clarify whether the physical parameters and particulate matter within sea ice varied across sampling periods, the above components were plotted separately for December, January, and February (Fig. 4), excluding newly formed ice. The sea ice salinity showed significant differences between December and February. The average salinity peaked in December (6.3), declined to 2.8 in January, and increased slightly to 4.0 in February. The Chl a concentration in the sea ice increased from December to February. Although the average value in February was more than 2.1 times higher than that in the preceding months, no significant monthly differences were observed (Kruskal-Wallis test, p=0.26). The mean monthly POC concentrations increased from December to February, with significantly higher concentration found in February than in other months. The PN concentration also tended to increase in February and significant differences were observed between February and the preceding months. However, the rate of increase from January to February was 1.6 times for POC and 1.8 times for PN, which was more moderate than the increase in Chl a concentration. POC : PN and POC : Chl a ratios, δ13C, and δ15N showed some monthly variations, but no significant differences were discerned (Kruskal-Wallis test, p>0.29).

To identify the factors affecting the sea ice POC and PN concentrations and their respective isotope ratios, we investigated their correlations with other physical and biogeochemical parameters (Fig. 5). Sea-ice POC and PN concentrations showed positive correlations with Chl a and *PO43- concentrations and Si(OH)₄ : NO3- (Spearman's rank test, r>0.30, p<0.05), while they correlated negatively with *NO3- and DSF (r<-0.20, p<0.05, Fig. 5a). Chl a concentration showed similar results to POC and PN regarding Chl a and *PO43-, but it also exhibited a weak positive correlation with salinity (Spearman's rank test, r=0.35, p<0.05). DSF showed negative correlations with Chl a concentration and salinity, similar to POC and PN, but a positive correlation was also observed between POC : Chl a ratio (Spearman's rank test, r=0.32, p<0.05).

The sea-ice POM δ13C and δ15N showed positive correlations with POC, PN, and Chl a concentration (Spearman's rank test, r>0.30, p<0.05), while they correlated negatively with POC : Chl a (r=-0.57 and r=-0.28 for δ13C and δ15N, respectively, p<0.05). Among each nutrient, both δ13C and δ15N showed a negative correlation with *NO3- and NO3-:PO43- (Spearman's rank test, r<-0.20, p<0.05), while only δ13C showed a positive correlation with *PO43- and Si(OH)₄ : NO3- (r>0.45, p<0.05).

The POC and PN concentrations in surface seawater also showed a strong positive correlation with the Chl a concentration (Spearman's rank test, r=0.90 and r=0.87 for POC and PN, respectively, p<0.05, Fig. 5b). However, unlike sea ice, they showed negative correlations with PO43- (Spearman's rank test, r<-0.78, p<0.05). δ13C showed a negative correlation with Si(OH)₄ concentration and the Si(OH)₄ : NO3- ratio (Spearman's rank test, r<-0.63, p < 0.05), while δ13C showed a positive one with sea surface temperature (r=0.62, p<0.05). δ15N showed a significant negative correlation only with salinity (Spearman's rank test, r=-0.66, p<0.05).

The POC_VWM concentration for sea ice (excluding newly formed ice) was 1137.2 µgL-1, with an average value of 1353.0 µgL-1 and a median of 1022.1 µgL-1. The POC supply from sea ice in the Indian sector, calculated from meltwater flux and the area of the SIZ, was 4.1±0.3 Tgyr-1 (Table 2). Assuming the primary production at the MIZ in this sector to be 27.9±5.0 Tgyr-1 (Arrigo et al., 2008), sea ice-derived carbon was found to account for 14.7±2.8 % of this production. The Chl_VWM concentration was 12.4 µgL-1. The POC_algae concentration estimated by Chl_VWM and the mean IAC : Chl a ratio (54.4±26.5, Sect. 2.6) yielded 674.6±328.6 µgL-1, which corresponds to 59.3±28.9 % of POC_VWM.

Sea ice salinity, excluding newly formed ice, was negatively correlated with the days since sea-ice formation: DSF; this is likely due to desalination associated with surface melting and brine discharge during summer, as well as snow accumulation (Eicken, 1992). Conversely, POC, PN, and Chl a concentrations gradually decreased as DSF increased (Fig. 6), although there is diversity among sea-ice samples. Our sea ice was collected during the summer season (December to February) when primary productivity is generally in a declining phase (net heterotrophy), as indicated by a decrease in total organic carbon (Dalman et al., 2025). In multi-year ice, nitrate concentration is much lower than first-year ice (Henley et al., 2023). In such environment, algal accumulation is slow, for the dissolution of diatom frustules (Si(OH)₄ accumulation) surpassed the consumption by diatoms (Sahashi et al., 2022). There was a weak but significant positive correlation between POC : Chl a and DSF, suggesting that in older (high DSF) samples, sea ice algae were at the senescent stage.

In accord with the previous studies, POC : PN ratio was generally higher than that in seawater and the median for Antarctic sea ice is around 10 (Dalman et al., 2025). This is partly owing to ice algae which have high C:N ratio even in nutrient-replete condition (Gleitz and Thomas, 1993). Under nutrient (NO3-) limitation, POC : PN could increase to >20 (McMinn et al., 1999) and it is more pronounced in productive spring season (Roukaerts et al., 2016; Dalman et al., 2025). EPS are rich in carbon and may accumulate in older sea ice as suggested by comparison of FYI and MYI (Meiners et al., 2003). Although there was no significant relationship between DSF and POC : PN (Spearman's rank correlation test, p=0.49), inferring the combination of nutrient status and EPS affected our results.

The maximum Chl a concentration in newly formed ice in the Southern Ocean is 260.0 µgL-1 (Ito et al., 2025), while the Chl a and POC concentrations in this study were 0.2–4.0 and 246.1–1354.0 µgL-1, respectively (Fig. 3b, c). Therefore, the samples in this study were considered to have formed without active frazil ice production or phytoplankton blooms (Ito et al., 2025; Takahashi et al., 2025). Despite the low Chl a concentration, the newly formed ice exhibited high average POC : PN and POC : Chl a ratios, which were significantly higher than those of some FYI/OI samples, such as those in SC (Fig. 3e, h). This difference is likely due to differences in ice-algal composition and the selective accumulation of organic matter during sea-ice formation and growth. The newly formed ice sample from St. S138, which had the highest POC : Chl a ratio, was dominated by nanoflagellates (Takahashi et al., 2025). The POC : Chl a ratio for Phaeocystis antarctica, the predominant nanoflagellate in the Southern Ocean, ranges from 41 to 284 (Arrigo et al., 2010), and this is significantly higher than that for diatoms (approximately 37, Whitaker, 1977). In East Antarctic sea ice, diatoms dominate over nanoflagellates as the sea ice grows (Takahashi et al., 2025), suggesting that differences in algal composition contribute to the relatively high POC : Chl a ratio in newly formed ice than older ice. Another process that influences the POC : Chl a ratio is the selective incorporation of particulates during sea ice formation (Gradinger and Ikävalko, 1998; Janssens et al., 2016; Takahashi et al., 2025). Since POC incorporation into sea ice is higher than PN and small (<10 µm) algal cells (Janssens et al., 2016), the dominance of small nanoflagellates in seawater (St. S138, Takahashi et al., 2025) could enhance POC : PN and POC : Chl a ratios in newly formed ice.

The δ13C of POC was -24.6±3.2 ‰ for all sea ice samples, which was higher than the previously reported δ13C of surface seawater in the SIZ (-31 ‰ to -24 ‰, Espinasse et al., 2019). The δ15N of sea-ice PN (1.5±2.2 ‰) was also generally higher than summer seawater (-5 ‰ to +7 ‰, Espinasse et al., 2019). Although spatiotemporally limited, our δ13C and δ15N of surface seawater obtained in January was lower than those of sea ice (-29±1.0 ‰ for δ13C and -0.5±1.3 ‰ for δ15N). The small variability in isotope ratios in seawater is likely owing to the narrower ranges of sampling region (50–64° S) and periods (January) compared to the previous data sets conducted in 37–70° S during June and May over several years (Espinasse et al., 2019). The isotope ratios of sea ice remained within the range previously reported for Antarctic sea ice (Rau et al., 1991; Kennedy et al., 2002; Munro et al., 2010). The stable isotope ratios of POC, in particular, varied across the observed regions, likely due to differences in sea ice organic matter and nutrient concentrations among regions (Fig. 3). This section discusses drivers of stable isotope ratios in the context of sea-ice biogeochemical parameters.

Microalgae incorporate ¹³C when CO₂ becomes limiting; hence, δ13C of POM in sea-ice could increase under closed conditions and high algal CO₂ demand (i.e., when high ice-algal biomass is reached) (Kennedy et al., 2002). According to Nomura et al. (2023), the brine volume fraction in brash ice is 16.2 %. Since this exceeds the threshold for gas permeability to the atmosphere (5 %; Golden et al., 1998), δ13C values could be distinct from those of more intact ice floes. However, our samples showed a significant positive correlation between sea-ice POC and δ13C, and high δ13C values (>-20 ‰) were limited to high POC concentration (>2000 µgL-1), generally consistent with the previous sea-ice core observations (Kennedy et al., 2002; Munro et al., 2010) (Fig. 7). Three possible reasons for the similar isotope ratios between brash ice and ice floes are as follows: (1) brash ice reflects the history of the ice floes before they are broken, (2) there is limited exchange between sea ice and the atmosphere at the microscale, and (3) dilution of brine DIC by surface melting of brash ice. For reason 1, although there are no direct measurements of productivity in brash ice, reports indicate net heterotrophic activity in late-summer (February) sea ice (Dalman et al., 2025). Thus, the isotope ratios during earlier periods (i.e., when sea ice was not broken into brash ice) may have been recorded in our brash sea ice samples. According to the field observation in spring Antarctic sea ice, δ13C increases with algal accumulation in the very surface layer of sea ice (5 cm inside sea ice from ice–water interface, McMinn et al., 1999), which could be the reason 2). For the reason 3, pCO₂ in sea ice declines with increasing brine volume fraction (Nomura et al., 2010; Geilfus et al., 2014; van der Linden et al., 2020). As pack ice breaks up into smaller pieces, it warms by approaching the temperature of seawater. As brine salinity decreases and DIC dilution in the brine progresses, CO₂ can become rate-limiting in brash ice (Nomura et al., 2010). However, the brine salinity of brash ice derived from ice temperature is 27.6±4.8 (Nomura et al., 2023; Vancoppenolle et al., 2019), indicating the dilution effect is not significant to determine brash ice δ13C.

Within sea ice, PO43- is regenerated alongside diatom proliferation, resulting in extremely low summer NO3-:PO43- ratios (Fripiat et al., 2017; Takahashi et al., 2022; Nomura et al., 2023). Due to PO43- regeneration, a negative correlation exists between ice-algal biomass, in particular diatoms, and NO3-:PO43- ratios in brash ice (Takahashi et al., 2022). In this study, a negative correlation between δ13C and NO3-:PO43- was observed, suggesting that diatoms contributed to the δ13C increase. Furthermore, previous microscopic observations of East Antarctic sea ice have confirmed that diatoms dominate during late spring and summer (Meiners et al., 2011; Takahashi et al., 2022, 2023, 2025). On the other hand, the negative correlation between δ13C and Si(OH)₄ : NO3- suggests that different diatom species in sea ice may exhibit distinct ¹³C isotope fractionation (Burkhardt et al., 1999). Fischer (1991) investigated the surface POC δ13C distribution in the Scotia Sea and reported a positive correlation between the particulate Si:N ratio and δ13C. Diatoms have different silica requirements depending on the species. For example, the Si(OH)₄ half-saturation constant of Corethron pennatum is more than ten times that of small diatoms (Sommer and Stabel, 1986). Large diatoms with high particulate Si:N ratios (i.e., those that induce low Si(OH)₄ : NO3-) incorporate relatively little ¹³C (Burkhardt et al., 1999; Fischer, 1991), suggesting that differences in diatom taxa within sea ice may be reflected in the variation of δ13C values. However, the correlation coefficient between δ13C and Si(OH)₄ : NO3- (r=0.46) is lower than those with POC, PN, or Chl a concentrations (r=0.60–0.69), suggesting a smaller effect on the increase in δ13C than algal accumulation.

According to studies on the stable isotope ratios of POM in seawater, nitrogen nutrient utilization reflects variations in PN δ15N (Saino and Hattori, 1980). Nomura et al. (2023) reported that despite brash ice having high permeability and exchange with seawater (>5 % brine volume fraction), the biological activity in sea ice influences the nutrient concentration. In this study, PO43- was highly enriched in sea ice, while Si(OH)₄, known for its slow decomposition rate, was lower than in seawater (Fig. 2e, f). Due to these characteristics, even brash ice subject to melting has a distinct nutrient status within the ice compared to seawater, which likely contributed to the observed δ15N values. In this study, δ15N showed a negative correlation with *NO3-, suggesting that nitrogen consumption contributed to the high isotope ratio. Diatoms selectively incorporate ¹⁴N (Wada and Hattori, 1978), hence, in a closed system where NO3- is actively consumed, the δ15N of remaining (unconsumed) NO3- increases (Wu et al., 1997). The uptake of NO3- is supported by the negative correlations between POC, PN, and Chl a concentrations and *NO3- (Fig. 5).

Conversely, it is known that δ15N values of PN or dissolved inorganic nitrogen also change during organic matter decomposition and nitrification (oxidation from NH4+ to oxidized nitrogen species) (Saino and Hattori, 1980). The POC : Chl a ratio in this study showed a negative correlation with δ15N. An increase in the POC : Chl a ratio is linked to NH4+ accumulation from predation and excretion by heterotrophic organisms (Schnack-Schiel et al., 2004) and remineralization of particulate organic matter (Roukaerts et al., 2021). Although δ15N decreases during the oxidation of NH4+ to NO2- through the fractionation by nitrifying bacteria, ¹⁵N accumulates in NO3- during the subsequent oxidation of NO2- to NO3- (Casciotti et al., 2003, Casciotti, 2009). Therefore, NO3- produced by nitrification is considered to have δ15N values similar to or lower than those of NH4+ produced by decomposition (i.e., NH4+ before its nitrification) (Henley et al., 2018). On the other hand, when algae directly assimilate NH4+, the δ15N of PN decreases due to the lower fractionation factor, -25 ‰ and -15 ‰ (negative values indicate greater incorporation of ¹⁴N into the algae) than nitrification (Waser et al., 1998). The nitrification rate in sea ice is known to correlate positively with Chl a concentration (Priscu et al., 1990). In this study, POC : Chl a was negatively correlated with Chl a concentration (Fig. 5a). That is, higher POC : Chl a ratios in sea ice may have been associated with reduced nitrification activity. The negative correlation between POC : Chl a ratio and δ15N (Fig. 7d) suggests that even if NH4+ may be re-mineralized due to the shift from autotrophy to heterotrophy, direct NH4+ utilization by ice algae likely outweighed assimilation after nitrification of NH4+ (conversion of NH4+ to NO3-) and resulted in low PN δ15N. This hypothesis is supported by observations showing relatively high NH4+ uptake rates in the upper layer of Antarctic drift ice, where the POC : Chl a ratio is higher than that in the interior or bottom sections of sea ice (Roukaerts et al., 2016).

The POC : Chl a ratio varied significantly across regions, seasons, and sea-ice types (FY/OI and newly formed ice). This result indicates that carbon sources other than ice algae are also important for the POC consumed by pelagic heterotrophs in the water column during sea ice melt. The relationship between POC and Chl a concentrations (Fig. 2c) showed that sea ice with high C_algae contributions (sample plots above the IAC : Chl a = 54.4 line) was most common in sea ice with low to moderate Chl a concentrations (<30 µgL-1); this aligns with previous reports that ice algae contribute most of the carbon in sea ice with extremely high Chl a concentrations (Takahashi et al., 2023). The POC : Chl a ratio showed a tendency to increase from December to February, and the POC : Chl a ratio was found to have a significant positive correlation with days since formation (DSF). These findings indicate that from early to late summer, older sea ice contributes to a greater proportion of non-algal-derived POC.

Although the types of non-algal-derived POC were not identified, one possibility is the presence of extracellular polymeric substances (EPS) or detritus within ice (Dumont et al., 2009). Studies examining temporal changes in sea-ice EPS and Chl a concentrations have reported that EPS tends to remain within the ice longer than Chl a during the melting season (Riedel et al., 2008). In general, EPS forms biofilms that cause ice algae and bacteria to attach within brine pockets, preventing particulate matter from following brine dynamics (i.e., salinity decreases during melting). In this study, although a weak positive correlation existed between Chl a concentration and salinity, POC and PN showed no relationship with salinity. Therefore, in older sea ice, some portions of ice algae (Chl a) may have been released through brine expulsion, whereas POC likely remained on the ice. The negative correlation between the POC : Chl a ratio and Chl a concentration supports the notion that Chl a is more readily decomposed or removed from sea ice than POC or PN.

Based on the measured sea ice POC concentrations and sea ice melt fluxes, the impact of sea ice-derived carbon on marine zooplankton as a food resource during the melt season was estimated (Table 2). When combined with primary production after ice retreat, POC from sea ice accounted for 14.7±2.8 % of primary production in the MIZ of the studied sector. It should be noted that since POC concentrations in sea ice during October–November – before the observation period of this study – were not included, calculations solely using summer sea-ice data may have resulted in an overestimation of the POC supply. This is primarily due to seasonal accumulation of organic carbon from September to February in Antarctic sea ice (Dalman et al., 2025). Nevertheless, according to satellite-based estimates by Bocquet et al. (2024), approximately 85 % of sea-ice melt (sea-ice volume reduction) occurs between December and February. Therefore, even if we excluded the POC supply from October to November, the summer POC from melted sea ice is equivalent to 12.5 % of production in the MIZ. It should be acknowledged that these estimates are averaged ones and the spatial or temporal differences are expected owing to variations in ice melt thickness (0.2–1.6 m, Komatsu et al., 2025), sea-ice POC concentration (Fig. 3c), and primary production in the MIZ (21.4–37.2 Tgyr-1, Arrigo et al., 2008).

The previous estimate of POC supply from sea ice was 0.253 TgCyr-1 in the Weddell Sea (60° W–30° E), calculated from sea ice extent and POC concentrations converted from sea-ice Chl a (Ackley et al., 1979). When estimated per unit area, this corresponds to 35.6 mgm-2yr-1, which is significantly lower than our estimate of 839.3 mgm-2yr-1 (Table 2). This difference may be attributed to the conservative assumptions used by Ackley et al. (1979), who averaged the low sea-ice Chl a concentration (1.2 mgm-2) over February and March. However, recent reports found an average sea-ice Chl a concentration of 6.4 mgm-2 across the Antarctic SIZ, which increases to 12.8 mgm-2 during February to March (Meiners et al., 2012). It should also be noted that the POC : Chl a ratio was 37 in the previous estimate, which is close to the lower limit observed during ice algal blooms (Whitaker, 1977). The POC : Chl a ratio can exceed 100 in spring sea ice (Roukaerts et al., 2016), and the median FYI/OI in this study was 170. If we use our Chl_VWM value and POC : Chl a ratio (37) used in Ackley et al. (1979), the POC_Input was estimated to be 338.6 mgm-2yr-1. The input of ice-algal carbon (IAC_Input) using Chl_VWM and IAC : Chl a was 674.6±242.5 (297.4–839.2) mgm-2yr-1 and our most conservative estimate was greater than that by Ackley et al. (1979), suggesting a significant underestimation caused by the measured sea-ice Chl a concentrations and the conversion factor from Chl a to POC. Given the differences between these assumptions, the POC supply from sea ice is one to two orders of magnitude higher than that estimated by Ackley et al. (1979), which is closer to our estimate.

Finally, we discuss the impact of POC supply from sea ice on the carbon content in the water column from the perspective of primary consumers. Antarctic filter-feeding euphausiids have been reported to consume ice algae which constitute 5 %–36 % (11 % on average) of the carbon in their diet in the summer MIZ (Kohlbach et al., 2019). Although it remains unknown if all sea-ice derived POM are consumed or settle to the depth or remineralized in the water column, our estimate (>14 % of MIZ production), which was temporally integrated, fell within this range, yet. In this study, the POC : PN ratio was >10 on average, indicating lower food quality than phytoplankton (surface seawater POM, 5.2±0.5 POC : PN ratio in this study). It should be noted, however, that surface seawater POC : PN ratios in this study are conservative values for they were determined in the most productive season (Janurary), when POC : PN usually reaches minimum during spring and autumn (Smith et al., 2000). Daase et al. (2011) found the growth time and body size during the nauplius stage were equivalent to those when fed high-quality (low POC : PN ratio) food (Daase et al., 2011). Therefore, POM from sea ice during the summer have a potential to support the growth of zooplankton larvae (such as copepod nauplii) and needs to be tested by grazer's isotopic signatures.