The study site was Malene Bight (64.179 N, 51.706 W) near the Greenland capital of Nuuk, southwest Greenland ( Figure 1 ). First-year sea ice covers the bight from about mid-January and lasts until about mid-April, reaching a thickness of 30 to 40 cm, with snow periodically covering the ice in depths of 10 to 20 cm (Søgaard et al., 2010). Air temperatures vary from an average of −14°C in February to −5°C in the beginning of April (www.asiaq-greenlandsurvey.gl). As part of a teaching program we (LCLH, BS, and DHS) have carried out sea ice studies in the area every year since 2015 and sea ice and snow conditions have not changed.

An area of 100 × 100 m was marked with wooden stakes on the ice in February 2022, and sampling was carried out between 26 February and 4 April 2022. Snow depths were first measured (n = 10) with a ruler at sampling sites within a square area of 1 m², which were then cleared of snow. Ice cores were extracted within the 1 m² area using a Kovacs Mark II 90 mm diameter ice auger, to secure minimum horizontal and vertical variability in sea ice parameters between cores (Lund-Hansen et al., 2021). Ten cores were collected during each sampling event with triplicate cores for sea ice physics, triplicates for nutrients and Chl a, triplicates for algal photobiology, and one core for DNA extraction. DNA sampling for species composition commenced on 15 March 2022 and continued after Chl a concentrations began to increase. Cores were immediately placed in a cradle and shielded from light using a photographer’s black cloth to prevent any photodamage of the ice algae. Ice thickness (core length) was measured with a ruler to the nearest cm, and ice core temperatures were measured in the physics cores by drilling a 3 mm hole to the center of the ice core and inserting a temperature sensor (Testo 110 Thermometer).

All cores were cut into 3 cm thick sections using a stainless-steel saw except the DNA core. It was placed in a separate sterilized polyethylene bag in a darkened and insulated cooler. All ice sections from the other cores were placed in separate polyethylene containers and immediately transferred to a second darkened and insulated cooler and brought to the laboratory for analysis. Under-ice water samples were collected at 1 m depth below the bottom of the sea ice through one of the core holes and kept in brown glass bottles placed in a third darkened and insulated cooler. Water salinity below the bottom of the sea ice was measured in situ with a YSI Pro30 CTD. All samples were returned to the laboratory cold room (4°C) within 1 h, and ice samples were thawed overnight at 4°C and processed the following day. Cores for DNA extraction were stored at −20°C and kept frozen during transport back to Denmark until sample processing.

Continuous (5 min) recordings of photosynthetically active radiation (PAR) and air temperature (30 min) were obtained from a permanent climate station in Nuuk, maintained and administered by ASIAQ-Greenland Survey (www.asiaq-greenlandsurvey.gl). Surface PAR, PAR albedo, and under-ice PAR were measured during sampling on the ice with Li-Cor LI-192 PAR sensors. The under-ice PAR sensor was mounted on a stick with an L-arm and lowered through a hole drilled in the ice, which was filled with ice and snow to prevent false light from above reaching the sensor. Bulk salinity of thawed sea ice samples was measured with a YSI Pro30 CTD. Brine volumes, based on bulk salinity and temperature of the ice, were calculated following Cox and Weeks (1983) for ice temperatures < −2°C and the Leppäranta and Manninen (1988) relations for ice temperatures > −2°C. Brine salinity was derived following Cox and Weeks (1983).

Samples for Chl a and nutrients were determined using standard methods described in detail by Søgaard et al. (2010). In brief, Chl a analyses were carried out at 3 cm vertical resolution and nutrients at 6 cm vertical resolution in the ice cores. Sea ice and under-ice water samples were analyzed for concentrations of NH₄ ⁺ (ammonium), NO₃ ^- (nitrate), NO₂ ^- (nitrite), PO₄ ^3- (phosphate) and Si(OH)₄ (silicate) by filtering the thawed sea ice and water samples through 0.45 µm Q-Max CA syringe filters into 12 mL test tubes and freezing (−19°C) for further analysis. The concentrations were measured on a Seal QuAAtro autoanalyzer using standard colorimetric methods. Mean values of triplicate samples are reported.

Photobiological parameters were derived from variable chlorophyll fluorescence (Schreiber et al., 1986) using a Walz Phyto-PAM (Pulse Amplitude Modulated) variable fluorometer fitted with a Phyto-ED emitter-detector unit (Heinz Walz GmbH, Effeltrich, Germany) as described previously (Lund-Hansen et al., 2020a; Sorrell et al., 2021). The Phyto-PAM instrument was operated in a darkened and cooled laboratory, and thawed sea ice samples with ice algae were dark-acclimated (12 h) before measurements. The instrument provides Φ_{PSII_max} (the maximum dark-acclimated quantum yield of Photosystem II), and its Rapid Light Curve (RLC) protocol (Schreiber, 2004; Ralph and Gademann, 2005) models the standard light response cardinal points, α (initial slope of the light response curve), E _k (onset of light saturation), and rETR_max (maximum light-saturated relative electron transport rate), using algorithms within the Phyto-PAM software. Differences in Chl a and photobiological parameters over time were tested with repeated measures analyses of variance after testing for, and if necessary, correcting, heterogeneity of variance, with differences between dates identified with post hoc Tukey tests. Isopleths of brine volume, temperature, Φ_{PSII_max}, and Chl a were developed using SURFER ver. 2.3.4 (Golden Software).

The intact ice cores used for DNA extraction were washed with autoclaved, sterile 0.2 µm filtered Milli-Q water, before 5 cm sections were cut using a pre-sterilized stainless-steel saw. All 5 cm sections were transferred to sterile Whirl-Pak^® sample bags and thawed at 4°C. Between 157–276 ml of sample was concentrated on 0.2 µm pre-sterilized polycarbonate filters (see Supplementary Table S1 for information on individual samples). Different sample sizes occurred as some filters became clogged due to high biomass accumulation; here it was decided that >150 mL samples were sufficient to describe the community structure. Sample processing was carried out in two rounds where the negative handling control (filter which was subjected to the same treatment without sample addition) was included in both rounds. Filters were immediately placed in sterile tubes and DNA was extracted directly from the filter, following the power soil pro kit protocol (Qiagen), including a DNA extraction negative control. 18S rRNA gene copies were quantified using primers Euk345F (5’- AAGGAAGGCAGCAGGCG-3’) and Euk499R (5’- CACCAGACTTGCCCTCYAAT-’3) (Zhu et al., 2005). The qPCR mixture contained 2 µL template DNA, 1 µL of each primer (10 pmol µL^-1), 1 µL BSA (10 ng µL^-1), 10 µL LightCycler Mastermix (Roche) and 4 µL dH₂O to a final volume of 20 µL. Samples were run in triplicate including triplicate negative qPCR controls without addition of template DNA. Samples were run on a Stratagene Mx3005P (Agilent technologies) at the following temperatures: 95°C for 5 min, 40 cycles with denaturation at 95°C for 30 s, annealing at 60°C for 30 s, elongation at 72°C for 20 s, followed by fluorescent acquisition at 79°C for 15 s. A melting curve was produced using 1 cycle at 95°C for 30 s, and 55°C at 1 min and 95°C for 30 s. 18S rRNA library preparation of DNA was performed as follows: Primers TareukFWD1 (5’- CCAGCASCYGCGGTAATTCC-3’) and TAReukREV3 (5’- ACTTTCGTTCTTGATYRA-3’) were used to amplify the V4 RNA loop (Stoeck et al., 2010). The PCR mixture contained 4 µL template DNA. The thermal cycling was run with the touchdown PCR approach with an initial denaturation step at 95°C for 3 min, 10 cycles with denaturation at 95°C for 30 s, annealing at 57°C for 30 s, elongation at 72°C for 30 s, followed by 15 cycles with denaturation at 95°C for 30 s, annealing at 47°C for 30 s, elongation at 72°C and a final elongation at 72°C for 5 min. The PCR products were cleaned using 30 μL AMPure XP magnetic beads. The second round of PCR was run for 10 cycles to incorporate overhang adapters and was run with the same conditions as the previous PCR with an annealing temperature of 57°C. Products were cleaned, and the Nextera XT Index primers were incorporated in a third PCR reaction which was run for 8 cycles following the previous condition and an annealing temperature of 55°C. The PCR products were quantified using a Quant-iT™ dsDNA BR assay kit on a FLUOstar Omega fluorometric microplate reader (BMG LABTECH, Ortenberg, Germany), diluted and pooled together in equimolar ratios. The pool was quantified using the Quant-iT™ dsDNA BR assay kit on a Qubit fluorometer (Thermo Fisher Scientific, Waltham MA) and sequenced on the Illumina MiSeq platform (Illumina, San Diego, CA) which produces two 300-bp long paired-end reads.

Bioinformatic analyses were performed in RStudio 4.3.1. Primer and adapter sequences were trimmed from the raw reads using cutadapt 0.0.1 (Martin, 2011). Forward and reverse read quality were plotted with the plotQualityProfile function from DADA2 1.21.0 (Callahan et al., 2017). Based on the read quality, a trimming of 280 bp and 200 bp were set for the forward and reverse reads respectively using FilterAndTrim, according to their quality (Callahan et al., 2017). The fastq files were randomly subsampled to the lowest read number using the ShortRead package 1.48.0 (Morgan et al., 2009), resulting in 51586 reads per sample. The subsampling allows for a more accurate comparison of the richness of the different samples. Error models were built for the forward and reverse reads, followed by dereplication and clustering into amplicon sequence variants (ASVs) (Callahan et al., 2017) with DADA2. The denoised forward and reverse reads were merged using the function mergePairs of default parameters with a minimum overlap of 12 nucleotides, allowing 0 mismatches. Sequence tables were made with the function makeSequenceTable. ASVs shorter than 360 and longer than 400 nucleotides were removed from the 18S dataset followed by chimeric sequence removal using the removeBimeraDenovo function ( Supplementary Table S1 ). Taxonomic assignment was accomplished using the naive Bayesian classifier against the Protist Ribosomal Reference (PR2) database v5.0.1 (Guillou et al., 2013) with the assignTaxonomy function from DADA2. ASVs mapped to mitochondria, chloroplasts and metazoa were removed from the dataset. Samples were decontaminated using the prevalence method (Threshold = 0.1) from the Decontam package (Davis et al., 2018). One ASV (ASV593) classified to the genera Exobasidiomycetes (formerly Malassezia) was detected and removed from all samples as it is a known skin associated fungus (Spatz and Richard, 2020) which has been associated with contamination in microbiome sequencing dataset. Statistical tests and visualization of the data was performed with phyloseq (McMurdie and Holmes, 2013), Vegan (Oksanen et al., 2019) and microeco (Liu et al., 2021). Redundancy analysis (RDA) was performed using the following explanatory variables: under-ice PAR (Par.Under), sea ice temperature (Ice.Temperature), air temperature (Air.Temperature), nutrient concentrations (ammonium, nitrate, nitrite, phosphate and silicate) in the ice, and downwelling PAR (PAR.Down). Genera with a relative abundance of <0.5% were removed from the dataset before calculating the ordination.

Ice thickness was effectively constant through the study period at an average of 36.9 ± 2.9 cm. Snow depths were also constant at an average of 15.1 cm during the first five sampling dates, but halved to 7.0 ± 0.8 cm on 4 April with the arrival of warmer air temperatures ( Table 1 ). Last sampling was commenced on 4 April as the ice was broken a few days later by vessels. Albedo was high (0.85–0.90) during the beginning and middle of the study period, but lower (0.63–0.72) during the last two samplings. Downwelling PAR at the surface increased from 205 to 1506 µmol photons m^-2 s^-1 during the study period, and PAR below ice increased from 2.2 µmol photons m^-2 s^-1 on 10 March to 65.5 µmol photons m^-2 s^-1 on 4 April. Note that all PAR readings refer to readings at 12 noon for comparison. Under-ice Chl a was very low (< 0.5 µg Chl a L^-1) except for the final sampling, with a concentration of 3.15 µg Chl a L^-1. Sampling of water for nutrients below ice was not carried out at all ice sampling dates, from 10 March to 4 April nitrate decreased from 11.44 to 7.92 µmol L^-1, silicate from 30.8 to 21.6 µmol L^-1, and phosphate from 0.80 to 0.61 µmol L^-1, whereas ammonium increased from 1.71 to 5.02 µmol L^-1 ( Table 1 ).

Sea ice temperatures were relatively low in the first part of the study period, but increased with the increasing air temperatures ( Table 1 ) during the last three samplings to approx. −1 ˚C to −1.5°C ( Figure 2B ). This warm period was characterized by relatively high brine volumes ( Figure 2A ) of 15–20%, as brine volumes are positively related to both ice temperatures and bulk salinity. Φ_{PSII_max} was relatively high (0.4 to 0.5) in the center of the ice but low (< 0.25) near the surface and bottom of the ice, except during the last two samplings ( Figure 2C ). Here, Φ_{PSII_max} increased vertically in the ice, rising as high as 0.61 at 5 to 10 cm depth.

Chl a concentrations in the ice increased significantly through the study (F = 17.35, p < 0.001). They remained low (around 0.5 µg Chl a L^-1) at first sampling (26 February), but increased to about 1.5 µg Chl a L^-1 during the third sampling (20 March) ( Figure 3A ). Average Chl a concentration in the ice increased from 2.3 to 6.4 µg Chl a L^-1 between 30 March and 4 April, with a maximum concentration of 13.4 µg Chl a L^-1 in the surface (0–5 cm) layer.

The bulk sea ice nitrate concentrations decreased from 2.0 to 0.05 µmol L^-1 at 15 cm depth from the beginning of the sea ice season to just before sea ice break-up ( Figure 3B ). Uptake rate was estimated as the difference in concentration at specific dates divided by the time between dates, which reached 0.06 µmol L^-1 d^-1 for nitrate at 15 cm depth in the ice between 26 February and 4 April (38 days).The estimated uptake rates for phosphate (0.006 µmol L^-1 d^-1) and silicate (0.04 µmol L^-1 d^-1) were much lower. Supplementary Figures S1, S2 show phosphate and silicate concentrations over time. The N:P ratio at 15 cm depth decreased from 9.6 to 0.25 during the study period.

The photobiological data from Phyto-PAM analyses are shown in Figure 4 . Φ_{PSII_max} did not differ significantly during the study ( Figure 4A ). However, ANOVA (F = 3.36, p < 0.01) and the post hoc tests revealed two statistically distinct time periods for rETR_max (marked as Periods I and II in Figure 4B ). During Period I, rETR_max remained constant and relatively low from 26 February to 20 March, but then suddenly increased, more than doubling, during the last two sampling dates (Period II in Figure 4B ), just before sea ice break-up. ANOVA and post hoc tests revealed that E _K also increased significantly (F = 7.19, p < 0.005) between these two periods, from 193.3 ± 79.6 µmol photons m^-2 s^-1 averaged over Period I, to 412.8 ± 181.7 µmol photons m^-2 s^-1 averaged over Period II, just before sea ice break-up. There were no significant differences in α over time, remaining constant from the beginning of the sea ice season (0.18 ± 0.04 µmol é mol^-1 photons) to just before sea ice break-up (0.20 ± 0.04 µmol é mol^-1 photons).

The relative abundance of taxa was similar across sampling dates and between the top and bottom of the ice and was clearly dominated by Dinophyceae and Ulvophyceae, with an average relative abundance of 31.2 ± 9.9% and 16.5 ± 12.7% for all depths and dates, respectively ( Figure 5 ). Relative abundances of the other common taxa were: Cryptophyceae 9.0 ± 3.6%, Filosa-Thecofilosea 6.9 ± 2.4%, and the two classes of diatoms, Mediophyceae 5.2 ± 3.4% and Bacillariophyceae 4.3 ± 3.9%. Ulvophyceae became more abundant over time in the bottom section of the ice, while the diatoms (Mediophyceae and Bacillariophyceae) remained rather constant during the sampling period.

The relative abundance data ( Figure 6 ) show that the most abundant genera were the two green algae Ulothrix and Chlamydomonas, the two dinoflagellates Chytriodinium and P. glacialis, together with taxa from the Protaspa-lineage_X. Overall we did not observe a large change in the dominant genera throughout the time-series. However, we did observe a 5.6-fold increase in the relative abundance of P. glacialis with a relative abundance of 1.8% on the 15/3 to 10.1% on the 4/4.

The redundancy analysis shows that a total of 75.2% of the constrained variation in the data could be explained by the explanatory variables on the first two axes, with RDA1 explaining most of the constrained variation (52.9%) and the second axis explaining 22.3% of the constrained variation ( Figure 7 ). The RDA analyses, comprising explanatory variables, show that the occurrence of the dinoflagellate genera Chytridinium was closely related to concentrations of nutrients (nitrate, silicate and phosphate). The green algae Ulothrix showed a negative correlation with the ice temperature and ammonium concentration. Additionally, the occurrence of Chlamydomonas Biecheleriopsis and P. glacialis was especially related to air temperature and downwelling PAR, together with the nitrite concentration on 20 March and 4 April, and specifically in the upper sections (5–10 cm) of the ice ( Figure 7 ).

The bloom of dinoflagellates was initiated during a period of decreasing snow depths and increased PAR in Malene Bight. We noted a continuous increase in under-ice PAR from 26.9 µmol photons m^-2 s^-1 pre-bloom to 65.5 µmol photons m^-2 s^-1 just before ice break-up, and this PAR increase (representing light availability for ice algae photosynthesis) was accompanied by an upregulation of photosynthetic activity (increased rETR_max) and higher photosynthetic light demand (increased E _k). This relation between increased PAR and upregulation of photosynthetic activity was also observed by Manes and Gradinger (2009) in a similar photobiological study at the Chukchi Sea, Alaska. However, a feature of our data is the already high Φ_{PSII_max} in the pre-bloom population, which remained constant and high throughout the study. This shows that the pre-bloom community was already photosynthetically viable and active at the start of the study, consistent with previous studies showing maintenance and activation of photosystems in sea ice algae at extreme low light intensities in the polar night and early season (Hancke et al., 2018; Kennedy et al., 2019). Indeed, a PAR flux of 26.9 µmol photons m^-2 s^-1 is considerably higher than minimum PAR for growth of sea ice algae (Mock and Gradinger, 2000; Hancke et al., 2018), and of a similar magnitude to those observed during diatom-dominated blooms elsewhere (Hawes et al., 2012; Leu et al., 2015). Higher ice temperatures and thus higher metabolic rates in the latter part of the study period may also have contributed to the increase in rETR_max.

The upregulation of photosynthetic activity in this study was accompanied by higher algal biomass, which clearly indicates a photobiological and growth response to the higher sea ice light availability (Schreiber, 2004). Although we did not measure C fixation directly in this study, e.g., as ¹⁴C fixation (Stemann-Nielsen, 1952) photosynthetic activity as determined by PAM methods correlates broadly with C fixation rates in microalgae, including ice algae, in many studies (Glud et al., 2002; Hancke et al., 2015; Lund-Hansen et al., 2017), although this is species-specific, preferably involving conversion of rETR to absolute rates of electron transport (Hancke et al., 2015); see also Camargo et al. (2022) for other complications. Nevertheless, the observed increase in rETR_max strongly indicates corresponding increase in primary production, given the broad correspondence of ¹⁴C-based carbon uptake and rETR_max values (Barranguet and Kromkamp, 2000). During a previous ice algae spring bloom at the same site, primary production reached 12 mgC m^-2 d^-1 at a maximum biomass of 1.5 µg Chl a L^-1 (Søgaard et al., 2010). Based on the maximum biomass of 13.0 µg Chl a L^-1 in the present study, this would equate to a production rate of 80 to 90 mgC m^-2 d^-1, which is high compared to other studies in Greenland sea ice (García et al., 2023). This high production appears to have been reflected in steadily decreasing bulk nutrient concentrations in the ice over time. Assuming that nutrient uptake followed the standard Redfield-Brzezinski ratio of 106C:16N:15Si:1P, nitrate appears to have been depleted in the sea ice just before ice break-up. This strongly suggests that the observed gradual reduction in bulk nutrient concentrations was related to nutrient take up by the ice algal photosynthesis and growth inside the ice during the bloom. Theoretically, the relatively high (20–26%) brine volumes at the end of the study period could have allowed nutrient-rich under-ice water to enter the ice and reinforce the bloom (Golden et al., 2007). This was apparently not the case, given the very low (0.05 µmol L^-1) nitrate concentrations in the lower sea ice. The sea ice broke in early April 2022. It can be speculated that a second bloom at the bottom of the ice would have followed the Dinoflagellate bloom, which, however, appears very unlikely as such a bloom would have required nitrate of which concentrations were close to detection limit beginning April in the bottom section of the ice. The under-ice bloom in Malene Bight described by Søgaard et al. (2010) extended 15–25 cm upwards into the ice and was not confined to the bottom per se as also observed in several other studies (Lund-Hansen et al., 2020).

This study showed a distinct dinoflagellate-dominated bloom (18–57%), with a much lower contribution of the two diatom classes Mediophyceae (0.7–18%) and Bacillariophyceae (0.4–16%). All samples showed at all dates and all depths showed a consistent higher relative abundance of the Dinophyceae compared to the two classes of diatoms combined. Hence, our observations challenge the classical view that Arctic sea ice algae blooms reach a climax with large diatoms (e.g. Mikkelsen et al., 2008; Leeuwe et al., 2018), as we observed a very high peak dinoflagellate relative abundance (18–57%) during this sea ice spring bloom in April. We do note that dinoflagellates tend to have more 18S rRNA copies (Martin et al., 2022) compared with diatoms which potentially could skew our results towards a larger proportion of dinoflagellates compared to diatoms. However, we additionally note that both Dinoflagellates as well as diatom 18S rRNA copy number have been shown to be highly variable with standard deviations between 6.9 × 10^-4 - 2.8 · 10^-5 cell^-1. The bloom is further supported by the increase in observed biomass which changed from 0.5 µg Chl a L^-1 at the start of the campaign to 13 µg Chl a L^-1 on the last day of the campaign. We did not find a remarkable increase in the relative abundance of either of the diatom classes during the increase in biomass, which likely shows that the bloom can be attributed to the community which was already present within the ice. The observed biomass at the end of the campaign was about five times higher than the sea ice biomasses of 2.8 µg Chl a L^-1 previously observed in a diatom-dominated spring bloom in Malene Bight (Søgaard et al., 2010), and in the adjacent Kobbefjord with a diatom bloom biomass of 1.5 µg Chl a L^-1 (Mikkelsen et al., 2008). We are aware that the sea ice was broken by vessels shortly after last sampling 4 April, and that a second diatom bloom could have taken place at a later stage. A diatom bloom was observed in May in Kobbefjord (Mikkelsen et al., 2008) for instance, but nutrients in the ice in the present study were fully exploited beginning April to support any potential new blooms.

The eukaryotic community composition data showed a clear dominance of genera from the green algae and the Dinoflagellata and with increased abundances of P. glacialis, the bipolar (i.e., occurring in both the Arctic and Antarctic) dinoflagellate (Montresor et al., 1999, 2003). A dominant, strong bloom of P. glacialis has previously been described from Antarctica (Stoecker et al., 2000; Thomson et al., 2004), whereas the present study is likely the first to document a similar dinoflagellate-dominated sea ice algae bloom involving P. glacialis in the Arctic. Our RDA analyses also demonstrated that the blooming of P. glacialis specifically occurred in the upper 5 to 10 cm of the ice during the later part of the sea ice season, with reduced snow cover, increased downwelling and high under-ice PAR, and higher air temperatures. Neeley et al. (2018) have previously demonstrated a significant positive relationship between P. glacialis and both light availability and temperature, and Kauko et al. (2018) have also observed that blooms of P. glacialis tend to be located in the upper, and accordingly more light-exposed, sections of the ice. The high affinity between sea ice and P. glacialis was recently demonstrated by Hardardottir et al. (2024) in the Arctic promoting the microalgae as candidate in reconstructing sea ice extent. The Chytriodinium, present in relatively high abundances (12–15%) at all samplings, is an ectoparasitic dinoflagellate of copepod and copepod eggs (Gomez et al., 2009). The occurrence of the Chytriodinium could reflect the relatively higher ice algae (Chl a L^-1) concentrations measured in this study on which the zooplankton is grazing (Lund-Hansen et al., 2020a). On the other hand, being a parasite, the genus would likely not be observed in conventional analyses but only in molecular studies like ours. The genus occurs in both North and South Atlantic (Gomez and Skovgaard, 2015) but this is first record of this genus in sea ice, as to the best of the author’s knowledge. The green algae Ulothrix showed also high (10–25%) relative abundances. A study from the Canadian Basin observed the genus being attached to the bottom of the icebreaker (Melnikov et al., 2002), but in the present study the genus was detected in the entire sea ice core, and not limited to the bottom of the sea ice. It is a brackish-water green algae, though salinities in the water just below the ice were high (32.5–35.7) which calls for further studies. Anyway, this is to our best knowledge, the first record of this genus in sea ice.

Based on our observations, the thinner sea ice and therefore higher PAR associated with climate change is likely to mediate a greater frequency of dinoflagellate-dominated ice algal blooms at the expense of the more shade-adapted diatoms. The question therefore is, what are the ecological implications of a shift from a diatom-dominated to a dinoflagellate-dominated sea ice bloom? This we have addressed in a conceptual model ( Figure 8 ) that highlights the specific ecological conditions and consequences of diatom versus dinoflagellate blooms. In diatom-dominated ice algae blooms, most of the biomass is usually heavily concentrated at the bottom of the sea ice, but in dinoflagellate-driven blooms the biomass is more evenly distributed throughout the sea ice column. The shift that will therefore likely reduce the availability of the algal biomass for the grazing crustaceans and copepods (Bluhm et al., 2017). This could potentially imply a reduction in carbon available for higher trophic levels (Kohlbach et al., 2017). However, several studies have shown that fatty acid content, especially the nutritionally important omega-3 fatty acid, is significantly higher in dinoflagellates than diatoms (44.3 μg FA mg DW^-1 compared to 7.6 μg FA mg DW^-1, according to Peltomaa et al. (2019)). Hence, the nutritional quality of the algae will increase with a change to dinoflagellate dominance. Furthermore, diatoms are larger and often chain-forming in contrast to the smaller, unicellular dinoflagellates, giving diatoms a higher sinking velocity, proportional to the square of the cell size according to Stoke’s law (Laurenceau-Cornec et al., 2020). Miklasz and Denny (2010) determined diatom sinking rates from 4 to 7.6 m d^-1 compared to 0.3 to 3.7 m d^-1 for dinoflagellates (Kamykowski et al., 1992). The change from a diatom to a dinoflagellate dominated-bloom will therefore likely favor benthic organisms in terms of food quality, with a much higher nutritional quality of carbon reaching the sea floor. On the other hand, the rate of supply of this organic material to the benthos will be delayed by the slower sinking rates.