During the Polarstern expedition TRANSSIZ (‘Transitions in the Arctic Seasonal Sea-Ice Zone,’ PS92) from May 19 to June 28, 2015 the composition, abundance, and distribution of sea-ice meiofauna and under-ice fauna were examined. Samples were collected during eight ice stations and fourteen SUIT stations between 7.068–19.907°E and 81.007–82.211°N. Two of the eight sea-ice stations were located on the marginal shelf of north Svalbard (19 and 32), four in the Sophia Basin and on its slope (27, 31, 39, and 47), and two at the Yermak Plateau (43 and 46). In close proximity to each ice station a SUIT station was conducted. Five of the fourteen SUIT stations were located on the marginal shelf of north Svalbard (19 and 32), eight were located in the Sophia Basin and on its slope (27, 28, 31, 38, 39, 47, 48, and 49), and four on the Yermak Plateau (43, 44, 45, and 56) (Figure 1).

This region of the Arctic Ocean is characterized by a pronounced inflow of Atlantic Water along the West Spitsbergen Current and the Fram Strait branch. The latter is assumed to carry most of the oceanic heat into the Arctic Ocean (Rudels, 1987; Beszczynska-Möller et al., 2012; Rudels et al., 2013). Around 80°N the Atlantic Water inflow bifurcates due to topographic steering. Here, one part of the current propagates eastward while another part follows the topography around the Yermak Plateau. North of Svalbard the Atlantic Water is flowing close to the surface and thus contributing strongly to seasonal sea-ice melt and delays in the refreezing of the ice during fall (Rudels et al., 2004, 2013). In the past decades, observations in Fram Strait have revealed a warming of the Atlantic Water inflow into the Arctic Ocean. This warming trend reached its maximum in 2006 and has slightly decreased since then (Hughes et al., 2011; Beszczynska-Möller et al., 2012).

At each ice station several ice cores were drilled for determination of environmental properties with a Kovacs corer (Kovacs Enterprise, Roseburg, OR, United States; inner diameter: 9 cm). To determine the chlorophyll a (chl a) concentration, one ice core was taken and the bottom 10 cm were cut into two sections of approximately 5 cm length. Those ice-core sections were then put into polyvinyl chloride (PVC) jars and 200 ml of 0.2 μm filtered seawater were added per 1 cm of ice core. Melting took place in a dark room at 4°C. After melting, the volume of the meltwater was determined and subsamples were filtered through Whatman GF/F filters. The filters were put into liquid nitrogen and kept at −80°C for later analysis. In the laboratories of the Alfred Wegener Institute, pigments including chl a were extracted from the filters with 100% Acetone and homogenized. The chl a concentration was then measured with high performance liquid chromatography (HPLC) as described in Tran et al. (2013) and calculated mean values of chl a concentration were used for further analyses. For ice-temperature measurements, one ice core was put into a PVC halfpipe immediately after coring and the temperature of the bottom 10 cm of the ice core was measured with a temperature probe (Testo 720) in 5 cm intervals. Mean values of those measurements were then calculated. In order to determine bulk salinity, one ice core was taken at each ice station and the bottom 10 cm section was cut off. The segment was put into a PVC jar and melted at 4°C in the dark. The salinity of the melted ice section was then measured with a salinometer (WTW Cond.3110) as described by Peeken (2016). Snow thickness was measured at 5 different points on each coring site by using a meter stick. Mean values were calculated and used for further analyses. Sea-ice thickness was measured using an ice thickness gauge (Kovacs Enterprise, Roseburg, OR, United States), which was lowered through one sea-ice meiofauna hole at each ice station.

For sampling of sea-ice meiofauna, two replicate ice cores were drilled at each ice station with a Kovacs corer (Kovacs Enterprise, Roseburg, OR, United States; inner diameter: 9 cm) and the lowermost 10 cm of each ice core were cut off for further examination. The 10 cm ice-core sections were put separately into PVC jars. In the ship’s laboratory 200 ml 0.2 μm filtered seawater were added per 1 cm of ice core to prevent the fauna from osmotic stress during melting (Garrison and Buck, 1986). Melting took place in a dark room at 4°C. After melting, the total volume was determined and the sample was then concentrated over 10 μm gauze and fixed in 4% buffered formaldehyde solution until later quantitative analysis in the laboratories of The Arctic University of Norway (UiT). The replicates of all eight ice stations were sorted under a stereo- microscope (Zeiss Discovery.V20) and identified to the lowest taxonomic level possible. Taxonomic names were verified for correctness and synonymy using the World Register of Marine Species (WoRMS¹).

The number of individuals per liter of melted sea ice was determined by dividing the number of individuals in each sample by the volume of the sample (minus the added filtered seawater). From that the number of individuals per m² (ind.m^–2) was then calculated by multiplying by 100, by the height of the ice-core section in meters (0.1 m) and by an ice-to-water density conversion factor of 0.95 (Bluhm et al., 2018). An average was then calculated for the two replicates per station.

An array of sensors was mounted on the SUIT frame in order to collect data on sea-ice and water properties (David et al., 2015; Lange et al., 2016; Castellani et al., in press). It contained an Acoustic Doppler Current Profiler (ADCP, Nortek Aquadopp^®, Norway), which measured the velocity and direction of water passing through the net at a frequency of 2 MHZ, and a sampling interval of 1 s, and a CTD (CTD75 M, Sea & Sun Technology, Germany) with built-in fluorometer (Cyclops, Turner Designs, United States), which measured water temperature, salinity and surface water chl a concentration every 0.1 s. An altimeter (PA500/6-E, Tritech, United Kingdom) mounted on the CTD probe measured the distance between the net and the sea-ice underside. In combination with pressure data from the CTD, the distance to the sea-ice underside was used to derive ice thickness profiles over the entire SUIT haul (Lange et al., 2016; Castellani et al., in press). For examination of the ridge density the keels of ridges were detected along each profile by using the Rayleigh criterion (Rabenstein et al., 2010; Castellani et al., 2014). For all of the above mentioned parameters mean values were calculated. An observer on deck estimated the snow thickness visually during trawling. These estimates are presented as approximate ranges if variable and are therefore reported in Table 1, but were not included in further analyses. Some environmental data were not available at SUIT stations 31 and 32 due to failure of sensors.

Sampling of under-ice fauna was performed with the SUIT (van Franeker et al., 2009), which consisted of a steel frame with a 2 m × 2 m opening and two parallel, 15 m long nets attached. One net was a 0.3 mm mesh plankton net covering 0.5 m of the net opening, the other was a 7 mm half mesh covering 1.5 m of the net opening. As this study focused on mesozooplankton (0.3–20 mm), the net used to calculate the abundances of the under-ice fauna was the 0.3 mm zooplankton net. van Franeker et al. (2009), Flores et al. (2012), and David et al. (2015) provided a more detailed description of the SUIT. The catch was concentrated over a 100 μm sieve and a defined fraction (mostly one half of the original sample) was put in 4% buffered formaldehyde solution for preservation and later quantitative analysis. If the density was still high, the sample was split again with a plankton splitter prior to analysis. The samples of fourteen SUIT stations were sorted and identified to the lowest taxonomic level possible under a stereo- microscope coupled to a digital image analysis system (Leica Model M 205C, image analysis software LAR 4.2 or a Leica Discovery V8 with Axiocam) in the laboratories of the Alfred Wegener Institute. Taxonomic names were verified using the current classification of WoRMS¹. Classification of Calanus spp. individuals to species level (C. hyperboreus, C. glacialis, and C. finmarchicus) was based on length measurements and stage determination according to Madsen et al. (2001). Abundances of ind.m^–2 were calculated by multiplying the total number of individuals of each taxon in the sample by the respective splitting factor and dividing the result by the water volume sampled by the SUIT at the respective station (measured by the ADCP). The result was then multiplied by the height of the net (2 m) (David et al., 2015).

A correlation based PCA (Mardia et al., 1979) with Euclidean distance measure was applied to both environmental datasets to reveal habitat typologies of the sea ice and the upper water column (Clarke and Warwick, 2001). Spearman’s rank correlation coefficients were used to identify environmental variables with high collinearity (Clarke and Warwick, 2001; Zuur et al., 2007) in order to decide which variables to include in the respective PCA. If pairs of environmental variables had a Spearman’s rank correlation coefficient >0.8, only one of the correlated variables was chosen for further analysis based on the relevance of this variable for the scientific question and the comparability to other studies (Zar, 1984). For the ice stations, six environmental variables were analyzed and four were retained for further statistical analysis (Supplementary Table S1). For the SUIT stations, seven environmental variables were analyzed and four were retained for further statistical analysis (Supplementary Table S2). The PCA was then applied to each normalized environmental dataset.

Three diversity indices were separately calculated for sea-ice meiofauna and under-ice fauna in order to investigate patterns of diversity over the sampled area: (1) Taxa richness (S) (the number of taxa observed per station), (2) Shannon diversity (H), and (3) Pielou’s evenness (J′) (Shannon and Weaver, 1963; Pielou, 1969). We note that these and subsequent analyses were done separately for sea-ice meiofauna and under-ice fauna communities, because gear types and mesh sizes are not directly comparable.

To visualize community similarity patterns of the sea-ice meiofauna and the under-ice fauna, NMDS (Shepard, 1962; Kruskal, 1964) based on a Bray–Curtis similarity matrix (Bray and Curtis, 1957) was applied to each data set. The four environmental variables selected for the PCA were fitted into the NMDS ordination. The generation of an NMDS is an iterative procedure. It compiles a plot by successively refining the positions of the points until they fit as closely as possible to the dissimilarity relations between the samples (Clarke and Warwick, 2001). Thus, short distances of sampling sites in the NMDS plot indicate a high similarity in community structure, whereas sampling sites with low similarity are further apart in the ordination plot. The goodness-of-fit of NMDS plots was assessed by Shepard plots and stress values (Clarke and Warwick, 2001; Legendre and Legendre, 2012). A stress value of <0.05 was considered a very good representation with no prospect of misinterpretation (Clarke and Warwick, 2001). We applied the NMDS analysis with different transformations (none to 4th root) of the abundance datasets. The lowest stress values for both sea-ice meiofauna and under-ice fauna datasets were achieved by untransformed abundance data. Based on these untransformed abundance data an analysis of similarity (ANOSIM; Clarke and Ainsworth, 1993) was conducted to test for significant differences of community similarity between a priori defined geographical regions (shelf, Yermak Plateau and Sophia Basin with adjacent slope). Again, this analysis was separately performed for sea-ice meiofauna and under-ice fauna. Geographical regions were defined by a combination of water depth and geomorphological structures. Statistical analyses were conducted with the R software Version 3.5.1² (R Core Team, 2018) by using the packages: vegan (Oksanen et al., 2018), ggplot2 (Wickham, 2016), dplyr (Wickham and Ruiz, 2018), and scales (Wickham, 2018).

The ice thickness of the sea-ice cores was very similar among stations and ranged between 1.07 m (station 19) and 1.25 m (station 39) with an average thickness of 1.16 m, reflecting first-year sea ice (Table 1). Snow thickness was highest at station 19 with 0.9 m while it ranged for the other stations between 0.13 and 0.4 m, with station 43 having the lowest snow thickness. On average, the stations had a snow thickness of 0.33 m across the sampled area (Table 1). The temperature of the sea-ice cores varied between −2.2°C (station 19) and −1.2°C (station 47) with an average temperature of −1.7°C (Table 1). Bulk salinity of the bottom 10 cm sea-ice sections stood out at station 47 with a PSU of 1.65. All the other stations ranged in their PSU between 3.88 (station 27) and 5.88 (station 39), with a mean salinity of 4.47 (Table 1). Sea-ice chl a concentration of the bottom 10 cm of the ice core varied from 199 μgm^–2 (station 43) to 790 μgm^–2 (station 32) with an average of 420 μgm^–2 (Table 1).

In the PCA of environmental variables, 67% of variance could be explained by the first two components (Figure 2). The first axis (PC1) explained 46% of the variance and was mainly driven by snow thickness. Along this axis there was a distinction between the two shelf stations 19 and 32, with station 19 being strongly influenced by snow thickness (Figure 2). The second axis (PC2) explained 22% of the variance and was mainly associated with bulk salinity. Along this axis there was a noticeable, albeit not very strong, regional distinction between the shelf, the Sophia Basin with adjacent slope, and the Yermak Plateau stations (Figure 2).

In total 10 sea-ice meiofauna taxa belonging to five phyla were identified in this study (Tables 2, 3 and Figure 3). Taxa richness (S) ranged from 2 to 8 and Shannon diversity index (H′) ranged from 0.60 to 1.53 across ice stations (Table 2). The lowest S and H’ were observed at stations in the Sophia Basin and on the Yermak Plateau (stations 39 to 47) (Table 2). The ice cores of these stations contained either only Harpacticoida and copepod nauplii (stations 39, 47) or these two taxa and Ciliophora (stations 43, 46) (Figure 4). Higher S and H’ were observed at the stations on the Sophia Basin slope and the shelf of Svalbard with station 27 being the station with the highest S (8) and H′ (1.53) (Table 2). Except for Amoebozoa, all phyla were present in the ice cores of this station (Figure 4). Pielou’s evenness (J′) ranged from 0.54 to 0.90 and did not strictly follow the pattern of S and H′, though values were mostly higher further offshore where stations had low taxa richness and were dominated by Harpacticoida and copepod nauplii (Table 2 and Figure 4). The lowest value of J′ was reached at stations closer to the shelf, which were dominated by Ciliophora (Table 2 and Figure 4).

Total abundances of sea-ice meiofauna ranged from 580 ind.m^–2 (station 39) to 17,156 ind.m^–2 (station 27) (Table 2 and Figure 4). The most abundant taxon across all stations was Ciliophora with a maximum abundance of 8,562 ind.m^–2 and an average contribution of 46% to the total sea-ice meiofauna abundance (Table 3). Second most abundant were copepod nauplii with a maximum abundance of 7,682 ind.m^–2 and a share of 29%, and Harpacticoida with a maximum abundance of 4,609 ind.m^–2 and a share of 20% to the total sea-ice meiofauna abundance (Table 3). The community structure showed some geographical or seasonal/temporal variability, since Rotifera, Dinophyceae, and Amoebozoa occurred solely at the four easternmost stations (stations 19–32) on the shelf of Svalbard and the slope of the Sophia Basin sampled early during the expedition (Figure 4). Those stations showed also a higher abundance of Ciliophora compared to the offshore stations (stations 39–47). The NMDS ordination of the sampling stations along ordination axis 1 largely resembled gradual changes in community structure along a quasi-bathymetric gradient from the shelf to the Sophia Basin with adjacent slope and the Yermak Plateau stations (Figure 5). However, the conducted ANOSIM showed no significant difference in community similarity between the three regions shelf, Sophia Basin with slope and Yermak Plateau (α > 0.05, R = −0.075, p = 0.41).

The bottom depth of the SUIT stations varied considerably during this study from 188 m on the shelf (station 19) to 2,249 m in the Sophia Basin (station 38) (Table 4). The average ice coverage in the areas of the under-ice hauls was 51% (Table 4). Ice thickness ranged between 1.05 m (station 19) and 3.84 m (station 44), with a mean of 1.70 m across the sampled stations (Table 4). The density of sea-ice ridges was highest at station 39 with 11.4 ridges km^–1 and lowest at station 19 with 1.4 ridges km^–1. The average ridge density was 4.9 ridges km^–1 (Table 4). The temperature of the 0–2 m surface water varied between 1.3 and 1.8°C with an average of 1.6°C. Stations on the Svalbard shelf and slope of the Sophia Basin (stations 19, 27, and 28) had lower temperatures (1.3–1.4°C) than stations in the Sophia Basin and on the Yermak Plateau (39–56: 1.6–1.8°C) (Table 4). Salinity of the sampled SUIT stations ranged from 33.4 PSU (station 47) to 34.3 PSU (station 45) with a mean salinity of 33.8 PSU (Table 4). Chlorophyll a concentrations of the surface water varied widely between 540 μgm^–2 at station 43 and 21,158 μgm^–2 at station 47 with an average of 6,967 μgm^–2 (Table 4).

The first two components (PC1 and PC2) of the PCA explained 75% of the variance and showed a regional pattern (Figure 6). Stations located in the Sophia Basin and adjacent slope grouped together as did those over the Yermak Plateau (Figure 6). PC1 explained 50% of the variance and was mainly driven by bottom depth. Along this ordination axis there was a regional distinction between the shelf, the Sophia Basin with adjacent slope, and the Yermak Plateau stations (Figure 6). PC2 explained 25% of the variance and was mainly associated with ice thickness. Along this axis there was a clear separation of station 44 from all the other stations, which had an appreciably higher ice thickness than the other stations (Figure 6).

In total we identified 32 taxa of under-ice fauna from 8 phyla in this study (Tables 5, 6 and Figure 7). Species richness (S) ranged between 7 and 21 with no clear pattern among stations. For example, the lowest and the highest S were reached at stations 48 (7 taxa) and 49 (21 taxa), which were geographically closest to each other. Both stations were located in the southern part of the Sophia Basin and dominated by C. finmarchicus and C. glacialis (Figure 8). H′ and J′ ranged from 0.58 to 1.73 and from 0.21 to 0.80, respectively, again with no clear geographic pattern (Table 5). The highest H′ and J′ were reached at stations 27 and 31 on the slope of the Sophia Basin. Both stations were dominated by equally abundant C. finmarchicus, C. glacialis, C. hyperboreus, and nauplii. The lowest H′ and J′ were estimated for the shelf stations 19 and 32, which were dominated by copepod nauplii (Table 5 and Figure 8).

Total abundances of under-ice fauna ranged from 15 ind.m^–2 (station 44) to 6,785 ind.m^–2 (station 32) (Table 5). The highest total abundances were found at stations in the center of the Sophia Basin (stations 38, 47, and 49) and at the single shelf station, with very high Appendicularia abundance (station 32) (Figure 8). The most abundant taxa were Appendicularia with a maximum abundance of 4,915 ind.m^–2 and an average contribution of 58% to the total under-ice fauna abundance. The high share of Appendicularia to the overall abundance was due to their exceptionally high abundance (4,915 ind.m^–2) at station 32. At all other stations, their abundance ranged between 0 and 4 ind.m^–2 (Figure 8). Except for station 32, all other stations were dominated by copepod nauplii with a maximum abundance of 1,747 ind.m^–2 and a share of 23% to the total under-ice fauna abundance (55% if Appendicularia abundance at station 32 was excluded) (Table 6 and Figure 8). Copepod nauplii were especially abundant at stations, which were sampled early during the expedition (stations 19–39). C. finmarchicus was the most abundant copepod species with abundances between 2 and 260 ind.m^–2 contributing 9% to the total under-ice fauna abundance (20% if Appendicularia at station 32 excluded) (Table 6 and Figure 8). C. glacialis ranged between 0.4 and 231 ind.m^–2 contributing 6% (14% if Appendicularia at station 32 excluded), and C. hyperboreus ranged between 0.2 and 49 ind.m^–2 contributing 2% (5% if Appendicularia at station 32 excluded) to total under-ice fauna abundance (Table 6 and Figure 8). The remaining 27 taxa together made up 2% (6% if Appendicularia at station 32 excluded) of the total under-ice fauna abundance (Table 6). The three Calanus species were abundant at all fourteen stations and had a constantly high contribution to the under-ice fauna (Figure 8). Another taxon with high frequency of occurrence, but mostly low abundances (0–56 ind.m^–2) were Cirripedia, which made up 1% of the entire under-ice fauna (3% if Appendicularia at station 32 excluded) (Table 6 and Figure 8). Cirripedia resembled the pattern of copepod nauplii that were mainly abundant at stations sampled early during the expedition (stations 27–39). The NMDS showed no clear separation of distinct communities based on taxa abundances (Figure 9). However, the ordination of the sampling stations along ordination axis 2 showed a gradual change in the community similarity from the shelf to the Sophia Basin with adjacent slope and the Yermak Plateau stations (Figure 9). This spatial community change was resembled in the conducted ANOSIM showed a significant difference in community similarity between the three regions shelf, Sophia Basin with slope and Yermak Plateau (α ≤ 0.05, R = 0.26, p = 0.04).

During this study, sea-ice meiofauna abundances ranged from 580 to 17,156 ind.m^–2 in the pack ice north of Svalbard. This abundance range is consistent with that reported from pack ice in the same region two decades earlier (Gradinger et al., 1999). When excluding Ciliophora, which most ice biota studies from the Arctic Ocean do not include, the total abundances of sea-ice meiofauna in this study ranged between 580 and 13,827 ind.m^–2. This wide range of abundances is again similar to other pack-ice studies from the geographic area of our study (Schünemann and Werner, 2004; Bluhm et al., 2018) and has also been reported from other locations such as Baffin Bay and the Amerasian Basin (Nozais et al., 2001; Gradinger et al., 2005).

The taxonomic composition showed both similarities and differences to earlier sea-ice meiofauna studies. After Ciliophora, Harpacticoida and copepod nauplii were the most abundant groups, both of which have been consistently reported to occur in sea ice. Harpacticoida are common sea-ice residents and tolerant to extreme conditions such as freezing into solid ice for short periods (Damgaard and Davenport, 1994). Their capability to reproduce several times a year with no interruption during winter (Carey, 1992; Friedrich, 1997) may contribute to their high abundance in our and other studies. For pack ice in the region north-west of Svalbard, similar relative abundances for Copepoda (including Harpacticoida) were reported by Gradinger et al. (1999) and Schünemann and Werner (2004), but nauplii abundances were much lower in those studies. For the pan-Arctic, Bluhm et al. (2018) synthesized 23 studies on sea-ice meiofauna and estimated a mean abundance across all ice types and seasons of 341 ind.m^–2 for Harpacticoida and 1,141 ind.m^–2 for nauplii, recognizing that nauplii abundances are highly seasonal. Calanoid copepod nauplii tend to be encountered in sea ice in highest abundances in springtime (Bluhm et al., 2018) matching the sampling time in this study. In addition, Nozais et al. (2001) suggested that copepods (including Harpacticoida and Cyclopoida) reproduce inside the bottom ice layer itself. Consistent with these findings we suggest that the high densities of nauplii were a result of successful reproduction of copepods inside the sea-ice in our study.

The taxonomic composition of the sea-ice meiofauna in this study, however, also shows some marked differences to earlier studies. Nematoda, Acoela, and platyhelminth flatworms (in earlier studies reported as ‘Turbellaria’) were among the most abundant taxa in earlier Arctic pack-ice studies (Gradinger, 1999; Nozais et al., 2001; Gradinger et al., 2005) including the region north of Svalbard (Gradinger et al., 1999; Schünemann and Werner, 2004), but neither of these taxa were observed in this study. Essentially the same pattern (absence of flatworms and nematodes except for a single ice core) was found in a parallel study in the same area that extended over six months comprising the entire ice-covered period (N-ICE data set in Bluhm et al., 2018; Granskog et al., 2018) suggesting our finding was not barely a bias of our limited sample size or the time point of our sampling. As Nematoda are thought to colonize the sea ice in shallow waters from the sediment, where they dominate the meiofauna of the benthic interstitial (Carey and Montagna, 1982), they are predominantly found in land-fast ice or pack ice formed on shallow shelves. So far, it was assumed that Nematoda are allochthonous and do not reproduce in sea ice (Riemann and Sime-Ngando, 1997). However, recently reproduction of Nematoda was reported from shallow Arctic land-fast ice of Alaska’s north coast (Gradinger and Bluhm, 2020). Colonization patterns of ice fauna of benthic origin may change as climate driven sea-ice formation shifts. In a recent analysis of sea-ice meiofauna communities in different types of Arctic sea ice, Kiko et al. (2017) suggested that pack ice of the new Arctic will favor pelagic-sympagic species (e.g., Ciliophora) over benthic-sympagic species (e.g., Nematoda), because the now dominant first-year pack ice tends to form farther offshore and multi-year ice vanishes. For our study region, recent backtracking analyses suggest that less sea ice formed on the Siberian shelves now actually reaches the Fram Strait area (Krumpen et al., 2019). In addition, much of the ice in our research area was formed in the central Arctic Ocean (Krumpen et al., 2019). As a consequence, less sediment and, we suggest, less benthic-derived ice biota from shallow Siberian shelf areas may reached our research area. Satellite-backtracking data, which showed the origin of the specific ice floes sampled during this study, support that hypothesis (Krumpen, pers. communication). Finally, Rotifera only made up 1% of the sea-ice meiofauna abundance compared to a pan-Arctic average share of 22% (Bluhm et al., 2018) and even 33% to the pack-ice meiofauna north of Svalbard (Schünemann and Werner, 2004). According to Bluhm et al. (2018), Rotifera tend to get more abundant the later the season. Correspondingly, the early sampling season could explain the low abundance of Rotifera during this study. The abundance of sea-ice meiofauna is known to vary at least as much among seasons than regions typically reaching its peak by late spring/early summer (Schünemann and Werner, 2004; Bluhm et al., 2018). Thus, differences in abundances between studies are in part due to different seasons of sampling. A small fraction of the difference may also be explained by the fact that this study reports data from the bottom 10 cm while on average one-third of the abundance may be found outside this layer (Bluhm et al., 2018). However, inspection of the melted top 10 cm of our ice cores showed sea-ice meiofauna was absent from this layer.

Sea-ice chl a concentration can be used as an indicator for the prevalent food conditions in the pack ice for the dominant sympagic grazers in this study, Harpacticoida and copepod nauplii, who are known to feed on ice algae (Grainger and Hsiao, 1990; Gradinger and Bluhm, 2020). With a mean of 420 μg chl a m^–2, the chl a concentration was markedly lower than in a summer pack-ice study from the same region almost two decades earlier when it reached a mean of 1,200 μg chl a m^–2 (Gradinger et al., 1999). It is possible that some of the ice algae were already flushed out of the ice, given the low bulk salinities (<5 PSU) and moderate temperatures of the sea ice, which indicated the onset of ice melting during this late spring expedition. Another explanation for the low chl a concentration could be the snow thickness, which resulted in extreme low irradiance under the ice as could be shown during this study for most of the ice stations (Massicotte et al., 2019). Sea-ice chl a concentration did not correlate with sea-ice meiofauna abundance, although in other studies this was the case (Gradinger, 1999; Nozais et al., 2001).

Both chl a concentration and sea-ice meiofauna abundance can vary by orders of magnitude on small spatial scales and representative sampling should encompass large sample sizes or sensor profiles (Lange et al., 2017). It is, therefore, possible that the sample size of this study was too small to disentangle patchiness effects from the potential relationship between ice algae and sea-ice meiofauna. Although ice coring can cause inherent biases, because of the potential to lose part of the bottom section and/or brine, it is the current standard method for assessing the community structure of sea-ice meiofauna (Gradinger and Bluhm, 2009; Eicken et al., 2014). We recommend morphological studies such as this one to be combined with molecular approaches that are increasingly used for enhanced taxonomic resolution (Hardge et al., 2017; Marquardt et al., 2018; Pitusi, 2019).

The overall mean abundance of under-ice fauna was 607 ind.m^–2 during this study, which is one order of magnitude higher than in the only other SUIT study from the central Arctic Ocean from summer 2012 (David et al., 2015). The diversity of the under-ice fauna, however, was very much in accordance with studies from the Laptev Sea (Werner and Arbizu, 1999) and the central Arctic Ocean (David et al., 2015). Highest taxa richness and abundance of the under-ice fauna were found at stations with highest surface chl a values, though overall phytoplankton concentrations of the surface waters were relatively low compared to other studies (David et al., 2015; Castellani et al., in press).

It is well known that Crustacea such as Copepoda and Amphipoda can live associated with Arctic sea ice (Arndt and Swadling, 2006; Søreide et al., 2010; Hop et al., 2011). Apart from one occurrence of remarkably high abundance of Appendicularia (station 32), copepod nauplii and Calanus species were the most abundant and most frequent taxa across the sampled area. C. finmarchicus, the dominant Calanus species found in this study (Figure 3), is a boreal Atlantic species and is characteristic of the zooplankton community in Arctic areas with a strong inflow of Atlantic Water (Auel and Hagen, 2002; Rudels et al., 2013; Ehrlich, 2015). It shows a drastic decrease in abundance to <1 ind.m^–2 toward the central deep basins (Kosobokova and Hirche, 2009; Kosobokova et al., 2011; David et al., 2015). C. glacialis and C. hyperboreus in turn are rather characteristic of Arctic conditions and dominated the under-ice fauna community in the central Arctic Ocean in other studies (Kosobokova and Hirche, 2000; David et al., 2015). As a well-known inhabitant of Arctic waters and an abundant species in first-year ice dominated environments (Werner and Auel, 2005) the high occurrence of C. glacialis in the present study is no surprise. In contrast to C. finmarchicus, this species reproduces in Arctic waters. Some studies predict a replacement of C. glacialis by the smaller and less energy-rich C. finmarchicus with increasing atlantification (Bonnet et al., 2005; Richardson, 2008; Polyakov et al., 2017). Potential consequences of that replacement for the Arctic ecosystem are not sufficiently assessed yet, but a recent study suggests it might not be as severe as previously assumed given the authors found copepod lipid content to be more related to habitat conditions than species identity (Renaud et al., 2018). The relatively low abundance of C. hyperboreus in this study can be explained by the fact that it is a high Arctic and deep water species, which occurs less in the meltwater layer underneath the ice (Hop et al., 2011). Both C. glacialis and C. hyperboreus, however, do use the under-ice environment as a nursing ground and time nauplii development with the springtime peak of primary production (Falk-Petersen et al., 2008; Kosobokova and Hirche, 2009; Søreide et al., 2010). High abundances of copepod nauplii under the ice, especially at the earlier sampled stations during this expedition, may have been related to spawning below the sea ice earlier in spring and to the release of nauplii from the sea ice.

Appendicularia are common in the Arctic Ocean (Kosobokova and Hirche, 2000; Auel and Hagen, 2002; Ershova and Kosobokova, 2019), but were not reported in such high abundance in the study region before. Mumm (1993) reported a relative abundance of 3% in the Nansen Basin in summer and similar ranges for Appendicularia were also reported from Kosobokova and Hirche (2000), and Ehrlich (2015) from the central Arctic Ocean. Similar to our study, David et al. (2015) reported a single station with a high abundance of Appendicularia in the ice-water interface layer from the central Arctic Ocean (52.6 ind.m^–2). Such isolated high abundances indicate a high patchiness under the sea ice. Maybe Appendicularia can make use of the increased primary production in the Arctic measured by Arrigo et al. (2008) and others, including under-ice blooms, which occur beneath annual sea ice (Assmy et al., 2017; Wollenburg et al., 2018). As Appendicularia are soft-bodied filter feeders and capable of responding faster to shifts in primary production than crustaceans would (Hopcroft et al., 2005), their high abundance at one station might be a consequence of following a highly productive patch of water.

Amphipoda were rare in this study in comparison with their abundances in the under-ice fauna north of Svalbard almost three decades earlier (Lønne and Gulliksen, 1991). A more recent study from 2012 from the central Arctic Ocean, however, showed a similar range of sympagic Amphipoda abundance as the present study (David et al., 2015). An explanation for the decline of Amphipoda abundance could be that it seems to be positively related to the complexity of ice structures and the age of the ice (Bluhm et al., 2010; Hop et al., 2011; Bluhm et al., 2017). In the area north of Svalbard, multi-year ice has been declining dramatically since the mid-2000s (Polyakov et al., 2012; Perovich and Richter-Menge, 2015) and with it the abundance of sympagic Amphipoda (Bluhm et al., 2017). In our study, mainly ∼1.5 year-old sea ice (Peeken, 2016) covered the sampling area and might have been the reason for the low abundance of sympagic Amphipoda. Apherusa glacialis was the most abundant among the amphipod species and occurred at every station. This is in accordance with other studies (Werner and Auel, 2005; Hop et al., 2011; David et al., 2015). The species occurs under any ice type through the Arctic Ocean, though it is also found occasionally in the water column (Kunisch et al., 2020). As surface chl a concentration was very low during this study, the low abundance of A. glacialis might be related to food limitation. The larger amphipod species Gammarus wilkitzkii and Eusirus sp. were absent or very rare in the analyzed samples, but were actually found regularly, albeit in very low abundances in the larger (7 mm mesh) shrimp net of the SUIT (Schaafsma, pers. communication).