Samples were collected in January 2018 aboard the R/V Helmer Hanssen, in June 2018 aboard the RRS James Clark Ross, and in August 2018 and 2019 and December 2019 aboard the R/V Kronprins Haakon. Samples were collected in the central and northern Barents Sea in open water and within the seasonal ice zone using trawls and box cores, targeting megafaunal epibenthos (see Figure 1 and Table 1 ). Organisms were selected to cover a wide range of taxa and feeding guilds, and frozen for subsequent identification and analysis. For further information regarding seafloor composition, please refer to Freitas et al., 2020, De Freitas et al., 2022a and De Freitas et al., 2022b.

In the laboratory, each organism was identified to the lowest practical taxonomic level (usually to species for most individuals) with taxon names standardized to the World Register of Marine Species (WoRMS Editorial Board, 2022) and assigned a feeding guild based on the Arctic Traits Database (Degen and Faulwetter, 2019). Five feeding guilds were used: surface deposit feeders, subsurface deposit feeder, suspension feeders, facultative surface deposit/suspension feeders, and carnivores/scavengers. Mollusc shells, polychaete tubes, and, where possible, stomach contents were removed prior to analysis. In the case of larger organisms, specific tissues were selected for further analysis (e.g., muscle in certain decapods and gastropods). Samples were then freeze dried and homogenised using a mortar and pestle. At least 0.2 g of freeze-dried material per sample was required for HBI analysis: in the case of small organisms, multiple individuals were pooled together to reach this mass (see Table S1 for details).

Extraction of lipids from dried organisms was carried out following (Brown et al. 2017b) methodology. After addition of an internal standard (9-octylheptadec-8-ene [9-OHD]; 10 μL; 2 μg mL⁻¹), samples were saponified in 4 ml methanolic potassium hydroxide (H₂O: MeOH, 1: 9; 20% KOH) at 70°C for 60 minutes. After addition of hexane (2 mL) each sample was vortexed for 1 min and centrifuged (2500 rpm: 2 minutes). The supernatant – containing hexane and non-saponifiable lipids (NSL) – was transferred to a new vial. Three aliquots of NSL-containing supernatant were extracted this way and extracts transferred to the same vial. The NSL-containing hexane was then dried under a gentle stream of N₂ gas, removing any remaining water and methanol. Hexane (1 mL) was added to the dried NSL, and the polar and non-polar fractions were separated using open column chromatography. Samples were run through a SiO₂ column (0.5 g), which was then flushed with hexane (5 mL). Polar lipids were retained on the column, whereas nonpolar lipids (e.g., HBIs) passed through, into a new vial. These were dried under an N₂ stream and resuspended in 50 µL of hexane for GC-MS analysis.

Nonpolar lipid fractions were analysed with a Shimadzu QP2020 GC-MS system using a 30 m Rxi-5Sil column (0.25 mm i.d., 0.25 µm film), operating according to Belt et al. (2012). Analysis was run in selective ion monitoring mode to increase sensitivity, and the ions targeted represented the three HBIs of interest (mass-to-charge (m/z) ratios of selected ions were m/z 350.3: IP₂₅, m/z 348.3: HBI II, m/z 346.3: HBI III). The mass spectral intensities of these HBIs were used to calculate the H-print (Equation 1), a ratio of pelagic-to-total HBIs (Brown et al., 2014c). A high H-print (>50%) indicates more pelagic HBIs, while a low H-print indicates more sympagic HBIs (Yunda-Guarin et al., 2020).

Using a previous H-print calibration, it was then possible to estimate the proportion of organic material from sympagic sources in the tissues of organisms, with Equation 2 (Brown and Belt, 2017).

Sea ice cover duration (SID) was calculated using gridded (resolution: 3.125 km) Advanced Microwave Scanning Radiometer 2 (AMSR 2) satellite data downloaded from https://seaice.uni-bremen.de/sea-ice-concentration/ (Spreen et al., 2008). Using MATLAB, SID was calculated as the number of days in a 365-day period where ice concentration was more than 15% in a grid cell. The commonly used 15% concentration threshold allowed for comparison with a similar study in the Pacific Arctic (Koch et al., 2020a). For each sampling event, SID was calculated for the grid cell containing the station, using the 365-day period preceding the sampling event. Figure 1 shows SID for the sampling area in the year preceding the first sampling event.

Numerical analyses were conducted in R (version 3.6.1). After examination, the data were transformed using the logit function (Equation 3), as they represent proportions (of sympagic OM). Regression models were used to examine the relationship between logit-transformed assimilated sympagic OM and SID. Due to the uneven distribution of sampling effort among seasons (see Section 3.3), with summer sampling being much more extensive, two analyses were conducted: (a) a model with only summer samples, to discern any effect of SID and feeding guild on sympagic OM assimilation across the whole study area and (b) a model restricted to stations south of (and including) NL4, but including both seasons to determine the effect of SID, season and feeding guild on sympagic OM assimilation. If a covariate was found to have no significant effect (significance level p<0.05) on sympagic OM assimilation, it was removed from the model, to have the simplest explanatory model for each analysis. In such cases, an ANOVA was used to compare the original and simplified models to confirm there was no significant difference between them.

Overall, 126 samples from 10 stations were analysed, and 12 were discarded due to a lack of detection of at least one HBI of interest, thus preventing H-print calculation. The proportion of OM from sympagic sources in these samples was highly variable, with values ranging from 0.4-95.8%. The proportion of sympagic OM was generally higher in the northern half (north of 78.5°N) of the transect but remained very variable: mean ± standard deviation of 61.5 ± 21.7% in the northern half, versus 8.9 ± 13.4% in the southern half, with ranges of 5.8-95.8% and 0.4-65.4%, respectively. Sampling effort was unequal among taxa and feeding guilds: out of 25 taxa, 16 only had one to two samples, while individual feeding guilds had between 6 and 41 samples ( Table S1 ; Figure S2 ).

Sea ice duration at the 10 stations lasted between 0-245 days ( Table 1 ) with mean and median SID at 110 and 123 days, respectively. The station with the lowest SID was NL1 (0 days in 2018, and 20 days in 2019) and the station with the longest SID was NL6 in winter 2019 (245 days in 2019). Regression models showed a significant effect of SID on proportion of sympagic OM assimilated in the summer, but no effect of feeding guild ( Figure 2 ; Table 2 ). Figures 2B–F shows the similar positive relationship with SID for all feeding guilds, except suspension feeders (though this is likely due to low sample size for this feeding guild). While there is a strong effect of SID on proportion of assimilated sympagic carbon, it is important to note the large variation, even at similar SID ( Figure 2A ). A similar regression analysis was conducted with Ctenodiscus crispatus samples, the taxon with the most samples (23 in summer), which covered most of the study area ( Figure 3 ). Within this taxon, there was a similarly strong relationship between SID and proportion of sympagic OM, again with large variability at similar SID.

Winter sampling was complicated by ice presence, and there were therefore more summer samples than winter samples (87 and 27 samples, respectively). Importantly, while the summer samples were from the whole study area, winter samples were mostly restricted to the southern half of the transect, and seasonal analysis was therefore restricted to this part of the transect. This included stations south of 79°N: NL1-NL4, and PR1-PR3, with SID from 0-160 days. There were 64 summer samples and 26 winter samples used for seasonal analysis. Regression models showed a significant effect of SID, season, and their interaction on proportion of assimilated sympagic OM, but no significant effect of feeding guild ( Figure 4 ; Table 3 ). However, note that both feeding guild as a covariate, and its interaction with SID had non-significant but low p-values (p=0.089 and p=0.070, respectively): this was largely driven by surface deposit feeders, which had higher proportions of sympagic OM at higher SID in summer than winter, and to some degree by suspension feeders, which had a low sample size and a different response to SID than other feeding guilds ( Figure 2D ). The regression presented does not include the feeding guild covariate, as an ANOVA reported no significant difference between the models with and without feeding guild ( Table 3 ). In summer, the proportion of sympagic OM increased more rapidly with SID than in winter, due to higher values at higher SID (more than 100 days). However, in stations where SID was between 0-57 days, the proportions of assimilated sympagic OM remained similar in both seasons.

NL4 only had summer samples and was located much further North than other stations used for seasonal analysis ( Figure 1 ) but had a similar SID to some of them (e.g., 160 days for NL4, 156 for PR3 – Table 1 ). It was therefore included in the seasonal analysis, to extend the dataset. A regression without NL4 (i.e., only stations south of 78°N) also showed a significant effect of season and confirmed that the presence of this northern station was not driving the seasonal difference (inset in Figure 4 ; Table S2 ). There was one winter sample from a station with higher SID, and it had a high proportion of assimilated sympagic OM (C. crispatus from P4: SID – 245 days year^-1; proportion of assimilated sympagic OM – 73.8%). However, it was not included in analyses as it did change the general seasonal response, and would have led to an unbalanced number of summer samples from higher SID ( Table S3 ; Figure S3 ).

The spatial distribution of sea ice duration followed a south-to-north gradient, generally increasing with latitude ( Figure 1 ). Previous studies have established a link between degree of ice cover and the input of sympagic material to the ecosystem. Studies on primary production found that sympagic production contributes a larger proportion of annual primary production where there is increased ice presence (Gosselin et al., 1997; Wassmann and Reigstad, 2011), while palaeontological studies have connected stronger levels of ice algae signals with higher sea ice concentrations (Stein et al., 2017). In this study, there was a significant increase in the contribution of sympagic OM to benthic diets with increasing duration of sea ice. This strong relationship indicates that sea ice cover plays an important role in the availability and uptake of sympagic versus pelagic OM by Barents Sea megabenthos.

The overall relationship of increasing proportion of sympagic OM assimilated with sea ice duration was observed across all feeding guilds (except in suspension feeders—see below). However, there is notably high variability in the assimilation of sympagic OM, even within the same feeding guild at locations with similar ice durations. While this could be due to the taxonomic differences between organisms within the same feeding guild, the well-sampled Ctenodiscus crispatus also show this variability within-species ( Figure 3 ), indicating there are other processes leading to this variability. One such process may be a patchy input of sympagic material greatly increasing the sympagic OM available to certain organisms.

Areas of particularly high sympagic OM sedimentation can be very localised due to the inhomogeneous deposition of algal aggregates (Boetius et al., 2013), leading to higher proportions of sympagic OM assimilated in organisms at these locations. Ice algal production and biomass in turn is very patchy because snow and ice thickness and sediment load strongly modulate production and themselves vary on small scales (e.g., Lee et al., 2008; Gradinger, 2009). Even ice-associated fauna can show highly variable uptake of sympagic OM depending on the availability of both sympagic and pelagic OM (Brown et al., 2017b; Kohlbach et al., 2022). Patchiness of OM deposits may also be caused (or further enhanced) by differences in bathymetry – such as depressions which concentrate OM – or near bottom currents redistributing sinking OM (Lovvorn et al., 2020).

In addition, individual organism responses to sympagic and pelagic OM input may further increase variability. The Arctic benthos exhibits high dietary plasticity and employs many strategies to adapt to variable food quality and quantity. These strategies include omnivory, switching sources of OM, and changing feeding behaviour by searching for fresh OM or selectively feeding on OM from specific sources (Sun et al., 2009; Boetius et al., 2013; Morata et al., 2015; Kędra et al., 2019). Benthic organisms also utilise different types of fatty acids from ice algae and phytoplankton, adapting their assimilation strategies to meet energetic requirements (Sun et al., 2007). Certain deposit-feeding organisms may have selectively fed on either sympagic or pelagic OM depending on the quality of OM (Sun et al., 2009), whereas mobile deposit feeders (e.g., Pontaster tenuispinus, Strongylocentrotus spp.) may have responded to concentrated deposits of fresh OM, such as ice algae aggregates (Boetius et al., 2013). As the sea ice extent in the Barents Sea has large interannual variability (Herbaut et al., 2015), flexibility in responding to variable inputs of different types of OM may be crucial for the megabenthos in this region, by allowing them to adapt their feeding and assimilation methods to make use of the OM reaching the seafloor. Therefore, while the predicted retreat of sea ice from the Barents Sea (Onarheim and Årthun, 2017) may result in less sympagic OM availability to the benthos, reduced ice cover often leads to higher pelagic primary productivity due to higher light availability (Arrigo & van Dijken, 2015). In the Barents Sea, the southwestern, ice-free areas are associated with higher primary production, generally have higher infaunal abundance (Cochrane et al., 2009) and higher megafaunal abundance and biomass (Jørgensen et al., 2015), indicating that benthic communities may benefit from reduced ice cover.

Suspension feeders were the only feeding guild which did not show a strong increase in proportion of assimilated sympagic OM with sea ice duration ( Figure 2D ). This trend is based on a very small number of samples: only four suspension feeders were analysed for the analysis of the effect of sea ice duration. With more samples, there most likely would have been a stronger increase in the proportion of sympagic carbon with longer sea ice duration – but possibly not as strong as other feeding guilds, reflecting a lower uptake of sympagic OM by suspension feeders. Previous studies have found that suspension feeders generally assimilated little sympagic OM compared to other benthic consumers, which these authors attributed to the short residence time of this organic material in the water column (McMahon et al., 2006; Schollmeier et al., 2018; Koch et al., 2020a). The higher uptake of sympagic material by Arctic deposit feeders, in contrast, is well-documented and attributed to the large, early season sedimentation pulse of this fast-sinking organic material to the seafloor at a time when zooplankton grazing pressure is low (McMahon et al., 2006; Roy et al., 2015; Schollmeier et al., 2018; Kędra et al., 2019). Carnivore/scavengers had the same trend as most other feeding guilds, indicating that their H-print was similar to the community of organisms they preyed upon. This suggests that the ratio of pelagic and sympagic HBIs remains largely unchanged as they move through the benthic food web.

Interestingly, surface deposit feeders and subsurface deposit feeders had similar proportions of assimilated sympagic OM, indicating mixing of OM to deeper layers of sediment. Arctic organisms respond rapidly to inputs of fresh organic material through increased oxygen demand and bioturbation activity (Morata et al., 2015). Bioturbation activity can rapidly mix organic matter into the sediment, for example freshly deposited detritus was mixed to a depth of 10 cm within nine days in the North Atlantic deep-sea (Graf, 1989). If mixing occurs on similar timescales in the Barents Sea, freshly deposited OM would rapidly be available to both surface deposit and subsurface deposit feeding organisms, leading to similar HBI ratios. Current knowledge suggests that HBI signals likely reflect recent (1-2 months) carbon inputs, as supported by HBI depuration rates [around 30 days in whole molluscs (Koch et al., 2020a)] indicating relatively rapid turnover, and HBI signals detecting seasonal differences in OM assimilation (see next section). Therefore, similarities observed between feeding guilds are likely due to ecological processes, rather than long-term stability of HBIs in the food web or environment.

Season affected the relationship between sea ice duration and proportion of sympagic OM: at longer sea ice durations, a higher proportion of sympagic OM was assimilated in summer than in winter. This is presumably due to a higher availability and more recent deposition of sympagic OM in the summer. However, the single winter data point from longer ice duration suggests that there can also be a high proportion of assimilated sympagic OM in this season. Sympagic production contributes to a higher proportion of total annual primary production at higher latitudes (Hegseth, 1998), and therefore the input and signal of sympagic OM likely remains in the system for longer. However, due to the variability observed in the summer, it is difficult to know how representative this single point is of the winter sympagic OM assimilation in the northern SIZ.

At shorter sea ice durations, the proportion of assimilated sympagic OM did not change seasonally, because there were low proportions in both seasons, presumably due to little sympagic production occurring in these areas. The higher summer values at longer sea ice durations indicate that sympagic OM can contribute to a large part of benthic diets in the summer in the ice-covered part of the Barents Sea shelf. Similarly high summer values have also been observed in the Chukchi sea, with both studies together jointly confirming the importance of sympagic OM to benthic food webs at this time on Arctic inflow shelves (Koch et al., 2020a). In contrast, sympagic OM contributed little to late summer pelagic food webs, even during an ice-heavy year in the Barents Sea (Søreide et al., 2006; Kohlbach et al., 2021a). This difference between benthic and pelagic food webs suggests that sympagic OM is present mostly in sediments in the late summer Barents Sea, and likely originated from the ice-algal spring bloom.

The lower winter contribution of sympagic OM to the benthos at higher sea ice durations is presumably due to a proportionally larger input of pelagic OM after the ice algal bloom [and increasingly from fall blooms (Ardyna et al., 2020)], as well as the summer consumption of sympagic OM (see Figure 5 ). For example, the stronger response of sympagic OM assimilation to SID in summer in surface deposit feeders suggests that these mobile deposit feeders responded to areas of higher sympagic OM deposition present only during the summer. Sea ice formation in the study area occurs in the autumn or winter, and there is therefore very little to no sympagic production which occurs outside of the spring bloom period. A seasonal difference in sympagic OM utilisation from summer to winter was not observed in zooplankton in the study area during the same period, because they generally did not feed on sympagic OM in the late summer (Kohlbach et al., 2021b). The presence (albeit low) of sympagic OM in the benthos in winter indicates that a repository of sympagic material remains in this season, although in low amounts. This is consistent with findings in the Chukchi sea, where sympagic material was found in surface sediments throughout the winter period (Koch et al., 2020b), and the sympagic OM assimilated by benthic invertebrates decreased from summer to winter, as was observed in this study.

As sympagic OM input has been shown to affect benthic food web structure by shifting ecological niche space of organisms and modifying feeding of higher trophic levels (Yunda-Guarin et al., 2020), the seasonal difference observed in sympagic OM assimilation may also cause similar seasonal changes in the benthic food web structure of the Barents Sea. Previous food web studies have, however, reported little seasonal change in Arctic benthic food web structure due to the dietary flexibility of benthic consumers (Renaud et al., 2011; Kędra et al., 2012; Bridier et al., 2021). However, these studies were conducted in areas with very little sympagic production, such as Young Sound (Greenland) and Kongsfjorden (Svalbard) and may have therefore not fully captured the impact of sympagic OM on benthic consumers. The large seasonal difference in assimilated sympagic OM observed here suggests there may a seasonal change in food web structure of the Barents Sea benthos where sympagic OM contributes to a large part of the diet in summer.

The ratio of sympagic-to-pelagic OM in benthic consumers was high relative to the contribution of sympagic production to total annual production in the Barents Sea SIZ. Sympagic production accounts for 16-22% of annual production (5.3 g C m^-2) in the Barents Sea SIZ (Hegseth, 1998), and 27% of daily primary production (11 mg C m^-2 d^-1) north of Svalbard (Ehrlich et al., 2021). In the northern half of our transect, the proportion of OM from sympagic sources assimilated by benthic consumers was 61.5%. While these numbers are not directly comparable due to what they represent (rates versus amount assimilated), the large discrepancy suggests a possible mismatch between supply and demand when extrapolating to total benthic demand. Another biomarker method also showed high contributions of ice-derived matter to benthic organisms in ice-covered waters, but the sympagic signal may have included feeding on refractory material (Søreide et al., 2013). Although there is less sympagic than pelagic production in the Barents Sea SIZ, phytoplankton blooms generally occur after the ascent of grazing copepods, leading to less sinking pelagic production reaching the benthos (Reigstad et al., 2008; Tamelander et al., 2008). Benthic utilisation of the OM that reaches the seafloor may also favour the uptake of sympagic OM through selective feeding or assimilation (e.g., through incorporation of the polyunsaturated fatty acids in ice algae) (McMahon et al., 2006). These processes would lead to higher-than-expected proportions of sympagic OM in benthic consumers when considering the proportion of annual primary production which occurs in sea ice. However, the high proportions of sympagic OM in the northern part of this study may also be due to the time of sampling, where ice retreat in the northern Barents Sea SIZ occurs in July to early August, with concurrent release of ice algae (Ji et al., 2013; Dong et al., 2020). We suggest therefore, that our sampling probably coincided with the period of maximum availability and uptake of sympagic OM by benthic consumers. This is supported by high levels of sympagic HBIs in flux-feeding zooplankton in the summer, indicating recent sympagic matter sedimentation (Kohlbach et al., 2021a). Sympagic OM values for the northern half of the transect may therefore reflect a short period of high uptake of sympagic OM by the benthos (see Figure 5 ).

The high proportion of sympagic OM observed in benthic consumers may also be an overestimation due to the HBI-based method used. HBIs are diatom-specific biomarkers and, while sea ice algae communities are usually dominated by diatoms, phytoplankton communities can also include a wide variety of other taxa which are not accounted for in HBI analysis (Wassmann et al., 1999; Hegseth and von Quillfeldt, 2022). In 2018, diatoms dominated the spring phytoplankton bloom (up to 270 µg C L^-1), but the community shifted to a mix of diatoms and nanoflagellates (mostly Dinobryon sp. and Phaeocystis pouchetii) in summer (maximum diatom biomass: 163 µg C L^-1, maximum nanoflagellate biomass: 86 µg C L^-1) (K. Davidson, SAMS, unpubl. data). In the summer of 2019, taxonomic and fatty acid analysis pointed to a mix of diatoms and flagellates in the phytoplankton community (Kohlbach et al., 2021a). While nanoflagellates were present in both study years, their contribution to benthic diets would not have been represented by HBI levels in benthic consumers. However, some of these small-celled producers are less likely to sink to the benthos than heavy, silicified diatoms, and therefore are more susceptible to grazing and disintegration in the water column (Heiskanen and Kononen, 1994; Fahnenstiel et al., 1995; Kiørboe et al., 1996). Still, it is documented that flagellates do contribute to benthic invertebrate diets (Napolitano et al., 1997; Paar et al., 2019), and the HBI analysis would therefore have overlooked this pelagic contribution to benthic diets.

Increased degradation of pelagic HBIs relative to sympagic HBIs could also have led to overestimation of the proportion of sympagic OM in benthic consumers. HBIs in senescent algal cells are susceptible to photooxidation, and the presumed longer residence time of pelagic OM in the water column may result in higher degradation of pelagic HBIs compared to sympagic HBIs (Rontani et al., 2016; Rontani et al., 2019). With these methodological limitations in mind, it is suggested that a multiple trophic marker approach may be the most effective in future studies (such as used by Kohlbach et al., 2016; Yunda-Guarin et al., 2020; Kohlbach et al., 2021a) when investigating the role of sympagic OM in the Arctic system (see Leu et al., 2020 for an in-depth investigation of various sympagic trophic markers). In addition, analysis of sympagic and pelagic particulate organic matter would improve confidence in the results obtained. As HBIs are produced by certain diatoms which represent only a small proportion of total algal abundance (Brown et al., 2014c) this would provide information on the presence of HBI-producing taxa during the study period, as well as give baseline values of the HBIs produced. The inclusion of more sympagic and pelagic HBIs in the H-print calculation on a study-by-study basis could make results even more robust, though it would also require adapting the calculation used to estimate sympagic OM in organisms (Equation 2).

While the absolute values presented here may represent an overestimation of the contribution of sympagic OM to benthic consumers, the spatial and seasonal data provide new insights into the role of sympagic production for Barents Sea benthic food webs and are consistent with results from the Chukchi sea and with sea ice patterns (Koch et al., 2020a). Based on our results and the literature, we present a figure illustrating the contribution of sympagic and pelagic OM to Arctic benthic food webs under short and long ice cover ( Figure 5 ). Our data do not reflect the magnitude of OM available to the system, but previous studies indicate a much lower feeding rates and activity during the winter season than in other seasons, due to less labile OM in the sediments (Renaud et al., 2007; Bourgeois et al., 2017). Even though current data do not cover the whole range of seasons, especially at higher latitudes, this conceptual figure represents a step toward a better understanding of how pelagic and sympagic production fuel benthic ecosystems.