Mesopelagic Sound Scattering Layers of the High Arctic: Seasonal Variations in Biomass, Species Assemblage, and Trophic Relationships

Mesopelagic sound scattering layers (SSL) are ubiquitous in all oceans. Pelagic organisms within the SSL play important roles as prey for higher trophic levels and in climate regulation through the biological carbon pump. Yet, the biomass and species composition of SSL in the Arctic Ocean remain poorly documented, particularly in winter. A multifrequency echosounder detected a SSL north of Svalbard, from 79.8°N to 81.4°N, in January 2016, August 2016, and January 2017. Midwater trawl sampling confirmed that the SSL comprised zooplankton and pelagic fish of boreal and Arctic origins. Juvenile beaked redfish dominated the fish assemblage in January and Arctic cod in August. The macrozooplankton community mainly comprised the medusa Cyanea capillata, the amphipod Themisto libellula, and the euphausiids Meganyctiphanes norvegica in August and Thysanoessa inermis in January. The SSL was located in the Atlantic Water mass, between 200-700 m in August and between 50-500 m in January. In January, the SSL was shallower and weaker above the deeper basin, where less Atlantic Water penetrated. The energy content available in the form of lipids within the SSL was significantly higher in summer than winter. The biomass within the SSL was >12-fold higher in summer, and the diversity of fish was slightly higher than in winter (12 vs. 9 species). We suggest that these differences are mainly related to life history and ontogenetic changes resulting in a descent towards the seafloor, outside the mesopelagic layer, in winter. In addition, some fish species of boreal origin, such as the spotted barracudina, did not seem to survive the polar night when advected from the Atlantic into the Arctic. Others, mainly juvenile beaked redfish, were abundant in both summer and winter, implying that the species can survive the polar night and possibly extend its range into the high Arctic. Fatty-acid trophic markers revealed that Arctic cod mainly fed on calanoid copepods while juvenile beaked redfish targeted krill (Thysanoessa spp.). The relatively high biomass of Arctic cod in August and of redfish in January thus suggests a shift within the SSL, from a Calanus-based food web in summer to a krill-based food web during winter.

Mesopelagic sound scattering layers (SSL) are ubiquitous in all oceans. Pelagic organisms within the SSL play important roles as prey for higher trophic levels and in climate regulation through the biological carbon pump. Yet, the biomass and species composition of SSL in the Arctic Ocean remain poorly documented, particularly in winter. A multifrequency echosounder detected a SSL north of Svalbard, from 79.8 to 81.4 • N, in January 2016, August 2016, and January 2017. Midwater trawl sampling confirmed that the SSL comprised zooplankton and pelagic fish of boreal and Arctic origins. Arctic cod dominated the fish assemblage in August and juvenile beaked redfish in January. The macrozooplankton community mainly comprised the medusa Cyanea capillata, the amphipod Themisto libellula, and the euphausiids Meganyctiphanes norvegica in August and Thysanoessa inermis in January. The SSL was located in the Atlantic Water mass, between 200-700 m in August and between 50-500 m in January. In January, the SSL was shallower and weaker above the deeper basin, where less Atlantic Water penetrated. The energy content available in the form of lipids within the SSL was significantly higher in summer than winter. The biomass within the SSL was >12-fold higher in summer, and the diversity of fish was slightly higher than in winter (12 vs. 9 species). We suggest that these differences are mainly related to life history and ontogenetic changes resulting in a descent toward the seafloor, outside the mesopelagic layer, in winter. In addition, some fish species of boreal origin, such as the spotted barracudina, did not seem to survive the polar night when advected from the Atlantic into the Arctic. Others, mainly juvenile beaked redfish, were abundant in both summer and winter, implying that the species can survive the polar night and possibly extend its range into the high Arctic. Fatty-acid

INTRODUCTION
Sound scattering layers (SSL) appear on active acoustic echograms as extensive echoes above background noise resulting from aggregations of organisms in the water-column (Proud et al., 2015). Mesopelagic SSL composed of a mixture of macrozooplankton and small pelagic fish are ubiquitous in the world's oceans (Irigoien et al., 2014;Proud et al., 2018). The global biomass of mesopelagic fish is estimated to 10 billion metric tons (Irigoien et al., 2014;St. John et al., 2016) and mesopelagic organisms represent important prey for higher trophic levels (Naito et al., 2013) and a key component of the biological carbon pump through diel vertical migrations (DVM; St. John et al., 2016;Aumont et al., 2018). Scientific interest in the global distribution and, to a lesser extent, composition of the SSL has increased over the last decade, partly due to recent projects of commercially harvesting mesopelagic fish. Large-scale studies were conducted in the tropical and subtropical regions (Irigoien et al., 2014;Aksnes et al., 2017), Antarctica (e.g., Jarvis et al., 2010), Pacific Ocean (Benoit-Bird and Au, 2004) and the North Atlantic (Pepin, 2013;Fennell and Rose, 2015). However, knowledge on species composition and total biomass within high latitude SSL remains scarce (Siegelman-Charbit and Planque, 2016).
Recent studies documented the composition  and diel vertical migrations  of SSL north of Svalbard in late summer, and reported high occurrences of fish and zooplankton of boreal origin. Wassmann et al. (2015) suggested that boreal organisms advected with Atlantic Water (AW) into the Arctic cannot survive and reproduce in their new environment, which represents a literal dead-end for most advected taxa. The polar night could limit the survival of pelagic species of boreal origin in the Arctic Ocean due to inferior feeding conditions imposed by the extreme light climate (Kaartvedt, 2008). Yet, high levels of biological activity prevail during the polar night and demersal fish of Atlantic origin such as haddock (Melanogrammus aeglefinus) and Atlantic cod (Gadus morhua) feed actively in winter . However, it is difficult to assess if pelagic organisms survive the polar night at high latitudes as the biomass and assemblage of the SSL are, at best, poorly documented in wintertime. Such information is nonetheless critical to understand the ongoing borealization of the Arctic fish community (Fossheim et al., 2015).
The winter diet of carnivorous zooplankton and pelagic fish represents an additional knowledge gap in the high Arctic. The herbivorous calanoid copepods efficiently store large amount of lipids from their algal diet during the spring bloom of diatoms (Falk-Petersen et al., 2007) and are at the base of the pelagic food web in summer (Falk-Petersen et al., 1990). In winter, flagellates dominate the algal community (van Leeuwe et al., 2018), and most Calanus spp. are in diapause at depths of hundreds or thousands of meters (Dale et al., 1999;Falk-Petersen et al., 2009b). Carnivorous zooplankton and pelagic fish must then switch their diet from copepods to different prey items. Further knowledge about seasonal variations in trophic ecology is needed to increase our understanding of the energy flow through pelagic food webs in Arctic marine ecosystems (Haug et al., 2017).
We conducted acoustic-trawl surveys in the European Arctic north of Svalbard (79.8 to 81.4 • N) in winters 2016 and 2017 and in summer 2016 to document seasonal and spatial variations in the distribution, species assemblage, and relative biomass of pelagic fish and macrozooplankton within the SSL. Here, we link these changes to environmental variables and assess seasonal variations in the flow of energy indicated by fatty-acid trophic markers.

Study Area
We conducted acoustic-trawl surveys in the northern Barents Sea, north of Svalbard, on board the R/V Helmer Hanssen from 14 to 16 January and 28 to 29 August 2016, and from 11 to 14 January 2017. The transects aimed northwards from Rijpfjorden (22.3 • E, 80.2 • N), crossed the shelf and slope, and stopped at >81 • N before heading south toward Smeerenburgfjorden (11.0 • E, 79.8 • N; Figure 1A). In January 2017, the ship followed the opposite direction (i.e., started close to Smeerenburgfjorden). Bottom depth ranged from <50 m close to shore to ∼2,000 m at the end of the transects.

Hydroacoustics
The keel-mounted Simrad EK60 R split-beam echosounder continuously recorded hydroacoustic data at 18, 38, and 120 kHz. The ping rate was set to maximum and pulse length to 1,024 µs. The echosounder was calibrated annually (Supplementary Table 1) using the standard sphere method (Demer et al., 2015). A Seabird 911 Plus CTD R recorded temperature and conductivity regularly throughout the transects from which we could derive profiles of temperature and salinity, sound speed (Mackenzie, 1981), and the coefficient of absorption at each frequency (Francois and Garrison, 1982).
Acoustic data were scrutinized and cleaned with Echoview R 8. We used Echoview's algorithms to remove background noise, impulse noise, and attenuated noise signals (De Robertis and Higginbottom, 2007;Ryan et al., 2015). A minimum signal to noise ratio threshold of 10 dB was applied. Samples with a lower signal to noise ratio were considered indistinguishable from background noise and were excluded from the analysis with the background noise algorithm. Korneliussen et al. (2018) recommend to average samples before classifying targets to reduce random variability of measurements of relative frequency responses and minimize biases introduced by differences in the volume sampled at the different frequencies. The choice of analysis cell size represents a trade-off between decreasing variability in the observed relative frequency response and minimizing violations of the assumption that backscatter is dominated by a single organism (De Robertis et al., 2010). Here, the 18 and 38 kHz samples were averaged over 5 × 5 m using Echoview's "resample" algorithm. We selected the size of the cells based on previous studies on mesopelagic SSL (e.g., 5 × 5 m;D'Elia et al., 2016) and pelagic organisms in the Arctic (e.g., 5 m × 5 pings; De Robertis et al., 2010).
Due to low SNR below 200 m at 120 kHz, only the 18 and 38 kHz datasets could be used to classify acoustic targets, which reduces the classification capacity as more frequencies increase the resolution of frequency responses (Korneliussen et al., 2018). Yet, these two frequencies can provide valuable information to classify functional groups of mesopelagic scatterers. Using the difference in Mean Volume Backscattering Strength ( MVBS in dB re 1 m −1 ) at 18 kHz and 38 kHz within each cell, D'Elia et al. (2016) defined a classification tree for mesopelagic organisms based on the size and categories of scatterers. We applied the same classification tree, where cells−14 dB < MVBS 18kHz−38kHz < −3 dB were assigned to crustaceans and small (1.5-15 cm) non-swimbladdered fish; −3 dB < MVBS 18kHz−38kHz < 0 dB to large (15-60 cm) non-swimbladdered fish; 0 dB < MVBS 18kHz−38kHz < 3 dB to gelatinous zooplankton, Cephalopoda and Pteropoda; and 3 dB < MVBS 18kHz−38kHz < 12 dB to swimbladdered fish between 2.5-30 cm. Because this classification tree is based on minimum and maximum sizes and models of frequency response for each functional group rather than specific species, we could apply it to classify targets in the Arctic as long as the scattering properties and size of the organisms forming each functional group were the same. The nautical area scattering coefficient (s A ; m 2 nmi −2 ) was then integrated over the depth of the SSL and 10-nmi wide windows for each functional group. The width of the windows was selected by plotting the standard deviation (SD) resulting from different window sizes to find the width at which the size of the windows did not impact the variance (i.e., the width at which SD reaches its asymptote). This method has the advantage of tending toward the real variance of the transects as it removes artificially high SD values resulting from a random selection of the size of the window.
For each season, the weight-based target strength function (TS W in dB kg −1 ; Table 1) of each fish and zooplankton species was calculated on the basis of its mean target strength (TS N ) and weighted mean weight in trawl samples (e.g., Benoit et al., 2008). The TS N was calculated by feeding the weighted mean length from trawl samples into TS to length relationships ( Table 1). We then estimated the mean integrated biomass based on the average s A of each scattering group and on the proportion and TS W (in the linear domain) of each species (Simmonds and MacLennan, 2005). To focus on mesopelagic SSL and avoid biases from shallower coastal areas, we discarded coastal areas <200 m and only used stations and acoustic data outside of Smeerenburgfjorden and Rijpfjorden for biomass calculations. We calculated the Weighted Mean Depth (WMD) of the SSL for each survey to document seasonal changes in vertical distribution (e.g., Knutsen et al., 2017).

Fish and Zooplankton Sampling
We deployed a Harstad pelagic trawl within the SSL to groundtruth the acoustic signal. Five to six pelagic trawl deployments were conducted for a period varying from 20-80 min at ca. three knots during each survey ( Figure 1A and Supplementary Table 2). The Harstad trawl had an opening of 18.28 × 18.28 m and an effective height of 9-11 m and width of 10-12 m at 3 knots. The mesh size of the inner liner of the cod end was 10 mm. All organisms were identified to species or genus onboard. We recorded the total number and weight of each species. When >30 specimens of a given species were collected, we measured and weighted a subsample of 30 individuals. To identify spatial and seasonal changes in the SSL community, a cluster analysis using the unweighted pairgroup average method with arithmetic means and the Bray-Curtis similarity index was conducted following the multivariate approach described by Darnis et al. (2008). Biomass of each taxon was expressed as a proportion of total catch at each station. Data were transformed (arcsin square-root) prior to the production of the similarity matrix to reduce the weight of highly abundant species and increase the impact of species likely to have been under-sampled (e.g., decapod species, gelatinous zooplankton such as cnidarians, and ctenophores). Verification of station groupings was strengthened using a nonmetric multidimensional scaling analysis (NMDS) applied to the same similarity matrix. A similarity of percentages (SIMPER) analysis conducted with the software PRIMER V.6 identified the taxa most responsible for the intragroup similarity and dissimilarity among pelagic assemblages. We used all stations for the multivariate analysis.
This study was conducted with permission from the Governor in Svalbard and followed the strict regulations regarding health, environment and safety enforced at UNIS and UiT The Arctic University of Norway. This study was carried out in accordance with the Norwegian animal welfare act and was approved by the Department of Arctic and Marine Biology at UiT.

Lipid Analysis
We analyzed the concentrations of Fatty Acid Trophic Markers (FATM) to document seasonal variations in energy flow between pelagic trophic levels (Dalsgaard et al., 2003). Subsamples of the most abundant vertebrate and invertebrate species (i.e., Sebastes mentella, Boreogadus saida, Leptoclinus maculatus, Benthosema glaciale, Arctozenus risso, Clupea harengus, Themisto libellula, Thysanoessa inermis, and Meganyctiphanes norvegica) were frozen at −80 • C for lipid analysis. We analyzed five to ten individuals of each of these species for 37 fatty acid (FA) and 8 fatty alcohol (Falc) compositions via gas chromatography (Kattner and Fricke, 1986). The sum of FA and Falc was expressed as total lipids. The concept of FATM is based on the knowledge that FA and Falc patterns characterize specific taxa of primary producers and zooplankton species, and that the patterns are transferred relatively unchanged through the food chain. Based on the FA and Falc composition, several biomarker ratios have been calculated. Here, the 18:1(n-9)/18:1(n-7) ratio indicated carnivory in zooplankton (Graeve et al., 1997;Auel et al., 2002), while 20:5(n-3)/22:6(n-3) and 16:1(n-7)/16:0 ratios differentiated diatom and flagellate based diets, respectively (Graeve et al., 1994). We used the highly polyunsaturated vs. saturated fatty acids ratio (PUFA/SAFA) to track the flow of energy, and the 22:1(n-11)/20:1(n-9) ratio to differentiate within diets based on Calanus species (∼1.8 or higher refers to C. hyperboreus, 1.1 to C. finmarchicus, and 0.7 or lower to C. glacialis) (Falk-Petersen et al., 1990. To assess differences in FATM composition among different species, we conducted a multivariate analysis of fatty acids using a correspondence analysis (CA) (Greenacre, 2016). Correspondence analyses are suitable for compositional data in the presence of a large number of zeros (Greenacre, 2016). A centroid discriminant analysis was conducted to discriminate between the species energy content and dietary derived fatty acids (i.e., structural fatty acids vs. copepod-derived long-chain fatty acids). Computations were conducted with the R package ca (Nenadić and Greenacre, 2007). Total biomass and number of individuals in midwater trawls are summed for each season and species. The mean acoustic biomass is calculated from the mean s A , the weight-based target strength (TS W ), and the proportion of each species for each season. References for the Target Strength (TS) to length relationships and the origins of all species are indicated. Total mean acoustic biomass and SD for each season are written in bold.
Frontiers in Marine Science | www.frontiersin.org   Figure 1B). The WMD of the SSL was 221 m in January 2016, 455 m in August 2016, and 251 m in January 2017, indicating that the vertical distribution of the SSL was >200 m deeper in summer than in winter. In January, the SSL was located in the AW, from ∼50 to ∼500 m above the slope and between ∼50 and ∼250 m above the deeper basin (Figure 1C), which corresponded to temperatures between 2.8 and 4 • C (Figure 2). Hence, the WMD of the SSL in January was mainly driven by organisms above the slope and the animals present above the deep basin were rather epipelagic ( Figure 1C). The average s A at 38 kHz within the SSL was 57 m 2 nmi −2 in January 2016 and 56 m 2 nmi −2 in January 2017. As for the WMD, the backscatter diminished above the deeper basin where less AW penetrated (≤42 m 2 nmi −2 in 2016 and ≤51 m 2 nmi −2 in 2017 for the region north of 81.9 • N). Pelagic organisms generally avoided the colder polar waters and MAW in January. In August, the average s A within the SSL was 351 m 2 nmi −2 . The SSL was concentrated in the lower part of the AW and in the MAW, between 200 m and 700 m ( Figure 1C) at temperatures between 1 and 4 • C (Figure 2).

Composition of the SSL
We sampled a total of nine fish species in the mesopelagic SSL (offshore regions) in January. Juvenile redfish (mainly the beaked redfish Sebastes mentella) dominated the pelagic fish assemblage (Figure 3). They represented 89% of the fish biomass with a mean standard length (SL) of 6.5 cm in 2016, and 83% of the biomass with a mean SL of 5.8 cm in 2017 (Table 1). Arctic cod (Boreogadus saida), referred to as polar cod in Europe, represented < 2% of the biomass in January, and had an average length of 7.6 cm (2016) and 7.0 cm (2017). The rest of the winter fish assemblage was composed of capelin (Mallotus villosus), juvenile haddock, Myctophidae (Benthosema glaciale and Notoscopelus kroyeri), juvenile Atlantic cod, and juvenile daubed shanny (Leptoclinus maculatus) in 2016, and of Myctophidae, juvenile daubed shanny, capelin, and juvenile gelatinous snailfish (Liparis fabricii) in 2017 (Table 1). Herring (Clupea harengus) were sampled in Smeerenburgfjorden and Rijpfjorden. Gelatinous zooplankton, mainly Cyanea capillata, dominated the zooplankton assemblage in January 2016 (93% of the biomass; Figure 3) and represented 22% in January 2017. Euphausiids (mainly Thysanoessa inermis), the cephalopod Gonatus fabricii, and the amphipod Themisto libellula comprised most of the remaining macrozooplankton ( Table 1).
The cluster analysis indicates three distinct station groups ( Figure 4A). Group 1 is composed of the two Smeerenburgfjorden stations sampled in January 2016 and 2017. Group 2 comprises the other stations (7) sampled in January 2016 and 2017, except for two stations located above the slope. The latter are included in group 3 which also includes all the stations (6) sampled in August. Group 3 presents an average low intragroup similarity compared to the other groups (38%, 54 and 64% in groups 3, 2, and 1, respectively; SIMPER) and divides into the sub-groups 3a,b,c. Groups 3a and b constitute two pairs of stations shallower than 500 m, whereas group 3c includes the northernmost stations at >1,500 m bottom depth. The NMDS analysis illustrates well the partitioning of groups, and also the wide extension of the stations making group 3 in a 2-dimensional ordination, which is indicative of comparatively low faunal similarity of its stations (Figure 4B). The dominance of the biomass by redfish and Thysanoessa spp. in winter and by T. libellula and M. norvegica in summer explains to a large extent the dissimilarity among the stations from January (group 2) and August (group 3a,b,c) (Supplementary Table 3). The main causes for differentiation of group 1 (Smeerenburgfjorden) are the high biomass of Arctic cod, capelin and Pandalus borealis present in that fjord. Cyanea capillata contributes heavily to the intragroup similarity of sub-groups 3a and b, whereas T. libellula and M. norvegica characterize sub-group 3c. There is no clear coastal-offshore gradient, except for the higher abundance of Arctic cod and northern shrimp in Smeerenburgfjorden, and the sub-group 3c made of stations deeper than 1,500 m.
Group 4 is composed of redfish, the euphausiid T. inermis and spotted barracudina (Figure 5). Redfish had a total lipid content varying from 3% of dry mass in January to 10.5% in August. In contrast, T. inermis had a higher lipid content in January (22.2%) than in August (13.4%). Spotted barracudina had the higher concentration of lipids of all species (44.7% of dry mass in August) ( Table 3)

Seasonal Variations in Biomass and Assemblage of Macrozooplankton and Pelagic Fish
The Svalbard branch of AW inflow advected and concentrated zooplankton and fish species of boreal origin from the North Atlantic into the AW mass north of Svalbard. The Svalbard branch follows the slope between the 600-1000 m isobaths (Meyer et al., 2017) with currents ∼0.5 Sverdrup (Koenig et al., 2017). These isobaths correspond to the location where the AW thickened in January ( Figure 1B). As most organisms followed the inflow of AW above the slope, in January the SSL was thicker and denser above the slope than above the colder and fresher deeper basin ( Figure 1C). The strong intrusion of AW up to the surface above the slope likely explains the absence of ice in the study area (Polyakov et al., 2017).
The s A values are proxies for the biomass of organisms within the SSL. The average s A value in summer (351 m 2 nmi −2 ) was more similar to values from boreal regions (e.g., 100-400 m 2 nmi −2 north west of the British Isles, 56 • -60 • N; Godø et al., 2009) than tropical and subtropical areas (mean of 1864 m 2 nmi −2 for the world's oceans between 40 • S and 40 • N; Irigoien et al., 2014). The SSL we detected in August was also weaker than SSL in the North Atlantic (e.g., 1165-1443 m 2 nmi −2 during daytime; Fennell and Rose, 2015). Our mean s A value in August was slightly higher and our mean s A values in January (56-57 m 2 nmi −2 ) slightly lower than observations from the same area in August-September 2014 (66-240 m 2 nmi −2 ; Knutsen et al., 2017), which supports our conclusion that the backscatter increases in summer, but also suggests interannual variations. Previous studies of SSL in the Barents Sea reported a decrease in backscatter from south to north [e.g., 215 to 81 m 2 nmi −2 (Siegelman-Charbit and Planque, 2016) and 365 to 66 m 2 nmi −2 ], which is consistent with our observations from January but not August. However, the latitudinal range we covered was smaller than in these studies and intraseasonal Only fatty acids and alcohols present in at least one species are listed.
variations might have been dominated by the shelf-basin gradient rather than a south-north gradient.
Calculating the biomass of mesopelagic organisms within SSL imposes several caveats and uncertainties. Biomass estimates from acoustic data are based on the species assemblage as well as average size and weight of each species in net samples. Midwater trawl nets, such as the Harstad trawl, are selective and do not sample the whole community (De Robertis et al., 2017). Despite a 10 mm-mesh in the codend, some of the smaller organisms might have passed through the large meshes of several cm in the first parts of the net. On the other hand, larger fish avoid nets more easily than smaller ones. Hence, net selectivity could have positively or negatively biased our estimates and a larger net with smaller mesh at the mouth would have made the catch more representative of the actual community. Midwater trawls are inefficient at sampling siphonophores because these fragile organisms are easily destroyed. Only one siphonophore was sampled in January 2016, but their abundance was likely higher as they can be abundant at high latitudes (Knutsen et al., 2018). Siphonophores have gas-filled pneumatophores that resonate at similar frequencies as swimbladdered fish, and it is possible that some of the signal classified as fish was in fact coming from siphonophores, thus resulting in an overestimation of the fish biomass. In other areas where SSL are mainly composed of Myctophidae with swimbladder characteristics changing with ontogeny and depth (i.e., changes in occurrence, shape, and composition of the swimbladder), variations in their TS-length relationship greatly complicate biomass calculations (Yasuma et al., 2010). Here, however, this bias was likely limited as Myctophidae only represented 2% or less of the biomass ( Table 1). In addition to net selectivity, TS dependency on depth and non-linearity in EK60 power measurements at lower SNR prevailing at mesopelagic depths may have biased biomass estimates either way.
Most caveats and biases were likely similar from one season to the other, and viewing our biomass estimates as relative rather than absolute values should not affect the interpretation and conclusions. Despite high uncertainties, these estimates provide rare quantification of seasonal variations in mesopelagic biomass in the high Arctic, which can be used as a baseline from which to monitor future changes in the pelagic ecosystem. They also demonstrate that macrozooplankton provided most of the biomass and lipids available within the SSL in summer, and that their biomass diminished by one order of magnitude in winter. Hence, macrozooplankton should be included in future surveys quantifying mesopelagic ecosystems.
The species assemblage varied more seasonally than spatially and, in contrast with many oceanographic regions (Irigoien et al., 2014), Myctophidae did not dominate the mesopelagic fish assemblage ( Table 1). The assemblage in August was significantly different from the assemblages in January, which were relatively consistent between January 2016 and 2017 (Figure 4). In August, macrozooplankton such as the amphipod T. libellula, the euphausiid M. norvegica, and the medusa C. capillata dominated the biomass. The high biomass of macrozooplankton and jellyfish and the occurrence of juvenile redfish, spotted barracudina, and some glacier lanternfish is consistent with previous observations of the SSL north of Svalbard Knutsen et al., 2017). However, in contrast with these studies which mainly sampled Atlantic fish species, Arctic cod dominated the biomass of pelagic fish in August (45%). The relatively high abundance of Arctic cod among Atlantic species is not surprising as the species is abundant in the northern Barents Sea where its distribution overlaps with that of Atlantic cod and haddock (Renaud et al., 2012), as well as that of juvenile redfish (Eriksen et al., 2015).
Year class strength of Arctic cod is highly variable, and since mainly age-0 and age-1 individuals occupied the SSL, the biomass of pelagic Arctic cod likely varies considerably from year to year.
Cyanea capillata and T. inermis dominated the macrozooplankton assemblage in January (G. fabricii were also abundant in January 2017), and juvenile beaked redfish dominated the pelagic fish assemblage (>83% of the biomass). Arctic cod only represented <2% of the biomass in January. Arctic cod sometimes remain under sea ice in the high Arctic (Lønne and Gulliksen, 1989), but no ice-associated Arctic cod were sampled nor observed acoustically in the same region in June 2017 (Pierre Priou, Memorial University, unpublished data). The average length of Arctic cod in winter (<10 cm) suggests that they were part of the age-1 cohort (Lønne and Gulliksen, 1989). Age-1 Arctic cod can be sexually mature (Nahrgang et al., 2014) and winter spawning migrations toward spawning areas in the fjords of Svalbard and the eastern Barents Sea (Ajiad et al., 2011) could explain the near absence of Arctic cod in January. Another plausible explanation is a change in preferred habitat with ontogenetic development, because Arctic cod leave their pelagic habitat for the demersal zone as they grow (Geoffroy et al., 2016).
The photoperiod constraint hypothesis suggests that inferior feeding conditions imposed by the extremely low irradiance prevailing during the polar night limit the abundance and northward range expansion of pelagic fishes of boreal origin and their zooplankton prey into the high Arctic (Kaartvedt, 2008). Here, however, the lower biomass in winter might be related to other factors than the inability to forage during the polar night. We observed a large decrease in the biomass of macrozooplankton and fish in January, but for species of both boreal and Arctic origins. Zooplankton and juvenile pelagic fish from the North Atlantic are more likely to be advected northwards in summer, when they colonize the pelagic realm, than in winter when they descend at greater depths. This most likely contributes to the lower biomass in January. Furthermore, the decrease in fish biodiversity mostly results from the ontogenetic descent at depth of benthic and demersal species. For instance, the wolffish caught in August were age-0 juveniles with a mean length of 6.29 cm, a length at which they descend to demersal depths before becoming exclusively benthic once they reach 10 cm (Moksness and Pavlov, 1996). Similarly, the mean lengths of the juvenile Atlantic cod and haddock caught in August indicate that they were on the verge of quitting the pelagic realm (Lough et al., 1989;Lomond, 1998). Most of these gadids likely descended to demersal depths by January, which explains their lower biomass in winter despite the presence of relatively warm AW. Although some pelagic boreal species might not survive the polar night, the only species for which we might infer such a phenomenon is the spotted barracudina, which remains pelagic at the adult stage and was absent in January. Juvenile beaked redfish, another boreal fish, displayed relatively constant biomass between seasons (Figure 3). The polar night may thus restrict the northward range expansion of some species, such as spotted barracudina, but others such as the beaked redfish can survive the extreme conditions prevailing during the high Arctic winter. Hollowed et al. (2013) predicted a northward range expansion of beaked redfish in the Arctic Ocean. Until the 1990s, juvenile Arctic cod dominated the fall pelagic fish assemblage in the offshore region of western Svalbard. Since then, juvenile redfish have become the dominant species because of high recruitment events, the northward retreat of the ice edge, and a longer icefree season resulting in surface temperatures reaching 2-5.5 • C (Hollowed et al., 2013;Eriksen et al., 2015). Here, we demonstrate that juvenile redfish can also dominate the northern Barents Sea pelagic assemblage in winter. In the Barents and Norwegian Seas, the ovoviviparous beaked redfish spawn in fall and larval extrusion occurs in March-April (Drevetnyak and Nedreaas, 2009;Planque et al., 2013). Based on an average flow of 0.1 m s −1 within the West Spitsbergen Current (Fahrbach et al., 2001;Cokelet et al., 2008), redfish captured on 28 August would have taken 96 days to drift the 830 km between their northernmost known extrusion zone close to Bjørnøya (Hollowed et al., 2013) and the northernmost sampling station, which implies an extrusion date prior to 24 May. The abundant age-0 redfish sampled up to 81.4 • N in August could thus have hatched in their known larval extrusion zone above the continental shelf break in the Barents and Norwegian Seas . Hence, for the moment, high abundances of age-0 redfish this far north does not necessarily imply a northward expansion of the extrusion zone. However, it suggests that they can survive and grow (e.g., mean length of 4 cm in August and 6 cm in January) after being advected into the Arctic.
Arctic cod, T. libellula, and M. norvegica dominated the fish/crustacean assemblage in August, while redfish and Thysanoessa inermis dominated in January. Hence, we conclude to a switch from a Calanus-based pelagic food web in summer to a krill-based pelagic food web in winter (Figure 6). This switch also explains why the few Arctic cod sampled in January fed more on krill than those sampled in August (Bore 6-10 vs. Bore 1-5; Figure 5). In the Arctic, primary production diminishes drastically after the spring bloom and the algal community changes from an abundance of diatoms in spring to a domination of flagellates in winter (van Leeuwe et al., 2018). Our sampling was conducted in August and January, after the spring bloom, which explains why most species contained relatively high concentrations of the flagellate marker 22:6(n-3) FA (Ackman et al., 1968) and why the concentrations of this marker doubled in Arctic cod and redfish between August and January. Despite lower algal abundance and a dominance by flagellates in winter (van Leeuwe et al., 2018), T. inermis likely continued to feed during that period as their total lipid/dry weight ratio slightly increased in winter (from 13% to 22%; Table 3). Huenerlage et al. (2015) concluded that lipid reserves in T. inermis are not sufficient to survive the Arctic winter without feeding and suggested that they shift to an opportunistic feeding mode during that season. A shift to detridal and benthic food sources during winter has also been suggested for Thysanoessa raschii (Sargent and Falk-Petersen, 1981). The ability to feed in winter may explain why Thysanoessa spp. dominated the macrozooplankton assemblage in January. In turn, their high biomass supported the juvenile redfish population in winter (Figure 3).
In August, Calanus spp. had started their descent to overwintering depths and were distributed throughout the water column down to 800 m, with a WMD of 408 m (M. Daase, unpublished data). They were thus accessible to Arctic cod, T. libellula and M. norvegica, distributed at similar depths, especially as these visual predators are adapted to the low light levels prevailing at these depths (Jönsson et al., 2014;Cohen et al., 2015). Surprisingly, the herbivorous Thysanoessa spp. were also located between 200 m and 700 m in August, below the productive euphotic zone. This behavior might be due to avoidance of visual predators, such as seabirds, in the top 200 m after the spring bloom, when irradiance increases and primary production diminishes.
In January, C. finmarchicus overwinters below the AW in the North Atlantic and Greenland Sea, and thus out of range for pelagic fish and zooplankton (Hirche, 1991;Gaardsted et al., 2010). However, in the waters north and west of Svalbard C. finmarchicus has recently been observed to concentrate in the upper 50 m in January (Daase et al., 2014;Berge et al., 2015;Basedow et al., 2018), an observation that has a great impact on estimates of the amount of C. finmarchicus advected in the Arctic with AW (Basedow et al., 2018). Calanus abundance integrated over the water column was, however, 5-10 times lower in January than in August (11,500-28,300 ind. m −2 and 111,640 ind. m −2 , respectively; Daase, unpublished data). Hence, Calanus spp. were less available for pelagic predators in January compared to August. In contrast, euphausiids and juvenile redfish remained abundant between 50 m and 500 m in winter, where they supported the winter krill-based food web (Figure 6). As AW progresses further east into the Eurasian Basin since the last decades (Polyakov et al., 2017), it is possible that more krill and redfish are advected in the European Arctic and that the krill-based food web becomes increasingly important in the region. Interestingly, Einarson (1945) showed that krill was also abundant in Svalbard waters during the last warm period in the 1930s.

CONCLUSIONS
This study provides rare information on the seasonal variations in biomass and composition of pelagic fish and macrozooplankton in the high Arctic. It shows that the diversity and total biomass of pelagic species forming the SSL diminishes during the polar night. Part of this decrease might be related to ontogenetic migrations toward the seafloor. Yet, some pelagic Atlantic-taxa such as the spotted barracudina could be on a trail to death when advected with AW into the high Arctic. Others like juvenile beaked redfish thrive during the polar night, possibly because they can exploit the lipid-rich krill-based food web prevailing during this period. Redfish have strong year classes followed by years with very low recruitment (Drevetnyak and Nedreaas, 2009). Future surveys should continue to monitor the fate of the strong 2016-2017 cohorts advected into the high Arctic to verify if they settle and reproduce in the region or if they migrate back south to their known spawning grounds. Despite a decrease in biomass during winter, pelagic fish and macrozooplankton advected in the high Arctic form a lipid-rich SSL which represents an important source of energy for the whales and other top predators spending the summer and fall north of Svalbard (Falk-Petersen et al., 2009a;Haug et al., 2017).

DATA AVAILABILITY
The datasets generated for this study are available on request to the corresponding author.

ETHICS STATEMENT
All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. This study was carried out with permission from the Governor on Svalbard and followed the strict regulations regarding health, environment and safety enforced at UNIS and UiT.

AUTHOR CONTRIBUTIONS
MGe designed the study and led sampling activities, analyses of the acoustic datasets, and writing of the manuscript. MD, MC, and NS contributed to sampling activities and analyses of the zooplankton and acoustic datasets. GD led species assemblage analyses. MGr conducted lipid analyses. FC contributed to interpret the physical oceanography results. JB, PR, and SF-P contributed to design the study and participated in sampling activities. All co-authors contributed to writing and reviewing the manuscript.

FUNDING
This study was supported by the Research Council of Norway (NFR) through the Arctic ABC project #244319 and the Marine Night project #226417. It is a contribution toward the ARCTOS research network (https://arctos.uit.no/), the Arctic Science Partnership (www.asp.org), and the Ocean Frontier Institute (oceanfrontierinstitute.com) funded through Canada's First Research Excellence Fund.