The hydroacoustic and hydrographic (temperature and salinity) data, used in the analysis, were collected in Kongsfjorden. This fjord, being oriented from Southeast to Northwest, is located in the northern part of the West Spitsbergen coast at about 79°N and 11°-12°E ( Figure 1 ). The absence of the sill at the fjord mouth allows for exchanging the fjord and shelf waters. Kongsfjorden is strongly influenced by warmer Atlantic waters, carried by the West Spitsbergen Current ( Figure 1 ) (Svendsen et al., 2002; Cottier et al., 2005; Promińska et al., 2017; Tverberg et al., 2019; De Rovere et al., 2022).

In all three years, the measurements were collecyed on board R/V Oceania in the first decade of August. Dates and times of the data collection on different transects are presented for different years in Table 1 .

Unfortunately, no fish samples from the water column were available to ground-truth the hydroacoustic data. However, in the interpretation of the results, we took into account that the two above-mentioned species belonging to the Gadidae family (Gadus morhua and Boreogadus saida) dominate the fish community in Kongsfjorden (Keskinen et al., 2004; Renaud et al., 2012; Berge et al., 2015a; Brand and Fischer, 2016; Kunz et al., 2016; Baschek et al., 2017; Fey and Węsławski, 2017; Fischer et al., 2017; Szczucka et al., 2017; Giacalone et al., 2021; Brand et al., 2022; Giacalone et al., 2022). Other less numerous species were listed in Berge et al. (2015b): capelin, haddock, herring, and mackerel.

A total of about 80.7 nautical miles of acoustic transects were sampled (27 Nm in 2013, 26.7 Nm in 2014 and about 27 Nm – in 2022). The hydroacoustic measurements have been conducted along four transects ( Figure 2A ). One of them was along the fjord axis extending from the mouth of the fjord to the easternmost glacier. Three additional transverse transects were perpendicular to the axial section. They were named as external (closest to the fjord’s mouth), middle, and internal (furthest from the fjord’s mouth) transects. Different colors of transects in the figure correspond to different study years: blue – 2013, red – 2014, and black – 2022.

The tracks followed varied slightly between the years. The curves indicating the external or middle transects coincide for all three years. The figure does not show transect along the axis in 2013 and the internal transect in 2014. This is due to the lack of hydroacoustic data for these transects, caused by unforeseen damage of data files at the data analysis stage. The gaps in hydroacoustic data are also listed in Table 2 , which shows the availability of data for further analysis.

Backscattering data were collected using Simrad EK60 echosounder equipped with a split beam transducer operating at a frequency of 70 kHz. Its beam width was 7 degrees from the acoustic axis at -3 dB level. The 256 ms pulse length and 2 s^-1 - ping rate were set in 2013 and 2014. In 2022, these parameters were respectively 1.024 ms and 1 s^-1. In all three years the power setting of 525 W was used. The acoustic transducer was mounted on a rigid frame attached to the broadside of the vessel, approximately 1 m below the sea surface.

Just prior to the survey, the echosounder was calibrated with the standard sphere method (Foote et al., 1987; Kongsberg Maritime, 2012). Hydroacoustic data were collected with dedicated Simrad ER60 software and stored digitally in a raw format for later analysis.

In 2013 and 2014, in parallel with hydroacoustic data sampling (along the same transects), temperature and salinity were continuously measured using a towed profiling system equipped with a Sea-Bird Scientific SBE49 FastCat CTD sensor (Promińska et al., 2017). In Figure 2B , the transects are marked with blue and red lines for years 2013 and 2014 respectively. The locations of CTD sections and a total number of the collected profiles varied between the two years ( Figure 2B ).

In 2013 and 2014, the CTD sensor was calibrated prior to the cruise and initially accurate to ±0.002°C for temperature and ±0.0003 S/m for salinity. CTD profiles were measured in the full water column with the spatial horizontal resolution of a few km when the vessel moved at a speed of about 3 knots.

In 2022, unlike previous years, vertical temperature and salinity profiles were not measured concurrently with the hydroacoustic measurements. Measurements were carried out at stations (black stars in Figure 2B ) located along the longitudinal transect (along the axis of the fjord) and the external transverse transect, where hydroacoustic data had previously been collected. CTD data was not collected along the middle and internal transects in 2022 (see Table 2 ).

In 2022, CTD data were obtained with the Sea-Bird Scientific 911+ CTD system. The double sets of temperature and salinity sensors were calibrated prior to the cruise and their initial accuracies were ±0.001°C and ±0.0003 S/m, respectively.

The target strength analysis was performed to compare the TS histograms of individual fish present in the study area and to examine the spatial distribution of fish of different target strengths. The raw data were processed using Echoview 4.9 (Myriax Software, Hobart, Tasmania). The analysis was performed following the methodology as described by Szczucka et al. (2017).

In the first step, the single target detection algorithm with the following parameters was used with: a minimum TS value of -70 dB a minimum and maximum echo length ratio of 0.7 and 1.3 respectively, a maximum beam compensation of 6 dB, and a maximum standard deviation of axis angle of 0.6°. Additionally, single target detections were filtered to exclude all targets >3° off the beam axis of transducer.

In the next step, a fish tracking algorithm was applied to reduce the stochastic error of the individual single target measurements (Simmonds and MacLennan, 2005). The algorithm identified the groups of extracted single targets that show similar patterns of systematic movement and are assumed to be generated by the same fish. The fish tracking criteria were as follows: minimum number of single targets in track: 3; minimum number of pings in track: 3; and maximum gap between single targets in track: 2. For each individual fish track, the mean TS value, mean depth, and geographical position were calculated.

The 12084 individual fish of different target strengths were selected for further analysis.

The temperature and salinity data collected with the SBE49 and SBE911+ CTD systems were processed with the SBE Data Processing Software following the recommended steps and parameters for each system and then vertically averaged into the 1 dbar bins. The vertical profiles of temperature and salinity were analyzed and visualized using the MATLAB software (version R2020b update 6).

In Figure 3 , the histograms of target strength are presented for three years: 2013 (plot A), 2014 (plot B), and 2022 (plot C). For each year the plot was generated for the data collected during the whole survey.

Based on the histograms, the three groups of fish could be selected: the first mode corresponds to a group of fish named “small fish” (lower TS - mode with TS< - 52 dB), the second one – “medium fish” (intermediate TS-mode with - 52 dB ≤ TS < - 37 dB), and the last mode – “large fish” (higher TS - mode with TS ≥ - 37 dB). The boundaries between the groups are indicated by vertical solid black lines for all years in Figure 3 . The choice of these dB - thresholds (limit values) is explained below in section 4.1.

The proposed names of groups are justified by the fact that the length of fish is one of the main parameters controlling fish TS (Simmonds and MacLennan, 2005) and that the larger the fish length is, the higher the TS. Throughout the manuscript, fish belonging to the groups “small fish”, “medium fish”, and “large fish” are named small, medium, and large fish respectively.

The figure demonstrates that, for fish with TS smaller than about -52 dB, the distribution is unimodal for all three years. However, for fish with TS larger than -52 dB, the bimodal distributions were observed in 2013 and 2022 and was unimodal in 2014.

We want to point out that the “medium fish” group was not numerous in 2013, was absent in 2014, and only in 2022 did its role grow significantly. On the other hand, the figure shows that the role of the “small fish” was the most prominent in 2013 ( Figure 3A ).

To generate Figure 4 , demonstrating the interannual variability of the size composition of the fish community in Kongsfjorden, the same data as for Figure 3 has been used: each year, the number of individuals in each class of fish was calculated for the entire study. A blue color in the figure corresponds to the “small fish” group, green color to “medium fish”, and brown color to “large fish”. The figure demonstrates that the percentage of small fish was the highest in 2013 (80%), while in 2014 and 2022 it was 19,6% and 48,9% respectively. This means that the total percentage of medium and large fish increased from 20% in 2013 to 80,4% in 2014 and to 51,1% in 2022. It is also worth emphasizing the growing trend in the share of medium fish: 14,8% in 2013, 26,6% in 2014, and 30,1% in 2022.

Figure 5 depicts the impact of the thermohaline structure on depth dependence of TS. Figures 5A–C present the results for years 2013, 2014, and 2022 respectively. Figures 5B, C show the transect along the fjord axis ( Figure 2 ), while Figure 5A presents the transverse external transect (we have no data for the transect along the fjord for this year, see Table 2 ). The left panels of each plot A - C show the spatial distribution of fish, belonging to the three groups: “small fish” (blue dots), “medium fish” (green dots), and “large fish” (brown dots). The right panels of the plots present the salinity spatial distribution along the transects. The boundary between Atlantic Water mass (AW) and Transformed Atlantic Water (TAW) is presented with a blue line, while the boundary between TAW and Intermediate Water (IW) is presented with a green line. The boundary between IW and Surface Water (SW) is indicated by a black line. The temperature and salinity ranges of different water masses have been selected following the classification employed by Svendsen et al. (2002); Nilsen et al. (2016), and Promińska et al. (2018).

A comparison of left and right panels of each plot demonstrates that in all years the small fish (blue dots) are not sensitive to water mass properties. For example, in 2013 and 2022, respectively, 34% and 28% of all small fish were above the boundary between waters of Atlantic and non-Atlantic origin. The boundary was at about 50 m depth (green lines in Figures 5A, C ). Unlike the “small fish” group, large and medium fish can only be found in the Atlantic origin waters (AW and TAW). Figure 5 also demonstrates that in all years a significant number of medium and large fish were in waters of Atlantic origin below the boundary between waters of the Atlantic and not Atlantic, both in the shallow and deep part of the fjord, e.g. in 2013 and 2022 only about 5% of all medium and large fish exceeded this boundary.

The presented effect of the type of water masses on the spatial distribution of fish of different size groups is also confirmed by the comparison of Figure 5B with Figures 5A and C . In 2014 ( Figure 5B ), medium and large fish were closer to the surface than they were in 2013 ( Figure 5A ) and 2022 ( Figure 5C ). The reason is that the boundary between waters of the Atlantic and non-Atlantic origin was significantly shallower in 2014. It was just below the surface, while in 2013 and 2022 it was about 50 meters below it. In 2014, 15% of all individuals from “medium fish” and “large fish” groups were found in shallow waters over 50 m deep. This percentage is 3 times higher than in 2013 and 2022.

In all years, large fish were found close to the bottom in the deep part of the fjord. However, in 2014 ( Figure 5B ), when AW filled the shallow part of the fjord, large fish were also found mainly near the bottom, but in both the shallow and deep parts of the fjord. Interestingly, unlike in other years, in 2022 ( Figure 5C ) medium fish were mainly found near the bottom in the shallow part of the fjord filled by TAW.

As mentioned above, the “predator-prey” relationship is one of the factors that could impact on the fish spatial distribution. Smaller fish (e.g., polar and juvenile Atlantic cod in Kongsfjorden) look for a spatial niche providing protection from predators and cannibalization by larger specimens, e.g., adult Atlantic cod (Brand et al., 2022). Taking this into account, we were curious how this might affect the mutual spatial distribution of three fish groups selected for the analysis (“small fish”, “medium fish”, and “large fish”).

The “small fish” group (blue dots in Figure 5 ) occupied different spatial areas compared to the areas occupied by large fish (brown dots in the figure). Figures 5A, C (for years 2013 and 2022 respectively) demonstrate that small fish (blue dots) occupied mainly the top part (not deeper than 150m) of the water column, while large fish were closer to the bottom (not shallower than 200m).

The “medium fish” and the “large fish” groups were spatially separated in 2022 ( Figure 5C ), while in previous years ( Figures 5A, B ) they occupied the same areas. In 2022, the medium fish were mainly present in the shallower part of the fjord near the bottom, while the large fish were found near the bottom, but in the deeper part. In 2013 and 2014 these two groups mixed in the same areas.

An additional comparison of fish spatial distributions has been conducted for years 2013 and 2022. In these two years, a similar vertical distribution of waters of Atlantic origin was observed: below 50 meters in both years (the right panels in Figures 5A, C ). This excluded the temperature and salinity structure as possible reasons for the difference in fish spatial distribution between these years. The difference, demonstrated below, should be controlled by other factors.

Comparisons were made for each group of fish (“small fish”, “medium fish”, and “large fish”) along each of the three transverse transects: external, middle, and internal ( Figure 2A ). The transect along the fjord was not considered due to a lack of hydroacoustic data in 2013 ( Table 2 ).

The difference between 2013 and 2022 was observed only for the “middle fish” group. It is presented in Figure 6 for the three selected transects: external (plot A), middle (plot B), and internal (plot C). In each plot the mean depth positions (black points) with the maximum and minimum depths (bars) and the 25% and 75% percentiles (rectangles) are presented for 2013 (on the left) and 2022 (on the right). Figure 6 shows that for all selected transects (plots A – C) mean depth of the spatial distributions of medium fish was shallower in 2022. The mean values in 2022 are smaller than 150 m depth, while in 2013 they are larger than this depth.

The presence of medium-size fish in the top water column in 2022 is also confirmed by Figure 7 . Plots A, B, and C in the figure are generated respectively for the external, middle, and internal transects. In each plot, the left column corresponds to 2013 and the right one to 2022. The percentage of medium-size fish in the first 150 meters under the surface is marked with a white color, while deeper than 150 meters is shown by a grey color. We can see that in 2013 about 70% - 81% of fish were observed deeper than 150 meters. These percentages were 77%, 70%, and 81% at external, middle, and internal transects respectively. In 2022, 51% to 70% of organisms were shallower than 150 meters. These percentages were 54,5%, 70%, and 51% at external, middle, and internal transects respectively.

A lack of sensitivity of “small fish” group to the water type, demonstrated in Subsection 3.2.1 ( Figure 5 ), confirms that this group is formed of polar cod, which as a native species is perfectly adapted to all types of fjord waters. The fact that small fish could be found both in the deeper and shallower parts of the fjord also indicates that the native polar cod belonged to this group (Brand et al., 2022).

The behavior of “medium fish” and “large fish” groups, presented in Subsection 3.2.1, means that the Atlantic cod belonged to these two groups. Firstly, it is confirmed by the presence of these two fish groups only in the waters of Atlantic origin: Atlantic Water and Transformed Atlantic Water (Renaud et al., 2012).

Secondly, the fact that the large fish were found mainly close to the bottom (brown points in Figure 5 ) confirms that the “large fish” group consisted of adult Atlantic cod. Adult Atlantic cod is a demersal fish as it lives and feeds on or near the bottom (e.g. Beamish, 1966; Rose and Leggett, 1990; Arnold, 1994; Giacalone et al., 2022).

Thirdly, the presence of medium-size fish both in the water column and near the bottom in the deeper and shallower parts of Kongsfjorden in 2014 ( Figure 5B , left panel) and around the bottom in the shallower part in 2022 ( Figure 5C , left panel) indicates that these may be juvenile Atlantic cod, which could be found in both the littoral and sub-littoral waters (Brand and Fischer, 2016; Brand et al., 2022).

To understand how to explain the spatial separation of different groups of fish, discussed in Section 3.2.2, we used not only the data we collected, but also the published results obtained in 2004 by Keskinen et al. (2004) and in 2016 by Giacalone et al. (2022). The comparison is presented in Table 3 . The study period and locations as well as information on the spatial separation of three fish groups (“small fish”, “medium fish”, and “large fish”) are presented in the table. The studies were conducted in the same season (mainly in August) and in close areas.

Keskinen et al. (2004) proved this by trawling the presence of two groups of fish, namely polar cod and adult Atlantic cod, which we name in Table 3 as “small fish” and “large fish”. Giacalone et al. (2022) observed three fish groups differing in their backscatter properties and their school morphometry. They suggested that these three groups were polar cod, juvenile Atlantic cod, and adult Atlantic cod (these groups are also named as “small fish”, “middle fish”, and “large fish” in the table).

The lack of the separation between polar cod and adult Atlantic cod in 2004 was confirmed by Keskinen et al. (2004) by trawling at middle depths and near the bottom. However, after 2004, we (in 2013, 2014, and 2022) and Giacalone et al. (2022) (in 2016) observed the separation between small and large fish.

Taking into account

it can be inferred that the separation confirms that the “small fish group” consisted of polar cod, while the “large fish” group included adult Atlantic cod. The appearance of the separation only after 2004 may have been due to the fact that a significant increase in the abundance of adult Atlantic cod, which could affect fish separation, was documented only after 2004 (Renaud et al., 2012; Berge et al., 2015b).

The spatial separation between “medium fish” and “large fish” groups has looked different over the years. The separation was first observed in Kongsfjorden in 2016 by Giacalone et al. (2022) and was confirmed in our study in 2022.

Taking into account that

it can be concluded that the clear spatial division between “medium fish” and “large fish” groups in 2022 strongly suggests that the group “medium fish” included juveniles of Atlantic cod.

The history of the separation between “medium fish” and “large fish” groups confirms this conclusion. The lack of the separation between medium fish and large fish till 2016 could be explained by the fact that the spawning of Atlantic cod inside the fjord only began in 2008 and after this year the Atlantic cod juveniles were regularly caught (Berge et al., 2015b; Brand et al., 2022). It could be that the juveniles in years 2013 and 2014 was not so abundant as in 2016 and 2022. The abundance increase prompted them to find the spatial niche to avoid cannibalism from adult Atlantic cod. An increasing share of medium-sized fish (juvenile Atlantic cod), demonstrated in subsection 3.1, confirms this.

Interpretation of the results presented in Figures 6 , 7 encountered difficulties due to the insufficient amount of the measured data collected. To explain the results, we put forward two hypotheses and presented some arguments justifying these hypotheses, but unfortunately their complete confirmation requires additional study.

The first hypothesis aims to explain the difference in vertical spatial distributions between 2013 and 2022 as a result of the separation between “medium fish” and “large fish” groups, discussed in the previous section. The difference consisted of the presence of medium fish at shallow depths in 2022. Perhaps the mentioned increase of medium fish abundance prompted them in 2022 to flee from the cannibalism of adult Atlantic cod, which prefer benthic conditions, to upper water layers.

Given that in 2013 and 2022, data were collected during the “day” (08.45 – 17.00 UTC) and “night” (00.10 – 04.00 UTC) periods ( Table 1 ), respectively, the second hypothesis proposes an interpretation of the results, taking into account the possible Diel Vertical Migration (DVM) of medium-size fish. Previous studies (Falk-Petersen et al., 2008; Last et al., 2020; Pettitt-Wade et al., 2021) have confirmed that biological clocks of marine organisms are important for shaping their daily activities at high latitudes, and there are some confirmations that organisms could perform DVM in the Arctic during the Polar Night or Polar Day.

It has been shown ( Figures 6 , 7 ) that during the “day” (2013) medium-size fish were closer to the seabed than during the “night” (2022). This is in agreement with the known DVM strategy of not large fish, just like juvenile Atlantic cod, which could be prey to many predators. This strategy is motivated by predation risk and includes descending into darker water masses during daylight (Clark and Levy, 1988; Rosland and Giske, 1994; Strand and Huse, 2007). It is in agreement with the confirmations presented above that the “medium fish” group could consist of juvenile Atlantic cod.

The presented difference between 2013 and 2022 for medium-size fish has not been observed for the “large fish” group. This is also in agreement with the previous conclusions that this group included adult Atlantic cod. The lack of the described DVM strategy for large-size fish is because that these fish have only a few predators (Pálsson, 1994) and predation risk is not the main motivation of their vertical movement. Diel variations of vertical distribution pattern of adult Atlantic cod, different than for medium-size fish, have been presented in Godø and Michalsen (2000) and Strand and Huse (2007).

The considered difference between the years has not been confirmed for the “small fish” group. It could be another argument that this group consisted of polar cod. There is no published evidence available that polar cod performs DVM at such high latitudes.

We carefully checked the echograms obtained along all transects in all three years. The inspection showed that, in 2013, several schools of fish were observed. It was the only year in which the research was conducted in “day” time. In the years 2014 and 2022 (“night” surveys), fish were dispersed in the water column in the echograms. Schooling fish during the day and their dispersal during the night are typical of many fish species that perform DVM (Pavlov and Kasumyan, 2000).

The presented arguments only partially justify the second hypothesis. For its reliable verification, the DVM study of fish in a selected location should be performed during multi-day continuous hydroacoustic observations, both during the day and at night (e.g. Simmonds and MacLennan, 2005; Benoit et al., 2010). Hydroacoustic measurements should be accompanied by biological sampling to correctly interpret the backscatter data obtained.