Sea trout Salmo trutta in the subarctic: home-bound but large variation in migratory behaviour between and within populations

Anadromous brown trout (sea trout), Salmo trutta , is currently in decline throughout its range, largely due to anthropogenic stressors in freshwater and marine habitats. Acoustic telmetry was utilized to study the marine migration of sea trout post-smolts from three populations in a relatively pristine subarctic fjord system. While at sea, the sea trout spent a substantial part of their time close to their natal river, preferred near shore over pelagic habitats and were strongly surface oriented. Despite a fidelity towards local areas, the sea trout utilized various parts of the fjord system, with maximum dispersion >30 km and total migration distance >300 km. Almost half of the sea trout (44%) migrated between river outlets, indicating that a metapopulation approach may be appropriate when managing neighbouring sea trout populations at high latitudes. Furthermore, the different populations displayed different migratory behaviours in terms of distance migrated, dispersion from origin and the likelihood of leaving their home area. This variation in migratory behaviour is likely influenced by spatiotemporal differences in habitat quality between sites, indicating that local habitat variations may promote population-specific behavioural responses even in relatively confined fjord systems.

. Sea trout can perform multiple feeding migrations into the marine environment, with fish re-entering fresh water to either spawn or overwinter . The duration of the marine migration is highly variable, and although many sea trout spend only the summer months in the marine environment (Flaten et al., 2016;Jensen et al., 2020), year-round migrations are frequently observed, occasionally also in the northern populations (Jensen & Rikardsen, 2012;Rikardsen et al., 2006). While at sea, sea trout spend most of their time close to the surface and typically reside within fjords and in coastal areas close to their natal river . Nonetheless, the distance of the marine migration varies extensively both among and within populations (Kallio-Nyberg et al., 2002;Pratten & Shearer, 1983). Examples of sea trout migrating several hundred kilometres from their origin have been documented throughout large parts of their distribution range (Birnie-Gauvin et al., 2019; Thorstad et al., 2016).
Since the 1980s, many sea trout populations have declined due to anthropogenic stressors occurring in fresh water and at sea (ICES, 2013). In the marine environment, habitat loss, overfishing, ecosystem changes, and pathogen spill-over from aquaculture have all contributed to the decline (ICES, 2013). The parasitic salmon louse Lepeophtheirus salmonis Krøyer 1837 is an example of a stressor that impacts anadromous salmonids on local and regional scales. In areas with open net cage farming of Atlantic salmon Salmo salar L. 1758, the density of salmon lice can increase by orders of magnitude, to levels that harm wild salmonids Krkošek et al., 2011;Thorstad et al., 2015). Although sea trout populations can cope with the parasite at natural levels, infections exceeding a certain threshold level increase mortality and reduce individual growth, either directly though physiological processes or indirectly via premature returns to fresh water Thorstad et al., 2015). In Norway, open net cage farming of salmon occurs along almost the entire coast, with the highest activity in the south-western and central parts where numerous sea trout populations are declining (Anon, 2019). In comparison, subarctic sea trout populations are more pristine, but with the expected northward shift of the Norwegian salmon farming industry, there are growing concerns for these populations (Vollset et al., 2021).
In recent years, there has been an increased effort in quantifying the marine habitat use of sea trout, and several studies using electronic tags have provided detailed documentation of their marine migration (e.g., Eldøy et al., 2015;Kristensen et al., 2019). Although these efforts have increased our knowledge of the marine migration of sea trout, they have also highlighted the extensive behavioural variability that exists within the species. This emphasizes the importance of quantifying the marine migration of sea trout throughout their range, particularly in areas where anthropogenic stressors are expected to increase .
This study describes the marine migratory behaviour of sea trout post-smolts from three populations within a subarctic Norwegian fjord system, using acoustic telemetry. The main aim is to provide a detailed description of the marine migration and habitat use of sea trout from different populations within a relatively pristine fjord system. Specifically, the horizontal distribution and depth use of individuals are quantified, to further the understanding of the behaviour of sea trout post-smolt at high latitudes and to investigate if population-specific migration patterns are present within the same fjord system. Based on a recent observation of the behaviour of sea trout post-smolts from one of the populations included in this study (Atencio et al., 2021), it is expected that fish will spend most of their time close to their natal river, primarily occupying near shore surface waters.

| MATERIALS AND METHODS
The handling of experimental animals complied with Norwegian animal welfare laws, guidelines and policies as approved by Norwegian Animal Research Authority (permit reference number 12267).

| Study area
The study was conducted from June to September in 2018 in the Alta Fjord system in northern Norway (70 N, 23 E, Figure 1). The Alta Fjord system refers to the Alta Fjord and the Stjernsund, Rognsund and Vargsund straits, which all connect the fjord to the Norwegian Sea ( Figure 1). The Alta Fjord itself is 38 km long, is 4-14 km wide and has a maximum depth of 488 m. The inner part of the fjord is categorized as a National Salmon Fjord (Figure 1), which is a conservation measure to protect important Atlantic salmon populations by preventing potentially harmful industrial activities, such as aquaculture facilities, to establish nearby . The fjord is considered subarctic despite its Arctic location due to the inflow of Atlantic Ocean waters (Skarðhamar et al., 2018). The summer surface temperature in the Alta Fjord varies between 5 and 16 C with an average temperature of 10 C in August (Skarðhamar et al., 2018). During summer, the upper layer (down to 5-10 m) is brackish, with relatively high salinities below (>33) (Skarðhamar et al., 2018).

| Study populations
A total of 92 sea trout were tagged at three sites in the Alta Fjord system: (a) in the fjord 2.5 km south of the Hals River outlet between 24 June and 5 July (n = 35), (b) in the fjord 0.8 km west of the Skillefjord River outlet between 2 and 9 July (n = 35) and (c) 8 km upstream the Alta River either in late April (n = 1) or between 13 and 16 July (n = 21) (Table 1; Figure 1). Fish caught in the marine envrionment were assumed to originate from the nearby rivers, and the sea trout were classified into three separate groups (hereafter termed populations): Hals River, Skillefjord River and Alta River.
The Hals River has a mean annual water flow of 4.3 ms À1 and an annual within-river angling catch of sea trout averaging 157 kg (www. ssb.no). The river has a 20 km stretch accessible to anadromous salmonids. The marine habitats north and south of the Hals River outlet are shallow and characterized by patches of sandy bottoms. The Skillefjord River is located at the base of a small fjord arm (Skillefjord) of the Alta Fjord. The river has a mean annual water flow of 3.1 ms À1 and an annual within-river angling catch of sea trout averaging 92 kg (ww.ssb.no). The river has a 13 km stretch accessible to anadromous salmonids. The Alta River is the largest river draining into the Alta Fjord, with a mean annual flow of 88 ms À1 and an annual within-river angling catch of sea trout averaging 2816 kg (www.ssb.no). The river has a 46 km stretch accessible to anadromous salmonids.

| Fish capture and tagging
Fish captured at sea were caught in bag-style fjord nets, which were inspected and cleaned at least once per day. Fish in good condition F I G U R E 1 Map of the Alta Fjord system. Shaded area depicts the part categorized as a National Salmon Fjord. Points show the positions of the acoustic receivers, with the number of hourly detections at each receiver coded by size and colour. Receivers with no detections are shown in dark grey. Yellow diamonds denote the location of the Hals River, the Alta River and the Skillefjord River outlets. Inserted map denotes the location of the Alta Fjord system (yellow square) in Fennoscandia T A B L E 1 Overview of Salmo trutta post-smolts tagged with acoustic tags in the Alta Fjord system and whether they were included or excluded in the analyses Note. Mean fork lengths L F and standard deviations of the different groups are given in parentheses. FW-return refers to the number of S. trutta assumed to end their migration in fresh water. Alta River refers to the number of S. trutta that entered the Alta River at some point during the migration. Other areas refer to the number of S. trutta detected within or in proximity to the other study rivers.
were selected for tagging and transported by boat to the nearest shore or marina in large (≥100 l) holding tubs, with continuous water exchange to ensure good water quality. Fish captured within the Alta River were caught using a fyke net or by electrofishing.
For surgery, the fish were anaesthetised using benzocaine (

| Receiver deployment
A total of 144 acoustic receivers (model TBR700, Thelma Biotel AS) were deployed in 15 arrangements, hereafter termed arrays, within the Alta Fjord system ( Figure 1). This included high-density receiver arrays placed near the Hals River outlet and in Skillefjord, nine acrossfjord receiver arrays positioned throughout the fjord system, three receiver arrays placed strategically within the Alta Fjord and one receiver array placed around the Alta River estuary ( Figure 1). In addition, two acoustic receivers were placed within the Alta River. In the Hals River and Skillefjord River, no receiver was placed in fresh water, but one receiver was positioned at the outlet of each river. Receivers

| Data filtering
The data set was filtered manually before analyses. Only tag numbers corresponding to tags included in the study were evaluated, and tags from other ongoing studies and acoustic noise were removed from the data set without evaluation. After the initial filtering, the data set consisted of 179,192 detections. Based on these, hourly positions, depths and habitats (near shore or pelagic) were estimated. This down sampling to hourly data was done to reduce the impact of multiple detections within a short period of time, i.e., a biased sampling distribution when the fish resided close to the receivers. Hourly positions were calculated using a weighted mean (Simpfendorfer et al., 2002), whereas arithmetic means were used to derive hourly swimming depths. Habitat use was only estimated for the hourly positions at the receiver arrays that provided complete coverage across the fjord ( Figure 1). Detections at the receivers located closest to land (approximately 200 m from the shore) were classified as near shore, whereas detections at all remaining receivers were classified as pelagic. If fish were located both in the pelagic and near shore habitats within the same hour, the habitat was set as missing.
Of the 92 tagged sea trout, 45 generated sufficient data to be included in the analyses. This included 13 fish tagged outside the Hals River Bay, 27 fish tagged in Skillefjord and 5 tagged in the Alta River (Table 1). The 47 fish omitted from the data set consisted of 23 sea F I G U R E 2 Maps of the area near the Hals River (a) and of Skillefjord (b). Points show the positions of the acoustic receivers, with the number of hourly detections at each receiver coded by size and colour. Blue lines represent the 10, 50 and 100 m bathymetry contours. Yellow diamonds denote the location of the river outlets, and crosses denote the release locations trout that were never detected at sea, 13 sea trout that were only registered for shorter periods (days) before disappearing and 11 sea trout that were only detected shortly after tagging before the tags were registered as stationary on the bottom (e.g., due to tag rejection, mortality or predation). The fork length L F of the 45 sea trout included in the analyses ranged between 140 and 216 mm (mean ± S. D. = 177 ± 17 mm). No significant difference in L F was detected between sea trout from the three populations (Fisher-Putman permutation test: P-value = 0.64) or between the sea trout included and excluded from the analyses (Fisher-Putman permutation test: Pvalue = 0.85).

| Data analyses
All statistical analyses were conducted using the R software version 4.0.2.

| Track estimation
Individual tracks were estimated by first executing a linear interpolation between all hourly positions. If the corresponding track resulted in fish crossing land between detections, the shortest possible inwater path was generated. This was done by first generating an adjacency matrix from a spatial grid encompassing the Alta Fjord system, weighted by the distance between grid cells. Subsequently, the shortest path was derived using Dijkstra's algorithm for weighted graphs using the shorthest_path function from the igraph package (Csardi & Nepusz, 2006). Track endpoints were set as the last detection at sea, or as the first detection in fresh water if none of the following detections were in the marine environment.

| Site fidelity
To investigate site fidelity, the probability of sea trout leaving their home areas was modelled using binomial generalized linear models (GLM), with Population, fork length L F and Track duration in days as fixed effects (Table 2). For each population, the home area was defined as the receiver array close to the tagging sites (Figure 1), and sea trout was considered to have left if they were detected at other arrays.

| Total migration distance
The total migration distance of individual fish was measured as the total distance travelled between (i.e., not within) receiver arrays ( Figure 1). This was done due to the nested spatial distribution of receivers in the study area and effectively prevented overestimating the distance travelled by sea trout that resided in areas with a high density of receivers for prolonged periods. The total migration distance was then modelled using linear models (LM) with Population, L F and Track duration in days as fixed effects ( Note. † denotes model terms included in the most parsimonious model. 95% C.I. denotes the 95% confidence intervals for regression coefficients. RE gives the random effects used on the models' intercept. R 2 gives the adjusted r-squared for the GLM, LM and GAM and the marginal R 2 for the LMM, with conditional R 2 in parenthesis. ΔAICc denotes the difference in AICc value between the most parsimonious model and the one providing the second-lowest AICc value. STRØM ET AL. 5 of the year as a population-specific smoothing term and Fish ID as random effect on the model's intercept to account for repeated observations of individual fish (Table 2). GAMMs were fitted using the mgcv package (Wood, 2011).

| Habitat use
When investigating habitat use (near shore vs. pelagic), only detections at the three complete receiver arrays within the Alta Fjord were included due to the low number of detections in the outer straits ( Figure 1). For each fish detected at these arrays, the proportional number of near shore and pelagic registrations were investigated, and potential differences in habitat use were tested using a χ 2 test.

| Depth use
To determine what influenced the depth use of sea trout, a set of and Fish ID as random effects on the model's intercept (Table 2).
Solar elevation in degrees was derived using the suncalc package (Thieurmel & Elmarhraoui, 2019) and used to determine if sea trout displayed diurnal variation in depth use. This was done because the study was conducted during parts of the year when the sun is constantly above the horizon at theses latitudes (70 N), which prevented assigning a diel period (i.e., day and night) to the depth observations. For Temperature and Salinity at the surface, numerical model data obtained on an hourly basis from the A12 model grid of the IMR NorFjords-160 hydrodynamical model which works with a 160 Â 160 m horizontal resolution and 35 vertical layers were used (for details see Myksvoll et al., 2020;Skarðhamar et al., 2018).
These data were also used to summarize the temperatures and salinities experienced by the sea trout. To prevent violations of model assumptions a log transformation was applied to the response variable.

| Model selection
In the models that included repeated observations of individual fish, temporal autocorrelation was investigated and, if necessary, corrected for using a first-order autoregressive process that accounts for the immediately preceding value. In the mixed-effect models, parameters were estimated by the restricted maximum likelihood to prevent potential biases (Zuur et al., 2013). For the linear models, the fit of all model combinations was assessed using the dredge function from the MuMIn package (Barton, 2020) and the models that provided the lowest conditional AIC (AICc) value were considered the most parsimonious. In the additive model, model terms were selected based on their significance.

| RESULTS
A total of 11,898 hourly observations were made from n = 45 sea trout post-smolts in the marine environment from 26 June to 2 September. Of these, 9563 (80%) and 2094 (18%) were made in July and August, respectively. During the marine migration, sea trout experienced temperatures ranging from 7.3 to 16.3 C (mean ± S.D. = 10.9 ± 1.4 C) and salinities ranging from 14.7 to 33.7 (mean ± S.D. = 30.7 ± 1.7).
Track duration ranged between 2 and 65 days (mean ± S.D. = 33 ± 19 days). Overall, 22 sea trout (49%) were assumed to have returned to the three study rivers, with migrations lasting from 8 to 65 days (mean ± S.D. = 44 ± 16 days). This included 6 sea trout from the Hals River, 13 from the Skillefjord River, and 3 from the Alta River (Table 1). Of the Hals River sea trout assumed to have returned to fresh water, three were last detected at the Hals River outlet, whereas the remaining three individuals were last observed within the Alta River. Of the Skillefjord River sea trout assumed to have returned to fresh water, five were last observed at the Skillefjord River outlet, whereas eight were last detected within the Alta River. In contrast, all sea trout from the Alta River that returned to freshwater were last detected within the Alta River (n = 3). The authors found that 16 of the 45 (36%) sea trout entered the Alta River at some point during the tracking period, with date of first river entry ranging from 26 July to 30 August. This included four sea trout from the Alta River (80% of the fish tagged there), nine from Skillefjord (33%) and three from the Hals River (23%) ( Table 1). Of the 16 fish that entered the Alta River, eight re-entered the marine environment after residing in freshwater from 1 to 10 days.

| Horizontal migration
During the tracking period, sea trout spent most of their time in the inner parts of the fjord system, and only 4 of the 45 sea trout (9%) were detected at the outer fjord straits and only for shorter periods ( Figure 1). The highest densities of detections were at the receiver arrays close to the tagging sites in Skillefjord (51%) and adjacent to the Hals River (31%) (Figure 1). When present in these areas the sea trout displayed no apparent attraction towards the receivers closest to the rivers (Figure 2), and the mean distance from the river outlet was 2.2 km (range = 0.5-3.8 km) in Skillefjord and 1.1 km (range = 0.1-3 km) in Hals.
Overall, there was a large variation in horizontal migration, both within and among populations, with some individuals from all three populations migrating between distant receiver arrays ( Figure 3). In all three populations, movements between river outlets were evident, and 44% (n = 20) of the fish were observed within or in proximity to rivers other than their natal river (Table 1). This movement trend was particularly high for sea trout from the Skillefjord River, where 52% (n = 14) were observed in proximity to the other rivers (Table 1). For the Hals and Alta River sea trout, 31% (n = 4) and 40% (n = 2) visited areas close to the other rivers in the study (Table 1) (Table 2).

| Total migration distance
The total migration distance (distance migrated between receiver arrays) ranged between 0 and 308.1 km ( ing terms were significant for both Skillefjord (EDF = 6.54, P < 10 À15 ) and Hals (EDF = 1.31, P = 10 À4 ), and revealed a strong seasonal difference with fish from the Skillefjord River dispersing further away for their origin than fish from the Hals River from mid-July onwards (Figure 4b). No effect of L F was evident (Table 2). In addition to the tendency to stay within the fjord, a substantial proportion of the sea trout (24%) displayed a spatial distribution limited to their home area. Sea trout are often observed to reside close to their natal river, and in some populations individuals are assumed to spend most of their time in estuaries, likely due to superior foraging conditions compared to adjacent marine habitats . Although a substantial utilization of local areas was evident in the current study, there was no indication of a particular attraction towards the receivers closest to the river for neither the Skillefjord nor the Hals fish. Furthermore, the tendency to utilize local areas varied significantly among populations. Fish from Skillefjord had a greater tendency to leave their home area, migrating longer distances and dispersing further from their origin, particularly from mid-July and onwards, than fish from Hals. It is possible that this observed difference could be a result of spatiotemporal variations in habitat quality between sites, and that habitats in proximity to the Hals River outlet may offer more suitable ecological conditions throughout the summer. This may indicate that even in relatively confined fjord systems, local habitat variations may promote population-specific migration patterns.
In addition to these described interpopulation differences, the sea trout post-smolts in this study displayed substantial intrapopulation variation in migratory behaviour. Individuals from all three populations displayed migrations throughout large parts of the Alta Fjord. Large variation in migration distance is common within sea trout populations, and for veteran migrants it has been documented that migratory behaviour is condition-dependent, with individuals in a poorer nutritional state performing longer migrations (Bordeleau et al., 2018;Eldøy et al., 2015). In this study, no attempt was made to attribute migratory behaviour to body condition. The only physical variable related to individual fish investigated was body length, which did not influence neither migration distance nor dispersion from origin. This absence of a size effect on the migratory behaviour of firsttime migrants concurs with a previous study, where sea trout postsmolts' tendency to either stay within Diel vertical movement is a widespread phenomenon among fishes (e.g., Righton et al., 2016;Walli et al., 2009), and for anadromous salmonids it is generally considered to reflect feeding behaviour or predator avoidance as a response to diurnal variation in light availability (Strøm et al., 2017). The current study was largely limited to the period of the year when the sun is constantly above the horizon at these latitudes (i.e., midnight sun); nonetheless, a diurnal effect on depth use was still evident with fish utilizing slightly deeper depths during periods with higher solar elevation. This coincided with a previous study on veteran sea trout from northern Norway, where a slight difference in depth use between day and night persisted through the year (Eldøy et al., 2017). This may indicate that either sea trout are able to adjust their vertical behaviour to subtle difference in light STRØM ET AL. 9 FISH intensity or diel variation in depth use may represent a more general behavioural pattern rather than an explicit response to daily variation in light (Eldøy et al., 2017).
For anadromous salmonids, the bulk of the lifetime growth is obtained at sea, and for sea trout, growth in the marine environment correlates with the duration of the summer feeding migration (Berg & Jonsson, 1990;Jensen et al., 2018). Currently, a major concern for sea trout populations is how increased infections by salmon lice influence the duration of the marine residency . In a recent infection experiment it was documented that the time spent at sea may be reduced by up to 80% under heavy salmon lice burdens . Although the marine residency times reported in the current study are slight underestimations, because most of the sea trout were tagged at sea, the true durations of the marine migrations are likely representative of the natural behaviour of sea trout at high latitudes, as salmon lice infections are low in subarctic areas . Nonetheless, with the projected expansion of aquaculture farming in subarctic areas (Vollset et al., 2021), a potentially dramatic increase in salmon lice burden and other stressors can be expected. This could lead to substantial negative effects on sea trout populations by altering the migratory behaviour of individuals, and reducing their marine growth and survival.
One major caveat of this study is the uncertain origin of sea trout from the Hals and Skillefjord Rivers. All individuals from these sites were tagged while at sea. The sampling of sea trout at sea (late June-early July) coincided with the peak smolt migration in the Hals River (Jensen et al., 2020), and therefore it is likely that most of the sea trout caught in the marine environment were correctly assigned. If this assumption is false, it could potentially weaken the study result regarding behavioural difference among and within populations. Nonetheless, it would not affect the conclusion that sea trout post-smolt from different populations utilized similar and overlapping parts of the fjord system, emphasizing the need for a metapopulation approach when managing the marine phase of these sea trout populations. Furthermore, it may be possible that the results are somewhat biased by the sampling methods.
Whereas the post-smolts caught in Skillefjord and Hals were all caught by bag-style fjord nets, during overlapping sampling periods, most sea trout from the Alta River were sampled with a fyke net, within the river, later in the year. Consequently, it is possible that the fish sampled in the Alta River represent a different constituent of the sea trout population, than the fish sampled in Skillefjord and Hals. Nonetheless, as the sea trout from the Alta River were excluded from most of the analyses, any sampling bias would have limited impact on the conclusions.
In summary, sea trout post-smolts spent most of their time close to their natal river. Nonetheless, longer migrations were observed in all three study populations and 44% of the sea trout migrated between river outlets. While at sea, substantial variation in horizontal migration was present both within and between populations. The observed interpopulation variation in migratory behaviour is likely influenced by differences in habitat quality between sites, indicating that local ecological conditions may determine the migration strategy of sea trout post-smolts.