key: cord-0283869-vhb01ww6
authors: Johannesen, Ása; Patursson, Øystein; Kristmundsson, Jóhannus; Pætursonur Dam, Signar; Mulelid, Mats; Klebert, Pascal
title: Waves and currents decrease the available space in a salmon cage
date: 2021-10-22
journal: bioRxiv
DOI: 10.1101/2021.07.23.453560
sha: 22283c11f675ecf362e5834a3c2c60cec69d42c7
doc_id: 283869
cord_uid: vhb01ww6

Due to increasing demand for salmon and environmental barriers preventing expansion in established sites, salmon farmers seek to move or expand their production to more exposed sites. In this study we investigate the effects of strong currents and waves on the behaviour of salmon and how they choose to use the space available to them. Observations are carried out in a site with strong tidal currents and well mixed water. Using video cameras and echo sounders, we show that salmon prefer to use the entire water column, narrowing their range only as a response to cage deformation, waves, or daylight. Conversely, salmon show strong horizontal preference, mostly occupying the portions of the cage exposed to currents. Additionally, waves cause salmon to disperse from the exposed side of the cage to the more sheltered side. Even when strong currents decrease the amount of available space, salmon choose to occupy the more exposed part of the cage. This indicates that at least with good water exchange, the high density caused by limited vertical space is not so aversive that salmon choose to move to less desirable areas of the cage. However, the dispersal throughout the entire available water column indicates that ensuring enough vertical space, even in strong currents, would be beneficial to salmon welfare.

Aquaculture is a major provider of fin fish protein consumed globally, accounting for 2 approximately 52% of all fish produced for human consumption [1] . In aquaculture, 3 Atlantic Salmon account for only 4.5% of production in weight, but 19% in value. While 4 aquaculture production is increasing globally, salmon production in the Atlantic is 5 stagnating. The causes mainly relate to complete exploitation of available farming sites, 6 with pollution and parasite infestations being the major factors limiting expansion in 7 near-shore sites [2] . 8 Salmon lice (Lepeophtheirus salmonis) are a major parasite in Atlantic salmon 9 farming. They spend their parasitic life stages on the salmon where they consume 10 mucous, blood, and skin, which leads to sores and in extreme cases, mortality [3, 4] . 11 Even sub-lethally, salmon lice are a cause of poor welfare due to the stress and pain 12 caused both by the parasites themselves and the removal methods [5, 6] . 13 October 22, 2021 1/22

Because of the ever escalating challenges posed by salmon lice and the limitations 14 put on biomass in near shore sites, the salmon aquaculture industry is making 15 investments in adapting current farming methods. One such adaptation is to move 16 farms out to more exposed locations where higher water exchange can mitigate 17 pollution and possibly dilute infectious sea lice [7] [8] [9] . At the current rate of 18 development, salmon will experience substantially larger waves in substantially different 19 farm constructions than those being used in industry for the last 20 years [10] . 20 Due to the forthcoming changes in salmon aquaculture, much work has been carried 21 out in order to determine how well salmon are able to cope with the currently most 22 extreme conditions [9] . From these studies, some information is available on swimming 23 speed capacity [11, 12] , sensitivity to variation in temperature and oxygen 24 saturation [13] [14] [15] [16] , swimming energy expenditure, and how a variety of these factors 25 affect growth and feed conversion [17] [18] [19] . 26 The behaviour of salmon in relation to currents and time of day has also been 27 extensively studied. Salmon change their swimming behaviour according to the 28 availability of light in their environment. When lights are deployed at night during 29 winter, salmon will maintain daytime swimming behaviours and navigate to the depth 30 of the lights [20] , while they disperse more and decrease swimming activity if no lights 31 are deployed [21] . This behaviours is modified by other environmental factors, such as 32 water temperature [22] . Current speed affects swimming behaviour in two major ways. 33 Strong current decreases aggressive interactions and increases shoal cohesion in 34 raceways [11] and in salmon cages a strong current changes swimming mode from 35 circular to standing on current, that is maintaining position in the cage while swimming 36 against the current [23] . 37 Currents do not only affect the swimming behaviour and energy expenditure of 38 salmon, they also affect the shape of salmon cages. Cage deformation due to current can 39 cause a decrease in cage volume of 20% at 0.5 m s -1 [24, 25] . While not as thoroughly 40 studied, there are indications that waves too can change the shape of a salmon cage [26] . 41 There is some recent work on how behaviour is affected by waves, mostly 42 investigating vertical preference and swimming effort [27, 28] . However, while there is 43 data on how changes in the available space within a cage affect biomass and salmon 44 welfare [29] , the effects of waves on available space and salmon behaviour have not been 45 thoroughly investigated. 46 In addition to increasing demand for salmon driving industry to innovate, there is 47 also an increased awareness of fish welfare considerations [30, 31] . This means that there 48 is growing consumer pressure for not only environmental certifications such as ASC 49 (Aquaculture Stewardship Council) [32] , but also for assurances that the farms can 50 deliver a minimum welfare standard, such as the RSPCA assured scheme [33] . Being 51 able to farm fish in exposed locations without compromising on welfare requires 52 extensive knowledge of how these new conditions will affect the fish. While most salmon 53 farmers have intimate knowledge of already established sites, it is still necessary to be 54 able to generalise such knowledge to new and more exposed sites. This study is an 55 attempt at detailing how salmon are affected by the combined effects of waves and 56 currents. Particularly, how the hydrodynamic conditions change their preferred 57 positioning in the cage and how they change their behaviour. 58 Here, we monitor the behaviour of salmon and how they use the space available to 59 them in a cage exposed to both currents and waves throughout the winter months of 60 2019/2020. 61 We predict that currents cause salmon to move upwards in the water column, and 62 that this is at least partially caused by the cage bottom being pushed upwards by the 63 current. 64 We predict that daylight and waves will cause the salmon to move downwards in the 65 Work was carried out at Hiddenfjord's "Velbastaður" salmon farm on the Faroe Islands, 70 which is a site that is exposed to strong tidal currents as well as large waves in winter. 71 The cages are arranged in a straight line along the coast, which is oriented close to a 72 Northwest-Southeast axis (Fig 1) . A small island ("Hestur") west of the farm creates a 73 strait through which tidal currents move, alternating between northbound and 74 southbound currents. The location of the farm is quite shallow (25-30m depth) and the 75 tidal currents ensure good water exchange and a well mixed water column. Due to the 76 hydrodynamic conditions at the site, the side of the farm nearer the shore is more 77 sheltered than the outer side. Depending on wind direction, waves enter the strait either 78 from the south or north of Hestur. For the purposes of this study, the southernmost cage was selected for observation.

Being located at one end of the row of cages, it was highly exposed to currents, 81 especially from the south. Additionally, the cage was expected to be exposed to waves 82 entering the strait from either side of Hestur. Ideally, several cages would be included 83 from different sites in order to get a general picture of the relationship between waves, 84 currents, and salmon. However, for reasons of logistics, this was not possible with such 85 an extensive setup. Therefore, this study details the conditions in one cage that has 86 been thoroughly monitored.

The cage was stocked with 112 thousand salmon with a mean body length of 88 54.0 ± 4.1 cm, weighing 2.16 ± 0.6 kg, which amounted to an approximate biomass of 15 89 kg m -3 . Additionally, the cage housed approximately 10 thousand lumpfish (Cyclopterus 90 lumpus). The salmon were fed using air driven surface feed spreaders regulated using a 91 feed camera, which detected uneaten pellets. Fig 1) . 96 Both profilers were set up to collect hourly wave data from a 20 minute burst. The

Sentinel V measured current data at one ping every two seconds and the AWAC 98 measured currents in bursts of 120 pings at one ping per second every five minutes.

Directly attached to the cage were several different sensors. The layout is presented 100 in Figure 2 .

To monitor vertical distribution of fish within the cage, two echo sounders (Table 1) surface as well as any fish within the echo sounder beam. This allowed us to measure 106 the depth of the cage as well as which parts of the water column were occupied by fish. 107 The echo sounders were set up to ping once every four seconds for the duration of the 108 experiment. Lumpfish do not have swim bladders, so they do not show up clearly in the 109 echo sounder. However, salmon cages on the Faroe Islands often contain a small amount 110 of saithe (Pollachius virens), which have entered the cage when they were small enough 111 to get through the net. These do have swim bladders, so the data from echo sounders 112 include both salmon and saithe.

The echo sounders were equipped with sensors measuring tilt and pressure (Table 1) . 114 These were used to detect any instances where deviations from vertical would invalidate 115 distance data and to validate the distance to the surface measured by the echo sounders. 116 Tilt sensors recorded tilt and pressure every five seconds.

In addition to the echo sounders, six pressure sensors (Table 1) were attached to the 118 cage in order to properly account for any cage deformation that may occur. These were 119 set up to record pressure every five seconds.

To monitor salmon behaviour, five video cameras were positioned in the cage 121 ( Table 1 ). The camera that the salmon farmers used for monitoring feed consumption 122 ("Feed camera") was included in our recording setup and was located approximately in 123 the centre of the cage and nearer the surface (at 7 m depth) than our other upward 124 facing cameras. Three upward facing cameras were located approximately equidistantly 125 around the edge of the cage, one towards the south where northbound tidal currents 126 entered the cage ("South"), one towards the north where southbound tidal currents 127 entered the cage ("North") and one towards the east, which was the most sheltered 128 location in the cage ("East"). The cameras were attached to the bottom of the cage and 129 held up using a buoy (at 9 m depth). One camera was placed on the side of the cage at 130 five metres depth looking inward ("Inwards") near the "South" camera. Camera 131 "North" and "South" were positioned such that they would capture current related 132 swimming behaviour, such as standing on current as well as any potential changes 133 caused by waves. The more sheltered camera was used to record salmon avoiding 134 currents or waves and whether consistent swimming behaviour (for example circling the 135 cage) persisted throughout the cage. The camera looking inwards was used to capture 136 close up video of salmon either orienting towards the current or swimming alongside the 137 edge of the cage as well as capturing salmon near the surface of the water. Finally, the 138 feed camera was used to capture presence within the centre of the cage outside of 139 feeding times. Details about the instruments can be found in Table 1 . 

Video cameras were remotely controlled using iSpy [34] and scheduled to record for 145 five minutes each once per week. In addition to these baseline recordings, alternative 146 schedules were enacted for bad weather events to record for five minutes three to four 147 times each day during bad weather to capture behaviours in large wave conditions. At 148 the end of the trial, the recorded wave data were sorted by wave height and period Unfortunately, when the weather was very bad, power did sometimes cut out, so it was 155 not possible to assemble a perfectly balanced number of videos for all conditions. For 156 swimming effort, most of the videos selected for behavioural observations were used, but 157 in some videos, no fish were present, so these two data sets are not the same size.

Echo data were recorded continuously on a local hard drive and then uploaded 159 remotely to a cloud server.

Current and wave data was recorded locally and downloaded when the ADCPs were 161 recovered.

Tilt and pressure sensors stored data locally and data were downloaded when the 163 sensors were recovered.

Welfare monitoring 165 In order to ascertain the general welfare of salmon on the farm during the study period, 166 we carried out Operational Welfare Indicator registrations. During the sampling period, 167 Operational Welfare Indicators (OWIs) were recorded every two weeks when weather 168 allowed ( Table 2) . Prior to this, welfare indicators were collected on a more ad-hoc basis 169 to establish a baseline, and finally again at harvest. The large gap in data from March 170 until harvest is due to the COVID-19 pandemic preventing fieldwork. The OWIs were 171 adapted by Hiddenfjord from SWIM 1 and 2 [35, 36] to make them more practical to use 172 as part of regular farm management practices. The OWIs were collected from 10 salmon 173 from each cage in connection with routine louse counting. The salmon were caught 174 using a dip net, anaesthetised in 60 mg L -1 Finquel (MS-222) and lice numbers and gill 175 condition were recorded before OWIs were recorded. After regaining consciousness in 176 fresh seawater, the salmon were released back into the cages. For the purposes of this 177 study, a sum of scores for each salmon excluding sclera colour was used to determine the 178 overall welfare of the salmon. Sclera colour or eye darkening was not included here 179 because while there is evidence that they can be used as an indicator of stress [37] , this 180 has not yet been well established in salmon. The total score excluding sclera can range 181 from 0 to 10 with a low number indicating good welfare and a high number indicating 182 poor welfare. Many individual based indicators in SWIM 1 and 2 are not included 183 because there was either no variation in them (e.g. sexual maturation did not occur) or 184 they did not relate to conditions on the farm at the time (e.g. deformities). Each salmon was scored on their overall condition, so both eyes and all fins were checked and the greatest score recorded for each.

Due to the nature of salmon moving in and out of camera, no attempt at counting the 187 salmon was made. Instead, videos were coded in one of three qualitative categories 188 throughout; "No salmon" (less than five salmon visible), "Some salmon" (more than 189 five and less than 50 salmon visible), and "Many salmon" (More than 50 salmon visible). 190 Finally, the presence of salmon near the surface was recorded for reasons of validating 191 the near surface echo sounder data. When salmon were recorded in cameras with a view 192 of the net, collisions with the net were also recorded.

In addition to the presence of salmon, the general behaviour of the salmon was Behaviours were extracted from videos manually using BORIS [38] and swimming effort 208 was recorded using VLC [39] with the "Time" extension [40] .

Echo sounder data were extracted from the raw data files using the "oce" package in 210 R [41] and exported for further processing in Python [42] . First, the water surface was 211 found. Second, data above the surface was removed, and depth adjusted to the surface 212 rather than the distance from the echo sounder. Third, the data were binned into 5 213 minute intervals and 16.5 cm depths and S v (acoustic back scattering strength) averages 214 were exported. Once data had been exported, the exported files were read into R, where 215 the surface and the lowest 4.5 metres (below the cage bottom) were removed. The code 216 for the echo sounder data processing can be found on Github [43] .

218 Analysis was carried out in R [44] and tidyverse [45] using the packages lubridate [46] 219 and circular [47] for data cleanup, lme4 [48] and lmerTest [49] for statistical inference, 220 performance [50] for model fit assessment, and ggplot2 [51] with colorspace [52] was 221 used for plotting. All data and code used for the analysis is available on Github [43] .

For video data, the following methods were used; to analyse swimming mode, a 223 general linearised mixed effects model was used with a binomial (log link) family where 224 each video was classified as salmon being mostly in either swimming mode, with current 225 strength and current direction as predictors, and with camera as the random intercept 226 term. The effects of environmental conditions on swimming effort was analysed using a 227 linear mixed effects model with tail beats per second as the dependent variable, current 228 speed, current direction, and wave period as predictors as well as camera as random 229 intercept term. The reason for including camera as random intercept term in these 230 models is that hydrodynamic conditions are not uniform throughout the cage, so the 231 salmon in the different cameras will be affected differently by the conditions measured 232 outside of the cage. The models described are the minimal adequate models, where 233 variables that did not significantly affect the fit of the model have been removed. The 234 amount of time where "Many" salmon appear in a camera (proxy for proximity to sides 235 and surface) was analysed using a general linearised mixed effects model with time investigated using a binary classification of videos to "shoaling" and "not shoaling" in a 245 binomial family glm due to the highly bimodal nature of this variable where salmon 246 were either mostly shoaling or not at all. Predictors were wave height, current direction 247 and camera. Model fit for each model was assessed using diagnostic plots.

For echo data, linear models were used to estimate the effects of environmental 249 variables on the "evenness" of fish dispersal within the water column using raw S v as a 250 proxy for relative fish density. The residuals from these models indicated how variable 251 S v was, so large residuals indicated "clumping" and small residuals indicated evenly 252 dispersed fish. We also used linear models to how estimate how environmental variables 253 affected fish swimming depth, weighing the depth variable by S v . While it is possible to 254 estimate real biomass from the back scatter [53] , we did not calibrate gain to do this, as 255 we were more interested in relative changes rather than biomass estimation.

This study was not a manipulative experiment. However, we did install video cameras 258 within the cage as well as echo sounders underneath it. We do not have reason to 259 believe that these affected the salmon, but they were a deviation from the regular 260 routines on the farm. We also handled the salmon in order to perform OWI monitoring, 261 but this was already part of the management routine at Hiddenfjord farms, so didn't 262 deviate from normal practices. Regardless, ethical approval was still applied for and 263 given by Fiskaaling's Ethical Board (Approval number 007). (Fig 3) . Flow direction was mostly bimodal switching between a 270 north-westerly current and a south-easterly current, hereafter referred to as Northbound 271 and Southbound current (Fig 3) . Due to the difference in particularly current direction 272 between the two ADCPs, the Sentinel V or a tidal analysis built on the data from the 273 Sentinel V were used for further data analysis using current as a predictor. m. South of the farm it measured (Hs) 3.24 m (Fig 4) . Wave period (Tp) ranged from 276 2.07 to 22.55 seconds. Low and high wave heights were recorded at the same time in 277 both ADCPs, indicating that they were exposed to similar wave heights (Fig 3) . The 278 wave data from the AWAC was used in analyses going forward due to the longer 279 measurement period. Fig 6 ) , which is consistent with the indication from the pressure 297 sensors that northbound current affected the cage more than southbound current.

Swimming effort 299 Because salmon recorded for swimming effort were more likely to be fish that spent a 300 longer time in frame than those who spent a short time in frame, there are likely to be 301 differences between these fish and the general population within the cage. Because of this change in swimming effort related to swimming mode, data were 305 analysed with swimming mode as a random predictor in a linear mixed effects model. 306 Swimming effort increased with current speed in northbound current, but this 307 connection was not present in southbound current (t = -2.358, df = 440, P = 0.019, 308 Fig 7) . Wave length interacted with current speed in such a way that in weak current, 309 salmon had slower tail beats in longer waves, but in stronger currents, they had faster 310 tail beats in longer waves (t = 2.309, DF = 440, P = 0.021, Fig 7) . Swimming effort in salmon standing on current in northbound and southbound currents. Shading indicates wave length less than 12 seconds (lighter colour) and more than 12 seconds (darker colour).

Salmon showed great horizontal preference (F 4,160 = 8.641, P < 0.001) with the "South" 313 camera recording "Many" salmon 36% of the time compared to the "East" camera, 314 which only recorded "Many" salmon 4.5 % of the time (Fig 8) . The amount of time 315 where "Many" salmon were recorded in cameras decreased overall in large waves 316 (combined high Hm0 and Tp) (F 3,161 = 3.398, P = 0.019, residual df = 159), though 317 this change varied by camera with the "North" and "South" cameras showing little or 318 even the opposite change (Fig 8) . However, in taller waves the total amount of time 319 where any salmon were observed in the "East" camera increased, with the proportion of 320 time when any ("Some","Many", or "Shoal") salmon were visible in the "East" camera 321 increasing from 50% in waves up to and including 1.1 m tall to approximately 75% in 322 waves taller than 1.1 m (F 4,155 = 7.545, P < 0.001).

Collisions with the net decreased in taller waves, but less so if current was strong too 324 (z = 2.431, DF = 41, P < 0.015, Fig 9) . 325 Shoal cohesion and position 326 Salmon were less likely to shoal in larger waves (z = -2.023, DF = 74, P = 0.043) and 327 when the current was southbound (z = -2.185, DF = 74, P = 0.029). The occurrence of 328 shoaling differed between cameras with no shoaling observed in the "East", "North", 329 and "Feed" cameras and with more shoaling observed in the "South" camera compared 330 to the "In" camera (z = 2.46, DF = 74, P = 0.014).

Fish avoided the surface during the day (Fig 10) and this pattern persisted 332 regardless of hydrodynamic conditions.

As current increased, the fish narrowed their vertical distribution within the water 334 column resulting in stronger localised S v (South echo sounder; F 3,250864 = 2993, P < 335 0.001, North echo sounder; F 3,239438 = 1836, P < 0.001) and greater residuals (South 336 echo sounder; F 3,250864 = 759.1, P < 0.001, North echo sounder; F 3,239438 = 1265, P < 337 0.001) as opposed to a lower signal with more even dispersal in weaker currents (Fig 11) . 338 Fish moved upwards in the water column in stronger current. The side of the cage 339 and direction of current affected the degree to which the shoals moved up (South echo 340 sounder; F 3,218622 = 8665, P < 0.001, North echo sounder; F 3,206399 = 7170, P < 0.001, 341 Light boxes indicate small waves and dark boxes indicate large waves. Dots are raw data and boxes and whiskers represent quartiles. Wave height was split at 1.1 m with "Small" waves being up to and including 1.1 m tall and "Large" waves being taller than 1.1 m. This was an approximate 50/50 split of the available data, and does not necessarily represent biological significance. Statistics are carried out using continuous wave parameters. Fig 11) . They still avoided the surface during the day, resulting in a greater 342 concentration of fish below 5m depth (Fig 11) .

In an effect similar to that of daylight, waves caused fish to move away from the 344 surface (South echo sounder; F 4,218621 = 6307, P < 0.001, North echo sounder; F 3,206399 345 = 6026, P < 0.001). However, at the side of the cage where current entered the cage, 346 fish moved upward, countering the effect of waves (Fig 12) . 

While injury scores were low throughout the entire production cycle with 90% of salmon 349 scoring three or lower in injuries most of the time, there is significant variation between 350 sampling times with a period in January and February where more than 25% of the Fig 13) . In late June, when the salmon were harvested, higher scoring fish had 353 decreased again to less than 20% (Fig 13) . Throughout the entire observation period, 354 only one fish scored more than 6 and only during the harshest winter months were a 355 substantial number of salmon scoring 5 and 6 found.

At the study site, currents caused the cage to deform, which is consistent with 358 measurements elsewhere indicating a minimum volume loss of 20% at 0.5 m s -1 [24, 25] . 359 The bottom of the salmon cage moved upwards on the side of incoming current from current enters the cage, and even in the sheltered side, the cone moves upwards in very 363 strong currents.

Because of surface avoidance during daylight hours [54] , salmon at this study site 365 preferred to occupy the cone portion of the cage during the day, while they occupied the 366 entire water column during the night. This caused a concentration of fish above the 367 cone (upwards of 10 m) in stronger currents, because a major component of cage 368 deformation is the bottom of the cage being pushed upwards [25] . When strong currents 369 peaked during daylight hours, this caused a greater concentration of fish in the 5-10 m 370 depth range above the cone and away from the surface. In our previous paper [27] , it there is no evidence to suggest that the salmon move away from the bottom of the cage 375 at any time, rather that they adjust to the change in space available to them.

In addition to the effect of daylight and current, waves also affected vertical shoal 377 positioning. At night, salmon avoided the surface in tall waves similarly to how they 378 respond to daylight. Due to the effect of cage deformation, this preference was not so 379 clear on the side of the cage affected by oncoming current. Whether this is caused by an 380 interaction between current and waves, or simply that surface aversion due to daylight 381 is stronger than that caused by waves is uncertain. However, daylight combined with 382 large waves caused fish to move further down as evidenced by both echo sounder data 383 and video data. This corresponds well with our previous paper [27] as well as what was 384 seen in Signar P. Dam's study [28] . Very little other work has been done directly on 385 salmon and waves in farming, but there is some theoretical work indicating that in long 386 period waves out at sea, the horizontal movement of the waves may exceed the 387 swimming capacity of salmon [55] , particularly if the cages are not deep enough for the 388 salmon to dive below the waves.

In terms of association with the net, there was some indication that there were fewer 390 salmon near the net in waves that were both tall and long, but the most decisive change 391 was in the more sheltered part of the cage, where the median proportion of time where 392 salmon were seen in the video increased by 50 percentage points. This indicates that 393 there was more dispersal of salmon from the more popular exposed side of the cage, to 394 the more sheltered side when the waves were large. Despite little to no change in "East" camera, collisions with the net generally decreased as wave height increased, but 397 only in weak current. It is possible that strong current negatively affects salmon's 398 ability to adjust swimming speed to accommodate waves, or this difference might be 399 due to a greater proportion of salmon standing on current near the net in strong current. 400 Free swimming salmon do not necessarily associate as strongly with the net.

As expected, salmon swimming mode was affected by current speed with fish 402 changing from swimming freely to standing on current as current increased [23] . 403 Perhaps unintuitively, the salmon beat slower with their tails once standing on current. 404 Literature suggests that salmon change from circular swimming to standing on current 405 when the current exceeds their preferred swimming speed [56] . However, at this 406 location, salmon changed swimming mode at fairly low currents, so the swimming effort 407 was actually lower at that current speed than their cruising speed. While there were 408 some effects of waves and currents on swimming effort, neither had a major effect, and 409 the low speed at which the salmon switched to standing on current might indicate that 410 they prefer to do so at this site. Literature suggests that forcing salmon to exceed their 411 preferred swimming speed might have a negative influence on salmon [57] , but 412 considering that current speeds rarely reached one body length per second (1 BL s -1 ) at 413 this site, this is unlikely to be a problem for these salmon, and might in fact perform 414 the role of environmental enrichment instead. Considering that salmon were also seen 415 less in the "East" camera, there is clear indication that even at low currents, salmon in 416 this cage do not perform a circular swimming pattern using the entire perimeter of the 417 cage as is seen in other farms [23, 58] . While currents were weak by 50 cm long salmon 418 standards, these cages were also inhabited by lumpfish, and it is possible that the 419 strongest currents at this site exceed lumpfish swimming capacity [59] . Lumpfish 420 welfare data were not collected for this study. However, the farmers at the site were 421 aware of potential implications of currents, and had deployed shelters that were adapted 422 for use in strong currents, providing firm surfaces to attach to as well as shelter from 423 the current. As opposed to our previous paper [27] , swimming effort did not decrease in 424 larger waves, but rather increased (though only in long period waves). The effect of 425 wave period was most apparent in stronger currents, which were not present at the 426 previous study site. It may also be that the larger waves seen in the video footage in 427 that study were tall waves, but with a shorter period than what was seen in this study. 428 While wave period and height do correlate, this correlation is not perfect, so that too 429 could account for the differences seen.

Taken together, the results indicate that salmon prefer to use the entire water 431 column available to them, and only move upwards in the water column due to cage 432 deformation and downwards due to waves and daylight. However, the data suggest that 433 they do not use all of the available horizontal space. There are a few potential 434 explanations for this. Two potential candidates are; 1) water quality is undesirable in 435 the more sheltered areas of the cage [15, 29] , 2) salmon prefer to occupy a space with 436 current. We have no reason to suspect poor water quality in the sheltered parts of the 437 cage, particularly as this study was carried out in the winter months with a lot of 438 mixing in the water column and good general water exchange, even in the sheltered 439 areas. Therefore, the more likely explanation is that the salmon actively choose the 440 more exposed side of the cage rather than avoid the sheltered side. We have found no 441 literature investigating preferential swimming in currents in a salmon cage other than in 442 relation to water quality parameters, but it appears as though there is preference for 443 current regardless of water quality.

OWIs remained low for the majority of fish throughout the production cycle.

Welfare indicator scores were slightly elevated in January and February of 2020 directly 446 after large wave events. However, fewer collisions were seen in larger waves, so if these 447 elevated OWIs were caused by collisions, they must have happened off-camera. There 448 are three reasons this could be: 1) collisions happened at night, 2) collisions happened 449 elsewhere in the cage, 3) power cut out during bad weather preventing us from seeing 450 the collisions in the largest waves. That being said, the OWIs were only slightly 451 elevated during these months and there are numerous other explanatory variables, 452 which we were unable to control for. Therefore, one ought not necessarily conclude that 453 bad weather events caused poor welfare.

This study is limited by the available study locations. Particularly that only one 455 cage was monitored for salmon behaviour. This limits our conclusion to this one cage, 456 but considering that the current literature on salmon in commercial cages is also often 457 limited to one or a few units, these data add to what is already a limited pool of 458 information.

The strongest effects that currents and waves have on salmon relate to how they 461 decrease the space available to the fish. Salmon actively choose to occupy areas in the 462 cage exposed to stronger currents, but avoid the surface in large waves. For farming in 463 new sites exposed to strong currents and potentially also in currently used sites, one 464 ought to consider the limitation to vertical space when currents cause cage deformation. 465 Therefore, making the cage resistant to deformation and deep enough that even in the 466 strongest currents, there is enough vertical space, can be beneficial to salmon welfare. 467 This is particularly relevant during the day, when waves are large, and if the biomass is 468 high, for example close to harvest. At sea, waves are likely to have longer periods, 469 necessitating a deeper diving depth to avoid them. Alternatively, salmon will have to 470 move away from the sides of the cage to avoid collision, which effectively reduces the 471 diameter of the cage, greatly reducing the volume.

Supporting information 473 We would like to thank Hiddenfjord for cooperating with us in this project. Without 474 access to their farm and the help of their staff on site, this work would not be possible. 475 

The State of World Fisheries and Aquaculture

The global economic cost of sea lice to the salmonid farming industry

A Review of the Impact of Parasitic Copepods on Marine Aquaculture

Caligus elongatus and other sea lice of the genus Caligus as parasites of farmed salmonids: A review

Salmon lice treatments and salmon mortality in Norwegian aquaculture: a review

Does the thermal component of warm water treatment inflict acute lesions on Atlantic salmon (Salmo salar)?

Effect of exposure on salmon lice Lepeophtheirus salmonis population dynamics in Faroese salmon farms

Identifying salmon lice transmission characteristics between Faroese salmon farms

Fish welfare in offshore salmon aquaculture

Challenges and emerging technical solutions in on-growing salmon farming

Not too slow, not too fast: water currents affect group structure, aggression and welfare in post-smolt Atlantic salmon Salmo salar

What is the limit of sustained swimming in Atlantic salmon post smolts?

The oxygen threshold for maximal feed intake of Atlantic salmon post-smolts is highly temperature-dependent

Diving behaviour of Atlantic salmon at sea: effects of light regimes and temperature stratification

Oxygen gradients affect behaviour of caged Atlantic salmon Salmo salar

Response of Atlantic salmon Salmo salar to temperature and dissolved oxygen extremes established using animal-borne environmental sensors

Dissolved oxygen variability in a commercial sea-cage exposes farmed Atlantic salmon to growth limiting conditions

Effects of light source and intensity on sexual maturation, growth and swimming behaviour of Atlantic salmon in sea cages

Growth and welfare of submerged Atlantic salmon under continuous lighting

Atlantic salmon Salmo salar instantaneously follow vertical light movements in sea cages

Artificial light and season affects vertical distribution and swimming behaviour of post-smolt Atlantic salmon in sea cages

Thermo-and photoregulatory swimming behaviour of caged Atlantic salmon: Implications for photoperiod management and fish welfare

The Interaction between Water Currents and Salmon Swimming Behaviour in Sea Cages

Three-dimensional deformation of a large circular flexible sea cage in high currents: Field experiment and modeling

Current induced net deformations in full-scale sea-cages for Atlantic salmon (Salmo salar)

Experimental study of an aquaculture net cage in waves and current

How caged salmon respond to waves depends on time of day and currents

A study on Atlantic Salmon (Salmo salar) in aquaculture: moving into exposed waters, the effect of waves on behaviour and growth

Fluctuating sea-cage environments modify the effects of stocking densities on production and welfare parameters of Atlantic salmon (Salmo salar L.)

Animal welfare risks of global aquaculture

Aquaculture Stewardship Council

RSPCA welfare standards for farmed Atlantic salmon

Salmon Welfare Index Model (SWIM 1.0): a semantic model for overall welfare assessment of caged Atlantic salmon: review of the selected welfare indicators and model presentation: Salmon Welfare Index Model (SWIM 1.0)

Salmon welfare index model 2.0: an extended model for overall welfare assessment of caged Atlantic salmon, based on a review of selected welfare indicators and intended for fish health professionals

Eye darkening as a reliable, easy and inexpensive indicator of stress in fish

BORIS: a free, versatile open-source event-logging software for video/audio coding and live observations

VLC media player

Time: VLC extension

Analysis of Oceanographic Data

Python 3 Reference Manual

Waves and currents decrease the available space in a salmon cage -repository

R: A language and environment for statistical computing

Welcome to the tidyverse

Dates and Times Made Easy with lubridate

Circular Statistics

Mixed-effects modeling with R

lmerTest Package: Tests in Linear Mixed Effects Models

performance: An R Package for Assessment, Comparison and Testing of Statistical Models

Elegant Graphics for Data Analysis

A Toolbox for Manipulating and Assessing Colors and Palettes

On the relation between echo intensity and fish density. Fiskeridirektoratets skrifter: serie havundersøkelser

Environmental drivers of Atlantic salmon behaviour in sea-cages: A review

A new criterion for multi-purpose platforms siting: Fish endurance to wave motion within offshore farming cages

Assessing swimming capacity and schooling behaviour in farmed Atlantic salmon Salmo salar with experimental push-cages

Fish welfare based classification method of ocean current speeds at aquaculture sites

Flow Fields Inside Stocked Fish Cages and the Near Environment

Metabolic rates, swimming capabilities, thermal niche and stress response of the lumpfish, Cyclopterus lumpus