Introduction
Norway is the largest exporter of Atlantic salmon (Salmo salar) in the world, with a production of 1,5 million tons in 2020 (FAO., 2022). Most of the grow out period happens in sea-cages which can host up to 200 000 individuals in a volume comprised between 18 500 and 80 000 m3. Sampling fish for size and biomass estimation at such scale is challenging and time-consuming for farmers. They usually sample around 20 fish per week for lice counting and use this sampling periodically to measure and weight fish, which may be not representative for the whole cage. This can have a strong impact on feeding strategies and slaughtering prevision, followed by effects on production cost and yields. New technologies, like automated measurements from images of stereo-paired cameras to estimate salmon size and biomass in sea-cages, are developing fast for more representative and non-invasive samplings. Nevertheless, previous studies suggested that salmon distribution might not be as homogenously thought. Folkedal et al. (2012) showed that the vertical distribution of salmon in 2000 m3 sea-cages is size-dependent, suggesting repercussions on biomass estimation depending on the sampling method. This needed to be confirmed at commercial scale. Similarly, Johannesen et al. (2022) suggested that horizontal distribution of fish in relation to depth might be very different in a commercial sea-cage. Both findings are important to consider for the deployment of automatic cameras to get representative results.
Materials and methods
Low-cost action cameras (GoPro Hero 8) were time-synchronized and deployed to cover 16 positions in a commercial sea-cage (18 850 m3) of the salmon sea-farm in Gudmunset-Norway (62.45366, 6.60326). Four depths (1 m, 5 m, 9 m, 14 m) were covered at four horizontal positions (North, East, South, West) in the sea-cage. Cameras were set to record for at least 1h, four times in one day (4 deployments, during: 1st feeding and 2nd feeding periods, between feedings, and after 2nd feeding). On the east side of the cage, homemade stereovision rigs with paired GoPros were deployed. 3776 fish through the different depths and deployments were manually fork-length measured with Eventmeasure software (SeaGis 2006-2023). Fish abundance was scored from 0 to 3 (absence of fish, low, medium, high) and growth-stunt fish counted at every position in the sea-cage on screenshots taken every minute for 60 minutes per deployment. Fish swimming speed was also estimated on a portion of the fish length measured using a vertical reference line and counting the time the fish took to pass the reference line from snout to fork-length (BL/sec). Average fish length was compared per depth and per deployment and the same procedure was applied for swimming speed. Cumulative means for fish size were created per depth to estimate when a sample was representative for its depth and for the whole cage.
Results
Salmon fork-length averages were significantly (p < 0.05) different between depths with longer salmon found deeper in the sea-cage (9 m and 14 m) compared to the shallowest depth (1 m) and intermediate one (5 m). This was the case for every deployment that day. A strong variation in size, from 24 cm up to 70 cm, was present and mostly showcased at 1 m than at deeper depths. Distribution of fish appeared quite homogeneous on the horizontal plan, with higher fish abundance at 9 m and 14 m, intermediate at 5 m and low at 1 m depth. No difference was found between feeding or none feeding periods and fish were more abundant at the deeper depths during the day. Fish mean swimming speed was around 1 BL/sec. Most growth-stunt fish were located at 1 m and some individuals could clearly be identified and observed several times at the same position. The cumulative means per depth indicated that average size for a particular depth could be reached with 50 individuals measured in some cases, but most of the time after 150 individuals measured.
Discussion and conclusion
These results are in accordance with the findings from Folkedal et al. (2012) and Nilsson et al. (2013) which found a vertical size-stratification with depth in smaller sea-cages. The use of stereovision to measure salmon in this study allowed to reduce potential bias in the selection of the fish to be measured as fish were not afraid of the stereo rig and were randomly selected to be measured. It appeared that this display did not disturb salmon in their “regular” sea-cage behavior. Results on fish abundance registered at the different positions in the cage suggest that in the absence of strong current or weather, salmon seem to use the whole cage quite homogeneously on the horizontal plan. The fact that high fish abundances were found at 9 m and 14 m, even during feeding times, suggest that the whole school of fish does not gather close to the feeding area at the same time. A different pattern would probably be observed in the center of the cage, not covered by our cameras, where feeding is occurring. From the different findings on vertical size stratification, horizontal and vertical distributions, and average size representativity per depths, several potential implications for an optimal deployment of automatic biomass estimating cameras can be mentioned. On one hand, if the primary objective of the farmers is to get the overall size/biomass variation in the cage, positioning the camera higher up appears more strategic as smaller fish tend to stay higher in the water column and might not be caught at deeper depths. This might be relevant for sea cages in which a large size stratification is present. On the other hand, if the objective is to get a close estimation of the mean size or biomass, it would be more interesting to stand were the larger part of the school is, in our case study: between 9 m and 14m. Anyhow, it seems difficult to obtain the exact average size if the camera is deployed at a single depth. Depending on the weather conditions, fish might use preferentially a part of the cage (Johannesen et al. 2022), and this should be accounted for when deploying equipment. Such studies should be conducted in different localities, during different seasons and production stages to be able to generalize, and automatization of measurements is necessary to conduct more experiments at that scale. This case study aimed to enrich what was previously known on salmon distribution at commercial scale and showed potential considerations for the use of innovative methods to measure salmon size.
References
FAO. (2022). In Brief to The State of World Fisheries and Aquaculture 2022. In In Brief to The State of World Fisheries and Aquaculture 2022. FAO. https://doi.org/10.4060/cc0461en
Folkedal, O., Stien, L. H., Nilsson, J., Torgersen, T., Fosseidengen, J. E., & Oppedal, F. (2012). Sea caged Atlantic salmon display size-dependent swimming depth. Aquatic Living Resources, 25(2), 143–149. https://doi.org/10.1051/alr/2012007
Johannesen, Á., Patursson, Ø., Kristmundsson, J., Dam, S. P., Mulelid, M., & Klebert, P. (2022). Waves and currents decrease the available space in a salmon cage. PLoS ONE, 17(2 February), 1–21. https://doi.org/10.1371/journal.pone.0263850
Nilsson, J., Folkedal, O., Fosseidengen, J. E., Stien, L. H., & Oppedal, F. (2013). PIT tagged individual Atlantic salmon registrered at static depth positions in a sea cage: Vertical size stratification and implications for fish sampling. Aquacultural Engineering, 55, 32–36. https://doi.org/10.1016/j.aquaeng.2013.02.001