Seasonal, diurnal and individual variation in Atlantic salmon (Salmo salar L.) swimming depth in commercial-scale sea cages

Atlantic salmon aquaculture may benefit from improved tools for behavioural monitoring and data-derived models that predict behaviour in commercial scale production cages. In the present study, timeseries depth registrations from individual farmed Atlantic salmon in commercial cages have been compiled from various projects, with the aim to resolve general swimming depth behaviour during the marine production stage. By implementing acoustic depth sensor tags, swimming depth data from 126 individuals of Atlantic salmon from ten different cages and five different farm sites were analysed over study periods lasting from 195 to 316 days. The study covers all seasons and were conducted at a latitude where there is hardly daylight during winter solstice nor nighttime or dusk during summer solstice. Average hourly swimming depth registrations for individual salmon were analysed by applying generalized additive modelling (GAM), where the main effects, hour of the day, temperature and sunlight hours were modelled as smooth terms in a non-linear way. The main effects were further modelled conditioned on the presence or absence of submerged lights and/or skirts used to shield the fish from salmon lice infestations. The results reveal that Atlantic salmon generally resided higher in the water column during nighttime than in daytime in all seasons except in the summer. The general tendency for diurnal depth patterns was most pronounced in winter and spring in cages without submerged lights, where average swimming depths varied from about 10–12 m during daytime to about 5–6 m during nighttime. The use of submerged lights in the cages disrupted the diurnal depth patterns and lead the fish to swim deeper in the water column during night-time in months with few sunlight hours. There was no substantial effect of the use of skirts on swimming depths. In addition to these general diurnal and day of the year trends, there were large individual variations in swimming depth. The study emphasises the utility of behavioural traits, particularly in modelling expected depth patterns as a surveillance tool in conjunction with real-time biomonitoring during the ongrowing phase at sea for Atlantic salmon.