As global aquaculture expands, intensifies and becomes increasingly automated, the monitoring of livestock for health and welfare status becomes increasingly challenging, particularly given fish are underwater, with population sizes many times higher than in most terrestrial farming (Huntingford et al. 2023).
There is thus a need to monitor the welfare status of the whole populations on farms and predict disease outbreaks, with potential for the latter demonstrated by Maloy et al. (2020). Subsequently, a hydroacoustic system has been developed which can monitor behavioural patterns of most of the fish population in large salmon cages 24/7 (see Kadri & Kvam 2023).
A welfare platform aggregates data from hydroacoustics, environmental sensors and the farm’s feeding systems. Welfare indicators are constructed from each of these sources and combined into an overall welfare score, to give farmers a warning system for changes in welfare status at the cage level. Using hydroacoustic measurements of the whole population we apply advanced pattern recognition and artificial intelligence algorithms to extract behavioural markers which are statistically calibrated over a large dataset to establish expected levels. Other parameters incorporated include temperature and oxygen, as well feeding; with appetite levels inferred from feeding system data for anomaly detection purposes.