Abstract
Intoduction With the expansion of the aquaculture industry, the need arises for scalable, reliable, and robust methods to assess fish behaviour in sea cages to guide operational management, which includes feeding optimisation and welfare assessments. Fish cage monitoring utilising either acoustic transmitters or underwater cameras is well-studied. However, the relationship between those two different measurement types seems to have not been explored, nor have they been evaluated together in one experimental site. Methods In our 1-month study, we compared the activity of 14 sentinel fish and the artificial intelligence (AI)-inferred speed of individuals from the European seabass ( Dicentrarchus labrax ) sea cage population in three feeding trials. Comparisons include a maximum activity comparison using persistent peaks, fish behavioural pattern establishment and retention, and periodical behavioural patterns. Results Our results demonstrate that under certain circumstances, both technologies are interchangeable from the perspective of persistent peaks and periodicity, but complementary when it comes to behaviour analysis such as food anticipatory behaviour (FAB). Discussion We anticipate that our findings will stimulate advances where multiple sensor types are in use to achieve a more holistic understanding of fish behaviour in the aquaculture sector using underwater technologies.