From signals to states: biologger-based classification of seabass welfare states in sea-cages

Aquaculture has grown significantly in recent years, increasing the need for advanced monitoring techniques to ensure fish welfare and optimise management practices. Understanding how fish respond to environmental and anthropogenic factors is key for improving welfare standards, and biologgers capable of measuring heart rate (HR) and external acceleration (ACC) provide valuable insights into physiological and behavioural dynamics. In this study, HR and ACC were recorded from adult European seabass implanted with biologgers and monitored in sea-cages for two 14-day periods in March and July. Feeding and routine cage maintenance occurred from Monday to Friday, whereas no aquaculture-related human activity took place during weekends. A Random Forest (RF) model was developed using labelled data from controlled stress-challenge experiments to classify four welfare states: resting, regular activity, reactive response, and proactive response. Standardized ACC was identified as the main predictor for proactive responses, whereas standardized HR contributed most strongly to resting and reactive states. Application of the model to sea-cage data revealed clear diel patterns: regular activity and resting predominated at night and early morning, while proactive responses increased from midday onwards and were closely related to feeding routines. Significant differences also emerged between weekdays and weekends, with stress-related states more frequent during weekdays and resting and regular activity dominating weekends, reflecting the influence of routine operations and human activity in the farming facilities. Seasonal patterns further revealed higher HR levels and a greater prevalence of proactive responses in July, likely driven by elevated water temperatures, increased anthropogenic pressure and enhanced behavioural alertness under summer conditions. Overall, the integration of biologgers with machine learning classification provides a robust framework for identifying welfare states in seabass reared in sea-cages, demonstrating how physiological, behavioural, and environmental data can be combined to inform management decisions, optimise operational protocols, and ultimately enhance welfare-oriented aquaculture practices.