Harnessing machine learning for the non-invasive assessment of wild fish welfare
Grantee: Thomas Pike
Institution: University of Lincoln, UK
Grant amount: $81,927
Grant type: Challenge grants
Focal species: Fish
Conservation status: Least concern
Disciplines: Animal behavior, cognition, climate science, ichthyology
Research locations: United Kingdom
Project summary
Wild fish are facing unprecedented challenges from climate-change-induced alterations to natural environments and anthropogenic stressors. Yet assessing their welfare is extremely challenging. To address this, this project aims to develop an open-source aquatic camera system capable of automatically and non-invasively quantifying a suite of behavioural and cognitive welfare indicators. Behavioural indicators include the degree of physical closeness between individuals within a group, the spatial arrangement of subgroups, movement behaviour, and agonistic interactions. The project will examine variation in these indicators along an urban–rural gradient to a range of putative stressors, before validating them using “challenge” tests involving ecologically relevant experimental interventions (e.g., acoustic disturbances such as boat noise). The overarching objective is to develop an open-source toolkit that integrates commercially available hardware with custom-designed software, allowing the methods to be readily used by others to easily, cheaply, and repeatedly assess the welfare of wild fish.
Why we funded this project
This project will provide proof of concept for a novel approach to monitoring fish welfare at scale that is designed to be widely generalisable and translatable across species, habitats, and contexts. This project will improve current strategies for assessing the welfare of wild fish, most notably by avoiding the need for capture, handling, or restraint, and by being entirely non-invasive and non-disruptive.