Behaviour Sense: A Machine Learning Framework for Predicting Wildlife Behavioural Adaptations Under Ecological Stress

Abstract

Wildlife populations all around the world have become increasingly prone to ecological stressors such as habitat deterioration, limited resources, temperature extremities, and increased predation. Behavioural adaptation is one of the quickest ways in which animals respond to such conditions. However, unlike physiological and demographic adaptations, behavioural changes are more likely to precede the development of any other adaptive responses and be relatively easy to reverse. In this paper, Behaviour Sense is introduced– a predictive model based on behavioural and ecological data that is capable of predicting how free-living and captive wildlife populations will adapt their activity patterns, foraging behaviour, movement patterns, and social structure in case of exposure to ecological stress. Instead of using solely retrospective observation data, this model relies on longitudinal behavioural data, physiological indicators of stress and environmental factors in order to produce early indicators of stress-related behavioural changes at the level of a whole species. Based on the literature in behavioural ecology and conservation biology, conceptual validation shows that such predictive approaches can significantly reduce the time needed for identification of ecological stress by wildlife populations.