Preliminary Findings of a Novel Thermal Drone-Based and AI Approach to Sampling Mesopredator Behaviour and Habitat Use

Habitat selection is often activity-specific, as animals may use different environments depending on whether they are foraging, breeding, or moving through the habitat. Behavioural studies of nocturnal species are challenging, and conventional methods are limited in their applicability. In this study, we tested a thermal drone in combination with Artificial Intelligence (AI) for focal animal sampling and habitat use of mesopredators. A drone mounted with a thermal video camera recorded the movements and behaviours of red foxes (Vulpes vulpes), European badgers (Meles meles), and Eurasian otters (Lutra lutra), while simultaneously geocoding their position. Additionally, we tested an AI-based analysis, LabGym for species and behaviour detection of video recordings. In Danish agricultural areas, both habitat separation and spatial overlap among the three mesopredators, were observed. Foxes showed a higher degree of versatility in both behaviour and habitat choice compared to badgers and otters. Otters were primarily found near water bodies, while badgers preferred foraging under tree cover and in meadows. The optimised LabGym achieved 80.4% mAP for species identification and successfully classified four behaviours with more than 80% accuracy. Using the thermal drone in combination with geolocation data and AI enables spatial mapping of mesopredator activities, adding valuable insights into predator ecology.