A biologically realistic model to predict wildlife-vehicle collision risks
Road networks have major ecological impacts on living organisms consequent to habitat loss and fragmentation, chemical and acoustic pollution, and direct mortality when wildlife-vehicle collisions are involved (WVC). The many past empirical studies revealed biological traits shared by species most vulnerable to roadkills (e.g. population density). Similarly, spatial locations of WVC hot-spots are associated to landscape features at large spatial scales, and to road characteristics at small spatial scale. We currently lack a comprehensive theoretical framework for WVC. Animal movement in relation to habitats is an essential driver of encounters with roads, but this remains largely ignored in studies, even when movement ecology provides the necessary tools to investigate the impact of animal movement on WVC. We built a general individual-based model incorporating recent knowledge in movement ecology (movement typology: roaming, migratory route crossing a road, active attraction and active repulsion of roads) to estimate WVC risks. We explored the relative effects of animal and vehicle movement parameters (speed, abundance, road sinuosity and animal movement pattern) on collision probability. We show that animal behaviour toward roads has major impacts on the number and risks of WVC, but also modulate the effects of other factors (animal traveling speed, species local abundance, road traffic volume) on WVC. Sensitivity analyses show that the movement and behaviour of the animal has more influence on WVC risks than any of the characteristics of roads and vehicles we tested. Our results suggest that (1) effective roadkill mitigation should be species-specific and could vary in efficiency depending on the targets movement pattern (mating and migratory seasons, foraging habits...) and (2) empirical studies of WVC should incorporate knowledge about the behavioural habits of the focal species in relation to roads.