Lessons from integrating behaviour and resource selection: activity‐specific responses ofAfrican wild dogs to roads

Abstract Understanding how anthropogenic features affect species' abilities to move within landscapes is essential to conservation planning and requires accurate assessment of resource selection for movement by focal species. Yet, the extent to which an individual's behavioural state (e.g. foraging, resting, commuting) influences resource selection has largely been ignored. Recent advances in Global Positioning System ( GPS ) tracking technology can fill this gap by associating distinct behavioural states with location data. We investigated the role of behaviour in determining the responses of an endangered species of carnivore, the A frican wild dog L ycaon pictus , to one of the most widespread forms of landscape alteration globally: road systems. We collected high‐resolution GPS and activity data from 13 wild dogs in northern B otswana over a 2‐year period. We employed a step selection framework to measure resource selection across three behavioural states identified from activity data (high‐speed running, resting and travelling) and across a gradient of habitats and seasons, and compared these outputs to a full model that did not parse for behaviour. The response of wild dogs to roads varied markedly with both the behavioural and the landscape contexts in which roads were encountered. Specifically, wild dogs selected roads when travelling, ignored roads when high‐speed running and avoided roads when resting. This distinction was not evident when all movement data were considered together in the full model. When travelling, selection for roads increased in denser vegetative environments, suggesting that roads may enhance movement for this species. Our findings indicate that including behavioural information in resource selection models is critical to understanding wildlife responses to landscape features and suggest that successful application of resource selection analyses to conservation planning requires explicit examination of the behavioural contexts in which movement occurs. Thus, behaviour‐specific step selection functions offer a powerful tool for identifying resource selection patterns for animal behaviours of conservation significance.