Estimating encounter‐habitat relationships with scale‐integrated resource selection functions

Abstract

Encounters between animals occur when animals are close in space and time. Encounters are important in many ecological processes including sociality, predation and disease transmission. Despite this, there is little theory regarding the spatial distribution of encounters and no formal framework to relate environmental characteristics to encounters. The probability of encounter could be estimated with resource selection functions (RSFs) by comparing locations where encounters occurred to available locations where they may have occurred, but this estimate is complicated by the hierarchical nature of habitat selection. We developed a method to relate resources to the relative probability of encounter based on a scale‐integrated habitat selection framework. This framework integrates habitat selection at multiple scales to obtain an appropriate estimate of availability for encounters. Using this approach, we related encounter probabilities to landscape resources. The RSFs describe habitat associations at four scales, home ranges within the study area, areas of overlap within home ranges, locations within areas of overlap, and encounters compared to other locations, which can be combined into a single scale‐integrated RSF. We apply this method to intraspecific encounter data from two species: white‐tailed deer (Odocoileus virginianus) and elk (Cervus elaphus) and interspecific encounter data from a two‐species system of caribou (Rangifer tarandus) and coyote (Canis latrans). Our method produced scale‐integrated RSFs that represented the relative probability of encounter. The predicted spatial distribution of encounters obtained based on this scale‐integrated approach produced distributions that more accurately predicted novel encounters than a naïve approach or any individual scale alone. Our results highlight the importance of accounting for the conditional nature of habitat selection in estimating the habitat associations of animal encounters as opposed to ‘naïve’ comparisons of encounter locations with general availability. This method has direct relevance for testing hypotheses about the relationship between habitat and social or predator–prey behaviour and generating spatial predictions of encounters. Such spatial predictions may be vital for understanding the distribution of encounters driving disease transmission, predation rates and other population and community‐level processes.