Road characteristics best predict the probability of vehicle collisions with a non-native ungulate

With roads encroaching into natural environments, there is an increased likelihood of wildlife coming into contact with vehicles, resulting in an increase in wildlife–vehicle collisions. Our goal was to investigate environmental correlates of moose–vehicle collisions (MVCs) on the island of Newfoundland, Canada. We developed predictive models to compare environmental variables at known MVC locations with environmental variables at random sites along the Newfoundland road network. The most supported generalized linear model explained ~36% of the variance in the probability of MVC occurrence. This top model predicted an increase in the probability of MVC occurrence: with decreasing distance to developed areas; on primary rather than secondary roads; on straight rather than curved roads; and in locations where roadside vegetation cutting has occurred. Our study highlights MVC predictors that are consistent with other wildlife–vehicle collision studies around the globe and which will serve as the basis for mitigation strategies on the island of Newfoundland with potential applications to other regions with high moose densities.