Identifying optimal road crossing locations for mammal species
In this paper we used a multi-scale approach to independently identify preferred road approach (up to 1km) and highway crossing areas, for a multi-species mammal community. This combined strategy could help improve road mitigation efforts, by prioritizing traditionally used areas of high potential for crossing and combine these with areas of preference for approaching the road. We further contrasted multiple modelling scales and found that an approach that explicitly incorporated several scales is in most cases superior to single scale approaches. Lastly, we compared predictive models using labour and cost intensive predictor variables derived from hand digitized air photos to freely available remote sensed predictor variables and found that using remote sensed variables outperformed costly ones in direct comparison.
Roads are a major cause of habitat fragmentation that can negatively affect many mammal populations. Mitigation measures such as crossing structures are a proposed method to reduce the negative effects of roads on wildlife, but the best methods for determining where such structures should be implemented, and how their effects might differ between species in mammal communities is largely unknown. We investigated the effects of a major highway through south-eastern British Columbia, Canada on several mammal species to determine how the highway may act as a barrier to animal movement, and how species may differ in their crossing-area preferences. We collected track data of eight mammal species across two winters, along both the highway and pre-marked transects, and used a multi-scale modeling approach to determine the scale at which habitat characteristics best predicted preferred crossing sites for each species. We found evidence for a severe barrier effect on all investigated species. Freely-available remotely-sensed habitat landscape data were better than more costly, manually-digitized microhabitat maps in supporting models that identified preferred crossing sites; however, models using both types of data were better yet. Further, in 6 of 8 cases models which incorporated multiple spatial scales were better at predicting preferred crossing sites than models utilizing any single scale. While each species differed in terms of the landscape variables associated with preferred/avoided crossing sites, we used a multi-model inference approach to identify locations along the highway where crossing structures may benefit all of the species considered. By specifically incorporating both highway and off-highway data and predictions we were able to show that landscape context plays an important role for maximizing mitigation measurement efficiency. Our results further highlight the need for mitigation measures along major highways to improve connectivity between mammal populations, and illustrate how multi-scale data can be used to identify preferred crossing sites for different species within a mammal community.