Wildlife-Vehicle Collisions and Wildlife Fences on Swedish Roads : A GIS-Based Regression Analysis

Sammanfattning

Wildlife-vehicle collisions (WVC) are a major problem in Sweden, with consequences for traffic safety, wildlife, and the economy. The aim of this study is to analyse the spatial distribution of WVC on Swedish road segments and to examine how these patterns relate to wildlife fences (WF) and selected road- and mitigation-related factors. The study uses a quantitative GIS-based approach based on secondary data. The analysis consists of a descriptive and an exploratory approach. First, WVC-related factors are compared descriptively between fence classes and counties. Second, a Poisson regression is used to analyse WVC counts on 1 km road segments. The descriptive results showed that road segments with Partial WF and Full WF had higher mean WVC per segment than segments with No WF. However, fenced segments also had higher mean traffic volumes, higher mean speed limits, and more complementary mitigation measures (CMM). When the included variables were analysed simultaneously in the Poisson regression, both Partial WF and Full WF were negatively associated with WVC compared with No WF. Full WF showed the strongest negative association. The model explained 33.72 % of the deviance in WVC counts, indicating that the included road- and mitigation-related variables accounted for a limited part of the observed variation. The residual analysis showed that 22.51 % of the analysed segments had positive residuals, and that high positive residual segments were mainly located in southern and central Sweden. These underpredicted segments may be relevant locations for further investigation before mitigation measures are planned. The results suggest that spatial planning should consider road characteristics and existing mitigation measures together, and that further analysis of landscape-related factors, animal species, and combinations of WF and CMM could develop the understanding of WVC.