Moose-vehicle collisions in Northern Norway: Causes, hotspot detection and mitigation

Abstract

Animal-vehicle collisions are a growing concern worldwide, from the perspectives of human health and animal welfare, and due to high socioeconomic costs. This has led to an intensive search for effective mitigation measures. However, underlying mechanisms increasing the collision risk is often unknown and hazardous road stretches can be difficult to detect. For the two counties Nordland and Troms in Northern Norway I have analysed the effect of an optic/acoustic mitigation measure. In order to conclude on the effect, I first looked at how temporal variation (i.e. population size, weather and traffic) correlate with the number of moose (Alces alces )-vehicle collisions (MVC), and whether the test sections for the mitigation measure were placed at objectively classified hotspot sections for MVCs. A total of 3,105 MVCs were recorded in the study area during the seven years long time series from 1 st of April 2009 to 31 st of March 2016. A large proportion of the accidents occurred during winter and the number ofMVCs were positively correlated to snow depth and population size. The predicted number ofMVCs for public roads in the study area was 0.46 MVCs/10 km/year. I used the novel kemel density estimation method KDE+ to objectively detect hotspots in the area. According to the KDE+ analysis, my MVC-data formed 77 significant clusters (hotspots) with three or more MVCs in each cluster. These hotspots contained 9.8 % of all the recorded MVCs. The hotspots were ranked by significance after their cluster strength. The optic/acoustic mitigation measure were put up on four road sections of various length in the study area in 2014. Two of four sections were classified as hotspots, although all sections had a higher number of MVCs than the prediction for the area. The mitigation system is supposed to scare away the moose using high frequency sound and blinking lights when cars are present, but I found no significant reduction in the number of MVCs after installation of the instruments. I conclude that with further improvement of the hotspot-detection method and by looking at the underlying mechanisms of variation in the number of MVCs such as snow depth and population size, the method can be used as a tool to select which road sections to mitigate. This can lead to a more cost-effective prevention of MVCs in the future. The optic/acoustic mitigation system did not show any significant reducing effect on the number of MVCs on the test sections.