Talk Are animal-vehicle collisions a random event – Analysis of the spatial distribution of accident data

Ungulate-vehicle collisions (UVC) in Sweden, Catalonia (NE Spain), the Czech republic as well as in many other European countries are an increasing traffic safety issue, causing an escalating loss in wildlife and a growing socio-economic burden. Conventional prevention methods appear as little cost-effective and both Transport and Wildlife administrations call for better targeted mitigation strategies. An essential requisite for this is a good understanding of how accidents are distributed and where they aggregate. Since 2010, in Sweden car drivers are legally obliged to report any UVC to the police and a majority of these reports is followed-up by contracted hunters who take care of the wounded or dead animal and report the exact accident location. In Catalonia and Czech republic, UVC are registered by traffic police and road management teams as drivers and not obliged to report the data. We present a thorough analysis of these  reports, using a modified kernel density estimation technique (KDE+) to identify significant clusters in accident frequencies and compare their spatial coherence between ungulate species and between years.  During the 5-year period of 2010 to 2014, some 79000 accidents with roe deer (Capreolus capreolus), 19000 with moose (Alces alces), 11000 with wild boar (Sus scrofa) and 7000 with fallow deer (Dama dama) or red deer (Cervus elaphus) have been reported by hunters in Sweden. Of these, 30% to 45% were distributed in a significantly aggregated pattern. UVC clusters covered less than 3% of the entire road network. Within these clusters, accident densities were on average 45 times higher than elsewhere on the roads. In Catalonia during 2007 to 2011, the analyses were performed on 2110 UVC of which 1987 were with wild boar. About 20% of these accidents were included in the significant clusters, covering less than 0,3% of the entire road network. In the Czech Republic, 2009 – 2013, 16 612 UVC (not species specific) were recorded by the police during 2009 – 2013. Of these, 33,2% were significantly aggregated in 2060 clusters covering 0,71 % (267 km) of the Czech road network.

There were important differences in aggregation and in cluster locations between species and some differences between time periods as well. Factors associated with clustering can be generalize as providing increased attractiveness and increased openness of the road to wildlife.

We conclude that, despite species specific and temporal differences in clusters, only a small portion of the road network needs to be mitigated by physical mitigation measures such as fences or fauna passages to affect a substantial part of UVC. Yet, for the remaining part of UVC that are not aggregated, a different mitigation approach is needed that corresponds to regional or global factors such as wildlife population densities or driver behaviour.


The study is stillon-going and will produce further results before the conference that shall be included in an updated version of the abstract.

ungulate-vehicle accidents, traffic safety, kernel density estimation, hotspot, spatial pattern, scale-effects