Crossings and collisions – Exploring how roe deer navigate the road network

Abstract

Context To investigate the major impact of roads on wildlife, most studies focus on hot-spots of wildlife-vehicle collisions (WVC) to identify areas in need of mitigation measures. However, on road stretches where the frequency of WVC is low, a question arises: is this because those locations are 'safe’ places for wildlife to cross the road with little risk of collisions; or is it because individuals avoid approaching and crossing the road in these locations? Objectives In this study, we addressed this gap by evaluating how roe deer crossings are related to WVC risk across the road network. Methods We used 56 076 WVC locations between 2013 and 2017 to predict the spatiotemporal risk zones in response to environmental, road-related and seasonal predictors using Species-Distribution Modelling (SDM). We compared the predictive WVC risk to the location of 20 744 road crossing by 46 GPS-collared roe deer individuals. Results We found that the risk of WVC with roe deer tends to be higher on federal roads that are present in a density of approximate 2.2 km/km2 and surrounded by broad-leafed forests and demonstrate that SDMs can be a powerful tool to predict the risk of WVC across the road network. Roe deer crossed roads more frequently in high WVC risk areas. Temporally, the number of WVC changed throughout the year, which can be linked to roe deer movement patterns rather than landscape features. Within this study, we did not identify any road segments that were a complete barrier to roe deer movement. Conclusions The absence of complete barriers to roe deer movement detected in the present study, might be due to the low spatial variation of the landscape, coupled with the high individual variation in movement behaviour. By applying our approach at greater spatial scales and in other landscape contexts, future studies can continue to explore the potential barrier impacts of roads on landscape connectivity. Exploring the relationship between crossing activity and collision risk can improve one’s ability to correctly identify road stretches that require mitigation measures to improve connectivity versus reduce collisions.