Wildlife-vehicle collision hotspots at US highway extents: scale and data source effects

Abstract

Highways provide commuter traffic and goods movement among regions and cities through wild, protected areas. Wildlife-vehicle collisions (WVC) can occur frequently when wildlife are present, impacting drivers and animals. Because collisions are often avoidable with constructed mitigation and reduced speeds, transportation agencies often want to know where they can act most effectively and what kinds of mitigation are cost-effective. For this study, WVC occurrences were obtained from two sources: 1) highway agencies that monitor carcass retrieval and disposal by agency maintenance staff and 2) opportunistic observations of carcasses by participants in two statewide systems, the California Roadkill Observation System (CROS; http://wildlifecrossing.net/california) and the Maine Audubon Wildlife Road Watch (MAWRW; http://wildlifecrossing.net/maine). Between September, 2009 and December 31, 2014, >33,700 independent observations of >450 vertebrate species had been recorded in these online, form-based informatics systems by >1,300 observers. We asked whether or not WVC observations collected by these extensive, volunteer-science networks could be used to inform transportation-mitigation planning. Cluster analyses of volunteer-observed WVC were performed using spatial autocorrelation tests for parts or all of 34 state highways and interstates. Statistically-significant WVC hotspots were modeled using the GetisOrd Gi* statistic. High density locations of WVC, that were not necessarily hotspots, were also visualized. Statistically-significant hotspots were identified along ~7,900 km of highways. These hotspots are shown to vary in position from year to year. For highways with frequent deer-vehicle collisions, annual costs from

Comparing Road-Kill Datasets from Hunters and Citizen Scientists in a Landscape Context

“…However, besides the dominating effect of arable land, significant differences in the percentage of different land cover classes between the two groups show that the two groups collect data in slightly different environments. Our analyses showed, that citizen scientists tended to report data of road-kills near developed areas, confirming others [9] and correlating with human population density as a key factor for spatial variation in citizen science data [39]. We conclude that this result should be considered when establishing a citizen science project with geo-referenced data or when analysing and interpreting geo-referenced citizen science data.…”

Influence of Landscape Factors on Amphibian Roadkills at the National Level

“…Malo's method has been suggested to be the best one for hotspot identification [33], in comparison with other statistical methods such as binary logistic regression (BLR), ecological niche factor analysis (ENFA), kernel density estimation and nearest-neighbor hierarchical clustering (NNHC). K-function [32,34,35] and Getis-Ord Gi* [36] have also been used as spatial clustering methods for roadkills.…”

“…Only three road sections in Romania were considered hotspots for tortoises in almost 4000 km of roads [68]. In California and Maine, using Getis-Ord Gi*, 10% of roads were identified as hotspots for mammals [36].…”

“…As resources are limited, mitigation measures must be applied at least in places with higher roadkill rates, the so-called hotspots [11,21]. Hotspots can be identified with spatial statistics and modelling procedures, such as probabilities based on Poisson distribution [32], binary logistic regression, ecological niche factor analysis, kernel density estimation, nearest-neighbour hierarchical clustering [33], Ripley's K function [8,34,35] or Getis-Ord Gi* [36]. Hotspots depend on: (1) points of animal crossings; (2) barriers for crossing roads safely, where the animal is blocked in the road; and (3) high roadkill rates [11].…”

Greener transport infrastructure – IENE 2014 International Conference

“…A growing number of countries is establishing reporting systems for citizen observations of road killed animals. Examples are presented from Belgium by Vercayie and Herremans (2015), and from California and Maine, USA, by Shilling and Waetjen (2015). Citizen science can be supported by new technologies, and provide road kill data with better extent in time and space than regular (often short-termed) scientific study, and with different extent and taxonomic accuracy (especially in smaller animals) than data collected by road maintenance or other officials.…”