Biological Conservation 126 (2005) 474–490 www.elsevier.com/locate/biocon 0006-3207/$ - see front matter  2005 Elsevier Ltd. All rights reserved. doi:10.1016/j.biocon.2005.07.001 Modelling of wildlife fatality hotspots along the Snowy Mountain Highway in New South Wales, Australia Daniel Ramp a,¤ , Joanne Caldwell b , Kathryn A. Edwards a , David Warton c , David B. Croft a a School of Biological, Earth and Environmental Sciences, University of New South Wales, Sydney, NSW 2052, Australia b New South Wales Department of Environment and Conservation, Tumut, 2720, New South Wales, Australia c Department of Statistics, School of Mathematics, University of New South Wales, Sydney 2052, Australia Received 9 February 2005 Available online 19 August 2005 Abstract The eVects of roads on the natural environment is of growing concern world-wide and foremost amongst these eVects are the fatalities of wildlife killed in collisions with vehicles. Aside from animal welfare and human safety considerations, fatalities may have signiWcant impacts on the population dynamics of species living adjacent to roads and thus can adversely aVect the viability of local populations. As such, the need to quantify and mitigate road-based fatalities is paramount. With a vast expanse of roads it is imper- ative to identify where animals are most likely to be killed (i.e. hotspots) and what are the contributing factors. In order to identify hotspots, we develop a modelling approach for both presence and presence/absence data. We use data collected from the Snowy Mountain Highway in southern New South Wales, Australia, to compare the eVectiveness of this approach for Wve species/groups of species. We observed that models of species killed in a clumped fashion were eVective at identifying hotspots, while for species where fatalities were distributed evenly along the road the models were less eVective. We recommend that where actual presence data exists spatial clustering is the preferred method of hotspot identiWcation. Predictive models of presence/absence date should be constructed if the intention is to extrapolate to additional areas. The added beneWt of predictive models are that they enable the identiWcation of explanatory factors and this knowledge enables species-speciWc management strategies to be developed and implemented at hotspot locations.  2005 Elsevier Ltd. All rights reserved. Keywords: Road-based fatalities; Hotspots; Predictive models; Kernel density estimates; Wildlife populations 1. Introduction The eVect of roads and traYc on habitat and wildlife is far-reaching. Community perception of this issue is not new, but consideration of roads as driving forces in ecology has only recently gained international awareness (Sherwood et al., 2002; Forman et al., 2003). Recent studies show that roads have a multitude of eVects on the natural environment such as impacts on microcli- mate (Ellenberg et al., 1981; cited in Forman et al., 2003), wind Xow (Ahrens, 1991), run-oV and water Xow (Fed- eral Highway Administration, 1996), addition of noise pollution (Reijnen et al., 1997) and facilitation of the dis- persal of both plants and animals (Tikka et al., 2001), including weeds (Ullmann, 1998), feral animals (Sea- brook and Dettmann, 1996; Stiles and Jones, 1998) and native species (van der Ree, 2002; Spooner et al., 2004). For wildlife, roads have numerous impacts on popula- tions (Forman and Alexander, 1998). Roads can form barriers to movement, fragmenting populations and * Corresponding author. Tel.: +61 29 385 2111; fax: +61 29 385 1558. E-mail address: d.ramp@unsw.edu.au (D. Ramp).