Predictive modelling of leopard predation using contextual Global Positioning System cluster analysis R. T. Pitman 1 , L. H. Swanepoel 2 & P. M. Ramsay 1 1 Marine Biology and Ecology Research Centre, Plymouth University, Plymouth, UK 2 Centre for Wildlife Management, University of Pretoria, Pretoria, South Africa Keywords GPS tracking; carnivore; felid; foraging; survey methods; kill site. Correspondence Ross T. Pitman, Marine Biology and Ecology Research Centre, Plymouth University, Plymouth PL4 8AA, UK. Tel: (+44) 7411 054667 Email: rosstyzackpitman@me.com Editor: Andrew Kitchener Received 23 November 2011; revised 14 May 2012; accepted 30 May 2012 doi:10.1111/j.1469-7998.2012.00945.x Abstract Apex predators are essential for the viability of healthy ecosystems. By studying carnivoran feeding ecology, we can obtain a better understanding of the ecological limits, resilience and predator–prey dynamics that govern these populations. However, monitoring elusive predators – like the leopard Panthera pardus – is often fraught with logistical and financial constraints, particularly in inaccessible terrain. In this study, we identified clusters of Global Positioning System (GPS) points from four GPS-collared leopards and investigated them in the field for potential kills. Environmental data from cluster sites were gathered alongside spatial and temporal data collected via GPS cluster analysis to develop statistical models capable of predicting the occurrence of leopard predatory events. Our results demonstrate that leopard predation can be accurately modelled either by using a combination of field data (i.e. collected at cluster sites) and remote data (i.e. obtained via GPS analysis), or simply remote data alone. Kills were more likely to be present at clusters where leopards exhibited longer handling times, at sites with dense vegetation cover, when leopards were more active 12 h before the cluster than 12 h after, where more tree refugia were present, in areas of higher elevation, at sites containing low levels of shrub cover, and when clusters began during diurnal or crepuscular hours. By using this modelling approach, comprehensive predation studies are not only efficient and cost-effective but also logistically feasible across a range of different habitats and may even be applied to other carnivorans. It provides a detailed analytical means of studying broader aspects of carnivoran feeding ecology, such as predation habits, carrying capacity, ecological hunting require- ments and species interactions, which are important aspects of carnivoran management. Introduction Carnivores were once thought to hold little ecological value, but are now known to be integral to healthy ecosystems. Their influence cascades through food webs and alters community structure through resource availability (Miller et al., 2001). Predatory activities control not only herbivore populations but others such as mesopredators, plants and invertebrates (Estes & Duggins, 1995; Henke & Bryant, 1999). Leopards Panthera pardus are apex predators with the widest distribu- tion of Africa’s large felids, but concern for their conservation is increasing as a result of habitat loss (Ray, Hunter & Zigouris, 2005), over-exploitation of prey by local people (Henschel et al., 2011), trophy hunting (Packer et al., 2011) and human-carnivore conflict (Balme, Slotow & Hunter, 2010). Studies of leopard feeding ecology – such as predation rates, feeding habitat selection and dietary preference – are a key to effective conservation efforts and management strate- gies through a better understanding of the ecological limits, resilience and predator–prey dynamics that govern leopard populations (Balme, Hunter & Slotow, 2007). However, there is a need to design efficient monitoring techniques, particu- larly in mountainous regions where data are lacking (Martins et al., 2011). Previous research on leopard feeding ecology has concen- trated on the analysis of faecal samples, direct observations using very-high-frequency (VHF) tracking collars and indi- rect observations using spoor (Bothma & Le Riche, 1984; Stander et al., 1997; Bailey, 2005). Although effective in other situations, these techniques are impractical in rugged, mountainous terrain, where locating faecal samples is challenging and VHF telemetry signals are inaccurate (Grimbeek, 1992; Martins et al., 2011). The hard substrate in these landscapes also makes it difficult to track leopards visually on foot. Journal of Zoology Journal of Zoology. Print ISSN 0952-8369 Journal of Zoology •• (2012) ••–•• © 2012 The Authors. Journal of Zoology © 2012 The Zoological Society of London 1