Robust mapping of human–wildlife conflict: controlling for livestock distribution in carnivore depredation models

Abstract Shifting human–wildlife conflict towards coexistence requires a robust understanding of where conflict happens and why. Spatial models of livestock depredation by wild predators commonly identify depredation hotspots in areas where livestock are most abundant (e.g. nearer villages or pasture). This may reflect underlying livestock distribution, rather than imply these areas are inherently risky for livestock. This limits the predictive power of these models and their usefulness for conflict mitigation and wild carnivore conservation. Here, we build spatial models of both cattle depredation (530 attacks mostly by lions and hyenas; 2009‐2013), and cattle presence (14 GPS‐collared herds; 2010‐2012) near Hwange National Park, Zimbabwe. We use Bayes’ theorem to combine the cattle depredation and presence models to quantify risk as the conditional probability of depredation given livestock presence. Our raw depredation models predicted higher depredation rates where cattle presence was more likely (near villages and in more open habitats). By contrast, our risk model predicted higher risk further from human activity and in more dense vegetation (where depredation rates were higher than expected given the low probability of cattle presence). Risk has also increased sharply towards protected areas (core carnivore habitat). Our formulation of risk captures high‐risk areas as those where livestock are most accessible (i.e. vulnerable) to predators as opposed to simply where they are most available (as in much previous work). We make recommendations for livestock protection and wild carnivore conservation based on our quantification of risk, such as where to avoid herding livestock and which areas to prioritize for livestock protection. Our approach may be profitably applied to guide safer livestock grazing or herding in other contexts where depredation and livestock movement data are available. We hope that the concepts and methods that we develop here will help advance the future study and mitigation of human–wildlife conflict more generally.