Bayesian belief network model to predict human-wildlife conflict in protected areas

Abstract

Human-wildlife conflict (HWC) poses a pervasive global challenge, affecting livelihoods and threatening biodiversity. To better anticipate and mitigate HWC risk, we developed a large-scale predictive model using a Bayesian Belief Network (BBN). We surveyed 1,011 park rangers across 135 terrestrial protected areas in three Andean countries, documenting recent HWC incidents involving wildlife persecution or killing, livestock depredation, crop damage, or threats to human safety and property. We identified key drivers of HWC risk, including governance, wildlife acceptance, participation, and habitat quality. A sensitivity analysis revealed that enhancing governance and improving wildlife acceptance could reduce HWC risk by > 85%. The BBN model demonstrated scalability, effectively identifying strategies to reduce HWC risk at multiple scales, from individual protected areas to national networks. Our findings highlight the importance of strengthening governance, increasing wildlife acceptance, and enhancing community participation in conservation efforts. BBNs provide a flexible, cost-effective, and data-driven tool to guide protected areas and wildlife managers in monitoring, anticipating, and making informed decisions to mitigate conflict and promote coexistence.