Addressing human–elephant conflicts in Taita Taveta County, Kenya: Integrating species distribution modeling into targeted conservation strategies

Increasing competition for space and resources at the agriculture-conservation interface poses critical challenges to wildlife conservation, often intensifying human–wildlife conflicts throughout the globe, including Kenya. With approximately 70 % of Kenya’s wildlife residing outside protected areas, land conversion for agriculture exacerbates human–wildlife conflicts, particularly involving African elephants (Loxodonta africana). Taita Taveta County in Kenya represents a hotspot for human–elephant conflict, where these incidents undermine both conservation efforts and livelihoods. This study assesses multiple distribution model algorithms and ensemble models, using Kenya Wildlife Service incident data and ten geospatial variables, to predict human–elephant conflicts in the county. The study extends the spatial pattern analysis to the comprehensive comparison of outputs, such as probability and risk maps, thus filling a critical gap by offering an innovative framework for human–elephant conflict modeling. Probability maps were reclassified into risk maps, and landscape metrics were derived to evaluate the spatial patterns of conflict risk. Results highlight that the ensemble model demonstrated superior consistency, predictive accuracy, and provided a more balanced representation of human–elephant conflict risk compared to single-algorithm models. The analysis identified proximity to houses and crops as key conflict predictors, with high-risk zones concentrated near human settlements and low-risk zones confined to protected areas. This study proposes that landscape metrics can further enhance the evaluation of risk map performance. By integrating ensemble modelling and landscape metrics, this research provides policymakers with actionable tools to balance human needs with conservation priorities, fostering sustainable human–elephant coexistence in Taita Taveta County and beyond.