UC Santa Barbara

Rapid land-use change is compressing the space between people and wildlife at an unprecedented pace, elevating concerns of conflict between people and keystone megafauna such as the African savanna elephant (Loxodonta africana). However, the underlying drivers of human-elephant conflict (HEC) are often inferred but not causally identified, and forecasting under future scenarios is limited, restricting our ability to anticipate where and why conflict will intensify. Using a dataset on conflict events from 2004 to 2020 across a transboundary landscape in Southern Africa, we integrate causal inference (panel data regressions) and machine learning (point process models) to link mechanisms with spatial forecasts. We find that human population growth, cropland expansion, and climate-driven aridity in core elephant-protected areas drive conflict risk. Furthermore, we show that landscape features such as wildlife fences and roads constrain elephant movement in ways that drive HEC. Applying these models to projected land use, population, and climate under future coupled climate-development scenarios, we find that the area at high risk of HEC increases by 33 to 100% by 2085 with corresponding rises in event frequency. Aggressive human land-use expansion leads to the most dramatic increases in conflict, with climate impacts playing a significant but smaller role. These results identify emerging conflict hotspots decades in advance, offering actionable foresight for land-use planning and mitigation in a region critical to elephant conservation.