Proximity to water shapes the distribution of natural elephant mortality in Hwange National Park, Zimbabwe.

While elephant poaching has received considerable attention, natural mortality can at times surpass human-induced deaths, especially under environmental stress. Understanding the ecological drivers of natural elephant mortality is therefore crucial for informing reintroduction efforts and preventing mass die-offs. In this study, we investigated environmental predictors of natural elephant mortality in Hwange National Park, Zimbabwe, using mortality records from 2020 to 2022. We applied four machine learning species distribution models, Random Forest, Gradient Boosting, Maximum Entropy, and Extreme Gradient Boosting, along with their ensemble to model mortality hotspots. The ensemble model outperformed individual models, achieving a True Skill Statistic of 0.54 and a Receiver Operating Characteristic of 0.83. Among all predictors, distance to water sources was the most influential variable (accounting for > 55% of model importance), with most mortalities occurring within 6 km of water points. Other key predictors included climate water deficit, normalized difference vegetation index (NDVI), tree cover percentage, and elephant density (each contributing > 5%). In contrast, maximum temperature of the warmest month and elevation had minimal predictive power (< 4%). Our results provide actionable insights for conservation planning. Areas close to water sources, particularly during dry periods, should be prioritized for monitoring and veterinary intervention. Meanwhile, regions with historically low mortality prevalences may serve as safer sites for reintroduction. This spatially explicit framework can help reduce post-release losses and enhance the long-term success of elephant conservation initiatives, especially in the face of ongoing environmental change.