An integrated machine learning framework to understand zoonotic spillover emergence across anthropogenically modified landscapes
(2025) Environmental Health Perspectives — p. 1-41 (2025)
Files
zhang-et-al-an-integrated-machine-learning-framework-to-understand-zoonotic-spillover-emergence-across2.pdf
Open Access - Adobe PDF
- 2.22 MB
Details
- Authors
Li, Sen 
- Abstract
- BACKGROUND: Anthropogenic land modification influences human-livestock-wildlife interactions and zoonotic spillover emergence. However, the extent of this impact remains unclear and could be better understood through the collaborative use of advanced predictive and explanatory analytical tools, alongside, an up-to-date dataset on zoonotic spillover events. OBJECTIVE: The main objective is to develop and evaluate an integrated modeling framework to predict and explain spatial patterns and nonlinear relationships of zoonotic spillover events, using updated datasets and the human modification index to differentiate anthropogenic pressures. METHODS: Our study expanded the historical datasets to include recent spillover events and a comprehensive set of predictors. By combining robustness of finely-tuned stacking algorithms with structural equation modeling, we considered global heterogeneity in relative reporting adequacy and mapped spillover patterns at different scales. Using the human modification index, we disentangled anthropogenic processes modifying natural ecosystem across land modification gradients, and described their linkages to spillover occurrence over the past three decades. RESULTS: This integrated approach effectively improved the model’s predictive and explanatory power. Our analysis reveals that the intermediate levels of human pressure facilitated the zoonotic spillover. The indirect effects of anthropogenic pressure, mediated by specific cropping intensity, are strongly associated with zoonoses emergence. Livestock distribution serves as an indicator of spillover hotspots, acting as effective proxies for distinctive landscapes. DISCUSSION: Our findings identify high zoonotic spillover risks present across geographically and socioeconomically diverse regions worldwide, extending beyond tropical areas, including extensive regions experiencing high-intensity human modification. These insights support targeted surveillance in areas with potentially high relative risk or uncertainty, and demonstrate how zoonotic spillover responds to complex human-environment interactions.
- Affiliations
Citations
Zhang, Y., Wang, J., Wang, L., Miao, L., Sun, Y., Yang, X., Fang, R., Guo, Y., Vanwambeke, S., & Li, S. (2025). An integrated machine learning framework to understand zoonotic spillover emergence across anthropogenically modified landscapes. Environmental Health Perspectives, 1-41. https://doi.org/10.1289/ehp15937 (Original work published 2025)