Assessing plague risk across seven decades of environmental change in Inner Mongolia, China: An interpretable machine learning.

Plague, caused by Yersinia pestis, persists as a public health threat, particularly in natural foci such as China's Inner Mongolia Autonomous Region. Understanding the distinct environmental drivers of pathogen transmission from wildlife reservoirs to humans is critical for targeted control. We integrated long-term human case data (1950-2023) with extensive rodent serological surveillance (2005-2023) to model these dynamics. Using a Generalized Additive Model and interpretable machine learning, we found that human cases were concentrated in specific foci and nonlinearly linked to proximity to railways and precipitation, but not rivers. In contrast, enzootic maintenance was predominantly driven by a negative correlation with nighttime light, nonlinear climatic effects, and positive associations with specific land uses (water, forest, residential). Our framework reveals that spillover and enzootic cycles are governed by differing environmental factors. This approach enables spatially refined risk assessment, guiding surveillance to incorporate the distinct anthropogenic and climatic drivers of plague transmission.