Understanding Multi‐Disaster Impacts on China's Grain Production for Food Security: An <scp>XGBoost</scp> ‐ <scp>SHAP</scp> Approach

ABSTRACT Agricultural natural disasters pose a growing threat to global food security by compromising the resilience and sustainability of farming systems. While previous studies have largely focused on assessing the specific impacts of individual disaster types on grain production, they have often overlooked the differential effects across various grain types and the underlying mechanisms driving these impacts. As a major global food producer, China critically needs to decipher the complex spatio‐temporal dynamics between these disasters and grain production output to safeguard food supplies. This study integrates spatio‐temporal pattern analysis with the XGBoost‐SHAP model to examine the heterogeneous and nonlinear impacts of five key disasters—drought, flood, hail, frost, and typhoon—on beans, cereals, and tubers in China from 2000 to 2022. Findings indicate a decline in the overall incidences of drought and flood, yet with growing spatial uncertainty. Grain production exhibited regionnal clustering: beans and cereals in the resource‐abundant northeast, and tubers in the stress‐tolerant southwest and northwest, signaling a “north‐to‐south” production shift. XGBoost‐SHAP revealed grain‐specific nonlinear impacts—beans were most sensitive to floods, cereals to droughts. All disasters reduced bean and cereal production, with losses accelerating beyond 5%–8% incidence thresholds. Notably, tuber production rose after floods, droughts, and frosts, likely due to substitution effects from grain competition. This work contributes novel insights into the nuanced impacts of various agricultural natural disasters on China's grain production, thereby laying a crucial foundation for developing data‐driven strategies to enhance agricultural resilience and ensure global food security.