Geo-XAI Reveals Wildfire Risk Driver Differences and Spatial Heterogeneity Between Drought and Non-Drought Periods in Southwest China Mountains

Global warming has increased the frequency of droughts, further complicating wildfire susceptibility assessment in topographically complex mountainous regions. However, the factors associated with wildfire occurrence and their nonlinear spatial responses under contrasting drought conditions remain poorly understood. To address this knowledge gap, we conducted a case study in a fire-prone mountainous region of southwestern China. Drought and non-drought periods were identified using the Standardized Precipitation Evapotranspiration Index (SPEI) and run theory. Using historical wildfire records from 2006 to 2020 and 16 wildfire drivers, we developed three machine-learning models and applied GeoShapley to quantify the contributions of key predictors and characterize their spatial dynamics under different drought conditions. The results showed that the Extreme Gradient Boosting (XGB) model achieved the best predictive performance (AUC = 0.85–0.91) and effectively captured the spatial patterns of wildfire susceptibility. Meteorological variables consistently emerged as the dominant controls on wildfire occurrence, although their relative importance and functional effects differed substantially between drought and non-drought conditions. GeoShapley analysis further revealed pronounced spatial heterogeneity in the effects of the major drivers, with these spatial patterns further modulated by drought conditions. In particular, during drought periods, the local geographical context amplified the spatial interaction effect of precipitation, resulting in a stronger risk-enhancing effect than during non-drought periods. By decomposing variable contributions into non-spatial main effects and spatially explicit interaction effects, this study reveals how the effects of wildfire drivers vary spatially under contrasting drought conditions and provides a robust framework for targeted wildfire mitigation in complex mountainous landscapes.