Abstract: To support the scientific and precise prevention and mitigation of flash floods, this study integrated remote sensing data and explainable artificial intelligence (XAI) techniques to conduct systematic research on flash flood occurrence probability prediction and disaster-causing factor analysis. Firstly, 18 influencing factors covering five dimensions—geographical location, topography, meteorology, human activities, and land cover—were extracted from remote sensing inversion datasets. Secondly, combined with historical flash flood survey data, the XGBoost machine learning algorithm was adopted to construct a flash flood occurrence probability prediction model. Finally, the SHAP (SHapley Additive exPlanations) interpretation method was introduced at both provincial and sub-regional scales to quantify the single-factor importance, impact effects, and multi-factor interaction effects of disaster-causing factors, and to reveal the regional heterogeneity in the influence and interaction of disaster-causing factors. Independent verification results showed that the model achieved a prediction accuracy of 0.90 and a precision of 0.87, demonstrating that it has the ability to effectively assess flash flood occurrence probability. At the provincial scale, the spatial distribution of predicted flash flood occurrence probability was highly consistent with the historical flash flood location pattern, which further confirmed that the model has strong extrapolation robustness. The disaster-causing factor importance ranking indicated that the impervious surface ratio is the most critical factor affecting flash flood occurrence probability, while the high importance of latitude and longitude coordinates highlights the necessity of considering geospatial relationships in flash flood prediction. At the sub-regional scale, multi-factor interaction analysis successfully identified the differentiated influence patterns of disaster-causing factors in different regions. This study verified the effectiveness of XAI in flash flood prediction and disaster-causing factor analysis, and provides a scientific reference for understanding the flash flood disaster-causing mechanism and accurately quantifying flash flood triggering conditions.