摘要
This study investigates the dynamic relationship between environmental gamma dose rate (GDR) and precipitation using minute-level observational data, which are critical for improving environmental radiation monitoring and early warning systems. Prior research has largely relied on hourly-averaged data, which tend to obscure rapid fluctuations during convective rainfall. To overcome this limitation, this study integrates high-temporal-resolution GDR measurements with spatiotemporal collocated precipitation records collected in Chiayi, Taiwan. An event-based dynamic response framework was applied to quantify the behavior of hundreds of rainfall events. To capture instantaneous responses, the time derivative of dose rate was analyzed, and advanced signal processing methods were combined with explainable artificial intelligence techniques. A LightGBM model was trained, and SHAP (Shapley Additive exPlanations) analysis was used to interpret feature importance across multiple scales. Results show that the instantaneous rate of change in GDR is tightly synchronized with rainfall intensity, while the total increase in GDR is more strongly correlated with cumulative rainfall than with peak intensity. The analysis further demonstrates a scale-dependent shift in dominant drivers: at the minute scale, rainfall and wind gusts are most influential, whereas at the six-hour scale, deep soil temperature and atmospheric pressure prevail. These findings highlight the value of minute-level data and advanced analytical methods for elucidating rainfall-radiation interactions and contribute to a framework for developing more accurate and physically grounded environmental radiation prediction models.