Explainable Artificial Intelligence Identifies Key Meteorological Drivers of Extreme Precipitation

Extreme precipitation (EP) poses significant threats to human societies across the Asian monsoon region, yet the complex interplay of meteorological factors driving these events remains poorly understood. This study develops and applies an explainable artificial intelligence (XAI) framework to systematically uncover the critical atmospheric variables influencing EP. By integrating machine learning with interpretability techniques, the framework not only achieves accurate EP prediction but also reveals the physical mechanisms underlying its occurrence. Results demonstrate that variables such as integrated water vapor, vertical velocity, and circulation patterns play dominant roles in shaping EP, with their combined extremes exerting stronger impacts than any single factor alone. This approach bridges the gap between data-driven prediction and process-based understanding, offering new insights into the dynamics of monsoon extremes and providing scientific support for improved forecasting and disaster risk reduction strategies in a region highly vulnerable to climate hazards.