Machine‐Learning (ML)‐Physics Fusion Model Accelerates the Paradigm Shift in Typhoon Forecasting With a CNOP‐Based Assimilation Framework

Abstract In this study, we develop an advanced hybrid forecasting system for short‐term (0–120 hr) typhoon predictions, seamlessly integrating the FuXi machine‐learning model with the physics‐based Shanghai Typhoon Model (SHTM). By employing spectral nudging, the hybrid FuXi‐SHTM model leverages FuXi's robust large‐scale forecasting capabilities alongside SHTM's mesoscale strengths, significantly enhancing track, intensity, and precipitation predictions for super typhoons Yagi (2024) and Krathon (2024). To further improve the forecasting capability for extreme typhoons, the Conditional Nonlinear Optimal Perturbation method is employed for the first time to identify sensitive regions for the hybrid model. Despite being constrained by FuXi's large‐scale forecast fields, the dense assimilation of satellite observations within these sensitive regions can further enhance typhoon forecasts. This study emphasizes the synergy between data‐driven strategies and established physical modeling, which can inspire further depth in understanding of extreme typhoon events.