FAST-ML: A Hybrid Physics-Machine Learning Framework for Tropical Cyclone Intensity Forecasting

This paper presents FAST-ML, a hybrid framework designed for tropical cyclone intensity forecasting. The framework addresses the challenge of rapid intensification (RI) prediction by combining the efficiency of data-driven approaches with physical constraints for improved interpretability. While full-physics numerical weather prediction models can represent RI-governing processes, they remain computationally expensive, and purely data-driven methods often lack physical grounding. FAST-ML bridges this gap by embedding physical priors into machine learning architectures for atmospheric fluid dynamics applications.