Short‐Term Forecasting of Cloud Physical Properties Based on Fourier Neural Operator Method

Abstract Accurately understanding the evolution and development of cloud physical properties (CPP) in advance is crucial for extreme weather forecasting and early warning. This study utilized the Fourier neural operator (FNO) method to develop a short‐term forecasting model of Cloud (Cloud‐FNO). Using the multi‐task learning framework and autoregression strategy, the model achieves accurate 6‐hr forecasting of cloud phase (CLP), cloud top height (CTH), cloud effective radius (CER), and cloud optical thickness (COT). Evaluation results on the independent testing data set show that the Cloud‐FNO model achieves an average CLP identification accuracy exceeding 74%, and the average root mean square errors for CTH, CER, and COT forecasts are 2.28 km, 6.52 μm, and 9.01, respectively. Importantly, the Cloud‐FNO model demonstrates strong forecasting capability and promising application potential for the CPP evolution under severe weather.