Deep Learning Improves GFS Wintertime Precipitation Forecast Over Southeastern China

Abstract Wintertime precipitation, especially snowstorms, significantly impacts people's lives. However, the current forecast skill of wintertime precipitation is still low. Based on data augmentation (DA) and deep learning, we propose a DABU‐Net which improves Global Forecast System over southeastern China. We build three independent models for lead times 24, 48, 72 hr, respectively. After using DABU‐Net, mean Root Mean Squared Errors (RMSEs) at are reduced by 19.08%, 25.00%, 22.37%, The threat scores (TS) all increased thresholds 1, 5, 10, 15, 20 mm day −1 times. During heavy days, RMSEs decreased 14% TS 7% within 48 hr. Therefore, combining DA learning has great prospects in forecasting.