Improving the heavy rainfall forecasting using a weighted deep learning model

Weather forecasting has been playing an important role in socio-economics. However, operational numerical weather prediction (NWP) is insufficiently accurate terms of precipitation forecasting, especially for heavy rainfalls. Previous works on NWP bias correction utilizing deep learning (DL) methods mostly focused a local region, and the China-wide forecast had not attempted. Meanwhile, earlier studies imposed no particular focus strong rainfalls despite their severe catastrophic impacts. In this study, we propose DL model called weighted U-Net (WU-Net) that incorporates sample weights various events to improve forecasts intensive China. It found WU-Net can further skill heaviest rainfall comparing with ordinary ECMWF-IFS. Further analysis shows improvement increases growing lead time, distributes mainly eastern parts This study suggests considering imbalance meteorological data could generated by prediction.