%0 Dataset
%T U-net based ensemble forecast precipitation dataset for the lower reaches of the Yangtze River and the Lixia River region (2021-2022)
%J National Cryosphere Desert Data Center
%I National Cryosphere Desert Data Center(www.ncdc.ac.cn)
%U http://www.ncdc.ac.cn/portal/metadata/a663c2cd-0c05-4e12-98dc-397f31314722
%W NCDC
%R 10.12072/ncdc.nhri.db6789.2025
%A Li Lingjie
%A Gao Rui
%K Numerical forecasting;precipitation;statistical post-processing;deep learning
%X Accurate weather forecasting is a critical guarantee for social development, urban safety operations, people's livelihoods, and the prevention of water-related and drought disasters. In recent years, despite significant progress in numerical precipitation forecasting techniques, its accuracy remains affected by uncertainties in initial conditions, structural limitations of models, and the constraints of parameterization methods. Additionally, the relatively coarse spatial resolution has hindered its widespread application in meteorological and hydrological fields. Therefore, this study conducted statistical post-processing research based on the U-net model using the multi-model superensemble data (CNE) from CMA, ECMWF, and NCEP. A statistical post-processing dataset for numerical forecasts was developed for the lower Yangtze River and the Lixiahe area for 2021–2022. After U-net correction, the spatial resolution was refined from 0.5° to 0.1°, and both deterministic and probabilistic accuracy metrics were comprehensively improved. Furthermore, the forecast lead time was effectively extended, providing high-resolution data support for model construction in the project's demonstration area.