In recent days, National Meteorological Information Center (NMIC) has made staged progresses in terms of severe convective weather AI application dataset construction. It can be directly applied in minute-scale short-range heavy rainfall, thunderstorm, and gale AI recognition and model training. After the trial operation, the precipitation estimation accuracy of the deep network model based on this dataset has outperformed the operational precipitation estimation product based on the radar.
The dataset is composed of high temperal-spatial resolution radar feature parameters, weather background feature parameters, and label data. It encompasses over 230 million short range heavy precipitation samples and over 10 thousand thunderstorm and gale samples.
Currently, AI application dataset provides data service based on meteorological big data cloud platform. It can furnish high quality data support for severe convective weather smart identification and forecast. (Sep. 15)
Editor: Liu Shuqiao