The Fengqing V1.0 AI global short-to-medium-range ensemble weather forecasting system has officially entered operational service in China. Developed by the National Meteorological Center in collaboration with various research institutions, the system utilizes the AI4S supercomputing cluster at the National Supercomputing Center in Tianjin for its training and iteration processes.
The system is built on a fully domestic technology stack, including core computing power, underlying algorithms, and data sources. It integrates a Vision Transformer architecture with physical constraints to model atmospheric uncertainty, which improves the accuracy of probabilistic forecasts for extreme weather events such as typhoons and heavy rainfall.
Fengqing achieves a global effective forecast horizon of 11 days. According to the developers, its core performance indicators, including probabilistic accuracy and dispersion error control, outperform the latest AI ensemble forecasting system from the European Centre for Medium-Range Weather Forecasts (ECMWF).
By leveraging the computing power of the Tianjin supercomputer, the system can perform global meteorological projections within minutes. This capability enhances the speed and stability of forecasts compared to traditional numerical weather prediction models, supporting sectors like transportation, energy, and disaster prevention.
