Chinese researchers recently presented a storm nowcasting system based on satellite data and AI techniques, achieving effective convection nowcasting up to four hours in advance.
This breakthrough, published in the journal Proceedings of the National Academy of Sciences, was jointly achieved by Wang Jingsong of the National Satellite Meteorological Center and researchers from various universities and research institutes.
Nowcasting is the process of obtaining current observations of the atmosphere and using them to generate rapid, short-term predictions of the future atmospheric state. The speed and practicality of nowcasting mean that it plays a crucial role in early warning systems of natural hazards around the world.
“Severe convective weather is characterized by strong suddenness, rapid evolution and great destructive power,” said Wang Jingsong. The lead time and coverage of it still leave much to be desired and hardly meet the needs of disaster emergency response.
The research team leverages the advantages of the Fengyun-4 series for large-scale, seamless monitoring to obtain long-duration data, enabling the extraction and prediction of the complex and random movements of convective cloud clusters.
The team proposes a deep diffusion model for satellite data (DDMS) to establish an AI-based convection nowcasting system. The system provides broad coverage of about 20,000,000 square kilometers, remarkable accuracy, and a high-resolution convective forecast every 15 minutes for up to four hours.
Its performance in convection nowcasting surpasses that of existing models, according to the research.
With the newly launched Fengyun-4 03 satellite, which enables inter-satellite coordination, high-speed data transmission, and broadcast functions, the three Fengyun-4 satellites will now form an integrated and coordinated observation network.
This enhanced network will significantly improve China’s capabilities in weather forecasting, meteorological disaster prevention, space weather monitoring, and ecological environment observation.
