China launches its first Fengyun satellite AI models to track typhoons and predict severe storms up to four hours ahead, improving early weather warnings
HUNAN PROVINCE: China has unveiled its first artificial intelligence models using Fengyun meteorological satellite data to improve extreme weather monitoring and early warnings. The China Meteorological Administration (CMA) introduced the technology on October 29 at the 2026 Fengyun Satellite User Conference in Changsha, Hunan Province.
The two models focus on typhoon monitoring and short-term forecasts of severe thunderstorms, heavy rainfall and other dangerous weather conditions.
AI Tracks Typhoons Every 15 Minutes
The first model uses satellite imagery and artificial intelligence to analyse typhoon locations, intensity, structure and development. It combines more than 20 satellite data products and examines seven categories of typhoon characteristics.
The system updates its analysis every 15 minutes and generates detailed interpretation reports within 27 seconds. These capabilities allow meteorologists to identify changes in storm behaviour and assess potential threats more quickly.
Severe Weather Warnings Up to Four Hours Ahead
The second AI model focuses on severe convective weather, including thunderstorms that produce damaging winds, hail and heavy rainfall. Using observations collected by the Fengyun-4 satellite over the previous two hours, the system predicts cloud development and identifies areas at risk. According to the CMA, the technology can provide warnings up to four hours before dangerous conditions develop.
Impact to Expect
The new technology could strengthen disaster preparedness by helping meteorologists detect dangerous weather earlier and support emergency response planning. China currently operates 10 Fengyun meteorological satellites, with satellite data available to 141 countries and regions.
The CMA plans to expand AI applications to other weather hazards, including sandstorms, heavy fog and cold waves, supporting more comprehensive forecasting capabilities.
