China has unveiled two artificial intelligence models for extreme weather forecasting, using data from its Fengyun meteorological satellites to move beyond monitoring current conditions and predict how weather systems will evolve.

The models — one for interpreting typhoon cloud images and the other for providing short-term warnings of severe convective weather — were introduced at the 2026 Fengyun Satellite User Conference in Changsha, Hunan province, on Thursday.

The Fengyun satellites continuously collect data on the formation, development and evolution of weather systems. However, the sheer volume of data has posed a challenge for forecasters seeking to extract key information quickly, said Han Bowei, an engineer at the National Satellite Meteorological Center, also known as the National Center for Space Weather.

As extreme weather events become increasingly frequent and intense worldwide, the need for faster and more accurate forecasts and warnings has grown. AI offers a way to process satellite data more efficiently and predict the evolution of weather systems, Han said.

ALSO READ: AI unlocks new future in parenting behavior

The Fengyun satellite-based intelligent typhoon cloud image interpretation model integrates more than 20 types of real-time satellite products. It analyzes cloud images every 15 minutes and generates a text report within 27 seconds, covering a typhoon's location, intensity, thermal and dynamic structure, wind and rainfall impacts, and likely evolution.

"The three-dimensional thermal structure of a typhoon is one of the key factors affecting changes in its intensity," Han said. "The AI model uses vast amounts of satellite data to analyze this structure and generate a clear, concise report to help forecasters quickly assess how a typhoon's intensity may change."

The model also displays the satellite data and product images underpinning each conclusion, allowing forecasters to verify its findings and conduct further analysis, Han said.

Last month, the model was used to monitor Typhoon Surigae, automatically identifying the storm and analyzing its intensity, wind radius and precipitation characteristics.

The second model focuses on short-term warnings of severe convective weather — small-scale, rapidly developing thunderstorm systems that can produce hazardous conditions, including damaging winds, hail and heavy rainfall that may trigger flash floods.

ALSO READ: AI crossroads: The philosophical rift in model development

Using two hours of continuous observation data from the Fengyun 4 satellite, the model predicts cloud evolution and identifies potential convective activity.

"With the new AI model, we can predict the occurrence, development and intensity of severe convective weather four hours in advance," Han said.

A study on the model was published last year in the journal Proceedings of the National Academy of Sciences, or PNAS. It highlighted the potential of combining satellite data with AI techniques to develop a nowcasting system with global applications, helping improve preparedness and mitigate the impact of severe storms worldwide.

Addressing concerns about inaccurate information generated by some AI models, Han emphasized that the new forecasting models produce verifiable results.

"All conclusions are based on satellite observations and can be verified," he said.

The two models are the first in a planned series designed to cover various types of hazardous weather, including typhoons, severe convective weather, rainstorms, sandstorms, heavy fog and cold waves, Han said.

"The goal is to build an intelligent monitoring system covering all these types of hazardous weather," he added.

Contact the writers at limuyun@chinadaily.com.cn