AI typhoon forecasting has taken a major step forward after Chinese researchers developed a new prediction system that combines artificial intelligence with atmospheric dynamics to improve the accuracy of typhoon track forecasts.

The breakthrough was detailed in a study published in the journal Advances in Atmospheric Sciences.

Researchers said one of the biggest challenges in weather forecasting is the atmosphere’s chaotic “butterfly effect,” where tiny changes in initial conditions grow over time and can significantly alter a typhoon’s projected path, especially in medium- and long-term forecasts.

Although AI has transformed weather forecasting in recent years by processing massive datasets more efficiently, most existing AI models depend largely on pattern recognition and lack integration with atmospheric physical laws. This limitation can reduce their reliability in longer-range forecasts.

To overcome this challenge, a research team led by Duan Wansuo of the Institute of Atmospheric Physics at the Chinese Academy of Sciences, working with Li Hao’s team from Fudan University, integrated nonlinear dynamics algorithms into China’s domestically developed FuXi meteorological large model.

Arshad Nadeem Was Not Fully Fit Before Glasgow, Says Former Coach

The resulting FuXi-CNOPs typhoon ensemble forecasting system identifies critical atmospheric regions that influence typhoon movement and detects initial weather disturbances most likely to amplify forecast errors. It then generates predictions that better reflect real atmospheric physics.

Researchers evaluated the system using 62 representative typhoons and 91 comparative experiments, with results showing consistent improvements in forecasting performance.

The system performs at a level comparable to leading international forecasting models in 24-hour short-term predictions. Its greatest advantage, however, appears in medium- and long-term forecasts ranging from 24 to 120 hours, where it delivers higher accuracy and greater stability.

The new forecasting model also improves efficiency. While conventional leading forecasting systems require 51 computational datasets to complete a forecast, the FuXi-CNOPs system achieves more accurate and stable predictions using only 31 datasets, significantly reducing computing resource requirements.

Researchers believe the technology could strengthen disaster preparedness by providing more reliable typhoon forecasts while lowering computational costs for weather agencies.