Chinese AI-physics system cuts typhoon track errors by up to 32% in tests

(ECNS) -- Chinese researchers have developed an AI-physics ensemble forecasting system that reduced typhoon track errors by up to 32.33% in comparative tests, the Institute of Atmospheric Physics (IAP) under the Chinese Academy of Sciences (CAS) said Wednesday.

Researchers from the Institute of Atmospheric Physics under the Chinese Academy of Sciences and Fudan University have developed an AI-physics ensemble forecasting system that reduced typhoon track errors by up to 32.33% in comparative tests, the institute said Wednesday.

FuXi-CNOPs, jointly developed by researchers from the IAP and Fudan University, combines the China-developed FuXi AI weather model with a dynamics-based method that generates physically meaningful variations in initial atmospheric conditions. This enables the system to forecast multiple plausible typhoon tracks and better assess uncertainty.

In 91 forecast experiments covering 62 typhoons, the Chinese system performed similarly to that of the European Centre for Medium-Range Weather Forecasts during the first 24 hours. However, its advantage widened between 24 and 120 hours, while the accuracy of uncertainty estimates improved by up to 29.2%.

The study was published in Advances in Atmospheric Sciences.

(By Intern Yang Hongran, Zhang Dongfang)