According to Xinhua, Chinese researchers have developed a new storm forecasting system that combines AI with atmospheric dynamic models, helping to improve accuracy in storm path forecasting. The research results have just been published in the journal Advances in Atmospheric Sciences.
Forecasting the path of the storm has long faced many difficulties due to the butterfly effect, the phenomenon that very small changes in initial atmospheric conditions can be amplified over time and lead to large deviations in forecast results.
In recent years, AI has helped accelerate the weather forecasting process. However, most current AI models mainly learn from historical data without fully incorporating the laws of atmospheric physics, causing accuracy to be reduced when forecasting for many days.
The research group led by Mr. Duan Wansuo (under the Institute of Atmospheric Physics, Chinese Academy of Sciences) in collaboration with Mr. Li Hao's group (Fudan University) developed the FuXi-CNOPs system by integrating non-linear dynamical algorithms into the FuXi weather forecasting model.
The research team said that the new system can accurately identify atmospheric areas that have a major impact on the direction of storm movement, while reducing the impact of initial disturbances that may distort forecast results.
Through assessment of 62 storms and 91 comparative experiments, FuXi-CNOPs achieved accuracy equivalent to the world's leading forecasting systems in 24-hour forecasting. For medium and long-term forecasts from 24 to 120 hours, the system gives more stable and accurate results.
Another advantage of FuXi-CNOPs is computing efficiency. While many traditional forecasting systems require processing 51 datasets to produce results, the new system only needs 31 datasets but still maintains high accuracy, thereby significantly reducing computing resource demand.
