Chinese scientists develop artificial intelligence system capable of accurately predicting typhoon tracks

Researchers from the Institute of Atmospheric Physics of the Chinese Academy of Sciences and Fudan University have successfully developed a new ensemble typhoon forecasting system. According to Science and Technology Daily, a TV BRICS partner, the system not only calculates where a typhoon is heading more quickly but also accurately explains why its track may deviate.

Typhoon tracks often differ from forecasts because of the chaotic nature of the atmosphere: even slight variations in the initial conditions can lead to significant changes in a storm's trajectory. Conventional forecasting methods use as many as 51 scenarios, known as ensemble forecasts, to account for all possible variations. The new system uses only 31 scenarios while achieving even greater forecasting accuracy.

The key to the system's efficiency lies in a new approach incorporated into its artificial intelligence model, enabling it to focus on the regions most critical to track uncertainty: the typhoon's core, spiral rainbands, and areas where it interacts with the subtropical anticyclone. This allows the system to identify where forecasting errors are most likely to increase.

In comparative trials involving 62 typhoon cases and 91 ensemble forecast experiments, the system delivered impressive results. During the first 24 hours, its accuracy was comparable with that of the world's leading forecasting centres. However, its advantages became even more evident in medium- and longer-range forecasts covering 24 to 120 hours: track errors were reduced by up to 32.33 per cent, while its ability to quantify forecast uncertainty improved significantly, increasing accuracy by as much as 29.2 per cent.

As explained by Duan Wansuo, a researcher at the Institute of Atmospheric Physics of the Chinese Academy of Sciences, the system not only predicts a typhoon's track but also identifies the reasons why it may change. This represents a major breakthrough beyond purely statistical models, which can generate results without explaining the underlying causes.

The advance paves the way for the development of reliable and interpretable intelligent forecasting systems, which are particularly important for improving disaster preparedness across the Pacific region.