Chinese AI Model Boosts Typhoon Forecast Accuracy for Hong Kong

A new machine learning model deployed at the Hong Kong Observatory and the National Meteorological Centre improves typhoon forecast accuracy. The observatory predicts four to seven typhoons for Hong Kong between now and October, with some potentially becoming super typhoons due to El Niño. The model, developed by the Shenzhen Institute of Advanced Technology (SIAT) affiliated with the Chinese Academy of Sciences, addresses the challenge of forecasting rapid intensification—when a cyclone's maximum sustained winds increase by 15 m/s within 24 hours or 10 m/s within 12 hours. Researcher Li stated that rapid intensification is rare and highly unpredictable, often delaying preventive measures. Traditional numerical weather prediction and statistical-dynamic methods have failed to accurately reflect typhoon intensity evolution or capture non-linear changes.