Hong Kong: AI Predicts Storms More Accurately Than Meteorologists

On January 28, 2026, HKUST researchers unveiled the Deep Diffusion Model (DDMS), which changes the game in meteorology. Trained on Fengyun-4 satellite data, the system predicts extreme weather events (downpours, convective storms) up to 4 hours in advance.

Forecast accuracy improved by ~15% compared to traditional methods. In Hong Kong's dense urban environment and frequent typhoons, this 15% means gained time for evacuation and infrastructure protection. The technology is already undergoing validation at the China Meteorological Administration.

Source: HKUST / Reuters