Hong Kong Scientists' AI Model Forecasts Severe Storms Four Hours Ahead

• Hong Kong University of Science and Technology (HKUST) researchers developed a new AI system, Deep Diffusion Model of Satellite Data (DDMS), for extreme weather forecasting.
• The DDMS model extends severe storm forecasting lead time from 20 minutes to up to four hours, significantly improving preparation time for emergency services and the public.
• Utilizing generative AI and diffusion modeling, DDMS is trained on Fengyun-4 satellite infrared data to detect early signs of convection and cloud formation.
• The system shows over 15% accuracy gain compared to current short-range prediction systems, with 3-16% accuracy improvement in the critical 2-4 hour window.
• Developed in collaboration with China Meteorological Administration, DDMS aims to enhance regional weather resilience and has potential for commercial applications in sectors like energy and insurance.

Why It Matters: HKUST's new AI model, DDMS, revolutionizes severe storm forecasting, providing up to four hours lead time.