China Showcases AI Weather Models With 5-Day Typhoon Forecast Accuracy
Updated
Updated · NBC News · Aug 11
China Showcases AI Weather Models With 5-Day Typhoon Forecast Accuracy
3 articles · Updated · NBC News · Aug 11
Summary
Fengwu, Pangu and Fuxi are being used alongside conventional weather systems as Typhoon Dolphin nears China, underscoring the country’s rise in AI forecasting.
AI models generate forecasts far faster and at lower computing cost than physics-based supercomputer models, while matching or beating them on some accuracy measures.
Five days before Dolphin’s landfall, Fengwu predicted the mainland China hit time and location within 30 minutes and 20 miles; developers also said it beat Google’s GraphCast on about 80% of tested variables.
The gains matter during East Asia’s typhoon season, when better track forecasts can sharpen evacuation, flood and transport planning, though AI still trails traditional models on storm intensity.
China’s push has opened a new contest with systems such as Google’s GraphCast and GenCast, FourCastNet and Europe’s AIFS, but researchers say AI is still more complement than replacement.