typhoon forecasting

Data Party THU

Nov 14, 2026 · Artificial Intelligence

MAESTRO: Multi-Agent LLM System Turns Disaster Forecasts into Coordinated Action

Researchers from Zhejiang University and partners created MAESTRO, a multi-agent LLM system that links real-time observation, prediction, impact assessment, and warning recommendations for typhoons, achieving 98.4% accuracy in 37.9 seconds versus 250 seconds for experts, and deployed in Zhejiang's emergency platform to cut action planning from hours to minutes.