Colorado State Develops AI System for Earlier Hail Forecasts

Colorado State University is developing an AI forecasting system that could warn of destructive hailstorms several hours or days in advance, according to a university research update. The project combines regional radar data with nine years of weather observations from across the United States.

Researchers are linking Hail-StormScope, a short term model developed with NVIDIA, to an existing university machine learning model for longer forecasts. Hail-StormScope uses NVIDIA Earth-2 technology and detailed data from multiple radar systems to produce hail forecasts with a lead time of two to three hours.

Current methods typically provide warnings 30 to 60 minutes before a hailstorm. The combined system will cover Colorado and Wyoming, with the short term model tracking storms as they change and the longer term model identifying risks further in advance.

The National Science Foundation funds the work through the Advanced Sensing and Computation for Environmental Decision making Engine. Researchers have not yet extensively tested the two AI approaches together.