AI-driven weather model could predict weather faster, more accurately
A critical weather model used by the National Oceanic and Atmospheric Administration is getting an AI-powered partner. The National Weather Service, part of NOAA, is beginning testing of an artificial intelligence version of the model to determine if it can more accurately forecast the weather and do so faster than traditional tools.
"This represents a significant leap forward in the application of artificial intelligence to environmental modeling," wrote project manager Isidora Jankov. "By leveraging new, high-resolution observations that help us better understand fine-scale physical processes, physical models can be improved, thereby improving the data on which AI-driven models are trained," she added.
NOAA recently announced it had turned over HRRR-Cast, a data-driven AI weather model, to the NWS for testing and evaluation.
The new model is NOAA's first experimental AI weather forecast system, and the agency said it's trained on three years of data from its widely used High-Resolution Rapid Refresh model, a short-term forecast model that has served as the agency's flagship for a decade.

The HRRR model and other traditional weather models combine real-time weather observations with complex atmospheric and physics-based equations to simulate weather conditions for a future period. Meteorologists and forecasters use these models to help predict events such as severe thunderstorms, heavy precipitation and other impactful weather phenomena on a global, regional and local level.
Unlike traditional weather models, which require a substantial amount of time and computing power to run, NOAA says HRRR-Cast can produce forecasts in a fraction of the time, on a single laptop, by learning to recognize specific weather patterns from historical data.
The developers found that the HRRR-Cast performed comparably to its traditional counterpart and even excelled in producing more realistic depictions of thunderstorm structure.
The developers also found that the AI model was 100 to 1,000 times more computationally efficient than the traditional model, meaning it can be run and deployed without relying on an energy and compute-intensive supercomputer.
Curtis Alexander, acting director of NOAA's Global Systems Laboratory, noted the AI model serves as a powerful proof of concept and the NWS will likely expand its capabilities by forecasting additional weather variables.
Both the AI version and the traditional version of the HRRR model will run together tangentially.
-ABC News meteorologist Kyle David Reiman







