According to Axios, NVIDIA (NVDA) Research has unveiled a new generative AI weather model that significantly enhances the accuracy of short-range weather forecasts, with the potential to predict extreme weather events such as flash floods and tornado outbreaks.
Nvidia (NASDAQ:NVDA) shared with Axios that this model is the first AI system capable of improving simulations of severe weather down to the kilometer scale.
Axios reports that if the model's success is replicated by other researchers, it could mark the beginning of a new era in precise short-term weather forecasting, potentially saving lives and reducing property damage.
Traditionally, AI models from tech giants like Nvidia, Microsoft, and Google have shown progress in generating medium-range, global weather forecasts, but Nvidia's new model, known as StormCast, is said to take this further by applying AI to local scales with greater accuracy.
Nvidia's StormCast, which is discussed in a paper submitted for peer review, written with the Lawrence Berkeley National Laboratory and University of Washington, reportedly operates at the mesoscale level, which means it can accurately resolve small-scale weather phenomena such as thunderstorms.
According to Axios, the model outperforms the leading U.S. short-range weather prediction system, known as HRRR, in predicting storm evolution on weather radar.
Despite these advancements, Nvidia's climate scientist Mike Pritchard told Axios that while AI's role in storm prediction is expanding, physics-based models should still be used to validate AI predictions.
As weather prediction increasingly incorporates AI tools, the combination of traditional methods and new AI technologies is expected to revolutionize the field, stated Axios.