NVIDIA Research has announced a new generative AI model called StormCast, designed to emulate high-fidelity atmospheric dynamics and improve weather prediction at the mesoscale level.
The model, introduced amid the current Atlantic hurricane season, aims to enhance disaster planning and mitigation efforts by enabling more reliable weather forecasting for events larger than storms but smaller than cyclones.
"StormCast represents a significant leap in our ability to predict and understand extreme weather events," said Mike Pritchard, a researcher at NVIDIA. "This technology could be crucial in saving lives and mitigating the billions of dollars in damage caused annually by severe weather."
Developed in collaboration with Lawrence Berkeley National Laboratory and the University of Washington, StormCast can predict over 100 variables, including temperature, moisture concentration, wind, and rainfall radar reflectivity at multiple altitudes. The model's predictions have shown up to 10% greater accuracy than the U.S. National Oceanic and Atmospheric Administration's current operational systems for forecasts up to six hours in advance.
NVIDIA trained StormCast on approximately three-and-a-half years of NOAA climate data from the central U.S., utilising NVIDIA's accelerated computing capabilities to expedite calculations.
The innovation has already garnered attention from climate scientists. Tom Hamill, head of innovation at The Weather Company, stated, "StormCast addresses major challenges in forecasting organised thunderstorms and winter precipitation. We're excited to collaborate with NVIDIA on developing and potentially using these deep learning forecast models."
StormCast is part of NVIDIA's broader Earth-2 initiative, a digital twin cloud platform combining AI, physical simulations, and computer graphics, to enable global-scale weather and climate predictions, with both accuracy and speed.
As extreme weather events continue to increase in frequency and severity, causing over $150 billion in annual damage in the U.S. alone, innovations like StormCast represent critical advancements in climate research and actionable weather prediction.