NASA COFFIES Team Uses AI to Forecast Sunspot Emergence Hours Early
A NASA-led research group has created a machine-learning system that can anticipate solar active regions up to 12 hours before they appear on the surface.
Researchers within NASA’s COFFIES program have built a novel AI model that predicts the birth of solar active regions up to half a day in advance. The interdisciplinary team, drawing on data from the Solar Dynamics Observatory and supercomputing power at NASA Ames, employed a sliding-window transformer architecture to spot subtle reductions in acoustic activity and magnetic fields that precede sunspot formation. Unlike existing methods that monitor only already visible spots, this system can estimate where new regions will surface, offering earlier insight into potential solar flares and coronal mass ejections.
The model is not yet ready for operational use, and the team plans extensive testing against historic events. If successful, the capability could bolster space-weather forecasts for NASA’s Artemis lunar program, upcoming Mars missions, and NOAA’s Space Weather Prediction Center. Officials such as Michelangelo Romano see the technology as a valuable addition to current monitoring tools.
Why it matters
Early detection of solar storms helps protect astronauts, satellites and Earth-based communications from disruptive space weather.
In this story
