Which achievement is attributed to NASA’s Surya AI model in testing against historical events and existing models?

Surya, developed by NASA with IBM and trained on nine years of Solar Dynamics Observatory data, successfully recreated the 2015 St. Patrick’s Day geomagnetic storm in tests. In predictive performance, it outperformed existing models by up to 16% in predicting solar flares. The model uses machine learning techniques (spectral block layers and long-short transformers) to forecast solar events—such as flares and coronal mass ejections (CMEs)—with lead times of up to two hours, improving forecast accuracy and operational preparedness for satellites, power grids, and aviation. Surya is also open-source, enabling international collaboration and model improvement.