IBM and NASA Develop a Digital Twin of the Sun to Predict Future Solar Storms

IBM and NASA have collaboratively developed Surya, an advanced AI foundation model that creates a digital twin of the Sun to improve predictions of solar storms. Trained on nine years of data from NASA’s Solar Dynamics Observatory (SDO), Surya processes high-resolution images and magnetic field measurements to better understand solar activity, particularly solar flares—bursts of electromagnetic radiation that can endanger astronauts and disrupt Earth’s technology infrastructure. By unifying diverse solar datasets and employing a long-range vision transformer architecture enhanced with spectral gating, Surya efficiently analyzes complex solar data with reduced noise and memory usage. Surya distinguishes itself by learning directly from raw data without extensive labeling, enabling faster adaptation and more accurate predictions. In tests, it integrated data from multiple solar observation instruments and outperformed traditional models by providing a two-hour advance warning of solar flares—double the typical one-hour lead time—and improved flare classification accuracy by 16 percent. NASA highlights Surya’s adaptable architecture, which could extend beyond heliophysics to