AI Predicting Solar Storms

A New Dawn for Space Weather Forecasting

NASA, in collaboration with IBM and several scientific institutions, has unveiled Surya, the first-of-its-kind AI foundation model dedicated to heliophysics. 

Named after the Sanskrit word for the Sun, Surya is engineered to analyze vast, high-resolution solar imagery and predict solar activity—such as flares and eruptions—with unmatched precision and lead time.

How Surya Works

• Massive Training with SDO Data

  Surya leverages nine years of continuous observations from NASA’s Solar Dynamics Observatory (SDO), including multi-band imagery and data on solar magnetic fields.

• Advanced AI Architecture

  The model adopts long-short vision transformer architectures and spectral gating, enabling efficient processing of high-resolution solar data without overwhelming memory demands.

• Enhanced Prediction Capabilities

  Surya offers up to 2 hours of advance warning for solar flares—double that of legacy forecasting tools—and improves classification accuracy by approximately 16%.

• Open-Source for Broad Collaboration

  The model and supporting datasets (SuryaBench) are publicly available via platforms like GitHub, Hugging Face, and IBM’s TerraTorch—inviting global scientists to contribute and iterate.

Why It Matters

• Protecting Earth’s Technologies
  Solar storms pose significant threats to satellites, GPS systems, power grids, and radio communications. Surya’s improved forecasts can provide crucial extra time for safeguards.

• A Versatile Foundation Model for Science
  Beyond flare forecasting, researchers can adapt Surya to other heliophysics tasks like solar wind speed prediction, active region mapping, and EUV spectra modeling.

• Surge in Scientific Infrastructure
  Funded and shaped by a diverse team that includes NASA’s data science unit and field experts like Dr.Andrés Muñoz-Jaramillo (Southwest Research Institute)→ Surya represents a multi-institution collaboration to advance space weather science.