IBM and NASA Launch Surya AI Model on Hugging Face for Solar Weather Prediction

IBM and NASA have released Surya, an open-source AI foundation model on Hugging Face, designed to analyze high-resolution solar observation data and predict solar activity impacting Earth and space-based technology. Named after the Sanskrit word for the Sun, Surya aims to support research in solar image interpretation and space weather forecasting to mitigate disruptions to GPS, power grids, telecommunications, satellites, airline navigation, and astronaut safety.

Solar flares and coronal mass ejections can cause satellite damage, navigation errors, power blackouts, and radiation risks. A Lloyd’s systemic risk scenario estimates global economic losses of $2.4 trillion over five years from a hypothetical solar storm, with $17 billion in expected losses. Surya, trained on nine years of high-resolution data from NASA’s Solar Dynamics Observatory, uses a custom multi-architecture solution to process images 10 times larger than typical AI training data. The model achieves a 16 percent improvement in solar flare classification accuracy and can visually predict flare locations up to two hours in advance.

The release includes the largest curated heliophysics dataset to support data-driven space weather research, covering tasks like solar flare prediction, solar wind speed,...