A pioneering artificial intelligence system developed by Nasa in partnership with IBM is capable of forecasting violent solar storms up to two hours before they erupt from the Sun, scientists have revealed. The system, named the Surya Heliophysics Foundational Model, was trained using nearly a decade’s worth of observations from Nasa’s Solar Dynamics Observatory. Such storms can disrupt satellite communications, interfere with GPS signals and pose risks to power grids on Earth, yet predicting them has long proved challenging due to their distance and rapid arrival once unleashed.
++ UK police to trial robotic dog for crime detection
Solar storms originate in dark, highly magnetised regions of the Sun known as active regions, which become more common as the star approaches the peak of its 11-year activity cycle. Researchers hope Surya will provide much-needed early warnings to satellite operators, energy providers and scientists tracking how solar radiation interacts with the Earth’s upper atmosphere. “Just as we use meteorology to forecast Earth’s weather, space weather forecasts predict the conditions and events in space that can affect our technologies,” explained Joseph Westlake, director of Nasa’s heliophysics division.
Early results suggest Surya – named after the Sanskrit word for the Sun – can generate accurate visual predictions of solar flares two hours into the future, setting a new benchmark in the application of AI for space weather. The system is now freely available to researchers worldwide through Hugging Face, GitHub and IBM’s TerraTorch library. Unlike traditional AI tools, which require extensive manual labelling of data, Surya takes advantage of the Solar Dynamics Observatory’s long-term, high-resolution dataset spanning nearly 15 years. This archive includes continuous imagery of the Sun every 12 seconds across multiple wavelengths, alongside precise magnetic field measurements.
++ Police probe reported rape of teenage girl near south-east London Golf course
Scientists believe this comprehensive record provides an ideal foundation for detecting subtle patterns in solar behaviour that shorter datasets might overlook. “Applying AI to heliophysics data is a vital step towards improving our defences against space weather and protecting astronauts, spacecraft, power grids and GPS,” said Dr Westlake. Project lead Andrés Muñoz-Jaramillo added: “Our hope is that the model has captured the critical processes behind the Sun’s evolution so that we can extract actionable insights. Ultimately, we want to give Earth the longest lead time possible.
