NASA and IBM Develop AI to Forecast Solar Flares Before They Strike Earth

Solar flares can be a threat to GPS and communications satellites.

 "An AI model trained on years of data from NASA’s Solar Dynamics Observatory can predict the sun’s future appearance and potentially flag dangerous solar flares."

NASA and IBM’s AI Predicts Solar Flares Hours Before They Hit Earth

An advanced artificial intelligence model developed by NASA and IBM can forecast the sun’s appearance hours into the future, even predicting solar flares that could signal dangerous space weather for Earth.

“I love to think of this model as an AI telescope where you can look at the sun and understand its moods,” says Juan Bernabé-Moreno of IBM Research Europe.

Why Solar Flares Matter

The sun’s “moods” are more than just a scientific curiosity. Solar flares unleash bursts of high-energy particles, X-rays, and extreme ultraviolet radiation that can disrupt GPS and communications satellites, endanger astronauts, and even affect passengers on commercial airlines.
Some flares are followed by coronal mass ejections, which can disturb Earth’s magnetic field and trigger geomagnetic storms capable of power grid outages.

Antarctic Climate Shocks Could Cause ‘Catastrophic Consequences’ for Future Generations

How the AI Model Works

The AI, named Surya after the Sanskrit word for sun, was trained on nine years of ultra-high-resolution images from NASA’s Solar Dynamics Observatory, capturing the sun in 13 different wavelengths. By analyzing this data, Surya learned to detect patterns and generate predictive images showing the sun’s state hours into the future.

In tests using historical flare data, Surya predicted solar flares up to 16% more accurately than standard machine learning models and could visually simulate flares two hours in advance.

“The power of AI is that it develops an intuition for the physics,” says Lisa Upton of Southwest Research Institute in Colorado.

Predicting the Sun’s Hidden Sides

Upton is particularly interested in whether Surya can anticipate activity on the far side and poles of the sun, areas that NASA’s instruments cannot directly observe. While Surya doesn’t explicitly model these hidden regions, it has shown success in forecasting how the sun will look once these areas rotate into view.

Scientists Discover Record-Breaking ‘Watermelon’ Nucleus, Challenging Atomic Theory

Limitations and Future Applications

Experts caution that predicting exact impacts on Earth remains challenging. Bernard Jackson of UC San Diego notes that the magnetic field configurations between the sun and Earth—critical for determining particle paths—cannot yet be directly observed.

For now, Bernabé-Moreno says the AI is designed for scientific research, but future integrations could allow power grid operators, satellite companies, and space agencies to use Surya as part of a real-time early warning system for solar activity.

“The potential is enormous,” Bernabé-Moreno says, “as we integrate Surya into systems that can answer practical questions about solar activity and protect critical infrastructure.”