NASA, IBM’s ‘Hot’ New AI Model Unlocks Secrets of Sun - UNIVERSE

This image from June 20, 2013 shows the bright light of a solar flare and an eruption of solar material shooting through the sun’s atmosphere, called a prominence eruption. Shortly thereafter, this same region of the sun sent a coronal mass ejection out into space -- a phenomenon which can cause magnetic storms that degrade communication signals and cause unexpected electrical surges in power grids on Earth. NASA's new heliophysics AI foundation model, Surya, can help predict these storms.

NASA/Goddard/SDO

Editor's Note: This article was updated Aug. 20, 2025, to correct the number of years of training data used and the model accuracy. The original article said the model was trained on 14 years of Solar Dynamics Observatory data and surpassed existing benchmarks by 15%; the model was actually trained on 9 years of data and surpassed existing benchmarks by 16%.

NASA is turning up the heat in solar science with the launch of the Surya Heliophysics Foundational Model, an artificial intelligence (AI) model trained on 9 years of observations from NASA’s Solar Dynamics Observatory. 

Developed by NASA in partnership with IBM and others, Surya uses advances in AI to analyze vast amounts of solar data, helping scientists better understand solar eruptions and predict space weather that threatens satellites, power grids, and communication systems. The model can be used to provide early warnings to satellite operators and helps scientists predict how the Sun’s ultraviolet output affects Earth’s upper atmosphere.

Preliminary results show Surya is making strides in solar flare forecasting, a long-standing challenge in heliophysics. Surya, with its ability to generate visual predictions of solar flares two hours into the future, marks a major step towards the use of AI for operational space weather prediction. These initial results surpass existing benchmarks by 16%. By providing open access to the model on HuggingFace and the code on GitHub, NASA encourages the science and applications community to test and explore this AI model for innovative solutions that leverage the unique value of continuous, stable, long-duration datasets from the Solar Dynamics Observatory. 

Illustrations of Solar Dynamics Observatory solar imagery used for training Surya: Solar coronal ultraviolet and extreme ultraviolet images from the Atmospheric Imaging Assembly (AIA) and solar surface velocity and magnetic field maps from the Helioseismic and Magnetic Imager (HMI).

NASA/SDO

The model’s success builds directly on the Solar Dynamics Observatory’s long-term database. Launched in 2010, NASA’s Solar Dynamics Observatory has provided an unbroken, high-resolution record of the Sun for nearly 15 years through capturing images every 12 seconds in multiple wavelengths, plus precise magnetic field measurements. This stable, well-calibrated dataset, spanning an entire solar cycle, is uniquely suited for training AI models like Surya, enabling them to detect subtle patterns in solar behavior that shorter datasets would miss.

Surya’s strength lies in its foundation model architecture, which learns directly from raw solar data. Unlike traditional AI systems that require extensive labeling, Surya can adapt quickly to new tasks and applications. Applications include tracking active regions, forecasting flare activity, predicting solar wind speed, and integrating data from other observatories including the joint NASA-ESA Solar and Heliospheric Observatory mission and NASA’s Parker Solar Probe.

“We are advancing data-driven science by embedding NASA’s deep scientific expertise into cutting-edge AI models,” said Kevin Murphy, chief science data officer at NASA Headquarters in Washington. “By developing a foundation model trained on NASA’s heliophysics data, we’re making it easier to analyze the complexities of the Sun’s behavior with unprecedented speed and precision. This model empowers broader understanding of how solar activity impacts critical systems and technologies that we all rely on here on Earth.”

These images compare the ground-truth data (right) with model output (center) for solar flares, which are the events behind most space weather. Surya's prediction is very close to what happened in reality (right). These preliminary results suggest that Surya has learned enough solar physics to predict the structure and evolution of a solar flare by looking at its beginning phase.

NASA/SDO/ODSI IMPACT AI Team

Solar storms pose significant risks to our technology-dependent society. Powerful solar events energize Earth's ionosphere, resulting in substantial GPS errors or complete signal loss to satellite communications. They also pose risks to power grids, as geomagnetically induced currents from coronal mass ejections can overload transformers and trigger widespread outages.

In commercial aviation, solar flares can disrupt radio communications and navigation systems while exposing high-altitude flights to increased radiation. The stakes are even higher for human spaceflight. Astronauts bound for the Moon or Mars may need to depend on precise predictions to shelter from intense radiation during solar particle events.

The Sun’s influence extends to the growing number of low Earth orbit satellites, including those that deliver global high-speed internet. As solar activity intensifies, it heats Earth’s upper atmosphere, increasing drag that slows satellites, pulls them from orbit, and causes premature reentry. Satellite operators often struggle to forecast where and when solar flares might affect these satellites.

The "ground truth" solar activity is shown on the top row. The bottom row shows solar activity predicted by Surya.

NASA/SDO/ODSI IMPACT AI Team

"Our society is built on technologies that are highly susceptible to space weather," said Joseph Westlake, Heliophysics Division director at NASA Headquarters. “Just as we use meteorology to forecast Earth's weather, space weather forecasts predict the conditions and events in the space environment that can affect Earth and our technologies. Applying AI to data from our heliophysics missions is a vital step in increasing our space weather defense to protect astronauts and spacecraft, power grids and GPS, and many other systems that power our modern world.”

While Surya is designed to study the Sun, its architecture and methodology are adaptable across scientific domains. From planetary science to Earth observation, the project lays the foundational infrastructure for similar AI efforts in diverse domains.

Surya is part of a broader NASA push to develop open-access, AI-powered science tools. Both the model and training datasets are freely available online to researchers, educators, and students worldwide, lowering barriers to participation and sparking new discoveries.

The process for creating Surya. Foundation models enhance the utility of NASA's Solar Dynamics Observatory datasets and create a base for building new applications.

NASA/ODSI IMPACT AI Team

Surya’s training was supported in part by the National Artificial Intelligence Research Resource (NAIRR) Pilot, a National Science Foundation (NSF)-led initiative that provides researchers with access to advanced computing, datasets, and AI tools. The NAIRR Pilot brings together federal and industry resources, such as computing power from NVIDIA, to expand access to the infrastructure needed for cutting-edge AI research.

“This project shows how the NAIRR Pilot is uniting federal and industry AI resources to accelerate scientific breakthroughs,” said Katie Antypas, director of NSF’s Office of Advanced Cyberinfrastructure. “With support from NVIDIA and NSF, we’re not only enabling today’s research, we’re laying the groundwork for a national AI network to drive tomorrow’s discoveries.”

Surya is part of a larger effort championed and supported by NASA’s Office of the Chief Science Data Officer and Heliophysics Division, the NSF , and partnering universities to advance NASA’s scientific missions through innovative data science and AI models. Surya’s AI architecture was jointly developed by the Interagency Implementation and Advanced Concepts Team (IMPACT) under the Office of Data Science and Informatics  at NASA’s Marshall Space Flight Center in Huntsville, Alabama; IBM; and a collaborative science team.

The science team, assembled by NASA Headquarters, consisted of experts from the Southwest Research Institute in San Antonio, Texas; the University of Alabama in Huntsville in Huntsville, Alabama; the University of Colorado Boulder in Boulder, Colorado; Georgia State University in Atlanta, Georgia; Princeton University in Princeton, New Jersey; NASA’s SMD’s Heliophysics Division; NASA’s Goddard Space Flight Center in Greenbelt, Maryland; NASA’s Jet Propulsion Laboratory in Pasadena, California; and the SETI Institute in Mountain View, California.

For a behind-the-scenes dive into Surya’s architecture, industry and academic collaborations, challenges behind developing the model, read the blog post on NASA’s Science Data Portal: https://science.data.nasa.gov/features-events/inside-surya-solar-ai-model

For more information about NASA’s strategy of developing foundation models for science, visit: https://science.nasa.gov/artificial-intelligence-science 

Source: NASA, IBM’s ‘Hot’ New AI Model Unlocks Secrets of Sun - NASA Science