Key Points
- IBM and NASA collaborated to create Surya, an open‑source solar‑weather AI model.
- Surya was trained on nine years of high‑resolution images from the Solar Dynamics Observatory.
- The 366‑million‑parameter model predicts solar flares with 16 percent higher accuracy than prior systems.
- It can generate visual forecasts up to two hours ahead of real observations.
- Accurate solar‑weather predictions help protect power grids from geomagnetic disturbances.
- Surya is lightweight enough to run on less powerful hardware.
- The model is freely available on Hugging Face for research and industry use.
- Its release coincides with discussions about potential cuts to NASA’s science budget.
Five different images of the Sun taken by the Solar Dynamics Observatory.
An graphic demonstrating the different effects of solar weather.
Background
Recent geomagnetic storms have reminded policymakers of the vulnerability of modern infrastructure. The 20‑year‑high geomagnetic event produced aurora displays far south of their usual range and underscored the challenges that solar eruptions pose to energy providers. In 1989, a series of plasma ejections caused a nine‑hour blackout in Québec, illustrating the stakes of accurate solar‑weather forecasting.
The Surya Model
In response, IBM and NASA have collaborated for years to create Surya, an open‑source foundation model named after the Sanskrit word for the Sun. IBM trained Surya on nine years of high‑resolution images captured by NASA’s Solar Dynamics Observatory, a satellite operating since 2010. The model, comprising 366 million parameters, is lightweight enough to run on modest hardware, making it accessible to a broad range of users.
Performance and Potential
Early testing shows Surya is 16 percent more accurate at answering the question “will there be a solar flare in the next 24 hours?” than prior systems. It can also generate visual predictions of what the Solar Dynamics Observatory might see, accurately forecasting the Sun’s appearance two hours ahead. Researchers are exploring longer lead times, which could be a game‑changer for power‑theft‑responsive grids that have been modernized over the past few decades.
Political Context
The announcement arrives as NASA’s science budget faces potential reductions, with proposals to cut operating funds for missions like the Solar Dynamics Observatory. While policymakers from both parties have voiced opposition, the budget’s final shape remains uncertain. IBM’s Juan Bernabe‑Moreno notes that the model would not exist without NASA’s scientific guidance, emphasizing that the AI platform has applications beyond the agency’s own missions.
Access
Surya is available for download on Hugging Face, inviting the scientific community and industry to leverage the model for improved solar‑weather prediction and related research.
Source: engadget.com