Surya: IBM and NASA’s AI Model for Solar Forecasting

IBM and NASA release Surya, an open-source AI model trained on solar data to improve space weather prediction and protect critical infrastructure.

Addressing the Growing Risks of Space Weather

IBM and NASA have jointly introduced Surya, a new open-source AI model designed to forecast solar activity and its potential impact on Earth and space-based systems. Named after the Sanskrit word for the Sun, Surya is trained on nine years of high-resolution solar imagery from NASA’s Solar Dynamics Observatory. The model aims to enhance understanding of solar flares and coronal mass ejections, which can disrupt satellites, navigation systems, and power grids. As reliance on space-based technologies increases, the need for accurate solar weather prediction becomes more urgent.

Solar storms pose a variety of risks, including damage to spacecraft, radiation exposure for astronauts, and interference with aviation and agriculture. A scenario developed by Lloyd’s estimates that a severe solar storm could result in global economic losses of up to $2.4 trillion over five years. Recent events have already caused GPS disruptions, flight rerouting, and satellite malfunctions. Surya is intended to support both scientific research and operational planning to mitigate these threats.

Technical Foundations and Performance Gains

Surya was trained on a dataset significantly larger than typical AI models, with solar images roughly ten times the size of standard training inputs. This required a custom multi-architecture approach to manage scale and maintain computational efficiency. The model delivers high spatial resolution, enabling it to detect solar features with greater precision than previous systems. Researchers report a 16% improvement in solar flare classification accuracy compared to earlier methods.

Beyond classification, Surya introduces visual forecasting capabilities, predicting the location and shape of solar flares up to two hours in advance. This feature marks a shift from traditional models that rely on partial satellite views and limited temporal data. The dataset used includes curated heliophysics observations, supporting tasks such as solar wind speed estimation and EUV spectra prediction. These capabilities offer scientists a more comprehensive toolset for studying solar dynamics.

Open Access and Broader Scientific Impact

Surya is available on Hugging Face, allowing global researchers to experiment, fine-tune, and extend its capabilities for various applications. IBM and NASA emphasize the importance of democratizing access to advanced AI tools, especially in fields like heliophysics where data complexity is high. The model joins the Prithvi family of foundation models, which includes AI systems for geospatial and climate forecasting. Last year’s release of the Prithvi weather model similarly aimed to support short- and long-term environmental predictions.

This collaboration reflects a broader trend of integrating scientific expertise with machine learning to accelerate discovery. NASA’s contribution includes not only data but domain knowledge essential for validating and refining the model. Surya’s release is part of IBM’s strategy to position AI as a driver of innovation in planetary science. Researchers can now build region-specific or industry-focused solutions using Surya as a foundation.

A Step Toward Predictive Space Infrastructure

Surya’s development coincides with increasing concerns about space weather’s impact on modern infrastructure. According to NASA, the Solar Dynamics Observatory has provided continuous solar monitoring since 2010, capturing images every 12 seconds across multiple wavelengths. This long-term dataset offers a rare opportunity to train models on consistent, high-resolution data. Experts believe that even modest improvements in forecast lead time—such as Surya’s two-hour window—can help protect satellites and adjust power grid loads before disruptions occur.

Interestingly, researchers are exploring connections between solar weather and terrestrial phenomena like lightning, suggesting broader implications for Earth-based forecasting. As AI models like Surya evolve, they may help uncover hidden correlations between solar activity and atmospheric behavior. This could lead to integrated forecasting systems that span both space and Earth environments. The release of Surya marks a meaningful step toward that future.