AI Headlines LogoAI Headlines

AI Research | 8/23/2025

IBM and NASA unveil Surya AI to predict solar flares hours ahead

IBM and NASA have introduced Surya, an open-source AI model that forecasts solar flares up to two hours ahead and pins down where on the sun’s surface an eruption might originate. Built on nine years of SDO data and released on Hugging Face as SuryaBench, early tests show meaningful accuracy gains and broader access to help protect satellites, power grids, and aviation.

Overview

IBM and NASA have teamed up to release Surya, a groundbreaking open-source AI model designed to forecast solar flares with unprecedented lead time. The system not only judges the likelihood of a flare but also visually predicts where on the Sun’s surface an eruption might occur, offering forecasters a new level of situational awareness. Think of Surya as a weather forecaster for space weather, but with the ability to point to a precise solar location two hours in advance.

How Surya works

  • Massive data, careful curation: Surya is trained on nine years of high-resolution solar observations from NASA’s Solar Dynamics Observatory (SDO). SDO has been quietly watching the Sun for more than a decade, accumulating petabytes of multi-channel imagery and magnetic field maps.
  • A compact yet powerful model: At its core is a 366-million-parameter transformer designed to process the SDO’s native 4096×4096 images. That’s not small, but it’s purpose-built to understand the subtle patterns that precede eruptions.
  • Open science, open data: IBM and NASA released Surya and its dataset, SuryaBench, on the Hugging Face platform, inviting researchers worldwide to test ideas, reproduce results, and push the field forward.

In addition to the flare-prediction capability, the team has demonstrated Surya’s versatility on related space-weather tasks, including predicting solar-wind speeds up to four days in advance, forecasting the emergence of active regions on the Sun, and predicting extreme ultraviolet (EUV) spectra that affect Earth’s upper atmosphere.

Why this matters

Space weather isn’t just a science puzzle; it’s a practical risk to the infrastructure we rely on every day. A severe solar storm can disrupt GPS navigation, affect airlines, damage satellites, and pose radiation risks to astronauts. Economic analyses have highlighted the potential scale of disruption: a major solar event could threaten trillions of dollars in value across the global economy over several years. That context helps explain why researchers are rushing to improve early warning.

The performance edge

  • In early tests, Surya showed a 16% improvement in the accuracy of solar-flare classification compared with existing methods. That’s not a slogan; it translates into more reliable alerts for operators managing satellites and ground systems.
  • Beyond classification, Surya’s ability to generate high-resolution, location-specific forecasts up to two hours ahead provides a real lead time advantage for protective actions.
  • The model has also been tested on a few other critical tasks, including wind-speed forecasting up to four days in advance, detection of new active regions, and EUV spectral predictions that influence atmospheric chemistry.

Open-source science, a broader reach

The release of Surya reflects a broader shift toward open AI in scientific discovery. By pairing deep NASA expertise with scalable AI, the project aims to democratize access to space-weather forecasting once reserved for a small set of operators. The implications ripple through satellite operators, power-grid managers, aviation planners, and beyond, giving them more time to adjust and safeguard the systems they’re responsible for.

What’s next

  • The Surya/Open-Source approach invites the global research community to extend, test, and validate the model across different data regimes.
  • As the scope of space missions expands, tools like Surya could become part of standard early-warning toolkits for critical infrastructure protection.
  • The collaboration aligns with a trend of releasing Earth- and space-weather models to accelerate scientific progress and improve resilience.

Context in the broader AI-for-science trend

This collaboration fits within a growing movement to apply large AI models to specialized scientific domains. Earlier iterations from the same teams have released climate and weather projection models under the Prithvi family, signaling a wider push to translate AI advances into practical, policy-relevant insights.

Key takeaways

  • Surya represents a meaningful leap in both predictive capability and accessibility for space weather.
  • The model’s open-source release lowers barriers to experimentation and adoption by researchers worldwide.
  • The work underscores how AI can augment human judgment in high-stakes, cross-disciplinary domains like heliophysics.

Sources

  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHKaYwAATxZIf_dTpLrOU6OECc8wE0a335EDdKxI5ludHL_4QM6MBsdSMPcw7xouFKekDEtHe1M-p9Nzhv9FjOI-osNQiikKOQLwNTXIoPRWi927lsNMcD-5rgoOfOd4DbogvZuSUbNbjweGA66vKlL7EYVBHzM-BZKPCL0hd3DH_kANzg697_M74M8VC4iUevj6gXXww7DmBe2dnpksdlvLtpmDBKRBydiYQt0VvmZ8k-xdIgha5tFmdM0kAgGyw5NL7gB0jq3EZ5_F2qcA6IkxFzWnskvGzm8TmdASMyxR25YDOVN8g==
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFBPE6jSD0CotusO4R5ATr1S5O5P5PV7sJvuCy-u5uKs-9i47mK4U0hDHSSChgkKxQ03p5kj8aB-0WiH2oHkwGK4wEHZjfGXm3_g4ozgrV9VuzR4uznNFS99I2Q_c-5fdraNfNnNeE9zCNXlBgRmcNmpXm9sxPTgsrHICgBfH1n9WkspvPTcfwS19xxbj7txglCwDOYJNjdD4c28fgHH-OEDdDXhzLCEz8Ix8M=
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGemd48xxCe1ntOSDVcQCXATJ2bULNpKSUv8054TN9kawhtD4Gc6_ZDrAGrATZf2Zu7dXkQCgVl2LoajujOG0ju-WdpDKYEuJK-kxKY7aPoc0xEzZ41olgTzkVUgmjNypXkW3ziDkDxvVb-yqa4QfXB1rd3qD_1j1jDnb7tib81x2q0pybxfIRjRRkxAckP6qsPB4hQIhrSF-IoqX0bpdL9BTewZwe30aGh3Hy_RyCqrlhtGyOqsQ==
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQEF_QPYyvFW9FHa1RAQ2KBmIBwm9gaz22KAC1KHvUfCJKUjQ0ZxicSL-9IElWYQih7X0xI9bF-m-u6q9NOI2BbK_-5IPh7gK1vDxxh-ZsSyuMoy0HstDlOqIM98h5_IdlIknXyhBAqjEF2uUYWnxRWGj53ITehva8RoK3f-8Ng2Ji3btDwY-m6GQ4RH8hnW9vUFWVgFwcPDFrB_RlZJfxRoWBMGXNRCru5gl4aBRbLGfkLjfySlep2giVGzdw==
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGztMgPrWs4YH1LepPOLBBaZy-goJFJCtjiQX7_OKYi3houqXJw7k0NHfU25NHdUi8ez71hLAsZn_Rc5y3c4zHc6yY-VxvlsJXNpsu78BDRANsJRK58eE1O6jSxpAeFbsxVHIYoJjaBUQnjDyUEcuS3dEsJ6fPgAHCSKQ91Jt3ilmBxQ8M=
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFO4vpyjazFEeljKbxvpVroGbOYd4ohgdifEziQHwz_7rDO7efUj5dMEq-RBAPWTogiOx4sRKTMJn7942nXexvRIrBF4iQMdINOojQA4KKMIw2NfosE6DDYqxOXKclmFdt-fdykaQqY8WS4V0xzp1lnzHYKnCvJ1_g=
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQFRM3th8k9ljXo2FtXdcgRux_Pu_bTFUirHH2LcGnTfQFEEvZe1Ww3F528EBBbPK04ln9ykMrvtxAiNmz32IXECTRFPLCv0xQPeQPOA0XftrnzAVOjSLRBPB8XW05MQvWZ_kdWrZyzKaNNhXdTzfjEx3EAyEpks2r840bqn2bP_gMAVXluBMr0T808MqJGn6cne8VAf4asuvaLL_M4kYkLULJEH2LQr2uyFI85Ljs3bEdk_e6_wnGVPIMmzwwA=
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQGkdjOL_WgY6Qd5JuOTqYIKUG8xQ9z6H9KUxaPbG0qTryHs3y3hyurEU6bdeqaKrWsmu0cvTkf49cbHCBJ9v1HdlHVUgMy3TspUMQfCklxZ_Htg9BkHxRRtBUM130qx5_a3d3J_ayvbKdgoS6XJFlE=
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQE_Zuqfj1CUrBbbC4pvXvabtDnSBBOWzi3v15Ei1Ja15Wu2l575MuEgtOqPtgHw9Y2Z75GasA7Dgk2hYlbj2ZFq07ajtKGOAxlm0PMt2tqefjkwxVKuh1YrcIUdDjIRkgKXCDgkqXEY8U8DosXEtqTBSD2WtuB-1BUR3qtnBUJg_nq2IKyLdUBTMcXH5pbjqS4c_witOZW74vcgUeREJLrlIBvhOE0yuyFotKbUo4SIWhQ5aSQYGYfmATX8ekqUy4Pf7oTGgNEry8spGNaR3G2koI-SmxuMX7EuYsfrF_4V8Cl2tg==
  • https://vertexaisearch.cloud.google.com/grounding-api-redirect/AUZIYQHjujH7sSAjbCAjfX2OtjYnc7siRtKhf-MkmjxNfAPz6uRHq4AiUrybxHr7L62yqLvXtzKMKUSOqQGbayl7QWkQdqM-HZvraLhOGJ_AKyc9GaGQAr9f2_gQiaJT8Mv-xfE17biXL097COh_jjXQMA0q2UAnN_HuI-hl-frxu19s7_oMXpbCA5xUCxCun1FRpGhBRwC234vqDgOlB69ZPT33mMFxRwmGXrleW96taQ==
Back to blog