IBM, NASA cook up AI model to predict solar tantrums
(2025/08/22)
- Reference: 1755855193
- News link: https://www.theregister.co.uk/2025/08/22/nasa_ibm_surya_model/
- Source link:
Boffins at IBM and NASA have concocted an AI model to help predict the weather, but this time it is taking on space weather that might disrupt satellites and spacecraft, possibly even terrestrial power grids and the internet.
IBM's hope is that the model has learned enough of the Sun's behavior so that it is able to generate actionable insights – in other words, warn of impending danger
Big Blue and the US space agency are open sourcing their latest AI model, named Surya, the Sanskrit word for the Sun. This has been developed to predict the kind of violent solar flare-ups that can send a torrent of charged particles toward the Earth, potentially damaging electronic infrastructure.
A famous example of solar activity causing havoc is the [1]Carrington Event of 1859, when a geomagnetic storm caused telegraph systems in Europe and North America to fail, sometimes catching fire because of induced currents. Modern-day electronic infrastructure is much more complex and sensitive.
The Surya model has been trained using nine years' worth of data from NASA's Solar Dynamics Observatory (SDO) satellite, which has been closely monitoring the Sun for some time. IBM said that its scientists, along with those from NASA and eight other research centers, used the data to build their foundation model of solar physics.
"Think of this as a weather forecast for space," said Juan Bernabe-Moreno, Director of IBM Research Europe for Ireland and UK. "Just as we work to prepare for hazardous weather events, we need to do the same for solar storms."
[2]
He added that Surya is intended to anticipate what will happen next. IBM's hope is that the model has learned enough of the Sun's behavior so that it is able to generate actionable insights – in other words, warn of impending danger.
[3]Tsunami forecasting about to get a lot faster thanks to El Capitan super
[4]US weather agency dangles $396M to run ops for its next space-watching fleet
[5]Microsoft-backed AI out-forecasts hurricane experts without crunching the physics
[6]Small ocean swirls may have an outsized effect on climate, NASA satellite shows
The researchers claim a 16 percent improvement in solar flare classification accuracy in early tests, and hail this as a substantial improvement compared to previous methods. Surya has also been developed to visually predict solar flares for the first time, providing a high-resolution image of where the flare is predicted to occur up to two hours ahead.
NASA and IBM previously worked together on an [7]open source AI climate model called Prithvi that was developed to predict weather patterns here on Earth, while using fewer compute resources.
[8]
Building on that model but trying something different, the researchers say they gave the Surya model sequential images from SDO, then had it envision what the observatory would see an hour in the future. They were then able to check the accuracy of its prediction against actual observed data. However, IBM says the solar images used to train the model are ten times larger than typical AI training data, requiring a specialized multi-architecture solution to handle it all.
Surya is available on [9]Hugging Face , [10]GitHub , and IBM's [11]TerraTorch library for fine-tuning geospatial AI models. The research team is also [12]open sourcing SuryaBench , a set of curated data sets and benchmarks intended to simplify building and evaluating applications for space weather forecasting. ®
Get our [13]Tech Resources
[1] https://en.wikipedia.org/wiki/Carrington_Event
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aKg_u97OWsXPNMCfV7KBCQAAAQ4&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2025/08/13/tsunami_forecasting_el_capitan/
[4] https://www.theregister.com/2025/08/13/noaa_nextgen_space_weather_program/
[5] https://www.theregister.com/2025/05/21/earth_system_model_hurricane_forecast/
[6] https://www.theregister.com/2025/04/17/nasa_swot_analysis_small_currents/
[7] https://www.theregister.com/2024/09/25/nasa_ibm_ai_weather/
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aKg_u97OWsXPNMCfV7KBCQAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://huggingface.co/nasa-ibm-ai4science
[10] https://github.com/NASA-IMPACT/Surya
[11] https://github.com/IBM/terratorch
[12] https://huggingface.co/collections/nasa-impact/suryabench-68265ce306fc2470c121af7b
[13] https://whitepapers.theregister.com/
IBM's hope is that the model has learned enough of the Sun's behavior so that it is able to generate actionable insights – in other words, warn of impending danger
Big Blue and the US space agency are open sourcing their latest AI model, named Surya, the Sanskrit word for the Sun. This has been developed to predict the kind of violent solar flare-ups that can send a torrent of charged particles toward the Earth, potentially damaging electronic infrastructure.
A famous example of solar activity causing havoc is the [1]Carrington Event of 1859, when a geomagnetic storm caused telegraph systems in Europe and North America to fail, sometimes catching fire because of induced currents. Modern-day electronic infrastructure is much more complex and sensitive.
The Surya model has been trained using nine years' worth of data from NASA's Solar Dynamics Observatory (SDO) satellite, which has been closely monitoring the Sun for some time. IBM said that its scientists, along with those from NASA and eight other research centers, used the data to build their foundation model of solar physics.
"Think of this as a weather forecast for space," said Juan Bernabe-Moreno, Director of IBM Research Europe for Ireland and UK. "Just as we work to prepare for hazardous weather events, we need to do the same for solar storms."
[2]
He added that Surya is intended to anticipate what will happen next. IBM's hope is that the model has learned enough of the Sun's behavior so that it is able to generate actionable insights – in other words, warn of impending danger.
[3]Tsunami forecasting about to get a lot faster thanks to El Capitan super
[4]US weather agency dangles $396M to run ops for its next space-watching fleet
[5]Microsoft-backed AI out-forecasts hurricane experts without crunching the physics
[6]Small ocean swirls may have an outsized effect on climate, NASA satellite shows
The researchers claim a 16 percent improvement in solar flare classification accuracy in early tests, and hail this as a substantial improvement compared to previous methods. Surya has also been developed to visually predict solar flares for the first time, providing a high-resolution image of where the flare is predicted to occur up to two hours ahead.
NASA and IBM previously worked together on an [7]open source AI climate model called Prithvi that was developed to predict weather patterns here on Earth, while using fewer compute resources.
[8]
Building on that model but trying something different, the researchers say they gave the Surya model sequential images from SDO, then had it envision what the observatory would see an hour in the future. They were then able to check the accuracy of its prediction against actual observed data. However, IBM says the solar images used to train the model are ten times larger than typical AI training data, requiring a specialized multi-architecture solution to handle it all.
Surya is available on [9]Hugging Face , [10]GitHub , and IBM's [11]TerraTorch library for fine-tuning geospatial AI models. The research team is also [12]open sourcing SuryaBench , a set of curated data sets and benchmarks intended to simplify building and evaluating applications for space weather forecasting. ®
Get our [13]Tech Resources
[1] https://en.wikipedia.org/wiki/Carrington_Event
[2] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=2&c=2aKg_u97OWsXPNMCfV7KBCQAAAQ4&t=ct%3Dns%26unitnum%3D2%26raptor%3Dcondor%26pos%3Dtop%26test%3D0
[3] https://www.theregister.com/2025/08/13/tsunami_forecasting_el_capitan/
[4] https://www.theregister.com/2025/08/13/noaa_nextgen_space_weather_program/
[5] https://www.theregister.com/2025/05/21/earth_system_model_hurricane_forecast/
[6] https://www.theregister.com/2025/04/17/nasa_swot_analysis_small_currents/
[7] https://www.theregister.com/2024/09/25/nasa_ibm_ai_weather/
[8] https://pubads.g.doubleclick.net/gampad/jump?co=1&iu=/6978/reg_software/aiml&sz=300x50%7C300x100%7C300x250%7C300x251%7C300x252%7C300x600%7C300x601&tile=4&c=44aKg_u97OWsXPNMCfV7KBCQAAAQ4&t=ct%3Dns%26unitnum%3D4%26raptor%3Dfalcon%26pos%3Dmid%26test%3D0
[9] https://huggingface.co/nasa-ibm-ai4science
[10] https://github.com/NASA-IMPACT/Surya
[11] https://github.com/IBM/terratorch
[12] https://huggingface.co/collections/nasa-impact/suryabench-68265ce306fc2470c121af7b
[13] https://whitepapers.theregister.com/