IBM and NASA have released an open-source artificial intelligence (AI) model designed to help scientists locate possible ice deposits, map craters, and study volcanic features on the Moon.
The NASA-IBM Lunar Foundation Model processes data collected by several instruments at different resolutions. It could reduce the time researchers spend manually studying lunar maps and images or building separate machine learning systems for individual research tasks.
The model is intended to support lunar research and preparations for a sustained human presence on the Moon. Scientists consider lunar ice particularly important because it could provide water and oxygen for a future Moon base. It could also potentially be used to produce rocket fuel for missions to Mars.
“NASA has spent decades building an extraordinary scientific record of the Moon, but collecting data is only part of the job,” said Kevin Murphy, chief science data officer and acting chief data and AI officer at NASA Headquarters in Washington. “We also have to make data easier for scientists to explore and use. The NASA-IBM Lunar Foundation Model shows what’s possible when we bring AI to NASA’s petabytes of scientific data. That’s a real opportunity we see with AI: turning large-scale data into new discoveries.”
In tests, the model lowered the error rate in identifying areas with high potential for lunar ice by as much as 22% compared with the SwinV2-B image model, according to an IBM-NASA technical paper.
It also improved the mapping of Irregular Mare Patches, volcanic features that help scientists study the Moon’s volcanic history and changing surface. The model delivered a 3% improvement over SwinV2-B while requiring less fine-tuning.
For crater detection, the NASA-IBM model produced results comparable to leading models at meter-scale resolution. At a broader resolution of about 100 meters, it outperformed SwinV2-B by nearly 19% while using half the training data. Detailed crater maps can help scientists estimate the age and geology of lunar terrain and allow NASA to assess landing sites and potential hazards.
Alongside the model, IBM and NASA released a machine learning-ready lunar dataset containing more than 30 aligned data layers from nine instruments across four missions. It includes tens of thousands of images and maps from NASA’s Lunar Reconnaissance Orbiter and GRAIL mission, along with data from the Japanese Aerospace Exploration Agency’s SELENE/Kaguya mission.
“The NASA-IBM Lunar Foundation Model gives scientists a foundation to explore the Moon at scale, connecting observations across instruments, revealing patterns that are difficult to see in isolation, and providing an open platform the global research community can build on,” said Juan Bernabe-Moreno, director of IBM Research Europe, UK and Ireland.
The model joins IBM and NASA’s Prithvi family of open foundation models covering geospatial data, weather, heliophysics, and now lunar science.