The National Aeronautics and Space Administration and IBM have launched an open-source artificial intelligence model designed to help researchers process petabytes of lunar observation data, according to a NASA announcement on Thursday, Sept. 10. Built in collaboration with IBM Research and several academic institutions, the NASA-IBM Lunar Foundation Model is publicly available on Hugging Face to streamline surface mapping, crater identification, and polar ice estimation.
NASA-IBM Lunar Foundation Model Deployed for Open-Source Planetary Science
Traditional algorithms require researchers to build and train specialized programs from scratch for individual tasks. In contrast, foundation models undergo pre-training on vast, unlabeled datasets. This architecture allows the NASA-IBM model to generalize across multiple scientific domains through rapid fine-tuning, according to NASA. The complete codebase is hosted on GitHub for testing and experimentation.
“NASA has spent decades building an extraordinary scientific record of the Moon, but collecting data is only part of the job,” Kevin Murphy, NASA chief science data officer and acting chief data and AI officer, said in the announcement. Murphy noted that the model demonstrates how artificial intelligence transforms large-scale data into new scientific discoveries.
Training Data From the Lunar Reconnaissance Orbiter
The foundation model was trained primarily on data compiled over 17 years by NASA’s Lunar Reconnaissance Orbiter, which has produced a high-resolution mosaic covering most of the lunar surface. According to NASA, the LRO mission has gathered more data than all other NASA planetary missions combined. The training dataset consists of roughly 2 million image tiles, including more than 1 million high-resolution camera images at one-meter resolution and nearly 964,000 multispectral images at 100-meter resolution.
Additional terrain and imagery data came from missions such as NASA’s Gravity Recovery and Interior Laboratory, NASA’s Lunar Prospector, and the Japan Aerospace Exploration Agency’s Selenological and Engineering Explorer. TechRadar reported that the unified public cache brings together over 30 spatially aligned layers from nine instruments across four missions, marking the first time such a dataset has been made fully available for machine learning.
Applications for Lunar Craters, Volcanism, and Polar Ice
Planetary scientists can adapt the pre-trained model to specific research tasks using minimal labeled data. According to NASA, the tool assists researchers in mapping craters, identifying young volcanic features, and modeling potential locations of water ice near the lunar poles. Juan Bernabe-Moreno, director of IBM Research Europe, UK and Ireland, stated that the system connects observations across multiple instruments to reveal patterns that remain difficult to detect in isolation.
For polar research, the model helps scientists estimate where ice patches remain stable on and below the surface within permanently shadowed regions. For geologists studying lunar volcanism, the AI accelerates the identification of irregular mare patches. These structures appear relatively young and challenge established timelines for lunar cooling, according to NASA.
Did you know? Data collected by NASA’s Lunar Reconnaissance Orbiter exceeds the total volume collected by all of the agency’s other planetary missions combined over decades of space exploration.
Frequently Asked Questions
Where can researchers access the NASA-IBM Lunar Foundation Model?
The open-source model is available for free on Hugging Face, while the complete codebase is hosted on GitHub for testing and experimentation.
What primary dataset was used to train the lunar AI model?
The model was primarily trained on over 17 years of data from NASA’s Lunar Reconnaissance Orbiter, supplemented by missions like GRAIL, Lunar Prospector, and JAXA’s SELENE.
What specific tasks can the AI model perform?
According to NASA, researchers use the model to map craters, identify young volcanic features, and model potential locations of water ice near the lunar poles.
Explore our latest coverage on space exploration and artificial intelligence applications in planetary science by subscribing to our newsletter or leaving a comment below.
Keep reading