Table of Contents
NASA and IBM have released the NASA-IBM Lunar Foundation Model, an open-source artificial intelligence model designed to help researchers analyze the Moon’s surface and study lunar science at scale.
The model combines decades of lunar observations from multiple instruments and missions into a foundation model that researchers can adapt for different scientific tasks. NASA describes it as one of the first open-source AI models specifically built for lunar science.
What Is the NASA IBM Lunar Foundation Model?
The NASA IBM model is designed to help scientists work with large amounts of lunar imagery and other scientific data without having to develop a separate machine-learning system from scratch for every research task.
The model was trained on roughly 2 million image tiles, including more than 1 million high-resolution camera images at 1-meter resolution and nearly 964,000 multispectral images at 100-meter resolution. Its training data primarily comes from NASA’s Lunar Reconnaissance Orbiter, with additional information from missions including GRAIL, Lunar Prospector and Japan’s SELENE/Kaguya.
More Than 30 Layers of Lunar Data
One of the project’s important components is a machine-learning-ready lunar dataset created by NASA and IBM.
IBM says the dataset combines more than 30 spatially aligned layers from nine instruments across four missions, bringing different types and resolutions of lunar observations into a common framework. This allows researchers to examine information from multiple sources rather than relying on an individual instrument or dataset.
| Area | How the NASA IBM Model Can Help |
| Lunar craters | Identify and map craters more efficiently |
| Potential ice | Estimate where ice may be stable |
| Volcanic features | Identify and study younger volcanic formations |
| Surface changes | Detect changes across different observations |
| Lunar research | Analyze large multi-source datasets |
Studying Potential Lunar Ice
One of the more interesting applications involves the Moon’s polar regions.
Researchers can use the model to estimate where lunar ice may be stable, particularly in and around permanently shadowed regions. However, the model does not independently confirm the presence of ice. Instead, it produces ice prospectivity estimates that can help researchers identify areas for further investigation.
This distinction is important because the model is a research tool rather than a system that directly discovers or verifies usable lunar resources.
Faster Mapping of Craters and Volcanic Features
The NASA IBM model can also assist with mapping lunar craters and identifying unusual volcanic structures known as irregular mare patches.
Craters are particularly important to planetary scientists because their distribution and characteristics can help researchers study the age and geological history of the Moon.
IBM also says the model can support crater analysis relevant to landing-site planning, including identifying potential hazards such as steep slopes and boulders. It should not, however, be viewed as an autonomous system for certifying a landing site as safe.
NASA IBM Reports Strong Benchmark Results
IBM reports that the model can exceed widely used methods by up to 23% on certain geographic-feature identification benchmarks. This figure should be understood as a task-specific benchmark result rather than a universal 23% improvement across all lunar research applications.
NASA reports that the model matched or exceeded several strong baseline models across its evaluated tasks. It delivered comparable results for crater mapping and irregular mare patch segmentation while showing a clearer advantage in estimating polar ice stability.
An Open-Source Model for Researchers
The NASA IBM project is being released as an open research resource.
The model is publicly available through Hugging Face, while its codebase and machine-learning-ready datasets are available for researchers to test, adapt and build upon. The model is also integrated into IBM’s open-source TerraTorch toolkit.
The release is part of the broader Prithvi family of foundation models developed through the NASA-IBM collaboration. The family includes models designed for areas such as Earth observation and space science.
What the NASA IBM Model Means for Lunar Exploration
The significance of the model is not that AI can independently explore the Moon. Instead, its value comes from helping scientists process and interpret enormous volumes of existing lunar data more efficiently.
By combining different observations into a reusable foundation model, researchers can start with an existing AI system and adapt it to specific lunar research questions.
That could make it easier to investigate the Moon’s geological history, map surface features, study potential polar ice and analyze changes across large datasets.
Frequently Asked Questions
What is the NASA IBM Lunar Foundation Model?
It is an open-source AI foundation model developed by NASA and IBM to help researchers analyze lunar data and perform different scientific tasks.
Can the model find water on the Moon?
It can help researchers estimate areas where ice may be stable, but it does not independently confirm the presence of water or ice.
How much data was used to train the model?
NASA says the model was trained on roughly 2 million image tiles, including more than 1 million high-resolution images and nearly 964,000 multispectral images.
Is the NASA IBM model open source?
Yes. NASA and IBM have made the model, code and machine-learning-ready datasets available to the research community.
Conclusion
The NASA IBM Lunar Foundation Model marks another step toward using foundation-model technology for scientific research beyond Earth.
Its ability to combine large amounts of lunar imagery and multi-instrument data could help researchers map craters, investigate volcanic features, estimate potential ice stability and study changes across the Moon’s surface.
Rather than replacing scientists or independently making exploration decisions, the model is designed to make large-scale lunar research more efficient. With its open-source release, NASA and IBM are also giving researchers around the world an opportunity to adapt the technology for new questions as lunar exploration continues.
