NASA and IBM have released the NASA-IBM Lunar Foundation Model, an open-source AI model designed to help researchers analyze the Moon’s surface and accelerate scientific exploration.
The foundation model was trained on a large lunar observation dataset assembled by NASA and IBM researchers. It combines multimodal and multi-resolution data from different instruments, allowing scientists to identify relationships across lunar observations that can be difficult to detect using conventional, task-specific machine learning models.
The model can support research into potential lunar ice deposits, volcanic features, and craters. Identifying ice in permanently shadowed regions could help researchers understand the distribution of water and oxygen resources that may be important for future lunar missions.
According to a NASA-IBM technical report, the model reduced error in identifying areas with high potential for lunar ice by up to 22% compared with the SwinV2-B ImageNet model. It also improved the detection of volcanic features by 3% and outperformed SwinV2-B by nearly 19% for context-scale crater detection while using half the training data.
Alongside the model, NASA and IBM released an open-source lunar dataset designed for machine learning. It combines more than 30 spatially aligned data layers from nine instruments across four missions, including NASA’s Lunar Reconnaissance Orbiter and GRAIL missions, as well as complementary data from Japan’s SELENE/Kaguya mission.
The dataset includes tens of thousands of lunar images and maps covering different geophysical characteristics of the Moon’s surface and subsurface.
The Lunar Foundation Model expands the Prithvi family of open foundation models, which includes AI models for geospatial science, weather, and heliophysics. The approach allows researchers to start with a shared foundation model and adapt it for different scientific tasks instead of developing separate machine-learning systems for each research question.
By making both the model and dataset openly available, NASA and IBM aim to give the global scientific community new AI tools for studying the Moon and supporting future lunar exploration.
