IBM, NASA launch AI model to map Moon’s ice, craters

### IBM and NASA Launch Open AI Model to Map Lunar Ice and Surface Features

IBM and NASA have introduced an open-source artificial intelligence model designed to help scientists analyze decades of lunar observations as the United States moves toward establishing a sustained human presence on the Moon.

The NASA-IBM Lunar Foundation Model is built to process large volumes of scientific data and assist researchers in identifying important features across the lunar surface, including potential ice deposits, craters and volcanic formations.

The model was trained using more than 30 layers of data gathered from nine scientific instruments operating across four NASA missions. One of the key sources of information is NASA’s Lunar Reconnaissance Orbiter, which has been studying the Moon’s surface for years.

The new system belongs to the broader Prithvi family of open foundation models developed jointly by IBM and NASA. Those models have also been designed for scientific applications such as geospatial analysis and weather-related research.

According to NASA and IBM, the lunar model demonstrated significant improvements during benchmark testing. It identified important surface features with as much as 23% greater accuracy than several commonly used existing approaches.

One potential application is the search for water ice in permanently shadowed areas of the Moon. Because these regions receive little or no direct sunlight, they can preserve deposits of ice that may have remained frozen for extremely long periods.

Lunar water is considered particularly valuable for future exploration. Scientists could potentially separate water into hydrogen and oxygen, providing materials that could be used for life-support systems as well as rocket propellant.

The ability to locate and map such resources could therefore become an important part of planning longer-duration lunar missions.

The model could also assist researchers in mapping craters and other terrain features that may affect the selection of safe landing locations. Detailed surface information is essential for evaluating potential landing zones and understanding the hazards astronauts could encounter.

Volcanic formations and other geological structures can similarly provide important clues about the Moon’s history while helping scientists better understand the characteristics of different regions.

Until now, many of these tasks have depended heavily on scientists manually examining large collections of lunar maps, photographs and other datasets. Some existing machine-learning tools can help, but they may operate with lower-resolution information or focus on individual types of observations.

By bringing information from multiple instruments and missions together, the new foundation model is intended to make lunar research faster and more efficient.

The development comes as NASA prepares for another phase of human lunar exploration through its Artemis program. NASA plans to return astronauts to the Moon in 2028 while developing technologies and infrastructure aimed at supporting a longer-term human presence.

The lunar model could eventually contribute to mission planning by helping researchers identify useful resources and assess the terrain surrounding potential exploration sites.

Improved maps could also help mission planners determine where scientific investigations should be conducted and where future equipment or habitats could potentially be deployed.

Resource mapping may have an additional benefit by reducing the amount of material future crews would need to transport from Earth.

The project reflects a broader effort by NASA and technology companies to apply artificial intelligence to space science as robotic missions continue to generate increasingly large and complex datasets.

Open-source AI models could make these capabilities more accessible to researchers by allowing scientists to adapt the technology for different lunar studies and other space-related applications.

As the United States works toward sustained exploration of the Moon and eventual missions to Mars, AI-powered analysis could become an increasingly important tool for turning vast amounts of space data into practical scientific and operational insights.

The NASA-IBM model represents another step toward combining advanced computing with planetary science, potentially helping researchers better understand the Moon while supporting preparations for future human exploration.