NASA and IBM Release Open-Source Tools to Map the Moon with Python and AI

NASA and IBM have released a collection of open source tools designed to improve how researchers create and analyze maps of the lunar surface. The project, announced through a collaboration between the space agency and the technology company, aims to give scientists, students, and developers easier access to high-quality lunar data processing capabilities that were previously restricted to specialized teams.

The tools build on years of orbital observations collected by NASA missions including the Lunar Reconnaissance Orbiter. That spacecraft has spent more than fifteen years circling the Moon, capturing detailed images and topographic measurements at resolutions that allow features as small as a few meters across to be distinguished. Processing those observations into consistent maps requires substantial computational resources and specialized software. By making portions of their workflow available on public repositories, the partners hope to lower the technical barriers that have kept many researchers from working directly with the raw data.

At the heart of the release is a set of Python libraries and Jupyter notebooks that handle common tasks such as image registration, photometric correction, digital elevation model generation, and map projection. These components were extracted from operational pipelines used at NASA centers and IBM research laboratories. The code is now hosted on GitHub under permissive licenses that allow both academic and commercial reuse. Documentation includes examples that walk users through downloading a subset of Lunar Reconnaissance Orbiter Camera data, aligning overlapping images taken under different lighting conditions, and producing a seamless mosaic that can be imported into standard geographic information system applications.

One practical benefit of the shared tools is the ability to combine data from multiple instruments. The Lunar Reconnaissance Orbiter carries both a high-resolution camera and a laser altimeter. Aligning the two datasets accurately has traditionally demanded custom scripts and significant manual adjustment. The new open source packages automate much of that alignment while preserving the precision of the original measurements. Researchers can therefore generate elevation models that are tied directly to the visual context of the surface images, making it easier to study slope stability, crater morphology, or potential landing sites for future missions.

IBM contributed machine learning models trained to detect and classify surface features such as boulders, ridges, and shadowed regions that can confuse traditional image processing algorithms. These models were developed using IBM’s Watsonx platform and have been converted to open formats compatible with popular deep learning frameworks. Users can fine-tune the pretrained networks on their own datasets or apply them immediately to new orbital imagery. The inclusion of these models reflects a growing recognition that artificial intelligence can accelerate the interpretation of planetary datasets that grow faster than human analysts can review them.

The release also includes utilities for handling the unique coordinate systems used on the Moon. Unlike Earth, the lunar surface lacks an internationally agreed-upon prime meridian until relatively recently, and different mapping projects have used slightly different reference frames. The provided libraries contain transformation routines that convert between several historical and modern lunar datums, reducing the risk of positional errors when data from disparate sources are combined. This standardization effort is particularly valuable for teams planning spacecraft operations where even small discrepancies in latitude and longitude can affect trajectory calculations.

Educational institutions stand to gain from the project as well. Lunar science courses often struggle to provide students with hands-on experience because the software used by professional cartographers is expensive or difficult to install. The open source notebooks run in standard web browsers and require only modest computing resources. Instructors can therefore assign laboratory exercises that involve generating their own lunar maps from authentic mission data. Several universities have already indicated they will incorporate the materials into upcoming semesters, potentially expanding the pool of young scientists familiar with planetary mapping techniques.

The collaboration between NASA and IBM began several years ago when the space agency sought ways to extract greater scientific value from its growing archives. IBM brought expertise in scalable data processing and cloud-based analytics, while NASA supplied domain knowledge about lunar remote sensing. Joint workshops helped identify which parts of the existing workflow would be most useful if released publicly. Rather than open sourcing entire mission pipelines, which contain sensitive calibration details or proprietary algorithms, the teams focused on modular components that could stand alone. This selective approach preserved necessary controls while still delivering practical value to the community.

Data accessibility remains a central theme. All of the source imagery used in the examples is available through NASA’s Planetary Data System, but downloading and organizing those files can be daunting for newcomers. The new tool set includes helper functions that query the archive, select appropriate images based on location and illumination criteria, and manage the resulting large files efficiently. By simplifying these initial steps, the project reduces the time between deciding to study a particular lunar region and producing publishable maps.

Accuracy and traceability receive careful attention throughout the code. Each processing step logs parameters and intermediate results so that other researchers can reproduce the exact same output or understand where differences arise. This emphasis on reproducibility addresses a common criticism of some earlier lunar mapping efforts that relied on one-off scripts without clear version control. The public repositories use standard software engineering practices including continuous integration tests that verify the tools produce consistent results across different computing environments.

Looking forward, the teams plan to expand the collection with additional capabilities. Future releases may include support for radar data from the Mini-RF instrument, thermal observations from the Diviner Lunar Radiometer, and hyperspectral cubes from the Moon Mineralogy Mapper that flew on India’s Chandrayaan-1 mission. Each new dataset will require its own preprocessing routines, but the modular design of the current libraries should make integration straightforward. Community contributions are also encouraged; the project maintainers have published contribution guidelines and a clear roadmap that lists desired enhancements.

The decision to open source these tools aligns with broader efforts to make space science more inclusive. In recent years NASA has increased its emphasis on providing open data and open code, recognizing that innovation often occurs when diverse groups can build upon the same foundation. Commercial space companies interested in lunar resource prospecting or base construction can use the tools to assess terrain hazards without needing to develop their own mapping software from scratch. Similarly, international partners who operate their own lunar orbiters can compare their results against standardized NASA products processed with the shared code.

Challenges remain. Lunar mapping still requires substantial storage and processing power for continent-scale products. While the open source tools run on laptops for small areas, generating global maps at the highest available resolution demands access to high-performance computing clusters. The partners are exploring ways to offer cloud-based versions of the workflow that would allow users without local supercomputers to tackle larger projects. Early prototypes hosted on IBM Cloud have shown promising results, though questions about long-term funding and data egress costs still need resolution.

Another consideration involves the balance between speed and precision. Some of the machine learning components trade a small amount of geometric accuracy for dramatically faster processing times. Documentation clearly explains these trade-offs so that users can choose the appropriate settings for their particular science goals. For studies that require the highest possible fidelity, such as planning crewed landings, the classical photogrammetric methods included in the package remain available and are thoroughly validated against existing official maps.

The release has already sparked interest across several research domains. Planetary geologists are using the tools to update crater catalogs with more consistent diameter and depth measurements. Volcanologists are examining sinuous rilles and dome fields with improved topographic context. Astrobiologists interested in permanently shadowed craters are better able to model illumination conditions over time. Even astronomers studying Earth’s satellite for calibration purposes have found value in the standardized lunar albedo maps produced by the software.

By placing these resources in the public domain, NASA and IBM have effectively invited the global scientific community to participate more directly in lunar exploration. Rather than simply consuming pre-made maps, researchers can now adjust every parameter and understand exactly how each pixel in a lunar mosaic was derived. That transparency builds confidence in the resulting products and encourages critical examination that can lead to further improvements.

The project also demonstrates how partnerships between government space agencies and private technology firms can accelerate progress. IBM’s experience with large-scale data infrastructure complemented NASA’s deep knowledge of planetary science, producing a final product that neither organization would likely have released on the same timeline working alone. Similar collaborations may become more common as the volume of data from upcoming missions to the Moon, Mars, and beyond continues to increase.

Users interested in exploring the tools can find the primary repository linked from both the NASA Planetary Data System and IBM’s open science portal. The documentation site offers installation instructions, tutorial videos, and a discussion forum where newcomers can ask questions. Regular updates will be posted as the code evolves and as new datasets are incorporated. For those who prefer a more guided introduction, several virtual workshops are scheduled over the coming months to walk participants through their first lunar map.

The availability of these open source lunar mapping tools marks a noticeable expansion in the resources available for studying Earth’s nearest neighbor. By lowering technical barriers and promoting reproducibility, the initiative stands to broaden participation in lunar science at a time when multiple nations and private companies are preparing to return to the Moon. The ultimate measure of success will be the number and quality of new discoveries that emerge from the wider community now equipped to work directly with the rich archive of lunar observations. As additional instruments reach lunar orbit in the years ahead, the same open framework can grow to accommodate them, ensuring that the scientific return on these expensive missions is shared as widely as possible.