Bennett reported at the end of July 2025 that AI is unlocking hidden lunar entrances that could provide crucial shelter for astronauts and reshape the future of moon exploration.
A known Example of Human Hideouts Credit: NASA/GSFC/Arizona State University
In a breakthrough study recently published in Icarus a team of international researchers has harnessed the power of artificial intelligence (AI) to unveil lunar pits and skylights—mysterious surface features that may serve as the entrances to hidden lava caves beneath the moon’s surface. These findings hold enormous potential for advancing future robotic and human exploration missions. The study outlines how AI and machine learning algorithms are being used to identify previously undetected lunar features, such as these pits and skylights, which could prove essential in understanding the moon’s volcanic history, as well as in protecting astronauts from the harsh environment of the moon.
AI-based models like ESSA (Entrances to Sub-Surface Areas) have been employed to process orbital images and identify new pits and skylights on the lunar maria, volcanic plains of the moon. The use of AI in this context opens the door to a new method of exploration, one that relies on the sophisticated processing capabilities of deep learning rather than traditional, slower, manual methods. With a growing database of lunar imagery, AI is now poised to uncover new sites of scientific interest, revealing not just the moon’s past but also offering potential refuge for astronauts in future missions.
What Are Lunar Pits and Skylights?
Lunar pits and skylights are depressions and openings found on the moon’s surface, often seen in volcanic regions known as lunar maria. These formations are not merely geological curiosities but represent the entrance points to ancient lava caves or tubes that stretch beneath the lunar surface. For scientists, these caves could provide shelter for astronauts in future lunar missions, a crucial consideration given the lack of an atmosphere, ozone layer, and magnetic field on the moon.
The threat of harmful solar and cosmic radiation makes it nearly impossible for astronauts to operate on the moon’s surface without some form of protection. Lava tubes could serve as natural shields, offering a safe haven from radiation and temperature extremes. These subsurface areas may also provide valuable insights into the moon’s volcanic past, helping scientists piece together a more detailed history of lunar geology.
With current technology, identifying these pits and skylights has been a slow process, relying heavily on visual analysis of orbital images. This is where AI steps in, offering an incredibly efficient alternative to traditional methods. AI models can process vast quantities of lunar data far more quickly, allowing researchers to identify key features that may have otherwise gone unnoticed.
How AI is Revolutionizing Lunar Exploration
Artificial intelligence has shown tremendous potential in recent years for planetary exploration, and the latest study demonstrates just how powerful AI can be in uncovering subsurface features on the moon. The researchers used several deep learning models to analyze images from lunar and Martian regions, with the goal of identifying pits and skylights on the moon’s surface. Among these models, the ESSA (Entrances to Sub-Surface Areas) algorithm stood out as the most effective, successfully identifying two new skylights.
The study highlights the enormous potential of AI, noting that:
“Since ESSA has surveyed just ≈0.23% of the moon’s surface so far, there are still vast amounts of data to which it can be applied.”
This indicates that the AI’s capabilities extend far beyond what has been achieved so far, with the potential to uncover even more lunar features as more data is processed. One of the major advantages of using AI for this task is the speed at which it can analyze data compared to traditional methods. What would have taken months or years of painstaking work by human researchers can now be completed in a fraction of the time.
ESSA’s Success in Identifying New Features
The ESSA model, in particular, has shown great promise in identifying new pits and skylights across the lunar surface. The team of researchers applied the model to data from the lunar maria, revealing two new skylights despite analyzing only a small fraction of the lunar surface. As the study notes,
“Up to now, ESSA has largely been applied to smaller mare deposits, which have well-defined boundaries, such as those within impact craters.”
However, the future applications of this AI model look promising.
With sufficient processing time, ESSA could extend its reach to larger regions of the moon, including expansive areas like Mare Frigoris, by iterating through latitude-longitude intervals. This capability to analyze vast stretches of the moon’s surface will only improve as more data is fed into the system, enabling ESSA to identify even more hidden features.
The Implications for Human Exploration and Shelter on the Moon
The identification of these lunar pits and skylights has significant implications for future lunar exploration missions. NASA’s Artemis program, which aims to return humans to the moon within the next few years, will greatly benefit from these findings. The program plans to land astronauts on the lunar south pole, far from the known locations of lava tubes. However, the ability to use AI to locate new sites of interest closer to the landing sites could provide astronauts with potential shelter options, adding a layer of safety to their missions.
The study not only helps uncover the moon’s geological history but also demonstrates how AI can be a key tool in identifying resources for in situ resource utilization (ISRU). For example, water ice deposits, often found in deep craters, could be vital for future lunar missions, providing both water and oxygen for astronauts. As AI continues to evolve, its application to planetary science is set to accelerate, helping scientists identify new features and resources on the moon and beyond.
Managed & Built by I2M Consulting, LLC | © 2010-
