AI Discovers Two Lunar Cave Entrances with Strategic Value

TL;DR:

  • AI model named Essa identified two previously overlooked lunar pits by scanning publicly available NASA images
  • South Marius Hills Pit is located in a lava tube-rich area; Bel’kovich A Pit is near the north pole where water ice is more likely
  • Pits may connect to underground cave networks offering natural protection from radiation and micrometeorite impacts

Researchers at the University of Kent have used artificial intelligence to identify two previously undiscovered lunar cave entrances that could support human survival during future space missions. The discovery demonstrates AI’s capability to accelerate the analysis of space data at unprecedented scales.

Context and Background

Daniel Le Corre, a PhD researcher at Kent, surveyed less than 0.3% of the Moon’s surface using an AI model trained to detect distinctive pit formations in NASA imagery. The system, named Essa (entrances to sub-surface areas), successfully identified the South Marius Hills Pit—previously overlooked despite being located in an area thought to be rich in lava tubes—and the Bel’kovich A Pit near the Moon’s north pole.

The strategic locations of these discoveries are significant. The South Marius Hills Pit’s proximity to known lava tube formations suggests potential access to extensive underground networks, whilst the Bel’kovich A Pit’s location near the lunar north pole increases the likelihood of water ice deposits—a critical resource for sustained human presence.

The pits may provide natural shelter from harmful radiation and micrometeorite impacts, addressing two of the most significant challenges for lunar habitation. These underground structures could also harbour water ice, making them particularly valuable for future exploration bases.

Looking Forward

Mr Le Corre emphasised that Essa enables analysis of space data volumes “at speeds that would have been unachievable manually, thus accelerating the search for the pits that will be most favourable for future exploration or habitation.” The model’s efficiency in scanning just 0.3% of the lunar surface to yield two significant discoveries suggests substantial potential for additional findings as analysis expands.

The University of Kent indicated that techniques developed for lunar exploration could eventually extend to Mars, where similar underground structures might provide protection for human settlements. This research demonstrates how AI systems can augment scientific discovery by processing vast datasets that would be impractical to analyse through manual methods alone.

Source Attribution:

Share this article