Duration: 36 months
Understanding the local-scale nature and evolution of the lunar regolith, at the surface and near-subsurface, is crucial for a variety of aspects: from the accumulation and preservation of volatiles, to planning and performing landing, roving and sampling operations, to in situ ressources utilizations. We have shown that rocks on the lunar surface record and preserve erosion events and -if coupled to models- can be used as a probe to determine regolith nature and evolution. These new rock models allow to retrieve time and compositional information from optical image analysis of boulders. Currently, high-resolution images of the NASA’s Lunar Reconnaissance Orbiter (LROC) Narrow Angle Camera (NAC) are the best source available to detect individual boulders. However, there are more than 1.6 million images potentially containing boulders, so that the detection and analysis of boulders by expert visual inspection is unfeasible. As a result, today the majority of boulders remains undetected. Some very recent studies, as well as our own preliminary investigation, indicate that machine learning on lunar rocks enables much faster and reproducible analyses. The novelty of the proposed research is to couple an automatic detection and characterization of boulders at the global scale with the new boulder models and with thermal data. This coupling will enable to answer a variety of key questions concerning the lunar regolith nature and evolution. We will understand subsurface regolith properties and thus volatile presence. This research will directly strength the investigations carried out by ESA on and around the Moon, like the ESA Prospect project. Products will be a know-how and software technology to produce ressources map, hazards map, subsurface properties map, and subsurface-robotic interactions scenarios for any point of the lunar surface.
