Physics Driven Machine Learning Based Multi-layer Surface Study of Faustini Crater Region Near the Lunar South Pole Using Chandrayaan-2 Dual Frequency Synthetic Aperture Radar (DFSAR)

Abstract

Despite the proximity to Earth, significant portions of the Moon lack quantitative exploration, particularly the frigid Permanently Shadowed Regions (PSRs) at the poles. Identified by LRO Diviner, these PSRs may hold valuable volatiles like water ice, ammonia, methane etc. deposited by solar winds and impacts. Precise mapping of their distribution is crucial for future robotic and human missions to planetary bodies of the Solar system. This study focuses on the south polar region’s ’Faustini Rim A’ (87.30S), a potential landing site for the Artemis 3 mission. To ensure a safe landing and facilitate future base development, Chandrayaan-2’s Dual-frequency Synthetic Aperture Radar data is employed to analyze the surface and subsurface dielectric, physical, geotechnical properties of the lunar regolith as well as volumetric rock abundance. A Multi-Layer Perceptron regressor, trained on backscattering coefficients simulated with the Integral Equation Model (IEM), estimated the dielectric constant of both surface and subsurface regolith layers. Established formulas derived from Apollo mission samples are applied to retrieve physical properties. Rock abundance analysis reveals maximum rock concentration on the surface of Shoemaker floor and on the wall of the small crater inside Shoemaker at its subsurface level. Comparatively lower dielectric constants are observed in the L-band than S-band across the entire Faustini Rim A. Notably, the PSR in Shoemaker exhibited dielectric permittivity of 3-4, while areas near the Faustini crater showed higher values when employing Gaussian surface correlation function. These findings suggest a link between dielectric constant and regolith density as well as porosity, with higher values corresponding to denser, less porous regions at the upper layer of regolith. © Indian Society of Remote Sensing 2025.