A Two-Step AI Topographic Correction Approach and Applications in the Lunar Reconnaissance Orbiter Diviner Christiansen Feature Image

We present a two-step, artificial intelligence (AI) approach for topographic correction of lunar thermal infrared Christiansen feature (CF) imagery. The method establishes a mapping between CF spectra from flat and rugged pixels that are closely matched in composition, space weathering, thermal gradients, and visible-wavelength albedo (differences ≤ 1%), and subsequently applies this mapping to the standard Christiansen feature dataset acquired by the Lunar Reconnaissance Orbiter (LRO) Diviner instrument. By mitigating topography-related distortions in CF signals, the systematic corrections enhance spectral contrasts associated with space weathering and achieving a correction accuracy of 0.0315 μm. Using the corrected CF imagery, we dentify occurrences of pure anorthosite and olivine in the Aristarchus region and distinguish between competing mechanisms proposed for swirl formation at Reiner Gamma. These results demonstrate that, after removal of topographic effects, thermal infrared CF imaging provides a powerful means to investigate the composition and evolutionary processes of the lunar surface.