Atmosphere/surface coupling in hyperspectral imaging of arid terrains: A comparison of atmospheric correction strategies from EnMAP (Earth) to CRISM (Mars)

Atmospheric correction of hyperspectral images is critical in geology because accurate retrieval of surface reflectance is necessary to identify diagnostic mineral absorption features and interpret geological units. Several atmospheric correction methods have been developed for terrestrial and planetary applications; however, comparisons between Earth and Mars-oriented approaches remain limited despite the similar radiative challenges being encountered in arid volcanic terrain. Environments dominated by dark basaltic surfaces, bright evaporite deposits, strong topographic contrasts, mineral aerosols, and limited or missing vegetation challenge many of the assumptions commonly used in operational hyperspectral processing chains. These assumptions include the Lambertian surface approximation and vegetation-based aerosol retrieval strategies.

This study presents a comparative analysis of atmospheric and photometric correction methods for hyperspectral data acquired over analog volcanic landscapes on Earth and Mars. The terrestrial component focuses on EnMAP Level-1C images of the Asal-Ghoubbet rift in Djibouti. The Martian component examines CRISM observations of Jezero Crater and related volcanic terrain. We aim to examine how atmospheric composition, aerosol properties, surface anisotropy, and photometric assumptions influence the retrieval of bidirectional reflectance, spectral fidelity, and mineralogical interpretation.

The Asal-Ghoubbet rift presents challenges as a terrestrial test site due to its combination of very dark basaltic lava flows, highly reflective evaporite deposits, steep fault escarpments, high atmospheric water vapor content, and moderate desert aerosol concentration. The absence of dense vegetation further complicates aerosol retrieval strategies commonly used in operational terrestrial processing chains. Similarly, the Martian volcanic terrains investigated in this study feature dark mafic surfaces, altered dusty deposits, strong topographic variations, and an atmosphere dominated by suspended mineral aerosols within a thin CO₂ envelope. In both terrestrial and Martian contexts, the combination of spectrally contrasting surfaces, anisotropic reflectance behavior, and aerosol scattering produces strong aerosol/surface coupling effects that complicate separating atmospheric and surface contributions in hyperspectral observations. Despite their environmental differences, Earth and Mars face similar challenges regarding radiative coupling between the atmosphere and the surface and satellite observation.

We compared several atmospheric correction strategies on EnMAP observations. They included physically based radiative transfer inversion using MODTRAN6 with ERA5 and CAMS atmospheric profiles; FLAASH; ISOFIT coupled with the sRTMnet emulator; ATCOR-S with and without topographic correction; the operational EnMAP Level-2A processor; and empirical approaches, such as QUAC, and dark-pixel correction methods. These strategies were compared with those applied to CRISM images, including dark-subtraction correction, spectral normalization using neutral regions, and the non-Lambertian MARS-ReCO framework. Unlike standard Lambertian approaches, the MARS-ReCO framework explicitly accounts for aerosol scattering and surface anisotropy by using CRISM multi-angular observations to retrieve the BRDF of the surface.

We evaluated the accuracy of the corrected reflectance products by comparing them with measurements from an ASD FieldSpec spectrometer and examining the stability of diagnostic absorption features and derived mineralogical maps. We paid particular attention to how uncertainties in atmospheric correction propagate into mineral mapping products and geological interpretations.

Preliminary results indicate that physically constrained and non-Lambertian approaches provide the most accurate spectral restitution over terrestrial and Martian volcanic landscapes. Conversely, empirical methods and Lambertian assumptions result in stronger spectral distortions in the shortwave infrared (SWIR) domain, especially over dark volcanic surfaces and highly reflective evaporites. A comparison of Earth and Mars underscores the pivotal role of aerosol/surface coupling and surface anisotropy in hyperspectral remote sensing. This comparison also demonstrates the importance of integrating photometric effects into atmospheric correction workflows for characterizing planetary surfaces.