APPROACHES TO MODELING THERMAL EMISSION AND REFLECTANCE IN CRISM AND OMEGA OBSERVATIONS FROM 1-5 μM

Mapping the water content of the Martian surface has implications both for the geologic history of the planet and as a resource for human explorers. Hydrated minerals, adsorbed and absorbed water on mineral surfaces, and water ice display broad absorption features near 3 μm. Retrieving surface spectral reflectance from MRO CRISM (0.36-3.92 μm) and Mars Express OMEGA data (0.38-5.1 μm) for wavelengths longer than ~2.5 μm is an underdetermined problem. At typical Mars temperatures observations at these long wavelengths will contain both solar reflection and surface emission. Surface emission is a hemispherical-directional phenomenon and solar reflection is bidirectional. Overall spectral radiance on sensor is influenced by emission, absorption, and scattering by atmospheric gases and aerosols, surface scattering, and surface emission. DISORT provides a method to model spectral radiance on sensor with atmospheric conditions, surface emission and reflection properly included. The problem is still underdetermined because mapping from spectral radiance to surface single scattering albedo (SSA) requires estimation of surface kinetic temperature. We have employed several machine learning approaches to solve simultaneously for SSA and T, using DISORT results and training with a dataset of hundreds of Mars analog spectra. For initial testing we focused on observations of Phobos and Deimos, which have relatively homogenous surfaces and no atmospheres, and on areas on Mars with independent ground-based temperature measurements from the Curiosity and Phoenix missions.