End‐member identification and spectral mixture analysis of CRISM hyperspectral data: A case study on southwest Melas Chasma, Mars

Abstract

We present spectral unmixing results over the southwest Melas Chasma region, where a variety of hydrated minerals were identified. We use the Discrete Ordinate Radiative Transfer radiative transfer model to simultaneously model Mars atmospheric gases, aerosols, and surface scattering and retrieve the single‐scattering albedos (SSAs) modeled by the Hapke bidirectional scattering function from Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) data. We employ a spectral unmixing algorithm to quantitatively analyze the mineral abundances by modeling the atmospherically corrected CRISM SSAs using a nonnegative least squares linear deconvolution algorithm. To build the spectral library used for spectral unmixing, we use the factor analysis and target transformation technique to recover spectral end‐members within the CRISM scenes. We investigate several distinct geologic units, including an interbedded polyhydrated and monohydrated sulfate unit (interbedded unit 1) and an interbedded phyllosilicate‐sulfate unit (interbedded unit 2). Our spectral unmixing results indicate that polyhydrated sulfates in the interbedded unit 1 have a much lower abundance (~10%) than that of the surrounding unit (~20%) and thus may have been partially dehydrated into kieserite to form the interbedded strata, supporting a two‐staged precipitation‐dehydration formation hypothesis. In the interbedded unit 2 phyllosilicates have an abundance of ~40% and are interbedded with ~20% sulfates. The results, in combination with thermodynamic calculations performed previously, suggest that the interbedded phyllosilicates and sulfates likely formed through coupled basalt weathering and evaporation. The methodology developed in this study provides a powerful tool to derive the mineral abundances, aiming to better constrain the formation processes of minerals and past aqueous environment on Mars.

A probabilistic approach to remote compositional analysis of planetary surfaces

“…Our model results show greater abundance errors than Mustard and Pieters [1987] for their mixtures because they constrain grain size while we do not. This emphasizes a key role that Journal of Geophysical Research: Planets 10.1002/2016JE005248 independent constraints on grain size-e.g., from thermal inertia [e.g., Liu et al, 2016] or from geologic context [e.g., Lapotre et al, 2017]-can have in effectively minimizing the errors in unmixing data. Altogether, our results highlight the importance of calculating uncertainties on unmixing model fits and considering the geological implications of the full range of permitted solutions, rather than interpretations relying on a sole acceptable solution.…”

“…Reflectance spectra allow the detection of key mineral and ice phases, and, when combined with quantitative semiempirical theories [e.g., Hapke, 1981;Hapke and Wells, 1981;Hapke, 1984Hapke, , 1986Shkuratov et al, 1999;Hapke, 2002Hapke, , 2008, enable the estimation of the composition and grain sizes of particulate surfaces ("spectral unmixing"). These models have been tested and used to invert for mineral abundances for laboratory particulate mixtures [e.g., Clark and Roush, 1984;Pieters, 1987, 1989;Hiroi and Pieters, 1994;Lucey, 1998;Poulet and Erard, 2004;Robertson et al, 2016] and for planetary surfaces from telescopic and orbiter-based spectroscopic data [e.g., McCord et al, 1998;Cruikshank et al, 2003;Poulet et al, 2008;Tirsch et al, 2011;Poulet et al, 2014;Edwards and Ehlmann, 2015;Goudge et al, 2015;Liu et al, 2016;De Sanctis et al, 2016].…”

Retrieval of Mineral Abundances of the Delta Region in Eberswalde, Mars

“…Hapke bidirectional reflectance distribution function relates the reflectance of a mixture to a linear combination of the SSA of its constituent endmembers (Hapke, 1981).It enables us to perform linear mixture analysis of the observed spectra (Goudge et al, 2015;Lin and Zhang 2017;Liu et al, 2016). The function is described as the following equation (Hapke, 1993):…”

“…In the VNIR, the observed signal of reflected light from a particulate surface is a non-linear combination of the spectral properties of the minerals (Goudge et al, 2015).The widely used non-linear mixing models in planetary spectra study are Hapke radiative transfer model (Hapke 1981) and the geometric optics model of Shkuratov (Shkuratov et al, 1999). Both the Hapke and Shkuratov models have been used to retrieve mineral abundances in VNIR spectra of Mars (Liu et al, 2016;Poulet et al, 2014).…”

“…Previous study mainly focus on the morphology, stratigraphy, and geology evolution of this place (Lewis & Oded 2006;Mangold et al, 2012;Pondrelli et al, 2008). Although these research provide clues of their geology history, mineral identification and abundance retrieval can further provide unique constraints on their formation environment (Liu et al, 2016). Poulet et al (2014) modeled the mineral abundances by average spectrum of Compact Reconnaissance Imaging Spectrometer for Mars (CRISM) 121 pixels over the Eberswalde deltaic deposits.…”