Hapke mixture modeling applied to VNIR spectra of mafic mineral mixtures and shergottites: Implications for quantitative analysis of satellite data

The mineralogy of Mars is well understood on a qualitative level at a global scale due to satellite data. Quantitative analysis of visible and near-infrared (VNIR) satellite data is a desirable but nontrivial task, due partly to the nonlinearity of VNIR reflectance spectra from the mineral mixtures of the Martian surface. In this study, we investigated the use of the Hapke radiative transfer model to generate linearly mixed single scattering albedo data from nonlinearly mixed VNIR reflectance data and then quantitatively analyzed them using the linear spectral mixture model. Simplifications to the Hapke equation were tested accounting for variables that would be unknown when using satellite data. Mineral mixture spectra from the RELAB spectral library were degraded to test the robustness of the unmixing technique in the face of data that mimic some of the complexities of satellite spectral data collected at Mars. A final test was performed on spectra from shergottite meteorites to assess the technique against real Martian mineral mixtures. The simplified Hapke routine produced robust abundance estimates within 5–10% accuracy when applied to laboratory standard spectra from the synthetic mixtures of igneous minerals in agreement with previous studies. The results of tests involving degraded data to mimic the low spectral contrast of the Martian surface and the lack of a priori knowledge of the constituent mineral spectral endmembers, however, were less encouraging, with errors in abundance estimation greater than 25%. These results cast doubt on the utility of Hapke unmixing for the quantitative analysis of VNIR data of the surface of Mars.