Identification of Terrestrial Alkalic Rocks Using Thermal Emission Spectroscopy: Applications to Martian Remote Sensing

We present a detailed study examining the use of laboratory thermal emission spectra (5-25 μm at 2 cm^-1 spectral sampling) for identification and classification of alkalic volcanic rocks. Modal mineralogies and derived bulk rock chemistries of a suite of terrestrial alkali basalts, trachyandesites, trachytes, and rhyolites were determined using linear spectral deconvolution. Model-derived mineral modes were compared to modes measured using an electron microprobe mapping technique to access the accuracy of linear deconvolution in determining mineral abundances. Standard deviations of 1σ of absolute differences between measured and modeled mineral abundances range from 0.68 to 15.02 vol %, with an average of 5.67 vol %. Bulk-rock chemistries were derived by combining modeled end-member compositions (wt %) in proportion to their abundances (recalculated to wt %). Derived oxide data compare well with measured oxide data, with 1σ standard deviations ranging from 0.12 to 2.48 wt %. Modeled mineralogies, derived mineral chemistries, and derived bulk chemistries were examined to assess their accuracy in classifying alkalic rocks. Derived bulk chemistry is the most effective classification tool, modal mineralogy is the least, and basalts and rhyolites are typically more accurately classified than trachyandesites and trachytes. However, no single classification scheme can accurately identify all rock samples, indicating that a combination of classification schemes is necessary to distinguish alkalic volcanic rocks using thermal emission spectra. To examine the accuracy of volcanic rock classification at lower spectral resolutions, thermal emission spectra were resampled to the resolution of the Thermal Emission Spectrometer (TES) instrument (10 cm ^-1), and a second series of linear deconvolutions was performed. Results from modeled mineralogies and derived bulk-rock chemistries at 10 cm ^-1 are comparable to results at 2 cm ^-1 spectral sampling, indicating that this amount of degradation of spectral sampling does not adversely affect modal mineralogies and chemistries derived from linear deconvolution. Data from 10 cm ^-1 spectral sampling were applied to derived mineral and bulk chemistry classification schemes with high degrees of success. Overall, results from this study suggest that linear deconvolution of thermal emissivity spectra can be successfully applied to the identification and classification of alkalic volcanic rocks on Mars.