COMPARISON OF DECONVOLVED MINERALOGY AND CHEMISTRY FROM THERMAL EMISSION SPECTRA OF VOLCANIC ROCKS

Bulk-rock chemistry is commonly used to classify volcanic rocks and constrain the petrologic processes that produced them. Recent studies have demonstrated that bulk-rock chemistry can be estimated remotely from thermal emission spectra. The modeled silica and alkali contents from spectra of the MGS Thermal Emission Spectrometer (TES) have been used to distinguish basalt and andesite terrains on Mars. Bulk-rock chemistry is derived by summing the analyzed compositions of spectrally deconvolved minerals in their relative proportions. Although these modeled chemistries classify rocks correctly, there is a problem with accuracy when trying to model oxides. Using a suite of volcanic rocks from Medicine Lake Volcano spanning the range from basalt to andesite, we find that major oxides derived from thermal emission spectra do not model accurately, although they still classify correctly. In trying to determine the cause for the low modeled silica abundances (which increase the abundances of other oxides), several kinds of analyses have been done. First, the rocks were remodeled with and without olivine to see if this mineral may be over estimated, affecting the calculated abundance of silica. This gave an overall better spectral fit but did not solve the problem. Point counts were made on samples with different silica contents, to compare the most abundant phases (plagioclase and glass) with deconvolved mineral proportions. Glass is underestimated in all samples compared to modal data, and plagioclase models within ~5% of the modes with the exception of one rock. However, we have not yet found a consistent variation in deconvolved mineral proportions that could explain the low calculated percentages of silica, suggesting that the problem may lie in the deconvolved compositions of plagioclase and/or glass.