Key points are not available for this paper at this time.
We have examined and applied existing classification schemes for volcanic rocks and developed new schemes using thermal emission spectra of terrestrial volcanic rocks. Laboratory thermal infrared spectra (5–25 μm, at 2 cm −1 spectral sampling), deconvolved modal mineralogies, and derived mineral and bulk rock chemistries were used to distinguish basalt, basaltic andesite, andesite, and dacite. Modal mineralogies derived from linear deconvolution of terrestrial volcanic rocks were compared to modes measured by an electron microprobe phase mapping technique to determine the accuracy of linear deconvolution in modeling specific mineral abundances. One σ standard deviations of the absolute differences between modeled and measured mineral abundances range from 2.4 to 12.2 vol %, with an average standard deviation of 4.8 vol % being in agreement with average uncertainties calculated in previous studies. Weighted average compositions of feldspars in the deconvolution generally overlap the measured ranges of plagioclase compositions and the presence of low‐calcium and high‐calcium pyroxenes was correctly identified. Bulk chemistries of volcanic rocks were derived with a relatively high degree of accuracy (1σ standard deviations ranging from 0.4 to 2.6 vol %) by combining the compositions of spectrally modeled phases in proportion to their relative abundances in a particular sample. These data were collectively used to examine existing and develop new volcanic rock classification schemes. However, no single classification scheme was effective in accurately classifying all samples. Multiple steps of classification were required to distinguish volcanic rocks, reflecting the mineralogic diversity and continuum of compositions that exists in volcanic rock types. In a companion paper Hamilton et al. , this issue these schemes are applied to the classification of Martian surface compositions.
Wyatt et al. (Sun,) studied this question.