A new systematic approach using the Modified Gaussian Model: Insight for the characterization of chemical composition of olivines, pyroxenes and olivine-pyroxene mixtures
Résumé
An automatic procedure has been implemented on the original MGM approach (Sunshine et al., 1990) in order to deal with an a priori unknown mafic mineralogy observed in the visible-near infrared by reflectance spectroscopy in the case of laboratory or natural rock spectra. We consider all the mixture possibilities involving orthopyroxene, clinopyroxene and olivine, and use accordingly for each configuration different numbers of Gaussians, depending on the potential complexity of the mixture. A key issue is to initialize the MGM procedure with a proper setting for the Gaussians parameters. An automatic analysis of the shape of the spectrum is first performed. The continuum is handled with a second order polynomial adjusted on the local maxima along the spectrum and Gaussians parameters initial settings are made on the basis of laboratory results available in the literature in the case of simple mixtures of mafic minerals. The returned MGM solutions are then assessed on spectroscopic grounds and either validated or discarded, on the basis of a mineralogical sorting. The results presented in this paper are a first quantitative step to characterize both modal and chemical compositions of pyroxenes and olivines. A demonstration of the methodology on specific examples of binary and ternary olivine-pyroxenes mixtures has been made, which shows that the different non-linear effects which affect the Gaussian parameters (center and strength) can be successfully handled. Of note is the fact that the band center positions associated with the different mafic minerals are not set here in the inverse problem, and thus the MGM outputs are truly informative of the chemical composition of pyroxenes and olivines. With the consideration of some limits on the detection thresholds, these results are quite promising for increasing the operational use of the Modified Gaussian Model with large hyperspectral data sets in view of establishing detailed mineralogical mappings of magmatic units.