Title

Advantages of first-derivative reflectance spectroscopy in the VNIR-SWIR for the quantification of olivine and hematite

Source of Publication

Planetary and Space Science

Abstract

© 2020 Elsevier Ltd The focus of this paper is to study the application of the first order derivative method for the estimation of minerals rates in different mineral mixtures. The primary goal with this is to find robust spectral features of specific minerals that are not severely influenced by the spectral features of the other minerals in a mixture. Results were used to select appropriate spectral features to be applied for quantifying the minerals in upcoming studies. Mixtures of different terrestrial minerals equivalent to those dominating the Martian surface with a grain size <0.25 ​mm were prepared and analyzed in the laboratory by reflectance spectroscopy in the VNIR-SWIR range. The first derivatives were computed and correlated with the mineral concentration at specific wavelengths using the Unscrambler X software. The results indicated the first derivatives near 2300 ​nm, that is a characteristic absorption feature of olivine rich in magnesium and iron silicate, correlate strongly to the olivine content for all the mixtures containing olivine, binary and ternary, with regression coefficients ranging between 0.93 and 0.98. Additionally, the main advantage of this work is that first derivative spectra of mixtures with different olivine ratio highlights in the overlapping regions of the spectra the wavelengths where the first derivative values correlate strongly to the amount of olivine in the mixtures. The region near 1050–1300 ​nm was identified as a promising one for hematite-olivine mixtures and 785–900 ​nm for magnetite olivine, with a regression coefficient mean of 0.97 and 0.98, respectively. The study of hematite-plagioclase mixtures demonstrates that wavelengths near 785–858 ​nm and 940–989 ​nm lying within the overlapping regions of hematite and plagioclase exhibit robust correlation to the hematite content with a regression coefficient mean of 0.98 for both areas.

Document Type

Article

Publication Date

9-1-2020

DOI

10.1016/j.pss.2020.104957

Share

COinS