PNNL

Abstract

Scattering artifacts adversely affect infrared reflectance measurements of powders and soils, and extended inverse scatter correction (EISC) is a flexible method useful for correcting for these artifacts. EISC was used to correct mid-infrared reflectance spectra of two different soils coated with dibutyl phosphate and the results were examined using regression analysis. To determine the correction, EISC fits a measured spectrum to a reference spectrum. However, if measured spectra contain features not included in the reference spectrum the fit can be biased resulting in poor correction. Weighted and robust least squares were used to account for these potential biases. Additionally, the present work demonstrates how analyte-free samples can be used to determine basis functions for an extended mixture model used in the correction. Corrected spectra resulted in partial least squares models that performed at least as well as 2nd derivative spectra and were more interpretable.