Pretreatment of near Infrared Spectral Data in Fast Biomass Analysis

Abstract

The ability to rapidly evaluate the chemical composition of biomass feedstock for purposes of process monitoring and optimisation is useful for gauging the potential applications and value of such feedstocks. Near infrared (NIR) spectroscopy, coupled with multivariate analysis and data pretreatment, was evaluated to remove interference from physical heterogeneity that could mask chemical property responses. Pretreatment methods included standard normal variate (SNV), multiplicative scattering correction (MSC), 1st derivative with the Savitzky-Golay algorithm (1st derivative), 2nd derivative with the Savitzky-Golay algorithm (2nd derivative), extended multiplicative signal correction (EMSC) and combinations of 1st derivative/2nd derivative with SNV. Results indicated that, of these methods, EMSC was most effective for diffuse reflectance NIR analysis of lignocellulosic biomass. The EMSC-pretreated data not only best accessed the chemical similarity of the probed feedstocks in our hierarchical cluster analysis but also consistently led to the overall best prediction of the chemical composition of the biomass.