Comparison of calibration models for rapid prediction of lignin content in lignocellulosic biomass based on infrared and near-infrared spectroscopy

Abstract

Lignocellulosic biomass has great potential as a renewable energy source due to abundance both as vegetation and in residues from agriculture and forestry. It can be utilized for biofuels and for value-added chemicals and materials. The challenge concerning lignocellulosic biomass lies in its heterogenous nature. The chemical composition may vary according to species, location, harvest season and botanic fractions. Therefore, it is crucial to assess the composition prior to any biofuel conversion method. Vibrational spectroscopy has been used extensively for rapid predictions of the chemical properties of biomass. In this study, calibration models based on infrared and near-infrared spectra has been compared using the same calibration set, 36 samples comprising wood, bark, needles, and twigs of three different wood species. PLS was performed to correlate EMSC pretreated infrared and near-infrared spectra to the chemical contents of the samples. R2 for the standard curve of the infrared model is 0,896 and for the near-infrared standard curve 0,921, using 3 PLS components. The effect of heterogeneity was tested by comparing calibration models based on finely and coarsely ground sample, where the R2 of the coarsely ground sample was 0,825, lower than 0,896, but still significant. The results also show that there is more variation in the lignin content between the fractions from a single tree than between similar fractions from different species. The calibration models that have been developed will be useful for frequent, rapid determination of lignin content in wood biorefining feedstocks.