Diffuse Reflectance, Near-Infrared Spectroscopic Estimation of Sugarcane Lignocellulose Components—Effect of Sample Preparation and Calibration Approach

Abstract

Near-infrared spectroscopic calibrations for sugarcane lignocellulose composition (ash, lignin, and structural carbohydrates) have been developed on a total dry weight and extractives-free dry weight basis. Reference spectra were measured from crude preparations of bagasse and fibrated sugarcane stalks. Spectral interference attributed to extractives in the crude preparations adversely affected the fitting of calibrations derived from extractives-free reference data. Extractives content varied between 5 and 68 % of total dry weight and depended on the nature of the sample preparation process. Applying DOSC (direct orthogonal signal correction) to the spectra circumvented the need to physically remove these extractives from the samples prior to spectroscopic measurement. DOSC enhanced the correlation between the extractives-free reference data and the measured absorbances. Meanwhile, the correlation between the extractives reference data and the measured absorbances was attenuated. This indicated that DOSC removed the spectral contribution of the extractives. This mathematical treatment of the spectra improved the cross-validation performance of the calibrations for extractives-free lignin and some structural carbohydrates (arabinan, glucan, and xylan). The DOSC-based models explained greater than 90 % of the variance in the calibration data and led to relative standard errors of prediction under cross-validation and independent testing that were less than 10 %. Although DOSC also improved the cross-validation performance of ash and galactan calibrations, their relative standard errors under independent testing were high (>10 %). DOSC was also advantageous for total dry weight calibrations as it reduced the number of latent variables employed in the models.