Use of NIR spectroscopy and partial least squares regression for prediction of chemical properties of Salix clones

Abstract

The chemical wood properties characterization of biomass feedstocks is important since it determines their potential for various chemical and energy applications. Willow (Salix sp.), a short-rotation hardwood species, is an interesting feedstock for such applications. It can represent an alternative raw material for non-sustainable resources, long-rotation trees, and agricultural biological materials. This research aimed to characterize the chemical properties of different wood tissues of willow and to develop a nondestructive method for their prediction. Wood, bark, and twig samples were sampled from six willow clones (Alberta, Canada). Extractives, cellulose, hemicellulose, and lignin contents were measured in sampled tissues by destructive chemical analysis tests using TAPPI standard test methods. The surface chemistry of the samples was analyzed by near-infrared spectroscopy. Partial least squares regression models were developed for quantitatively predicting wood chemical components. The coefficients of determination (R² calibration, R² cross-validation, and R² prediction) ranged from 0.66 to 0.99 for extractives, lignin, cellulose, and hemicelluloses. Carbohydrates (glucose, xylose, arabinose, and mannose) ranged between 0.98 and 0.99, demonstrating excellent predictive capability for these components. These results confirm the applicability of the developed models to predict the chemical properties of willow biomass. Furthermore, the prediction models apply to all investigated tissues, emphasizing consistency. These results provide the possibility to characterize the chemical properties of willow biomass using easily implemented, cost-effective, nondestructive, and rapid NIR equipment.