Abstract:
The proportions of lignin, and the structures and proportions of the polysaccharides that occur in plant cell walls vary with cell type and taxon. Monosaccharide compositions of cell walls are often determined by first acid hydrolysing the cell-wall polysaccharides, for example by a two-stage sulfuric acid hydrolysis, to release the component monosaccharides, which are then separated and quantified. In two-stage sulfuric acid hydrolyses of cell walls, Klason lignin forms an acid-insoluble residue that can be dried and weighed giving a value for lignin content. However, these methods are time-consuming and laborious. As a consequence, I investigated the use of mid- and near infrared (IR) spectroscopies, coupled with the multivariate data analytical technique, partial least squares (PLS) regression, to predict the lignin contents and monosaccharide compositions of Pinus radiata wood cell walls. IR spectroscopies are rapid, non-destructive, analytical techniques that can be used for compositional analysis of cell walls, but sample preparation and the development of a suitable training set remains the most time-consuming aspect of analysis using these techniques. In this study, a training set with a wide linear range of lignin contents and monosaccharide compositions was established and the wood cell-wall components were determined using a two-stage sulfuric acid hydrolysis and high-performance anion-exchange chromatography with pulsed amperometric detection (HPAEC-PAD). This training set was used to build PLS-1 models (where variables are modelled separately) that successfully predicted the compositions of cell walls in both milled and solid wood. I examined samples of milled wood with two different particle sizes, and of solid wood with two different surface finishes. The milled and solid wood samples were analysed in both dry and ambient moisture conditions. Of the samples prepared for this study, one type of each preparation was easier and quicker to prepare: 'large' (< 0.422 mm) particles of milled wood were easier to prepare than 'small' (< 0.178 mm) particles; solid wood with a 'rough' surface finish was easier to prepare than solid wood with a 'smooth' surface finish; and both milled and solid wood samples in ambient moisture conditions were quicker to prepare than those in dry conditions. Using these samples, I investigated the effects of particle size and moisture content of milled wood on the PLS-1 models and subsequent predictions. I also investigated the effects of surface finish and moisture content of solid wood by using PLS-1 models built with spectra of milled wood to predict the compositions of solid wood. Overall, the study showed that although particle size and moisture content have small effects on both the PLS-1 models and subsequent predictions, and surface finish has a small effect on predictions, the wood samples that were the easiest to prepare gave accurate compositional predictions. These samples were ambient 'large' particles to build PLS-1 models for compositional predictions of both ambient 'large' particles and ambient solid wood with a 'rough' surface finish.