Further classification of eucalypt pulpwoods using principal components analysis of near-infrared spectra

Models derived from principal components analysis (PCA) of the near infrared (NIR) spectra of eucalypt woods have been shown to be a good means of classifying the woods. However they do not give numerical measures of the distances between the classes, In this report the soft independent modelling of class analogy (SIMCA) approach is used to determine distances between the PCA models. The NIR spectra and pulp yields used in the models were derived from samples of woods from two provenances of each of E. globulus and E. nitens and one of E. grandis grown on four sites in the Esperance Valley in Tasmania, Some further E. globulus samples were a mixed age set from various Tasmanian provenances of native forest origin, The distances between models for species tended to be greater than the distances between models for provenances of the same species and age. The model for older aged native forest woods from E. globulus was a large distance from the model of the same species grown in plantation. The compliance of the data with the models was checked graphically and the discrimination and modelling powers of the NIR data determined across the whole range (1100-2500 nm).