Near infrared spectroscopy (NIRS) has been shown effective as a tool for identifying Swietenia when tested as laboratory-processed powder, but testing such powdered wood is not readily adaptable to the fieldidentification of wood. This study explored the efficacy of a fiber optic NIRS scan of solid wood surfaces to separate Swietenia macrophylla King, Carapa guianensis Aubl., Cedrela odorata L., and Micropholis melinoniana Pierre. Transverse, radial, and tangential surfaces were scanned to determine if the surface from which data were collected influenced the spectra recorded. Surfaces were scanned before and after removing the oxidized surface layer of the blocks to test effects of exposure on the spectra. Partial least squares for discriminant analysis models were developed for each taxon separately, based on a calibration set composed of at least 67 samples and a test set with at least 45 samples. The anatomical surface scanned, but not the presence of an oxidized layer, influenced the spectra for each species, necessitating the comparison of the same planes of section. The discriminant models showed small errors for each species, indicating that reliable identifications can be made with NIRS of solid wood surfaces in these species.

The resistance to decomposition of mahogany wood (Sweitenia macrophylla King) ranges from high to moderate level. Wood extractives, mainly due to the presence of phenol compounds are related to the natural durability of wood. The technique of near infrared spectroscopy (NIRS) coupled with multivariate analysis has been applied to assess the extractives and phenols of 41 samples of mahogany in powder form. The hot water-soluble extractives were quantitatively determined, and total phenol content was measured with the Folin-Denis colourimetric reagent. Models were developed with the NIRS data for each of the two variables. The results indicated that NIRS can be a useful tool to a rapid evaluation of the extractive contents and total phenolic compounds of mahogany wood. The method was able to predict the interesting properties with errors lower than 10% (w/w) and had the capability of detecting samples that have a minimum concentration of 2.4% (w/w) of extractives and total phenolic compounds, respectively. (Résumé d'auteur