Discrimination of amazonian forest species by NIR spectroscopy: wood surface effects

The need for fast and reliable methods to determine wood-producing species at risk of extinction in tropical forests has become increasingly evident. The aim of this study was to evaluate the potential of near infrared (NIR) technology to identify these types of wood species in Amazon forests by developing multivariate models for the spectral signatures of wood specimens processed by circular saws and chainsaws. Three trees of six species from the Legal Amazon were evaluated (Manilkara elata, Dinizia excelsa, Goupia glabra, Hymenaea sp., Micropholis melinoniana and Copaifera sp.), of which 350 wood specimens were cut with chainsaws and circular saws. The processing-type effect on the information quality recorded on wood surfaces was evaluated in terms of the performance of the classification of tropical wood species through cross validation and independent sets based on partial least squares-discriminant analysis (PLS-DA). The results indicated that PLS-DA models based on spectral signatures recorded on wood surfaces processed with a circular saw achieved a higher percentage of correct classifications (99.2%), while models based on spectral signatures taken from wood surfaces processed with a chainsaw correctly classified 95.2% of the specimens. This study suggests that unknown wood samples belonging to these species can be correctly and satisfactorily preclassified regardless of the tool used for sampling the wood before identification, inspection and commercialization operations.