Résumé
Dalbergia and Diospyros species are appreciated for the quality of their wood. They are illegally logged and traded due mainly to the difficulty of their identification. Current used identification technics can not reach species level for the case of some Dalbergia and Diospyros species. Thus, the aim of this study is to assess the potentiality of a portable Near InfraRed (NIR) spectrometer to discriminate some precious woods of Madagascar. NIR Spectroscopy, compared to others identification methods, is more suitable for field use because it does not need any specific scientific knowledge from users, it needs no sample preparation and it is low cost. Absorbance spectrum were acquired from 72 corewood samples belonging to the two genera (Dalbergia and Diospyros), and four Dalbergia species (D. chlorocarpa, D. abrahamii, D. greveana, and D. pseudobaronii). Several spectra preprocessing methods were tested and the best methods were used. Discrimination models were established using Partial Least Square Discriminant Analysis (PLSDA) method. Models were established to discriminate the two genera and to discriminate Dalbergia species. There are also models to discriminate species based on spectra from heartwood and that from sapwood. Results show that better discrimination can be obtained from spectra of heartwood. With a test set validation, the best models give a high classification rate of 93% for the discrimination of the two genera, and a high classification rate of 80% for the discrimination of Dalbergia species. This work highlighted the ability of the MicroNIR Spectrometer to be an efficient helpful tool to monitor timber trade. Deeper study should be carried out to consider more samples for each species from several origins and to use a wider spectral range to help for a legal and sustainable trade of these species.