Résumé
Madagascar's precious wood (Dalbergia and Diospyros) are highly valued and heavily traded products due to various interesting intrinsic qualities, which conducted to an important recrudescence of an illegal logging in recent years. Indeed, identifying these woods in field is challenging without their reproductive organs and this often leads to error permits issues. Developing an identification tool that is both simple and rapid could be the key solution. Recently, Near Infrared (NIR) spectroscopy has emerged as a powerful method for species discrimination, enabling rapid and multiple analysis. A low cost NIR device : NIRscan Nano spectrometer from Texas Instrument, suitable to the economic context of Madagascar was chosen for this study. To demonstrate its efficiency, it has been compared to another performant portable NIR device : VIAVI MicroNIR 1700. The main difference of the two devices are detector and signal noise ratio. To achieve this, wood samples (53 samples of 6 Dalbergia species and 74 samples of 6 Diospyros species) were collected in the field, stabilized at a relative humidity of 12%, and their spectra were acquired simultaneously with the 2 devices. The spectra were pre-treated with combinations of chemometric pre-treatments. ¾ of the samples were used for cross validation with PLS-DA and the best model issued for each instrument was selected which afterward validated with the remaining 1/4 of the samples in an independent validation. It has been demonstrated that the identification rate obtained by the two spectrometers is comparable. The models predicted over 55% of all Dalbergia samples and over 65% of all Diospyros samples. The lower identification rates are due to the limited wavelengths, which do not capture some infrared information. However this tool can be used as a field preliminatory prediction method for wood control, reducing part of the time consuming steps in lab analysis.