Natural durability, ethanol-toluene extractives and phenol content prediction of eight wood species from Madagascar using NIRS multispecific models

Madagascar has nearly 4000 species of trees and shrubs, but wood properties of only 200 species have been studied. Some properties, such as the natural durability or chemical composition are of importance for adequate and optimal use of these timber species. Since durability tests take long time and chemical analyzes can be very expensive, alternative methods such as near infrared spectroscopy (NIRS) and calibration, could be used to characterize these properties. Thus, the objective of this study is to analyze the NIRS potential to predict natural durability and chemical composition for eight (8) species from rainforest of eastern Madagascar. Natural durability tests in the laboratory were conducted on both leached and un-leached wood samples. Two types of fungi (Coniophora puteana and Coriolus versicolor) were used. The chemical composition analyzed were ethanol-toluene extractives and phenol content. The NIR spectra were acquired with a microNIR spectrometer (spectral range: 950-1650 nm, spectral resolution: 6.2 nm) on the solid unleached wood samples. Models were established with PLS regression, with wavelength selection, using the best pre-processing method and tested with repeated cross-validation method. Results showed that for each property, samples can be considered to have enough variability to allow the establishment of a good prediction model. Established models are evaluated satisfactory to good, with R2CV always higher than 0.69. Concerning the model for phenol content, R2CV is 0.89. For mass loss, R2CV ranges from 0.71 to 0.79. These models can be used to predict wood properties and can be improved by including new samples. In the future, it would be interesting to analyze these reference samples but using a more efficient spectrometer with a wider spectral range to establish NIRS models