Variabilité de la densité du bois de Ravenala madagascariensis Sonn. Prédite par la spectroscopie proche infrarouge.

Near infrared spectroscopy (NIRS) is a promising non-destructive method for wood analysis. In this study, the effectiveness of SPIR in predicting the wood density (WSG) of Ravenala madagascariensis, an endemic non-woody species of Madagascar, was evaluated. To do this, radial samples were taken from 3 compartments (stipe, leaf sheath, leaf), from 120 trunks of Ravenala madagascariensis collected from 3 sites. Near infrared spectra were taken from the samples. The results show that the optimal prediction model combines a "SNV (standard normal variation) + DT (detrend)" pre-treatment and uses 11 latent variables. The parameters of this model are : RMSEcv of 0.013 g.cm-3, R²cv of 0.73 and RPDcv of 2.76. In independent validation, the model showed an R² of 0.70 and an RPD of 2.17, with 11 numbers of latent variables. Application of the predictive model revealed significant radial variability in WSG in Ravenala madagascariensis. The central zone has a lower density (mean of 0.082 g.cm-³) than the peripheral zone (0.12 g.cm-³), with a highly significant difference (threshold >0.1%). In addition, there was a significant interaction effect between the radial part and the compartment on WSG. No significant effect was observed for the radial part × site interaction. This study provides valuable information on the wood properties of this endemic species, improving our understanding of its ecological and physical importance.

Key-words: Near InfraRed Spectroscopy, prediction, WSG (wood specific gravity), radial variability, Ravinala

02/05/2024