Prediction of the modulus of elasticity by Near Infra-Red Spectroscopy (NIRS) and direct measurement by ultrasonic device from cores of Liquidambar (Liquidambar styraciflua (L.)) from Madagascar

Standardized methods of wood modulus of elasticity (MOB) evaluation are destructive and require many replicates. To overcome these constraints, the research of new technologies to characterize the wood properties has been focused towards tools such as NIRS and ultrasonic measurement (Bailleres and al. 2002; Hein and al. 2010; Brancheriau, 2013). Thus, the aim of this study is to compare two measurement methods of the MOB using ultrasonic device (Method 1) and NIRS (Method 2) applied to cores of wood from standing trees. Liquidambar was used, as this timber has a rather un-conventional behavior (e.g. its MOB is not correlated with its density). 292 samples (360x20x20mm3) were extracted on 54 diametrical planks. The MOB was measured by acoustic method BING® (Brancheriau and Bailleres 2002). NIRS calibration model was established using spectra measured on these samples and regressed by Partial Least Square (PLS) with MOB. Meanwhile, 30 long samples (360x20x20mm3) were used for correlation between MOB estimated by ultrasonic device and by BING® respectively. From the middle of each long sample, a small sample (10x20x20mm3) was cut and used in the same way as described above. To confront the value of the MOB predicted by NIRS model (292 samples) and the MOE measured by ultrasonic device (30 small samples), 33 diametrical cores of Liquidambar were extracted from each-tree. The MOB estimated by BING with the 30 long samples and the MOE' measured by ultrasonic device with the 30 small samples are not statistically different (r =0.72,p<0,001). The NIRS PLS regression from the 292 samples showed a model with r2=O.8 and Ratio Performance Deviation (RPD) 2.57 for MOE. To evaluate the MOB values obtained with the method 1 and 2, linear regression has been computed with R using 61 7 measurements from pith to the bark on the 33 cores. Linear regression of the MOE predicted by NIRS on cores and the MOE measured with' ultrasonic device on the same core are closely linked (r2 =0.68, p<O.OOI).This work showed that it is possible to predict the modulus of elasticity from cores and whose values are in the same range as those obtained by vibration method (BING®). This work demonstrated the reliability and advantages of the methods 1 and 2, for 'conducting non-destructive MOB evaluation. These two methods can be used to measure the radial variability of the MOB even in core sampling, allowing then to preserve the standing trees (Résumé d'auteur