Contribution of Near Infrared Spectroscopy to identify CITES Appendix II woods: case of four Dalbergia species from Madagascar

Contribution of Near Infrared Spectroscopy to

identify CITES Appendix II woods: case of four

Dalbergia species from Madagascar

CONTEXT

Core sampling:

54 wood cores

Reference values:

Genus and species name (determined by botanical method)

NIRS measurements (950 nm - 1650 nm)

A total of 3 absorbance spectra were collected for each wood type (sapwood and heartwood) by using a VIAVI MicroNIR device. Then, these 3 spectra were averaged to one average spectrum.

Spectral mining and preprocessing :

- Principal Component Analysis, outliers identification

- Standard Normal Variate, Detrend, Derivation of Savitzki-Golay

Discriminantion models calibration (for sapwood and heartwood) :

- Discriminant analysis

(PLSDA) and cross

validation

- Identification of the number of Discriminant Variables

External validation:

- Identification of external samples

- Evaluation of discrimination models through the percentage of misclassifion

Field work

Laboratory work Calibration set (37)

Validation set (17)

OBJECTIVE

Precious wood of Madagascar, which includes rosewood, palissander and ebony tree species are seriously threatened because of the intensification

of their illegal logging and trade.

This frightening situation is worsened by the difficulty of species identification.

Need a tool to help control agent to identify precious wood.

This study aims to investigate the potentiality of using a Micro-nir spectrometer to discriminate Malagasy Dalbergia and Diospyros species.

RESULTS

First Derivates preprocessing of the spectral data improves the performance of the discrimination models.

The discrimination of the two genera (Dalbergia and Diospyros, Table 1, 2) gives better results. Discrimation models calibrated from the heartwood are more efficient than those which are calibrated with the sapwood (Table 1). For the best models, misclassification rate for the sapwood and heartwood models of the external validation step are respectively 12.9 % and 7.6% for the two genera, and 36.3 % and 20% for the four Dalbergia species (Table 1). Discrimination models of the four Dalbergia species show an encouraging result with an error rate of 20%, even if the number of samples is low.

Reference values Dalbergia Diospyros Pr edic te d va lu e s Dalbergia 23 2 Diospyros 1 Reference values

D. abrahamii D. chlorocarpa D. greveana D. pseudobaronii

Pr edic te d va lu e s D. abrahamii 3 1 D. chlorocarpa 3 1 D. greveana 1 D. pseudobaronii 1

CONCLUSION AND PERSPECTIVES

This first use of the Near InfraRed tool to discriminate malagasy precious woods brings an innovation in the field of wood science. Results highlight the potential of the portable MicroNir in the identification of Malagasy woods. Spectra measured on the heartwood allow to separate more efficiently the species, probably because of the exractives (Jebrane, 2009) which influence the absorbances. However, additional works should be carried out in order to use this tool effectively in the identification of Malagasy precious wood. Number of test samples per species should be increased (20 to 30 samples, Agelet and Hurburgh, 2010), and all existing Dalbergia and Diospyros species in Madagascar should be considered to meet CITES requirement.

Discrimination of the 2 genera Preprocessing Sapwood Heartwood

Derivation 12.9 % 7.6 %

Discrimination of the 4 Dalbergia species Preprocessing Aubier Duramen

Derivation 36.3 % 20 %

Table 1 : Percentage of misclassified

samples of the external validation step

Table 2 : Confusion matrix of the best discrimination models during the external validation step

References

Agelet, L.E. & Hurburgh, C.R. (2010), "A tutorial on near infrared spectroscopy and its calibration."

Critical Reviews in Analytical Chemistry, 40:246 260.

doi:10.1080/10408347.2010.515468 Jebrane, M., (2009),

Fonctionnalisation chimique du bois par transestérification des esters d’enol. Thèse de Doctorat - spécialité

Chimie Organique - Université Bordeaux 1. 22:53–78.

2

3

4

5

limit of the study areas

wavelength Ab so rb an ce

MATERIALS AND METHODS

Species name Number of  samples Dalbergia abrahamii 9 Dalbergia chlorocarpa 14 Dalbergia greveana 7 Dalbergia pseudobaronii 8 Diospyros sp. 16

1

Andry Clarel RAOBELINA1_{, Andriambelo Radonirina RAZAFIMAHATRATRA}1_{, Gilles CHAIX}2, 3, 4_, 

Tahiana RAMANANANTOANDRO1

1_{Université d’Antananarivo, ESSA‐Forêts,}  2_{ESALQ‐USP, Piracicaba, Brazil,} 3_{CIRAD, UMR AGAP, Montpellier, France} 4_{AGAP, Univ Montpellier, CIRAD, INRA, Montpellier SupAgro, Montpellier, France} 

[email protected]

Discrimination of the four Dalbergia species (1

_{derivate, 7 VDs)}

Discrimination model of the two genera 

(1

_{derivate, 9 VDs)}