Recognition of the geographical origin and sex of Patagonian forests of Araucaria araucana through the analysis of leaves by FT-NIR: a methodological approach

MOLINA, J.G.A.; FRANCO, M.P.; FRANZOL, S.D.; HADAD, M.; TOMAZELLO FILHO; MORENO, C.; ROIG JUÑET, F.A.; CHAIX, G.

Conferencia; 17th International Conference Near Infrared Spectroscopy; 2015

In the frame of this study, and because Araucaria araucana is a monoic tree species we try to use FT-NIR to detect signals in the spectrum related to both provenance (humid/xeric environments) and sex from direct lectures of NIR on leaves of adult trees. We grouped the samples in 3 groups according to the climate (from dry to humid conditions). For the adult trees we noted the sex of each tree when the flower and/or fruit were visible. Leaves were kept in cold and dry environment during the transport and were dehydrated by lyophilization in the lab. The absorbance spectra were obtained from Bruker MPA spectrometer by applying FT-NIR both by integration sphere (the leaves for each tree were putted in rotate cup) and with optic fiber (we measured one spectrum for each leave on inferior face). To verify and limit the anatomical and structural effects on NIR information, we compressed grounded leaves for each tree and we measured with integration sphere one spectrum for each pastille. We used the Unscrambler software 10.3 (Camo, AS, Norway) for NIR data analysis. Linear Discriminant Analysis (LDA) was used to classify leaf materials according their provenance and sex of the trees.For provenance discrimination for any type of material and/or measurement methodology, we obtained a good level of classification when using optic fibre, rotate cup or pastille samples, attaining 78, 95, and 88% respectively (Table I).For sex discrimination, the percentages of good classification were 75, 87, 81% when we used optic fibre, rotate cup (intactleaves) and pastille samples (grounded leaves), respectively (Table II). The best results were obtained using the integration sphere, because this methodological variant may result in a mean value relative to the whole leaves for each tree with rotate cup. The restricted information due to the smallsurface measured by the optic fibre could explain the lower accuracy of the LDA model. The NIR data obtained from pastille samples, lacking information of the leaf structure and anatomy, gave intermediate results. However, the resultedgood percentage confirmed our hypothesis on chemical differences due to region/forest provenances and sex of the trees. More sites should be included to confirm these preliminary results.