Assessing macro-elements contents in vine leaves and grape berries of Vitis vinifera using near-infrared spectroscopy coupled with chemometrics

Abstract

Context and purpose of the study – The cultivated vine (Vitis vinifera) is the main species cultivated in the world to make wine. In 2017, the world wine market represents 29 billion euros in exports, and France contributes 8.2 billion (28%) to this trade, making it a traditional market of strategic importance. Viticulture is therefore a key sector of the French agricultural economy. It is in this context that the nutritional diagnosis of the vine is of real strategic interest to winegrowers. Indeed, the fertilization of the vine is a tool for the winegrower that allows him to influence and regulate the quality of the wine. Nowadays, nutrition analysis is made with CHNS analyzer for elemental particles, and mass-spectroscopy for macro and microelements. Such methods are destructive and time consuming, then results could be obsolete for the vine grower. Near-infrared spectroscopy coupled with chemometrics tools allows to developed models of prediction that can provide accurate information about nutrition status of the vine in the field. In this study, we concentrate on the relative amount of Carbon [C], Hydrogen [H], Nitrogen [N], Sulphur [S] in dry matter (DM) and the C:N ratio.

Material and methods – 252 samples of different organs (leaves blade, leaves petioles, pea sized berries and berries at véraison) of 4 varieties (Muscat, Chasselas, Négrette and Sauvignon blanc) were analyzed. Spectrum were taken on both fresh material and dried ones with a reflectance spectrometer. The spectra were pre-processed using multiple scatter correction (MSC) and 1st and 2nd order Savitsky-Golay derivative (D1 and D2), before developing the cross-validation models using partial least square (PLS) regression and test it on a prediction set.

Results – The coefficient of determination in prediction (r²), the roots mean square error of prediction (RMSEP) and the ratio of performance of prediction (RPD) were obtained for C (0.49, 14.6% of DM and 1.33 on fresh material with MSC, 0.45, 15.4% of DM and 1.26 on dry material with MSC), H (0.56, 1.71% of DM and 1.45 on fresh material with D1, 0.49, 1.88% of DM and 1.32 on dry material with D1), N (0.91, 1.12% of DM, 3.32 on fresh material with raw spectra, 0.95, 0.84% of DM and 4.39 on dry material with MSC), S (0.47, 0.319% of DM and 1.31 on fresh material with MSC, 0.46, 0.322% of DM and 1.30 on dry material with D2) and C:N ratio (0.85, 8.20 and 2.58 on fresh material with raw spectra, 0.87, 7.55 and 2.80 on dry material with D2). Results show that the near-infrared reflectance spectroscopy can be used to assessing the level of nitrogen nutrition in vine and the C:N ratio. All model performance could be improved by increasing the number of samples.