Assessing macro- (P, K, Ca, Mg) and micronutrient (Mn, Fe, Cu, Zn, B) concentration in vine leaves and grape berries of vitis vinifera by using near-infrared spectroscopy and chemometrics

Macronutrients (phosphorus, potassium, calcium, and magnesium) and micronutrients (manganese, iron, copper, zinc, and boron) play an essential role not only in the general physiology of vines but also in the quality of wine produced. The quantity of each nutrient in the vine is generally determined by analyzing the leaf blades or petioles, but this approach imposes a typical delay of two weeks between sampling and receiving the results, which precludes real-time detection of nutritional deficiencies (e.g., boron deficiency at flowering). Therefore, a method to rapidly analyze vine organs is highly desirable. One candidate for such a method is near-infrared (NIR) reflectance spectroscopy coupled with chemometric methods, based on which winegrowers have already developed prediction models. This approach is widely used today in agriculture. The aim of the present study is to determine whether NIR spectroscopy can be used to obtain accurate information about the nutritional status of vines. In this study, we focus on the mass of phosphorus (P), potassium (K), calcium (Ca), magnesium (Mg), manganese (Mn), iron (Fe), copper (Cu), zinc (Zn), and boron (B) contained in different vine organs (leaf blades, petioles and berries) over the course of a year. The concentration of these elements was determined based on NIR absorbance spectra from 677 samples of various dried vine organs. Partial least square models for classification and prediction were then developed based on raw and pretreated spectra for each organ, following which the models were tested on an external prediction set. The results show that, for Ca and Mg, all organ models can be used routinely for classification or prediction. For prediction, the Ca (Mg), model produces r² = 0.88, 0.70, and 0.72 (0.60, 0.72, and 0.80) for leaf blades, petioles, and berries, respectively. Only for leaf blades (berries) is the Ca (Mg) model sufficiently accurate to be used for prediction. For berries, the P, K, and Zn models produce r² in prediction of 0.77, 0.79, and 0.82, respectively. For petioles, the K model proves reliable for prediction, with r² = 0.76. The Fe, Cu, and B models produce r² = 0.72, 0.71, and 0.52, which are suitable for classification but not for prediction. Finally, for leaf blades, the Fe and Cu models produce r² 0.58 and 0.61, respectively, in prediction and thus can be used routinely for classification.