3.1. Basic Parameters and NIR Analysis
Six samples of wine produced with pure or mixed fermentation from each of the three varieties were analyzed by NIR. In the unprocessed NIR spectra the main wine compounds, i.e., water and ethanol, dominate the spectrum and often overshadow minor compounds with similar functional groups, which show peaks in the same regions. Absorption bands with high absorption units (e.g., some water signals) are automatically cut out by the instrument. They should not be used since they are characterized by very small light intensities, and the resulting signals contain more noise [20
]. NIR spectra of fruits, vegetables, and their juices, with a high percentage of water, show absorption bands related to water around 6800 cm−1
and at 5000 cm−1
]. Water is the main molecule responsible for those signals, but it is not the only one. Other molecules with O-H groups contribute to the same (or very close) area of the spectrum, e.g., around 5000 cm−1
of glucose and ethanol also show peaks [22
The grouping in the PCA plot obtained from the spectra of all the three varieties (Figure 1
A) shows that Primitivo samples are grouped on the left side, while it is not possible to separate samples from the other two varieties. It is clear that samples are grouped by the variety and not by the different yeasts employed in the winemaking. The loadings (Figure 1
B) of this PCA analysis show how the wavelengths mainly responsible for the variance in the data set are around 7000 cm−1
, 5300 cm−1
, and 488–4200 cm−1
, which correspond respectively to the first overtones of COH groups, the first overtones of CH2
groups, and COH combination (stretching and bending) vibrations. In the NIR spectra, there was a strong overlap of signals of functional groups belonging to different molecules. Thus, a specific attribution is not an easy task. Nevertheless, the strong signal around 700 cm−1
can be mainly attributed to OH vibrations of two of the main compounds found in wine, which are water and ethanol. In the spectra of pure ethanol, it was observed that the absorbance changes with the ethanol concentration [9
]. Indeed, the three wines showed very different ethanol contents (similar for Negroamaro and Aleatico and higher for Primitivo) which might explain the grouping in the PCA plot. Clearly, due to this strong difference in one of the main components of the samples, ascertaining the differences in terms of other (minor) compounds was not possible. These hypotheses were confirmed by the groupings found in the PCA performed using chemical analysis data as variables (Figure 2
), in which Primitivo samples were grouped together due to their acidic composition and higher content of ethanol. Indeed, the signals linked to the main acids found in wine (tartaric, malic, acetic and lactic acid) in a NIR spectrum are too close to those of ethanol, which overshadows them with its higher content.
In order to minimize the influence of ethanol and water, we decided to investigate each variety on its own. The NIR analysis of the Negroamaro samples, with close attention to ethanol content, was able to discriminate between the wines obtained with S. cerevisiae
and those obtained with non-Saccharomyces
A). The difference between the NIR and chemical PCAs of Negroamaro suggest that the variables employed to build the chemical data PCA (Figure 4
) were not responsible for the grouping found in the NIR PCA. Negroamaro samples have similar ethanol content, thus, even if the loadings of NIR PCA are in similar regions of the spectra (Figure 3
B), compared to those of NIR PCA of all the samples together, the loadings in the regions 6500–7000 cm−1
and 4500–4300 cm−1
can be attributable to compounds other than ethanol. We attempted an interpretation based on the known NIR absorbance signals of metabolites commonly present in wine, focusing only on the NIR spectral regions responsible for wine differentiation (the regions with high loadings on PCA axes). It has been previously reported that wines from non-Saccharomyces
produce different metabolites (e.g., are found to be more fruity) or produce them in different concentration (e.g., glycerol) [23
]. Indeed, in the NIR PCA loadings, important spectral regions are those related to glycerol (4400, 5300, 7200 cm−1
]. This led us to hypothesize that glycerol content might be one of the factors responsible for the differentiation found between pure and mixed fermentation samples.
For Aleatico nero wines, there is not a clear separation between pure and mixed fermentation samples in the PCA plot obtained with NIR data (Figure 5
), while in the PCA plot built with chemical data, the pure and mixed fermentation samples are clearly separated (Figure 6
). Here, ethanol content is again not the main variable responsible for sample differentiation. Despite the specific nature of compounds responsible for differences among wines, it was not possible to discriminate between pure and mixed fermentation wines with NIR spectroscopy.
In both the PCAs of the Primitivo samples, there is not a clear separation between pure and mixed fermentation (Figure 7
and Figure 8
). Indeed, even considering only the chemical data, there is not a net difference among all the monovarietal Primitivo wines. The numerosity of samples is a key point in NIR analysis. As for Aleatico nero wine, a larger number of samples could have improved the separation among wines and helped to identify differences. In order to understand if these results were due to a lack of discriminant capacity of the method employed or rather due to a lack of significant differences in wines composition, we investigated the perceived difference among pure and mixed fermentation wines by performing a sensory analysis.