Measuring the levels of phenolic compounds during red wine fermentations is a challenging practice.
Phenolic compounds are key components of red wines. Anthocyanins are the group of compounds that give red wine its colour. The levels and composition of these molecules will define the wine colour’s tonality and intensity. Moreover, they are also involved in reactions and interactions with other compounds, with consequences on colour stability and mouthfeel. On the other hand, tannins are responsible for the bitterness and astringency perception. The levels and composition of this relevant group of phenolic compounds thus play a major role in a red wine’s mouthfeel properties and longevity (Aleixandre-Tudo et al., 2017).
The reflexion or transmission of infrared radiation through a sample causes chemical changes at molecular level that are captured in the spectral properties. The spectral data thus provide information on the chemical composition of wine samples (Aleixandre-Tudo et al., 2018a). The spectral information is then correlated with reference data (phenolic levels measured following the conventional methods) and prediction models can thus be established. Before the prediction models can be used in new samples, strict calibration and validation procedures are followed to ensure accurate and reliable predictions. The advantage of using infrared spectroscopy relies on the simplicity, cost-effectiveness and reliability of these techniques (Cozzolino, 2015; Dambergs et al., 2015). As mentioned, with a simple infrared spectral measurement a snapshot prediction is obtained.
The ability of three different infrared instruments to predict the levels of anthocyanins, tannins, total phenols and colour density in fermenting red wine samples and finished wines, was evaluated and reported. The accuracy and robustness of the calibrations were extensively investigated through model performance.
Materials and methods
Fermenting samples were collected from 13 different red wine fermentations at the Welgevallen Cellar (Stellenbosch) during the 2015 and 2016 vintages. Samples were collected daily on the first 15 days and afterwards every three days for a period up to two months. Shiraz (five), Cabernet Sauvignon (five), Pinotage (two) and Grenache (one) were the cultivars included in the study. Several winemaking practices were performed in order to increase phenolic levels and compositional variability. A total number of 391 samples were collected. Moreover, finished wines (178) spanning a wide range of vintages (2005 – 2016) and cultivars (12) were also selected and included in the models. A total number of 569 samples were tested in the calibrations. Models were built for total phenols index (TPI) (Iland et al., 2000), MCP tannins (mg/ℓ) (Sarneckis et al., 2006; Mercurio et al., 2007), anthocyanins (mg/ℓ) (Iland et al., 2000) and colour density (Glories, 1984).
Spectra of these wine samples were collected in transmission mode using a multi-purpose analyser (MPA) Fourier transform-near infrared instrument (FT-NIR) (Bruker Optics, Ettlingen, Germany) in the wavenumber range of 12 500 – 4 000 cm-1. Fourier transform-infrared (FT-IR) spectral measurements were also performed using a WineScan™ FT120 instrument (Foss Electric, Denmark). Samples were scanned from 5 011 – 929 cm-1, which includes a small section of the near-IR region. Finally, spectra were also measured on an Alpha attenuated total reflectance Fourier transform-mid infrared (ATR FT-MIR) spectrometer (Bruker Optics, Ettlingen, Germany), within the 4 000 – 600 cm−1 range.
OPUS software (OPUS v. 7.2 for Microsoft, Bruker Optics, Ettlingen, Germany) was used for data analysis and model performance evaluation. The accuracy and robustness of the prediction models were evaluated using various parameters. The prediction error measures the average difference between the values predicted and those obtained with the reference methods and is given in the same units as the reference values. To standardise the prediction accuracy, the residual predictive deviation statistic is reported. It is defined as ratio of the standard deviation (SD) and the standard error in prediction (RPD = SD/RMSEP). High SD and small RMSEP are thus required to ensure model accuracy. The higher the RPD, the better the ability of the model to accurately predict the levels of phenolic compounds in new samples.
Results and discussion
The average, standard deviation, minimum, maximum and coefficient of variation for the reference methods can be observed in Table 1. The variation of the concentration ranges for the four phenolic determinations seems to be large enough to widely cover variability normally found in different wines (Aleixandre-Tudo et al., 2015). The extraction of phenolics during the fermentation process (average of the 13 investigated fermentations) with logarithmic regressions is also shown in Figure 1. As can be observed, high coefficients of correlation (R2 > 0.9) were always noticed between the measured parameter and the logarithmic equations. In addition, and contrarily to what is shown for TPI and MCP tannins (with a continuous increase until pressing), anthocyanins and CD extraction patterns show a maximum more or less 10 days after crushing with a subsequent plateau where levels of anthocyanins and CD start decreasing (Ribéreau-Gayón et al., 2006).
FIGURE 1. Fermentation extraction curves with logarithmic equations and correlation coefficients. Average values and standard deviation of 13 fermentations.
Table 2 show the errors and RPD values for the three different instruments investigated. Moreover, it is reported and well accepted that RPD values higher than 2.5 indicate accurate models able to predict the levels of phenolic compounds in wine applications (Dambergs et al., 2012; Aleixandre-Tudo et al., 2015, 2018b; Cozzolino et al., 2008). RPD values higher than 2.5 were always observed for the three instruments and four parameters under study. NIR spectroscopy seems to provide the most accurate models with outstanding calibrations for MCP tannins and TPI levels. FT-IR showed slightly lower RPD values, with tannins and TPI again as best models. Finally, ATR-FT-MIR instrument was also able to accurately predict phenolic levels with a very good RPD value for the TPI model.
FIGURE 2. Relationship between total phenolics (TPI) measured by the reference method and predicted using NIR spectroscopy.
Significance of the study
The phenolic levels of some of the most relevant phenolic parameters, including total phenolic index, MCP tannin levels (mg/ℓ), anthocyanins (mg/ℓ) and colour density, can be accurately estimated during the fermentation process, using prediction models developed for three different infrared spectroscopy instruments (including NIR, MIR and IR spectroscopies). By using this technology, a single spectral measurement is thus suitable to obtain the phenolic information. Once the initial investment of the IR instrument is made, the phenolic information is achieved in a fast, simple, accurate and cost-effective manner. Measuring phenolic levels during the fermentation process has a number of possible applications. For example, assessing tannin levels during fermentations by tasting is especially challenging due to the high sugar content, but these methods pave the way for wine producers to do it more accurately. This can increase a winemaker’s ability to decide on a certain winemaking strategy by increasing mixing of the skins and juice to increase extraction or on deciding when to press. Moreover, colour evolution and phenolic extraction during the fermentation can be more closely monitored, which could lead to a better proportion of the different phenolic classes with a potential subsequent improvement in colour and phenolic stability during wine ageing.
During harvest time, it can be daunting for wine producers to measure phenolic extraction during red wine alcoholic fermentation due to time constraints. The conventional reference methods require proper laboratory facilities, which includes lab consumables and specific equipment and are often multi-step, time-consuming and difficult to perform, in some cases. Alternatively, novel techniques using the spectral information contained in the infrared region of the electromagnetic spectrum, are becoming popular. Thanks to previously developed and well established prediction models, the levels of phenolic compounds can be estimated by measuring only the infrared spectra of the samples. Once the instrument measures the spectra and with optimised models, the levels of phenolic compounds and/or parameters of interest can be automatically obtained. The suitability of three different infrared instruments for the prediction of phenolic levels during the red wine fermentation process, has been evaluated and confirmed.
The authors gratefully acknowledge Winetech SA for financial support.
Aleixandre-Tudo, J. L.; Nieuwoudt, H.; Aleixandre, J. L.; Du Toit, W. J. Robust ultraviolet-visible (UV-vis) partial least-squares (PLS) models for tannin quantification in red wine. J. Agric. Food Chem. 2015, 63 (4), 1088 – 1098.
Aleixandre-Tudo, J. L.; Nieuwoudt, H.; Aleixandre, J. L.; Du Toit, W. J. Chemometric compositional analysis of phenolic compounds in fermenting samples and wines using different infrared spectroscopy techniques. Talanta 2018, 176, 526 – 536
Aleixandre-Tudo, J. L.; Nieuwoudt, H.; Olivieri, A.; Aleixandre, J. L.; Du Toit, W. J. Phenolic profiling of grapes, fermenting samples and wines using UVVisible spectroscopy with chemometrics. Food Control 2018, 85, 11 – 22
Aleixandre-Tudo, J.L.; Buica, A.; Hieuwoudt, H.; Aleixandre, J.L.; du Toit, W.J. Spectrophotometric Analysis of Phenolic Compounds in Grapes and Wines. J. Agric. Food Chem. 2017, 65, 4009 – 4026
Cozzolino, D. The role of visible and infrared spectroscopy combined with chemometrics to measure phenolic compounds in grape and wine samples. Molecules, 2015, 20, 726 – 737.
Cozzolino, D.; Cynkar, W. U.; Dambergs, R. G.; Mercurio, M. D.; Smith, P. A. Measurement of condensed tannins and dry matter in red grape homogenates using near infrared spectroscopy and partial least squares. J. Agric. Food Chem. 2008, 56, 7631 – 7636.
Dambergs, R. G.; Mercurio, M. D.; Kassara, S.; Cozzolino, D.; Smith, P. A. Rapid measurement of methyl cellulose precipitable tannins using ultraviolet spectroscopy with chemometrics: application to red wine and inter-laboratory calibration transfer. Appl. Spectrosc. 2012, 66 (6), 656 – 664.
Dambergs, R.; Gishen, M.; Cozzolino, D. A review of the state of the art, limitations, and perspectives of infrared spectroscopy for the analysis of wine grapes, must, and grapevine tissue. Applied Spectroscopy Reviews, 2015, 50(3), 261 – 278.
Glories, Y. La couleur des vins rouges, 2eme partie. Connaissance de La Vigne et Du Vin, 1984, 18, 253 – 271.
Iland, P.; Ewart, A.; Sitters, J.; Markides, A.; Bruer, N. Techniques for chemical analysis and quality monitoring during winemaking (1st ed., pp. 1e111). Campbelltown, South Australia: Patrick Iland Wine Promotions. 2000.
Mercurio, M. D.; Dambergs, R. G.; Herderich, M. J.; Smith, P. A. High throughput analysis of red wine and grape phenolics – adaptation and validation of methyl cellulose precipitable tannin assay and modified Somers color assay to a rapid 96 well plate format. Journal of Agricultural and Food Chemistry, 2007, 55(12), 4651 – 4657.
Ribéreau-Gayon, P.; Glories, Y.; Maujean, A.; Dubourdieu, D. Handbook of Enology: The chemistry of wine and vinifications. Handbook of Enology, 1. 2006.
Sarneckis, C. J.; Dambergs, R. G.; Jones, P.; Mercurio, M.; Herderich, M. J.; Smith, P. A. Quantification of condensed tannins by precipitation with methyl cellulose: Development and validation of an optimised tool for grape and wine analysis. Australian Journal of Grape and Wine Research, 2006, 12(1), 39 – 49.
– For more information, contact Jose Luis Aleixandre-Tudo at firstname.lastname@example.org.