Department of Viticulture and Oenology

Permanent URI for this community

https://scholar.sun.ac.za/handle/10019.1/31

Browse

Browsing Department of Viticulture and Oenology by browse.metadata.advisor "Bosman, Gurthwin W."

Now showing 1 - 2 of 2
  • Thumbnail Image
    Item
    The quantification of red wine phenolics using fluorescence spectroscopy with chemometrics
    (Stellenbosch : Stellenbosch University, 2021-03) Dos Santos, Isabel Anne; Du Toit, Wessel J.; Aleixandre-Tudo, Jose Luis; Bosman, Gurthwin W.; Stellenbosch University. Faculty of AgriSciences. Dept. of Viticulture and Oenology.
    ENGLISH ABSTRACT: The organoleptic and perceived quality characteristics of red wine are largely influenced by important phenolic compounds extracted throughout fermentation from the grape berry to the final wine matrix. These complex secondary metabolites have resulted in numerous equally complex analysis methods, the implementation of which are yet to form part of routine phenolic analysis during winemaking. In this study, front-face fluorescence spectroscopy was investigated for its suitability in quantifying phenolic parameters of unaltered samples and the subsequent implications for non-invasive analysis throughout fermentation. A front-face accessory and fluorescence spectrophotometer were successfully optimised in order to analyse samples directly, eliminating the need for sample dilution as with conventional fluorescence spectroscopy. A diverse dataset comprising 289 fermenting musts and wine were analysed using the optimised fluorescence protocol and the most commonly used UV-Vis spectrophotometric methods for the following phenolic parameters; total phenolics, total condensed tannins, total anthocyanins, colour density and polymeric pigments. Different statistical analysis methods were explored for their suitability in model development, specifically Parallel Factor Analysis (PARAFAC) and a gradient boosting machine learning algorithm (XGBoost). Subsequent to the investigation of the most optimal chemometric method, a machine learning pipeline was generated to develop accurate regression models per phenolic parameter. Successful models were obtained for total phenolics, total condensed tannins and total anthocyanins while polymeric pigments and colour density require further investigation and refinement. Following model development and optimisation, an external validation experiment monitoring a Cabernet Sauvignon fermentation was used to examine prediction accuracy under fermentation conditions, specifically investigating the effect of carbon dioxide and must turbidity. No effect of sample preparation treatment was found and the potential for analysing unaltered samples directly during fermentation was possible. Fluorescent properties of fermenting musts and wines were explored and the responsible spectral regions of interest tentatively identified. Differences in fluorescence between musts and wines were found and upon closer inspection, unique changes were monitored and identified throughout fermentation using the Cabernet Sauvignon experiment. The unique fluorescent profiles of wines is widely accepted, and the classification of South African red wine cultivars was successfully conducted using Neighbourhood Component Analysis (NCA). These results may have beneficial implications for authentication and quality control by industry bodies. Overall, front-face fluorescence spectroscopy holds several advantages including it being non- invasive, user-friendly, relatively economical, rapid and accurate, and thus presents itself as a promising alternative to the current phenolic analysis methods with the added benefit of direct phenolic analysis throughout red wine fermentation. The potential for implementation within on- line automated systems or portable optical devices may be of interest to producers and allow for monitoring of phenolic content and extraction directly from the fermentation vessel throughout red wine production.
  • Thumbnail Image
    Item
    The quantification of white grape juice phenolics using various spectroscopic methods and chemometrics
    (Stellenbosch : Stellenbosch University, 2021-12) Clarke, Sarah; Aleixandre-Tudo, Jose Luis; Du Toit, Wessel J.; Bosman, Gurthwin W. ; Stellenbosch University. Faculty of AgriSciences. Dept. of Viticulture and Oenology.
    ENGLISH ABSTRACT: Phenolic compounds are aromatic, secondary metabolites found in plant tissues. They have a number of bioactive properties as well as positive effects on health. Phenolic compounds, although found at lower levels in Méthode Cap Classique (MCC) and white wines, contribute to the mouthfeel and flavour as well as having antimicrobial and antioxidant properties. The levels of the phenolic compounds present in a wine depend on the variety, ripeness of grapes at harvest, soil type and the vinification processes applied. During the pressing stages, juice is obtained from the grapes and skins in contact with the juice. This maceration time, although limited, allows for phenolic compounds to be extracted and dissolved in the juice. The ability to monitor the phenolic concentration during the pressing stages of MCC and white wine could potentially increase the yield recovery of quality juice, furthermore, allowing increased control in the vinification process and can lead to benefits such as improved and consistent wine quality. Current phenolic compound analysis methods can be outdated and unreliable due to interferences. In order for phenolic monitoring techniques to be useful in the wine industry they must be compatible with process control methods. Spectroscopy techniques, alongside chemometrics, for the quantification of phenolics have the potential to be implemented into wineries as in-line and on-line systems. These techniques provide increased accuracy and reliability. This research explores a range of analytical techniques which may be applied to the quantification of phenolic compounds with the use of calibration models. Infrared (IR), Raman and fluorescence spectroscopy were the analytical methods explored and the reference total phenolic index (TPI) data was collected using Ultra-Violet/Visible (UV/Vis) spectrophotometry. These spectroscopic techniques were chosen as they are suited for the implementation into portable devices and hence could be of use to the wine industry for process control analysis. The spectroscopic analyses performed are: - Attenuated Total Reflectance Mid-Infrared Spectrometer (ATR-MIR). - Multi-Purpose Analyser (MPA) Transmission Fourier Transform Near-Infrared Spectrometer (T-FT-NIR). - Matrix F Diffuse Reflectance Fourier Transform Near-Infrared Spectrometer (DF-FT-NIR). - Raman spectroscopy with a central wavelength of 532nm. - Fluorescence spectroscopy with emission spectra between 300nm and 575nm and excitation wavelengths between 300nm and 575nm. Partial Least Squares (PLS) regression models were built for all analytical methods explored and the robustness of these models were examined using a range of statistical parameters. Further techniques, such as machine learning, were explored for the data obtained in the fluorescence spectroscopy. T-FT-NIR provided the best model for TPI with 0.547 and 2.12 RMSEP and RPDval, respectively. Moreover, high prediction accuracy was observed with DF-FT-NIR for the MCC dataset with 0.457 RMSEP and 2.01 RPD. The models obtain form the Raman and fluorescence spectra underperformed those of the IR instruments. However, improvements in fluorescence model performance were achieved when the use of a machine learning analysis pipeline was explored. The statistical parameters used to determine model robustness did not indicate that all of the predication models constructed are of immediate use to the wine industry. Despite these results, it is believed that the aim of this research is worth further investigation. The observed models do indicate results which could be of potential use for screening purposes. Further research could be the key to unlocking the potential of these spectroscopic methods for phenolic quantification as this would reduce the number of variables which are believed to have caused the results observed.