Doctoral Degrees (Viticulture and Oenology)

Permanent URI for this collection

https://scholar.sun.ac.za/handle/10019.1/575

Browse

Browsing Doctoral Degrees (Viticulture and Oenology) by browse.metadata.advisor "Brand, Jeanne"

Now showing 1 - 1 of 1
  • Thumbnail Image
    Item
    A chemometric approach to investigating South African wine behaviour using chemical and sensory markers
    (Stellenbosch : Stellenbosch University, 2021-03) Mafata, Mpho; Buica, Astrid; Brand, Jeanne; Medvedovici, Andrei V.; Stellenbosch University. Faculty of AgriSciences. Dept. of Viticulture and Oenology.
    ENGLISH ABSTRACT: The aim of this dissertation was to demonstrate the value of comprehensive narratives and elucidate critical steps in data handling in Oenology, while highlighting some common misconceptions and misinterpretations related to the process. This compilation was a journey through different stages of dealing with oenological data, with increasing complexity in both the strategies and the techniques used (sensory, chemistry, and statistics). To achieve this aim, different strategies and multivariate tools were used under two prime objectives. Firstly, several multivariate descriptive approaches were used to investigate two oenological problems and lay out the contextual foundations for the statistics-focused work (Chapters 3 and 5). Secondly, in increasing levels of complexity, statistical strategies of constructing comprehensive data fusion as well as pattern recognition models were investigated (Chapters 4 and 6). A comprehensive literature review (Chapter 2) examined and addressed common misconceptions in the different stages of data handling Oenology. The first oenological problem, described in Chapter 3, investigated the evolution of the sensory perception of aroma, as well as the antioxidant-related parameters and volatile compound composition of Sauvignon Blanc and Chenin Blanc wines stored under different conditions and durations. The study applied an appropriate sensory method for this research question, namely, Pivot©Profiling. The study was able to show the evolution of Sauvignon Blanc from ‘fruity’ and ‘herbaceous’ and of Chenin Blanc from ‘fruity’ and ‘tropical’ both towards ‘toasted’, ‘oak’, and ‘honey’ attributes. Chemically, the volatile composition did not show any trends. However, wines stored at higher temperatures for longer periods had relatively higher UV-Vis absorbance, colour density as well as higher b* (yellow) values and lower clarity in terms of L* index, compared to the control. The second oenological problem, described in Chapter 5, investigated the typicality of South African old vine Chenin Blanc perceptually and conceptually using a typicality rating and a flexible sorting task. The sensory methodology followed published strategies for investigating typicality. This study did not find a unique sensory space of the old vine Chenin Blanc due to a lack of perceptual consensus among the industry professionals for the wines included in the study. However, it did find that the industry professionals had unified ideas about the attributes of an ideal old vine Chenin Blanc wine. The first of the statistics-focused studies, described in Chapter 4, explored data fusion at low and mid-level using principal component analysis - PCA (low and mid-level) and multiple factor analysis - MFA (mid-level). The study looked at data pre-processing and matrix compatibility, which are important data handling stages for data fusion. Like the contextual chapters (Chapter 3 and 5), and keeping with the aim of this compilation, this chapter gave a detailed descriptive narrative of the data handling. Through detailed examination of the process, the study found that MFA was the most appropriate data fusion strategy. The second statistics-focused study, described in Chapter 6, continued to exploit the multiple advantages of multiblock approach of MFA. Additionally, this chapter showed the reliability of fuzzy k-means clustering compared to agglomerative hierarchical clustering (AHC).