Natural white wine alcoholic fermentation: a focus on progression trajectories and sensory outcomes
| dc.contributor.advisor | Nieuwoudt, HĆ©lĆØne H | en_ZA |
| dc.contributor.author | Kruger, Marinda | en_ZA |
| dc.contributor.other | Stellenbosch University. Faculty of AgriSciences. Dept. of Viticulture and Oenology. | en_ZA |
| dc.date.accessioned | 2021-05-24T13:29:15Z | |
| dc.date.available | 2021-05-24T13:29:15Z | |
| dc.date.issued | 2021-03 | |
| dc.description | Thesis (PhDAgric)--Stellenbosch University, 2021. | en_ZA |
| dc.description.abstract | ENGLISH ABSTRACT: Wine is the result of the impact of collective production decisions made from the vineyard and throughout the steps taken in the winery. Nowadays, consumers have access to wines from across the world. This has resulted in an elevated consumer-demand for wines that demonstrate, in sensorial terms, individuality, exceptional quality, and provenance. Wine producers interpret this hugely competitive market-pull through alternative winemaking strategies, such as natural or spontaneous fermentations, in the belief that wines produced in this way reflect uniqueness and authenticity. The sensory profile of a wine is hugely influenced by the fermentation regime chosen by the oenologist or winemaker. Alcoholic fermentation is the bioprocess whereby grape sugars, which mainly consist of glucose and fructose, are converted by yeasts to ethanol, CO2, and secondary metabolites. The sensory profiles of some white wine styles are largely determined by the grape flavour compounds and those derived during alcoholic fermentation. Alcoholic fermentation (AF) is, arguably, the most important step in winemaking and, therefore, the control and monitoring of this bioprocess is of the utmost importance for a predictable duration and outcome, as well as reproducibility from one vintage to the next. Fourier Transform mid-infrared (FT-MIR) spectroscopy is well implemented, in wine laboratories, for routine chemical analysis of alcoholic fermentation parameters. Extensive research exists for quantification calibration development using FT-MIR spectroscopy but only few studies where the FT-MIR spectra are used for qualitative calibrations. No studies for explorative data mining of the information-rich FT-MIR spectra of AF could be found. The visualisation of big data is receiving much attention. Visualisation of data using multivariate data analysis techniques, gives a clear idea of what the information means, it highlights trends, patterns, and outliers. In this study, the visualisation of AF process data is novel. Using Chardonnay grape must (Data set 1), fermented at a constant temperature and a three- by-three experimental design, it was possible to visualise the variation of the progression trajectories between the fermentations. The data consisted of FT-MIR spectra and chemical parameters which aids the interpretation of the progression trajectories. Statistical data analysis of the chemical parameters correlated with the visualisation of the FT-MIR spectral data and chemical parameters using multiway partial least square regression (MPLS) or batch evolution modeling (the term used in this study). PCA of the PLS scores of the BEM, the fermentations could be visually compared on the PCA score plot. Further to determine class separation by orthogonal PLS discriminant analysis (OPLS-DA) confirmed correlation and variation between the fermentations. The second data set (Data set 2) was historical, Chenin blanc, Colombard and Chardonnay fermentation data, from a commercial winery. The fermentations all fermented to dry (residual sugars < 5g/L). No prior knowledge existed of which yeasts were used, only that best fermentation practises were applied. The time trajectory of alcoholic fermentations varies greatly. To compare and monitor fermentations effectively the biological state at a certain point needs to be the same in relation to the process. A relative time scale was introduced in this study to realign the data and put it on a generic time basis. The PLS model with X- FT-MIR spectra and Y-relative time demonstrated significant statistical indicators. Concluding that relative time implemented in the multivariate model, ensures correct interpretation of the progression trajectories of a given point in time and that the prediction of fermentation time is possible. Natural or uninoculated fermentation introduces more variation within the FT-MIR spectral data. The reason being that these fermentations are not inoculated with a Saccharomyces cerevisiae yeast, as in the first two data sets. In natural fermentation, the grape must is fermented by the indigenous S. cerevisiae and non-Saccharomyces (NS) yeasts. Yeast-to-yeast interaction occurs which introduces more variation than an inoculated commercial yeast fermentation. It was important to compare the between variation of the different fermentations. The third data set (Data set 3), comprised of Sauvignon blanc, this global popular cultivar was chosen as a solid foundation exists due to previous research. Inoculated and natural fermentations on a micro- and commercial scale was performed. S. cerevisiae and Torulaspora delbrueckii was each inoculated in one small-scale fermentation, respectively. Two co-inoculation fermentations, the sequential inoculation strategy of T. delbrueckii and S. cerevisiae, as well as two natural fermentations were performed on small- and commercial-scale. FT-MIR spectroscopy spectral data was acquired during each of the fermentations, chemical parameters were determined through the calibrations from the FT120 Winescan, the chemical aroma compounds were measured as well as sensory profiling using sensory descriptive analysis method of the finished wines. Natural fermentations had lower fermentation kinetics and thus has a longer duration time than inoculated fermentations. BLM, with relative time (Y-variable) and the FT-MIR spectra (X-variables), visualised the variances between the fermentations. The S. cerevisiae inoculated fermentation was more different in comparison to the other fermentations. This was seen on the multivariate statistical control charts (MSPC) or batch control charts, as the fermentation sat well above the other fermentations which indicated higher fermentation kinetics. The PCA score plot of BEM also confirmed the variation visually. A point-to-point function on the MSPC charts demonstrated the loading weights for glucose, fructose, and ethanol, responsible for this variation. It was established that natural fermentation can also be monitored and compared in the same model as inoculated fermentations. PLS-TreesĀ® , hierarchical classification method, was used to explore the within variation of the fermentation spectral data. Three data clusters were identified within the progression of inoculated fermentations. However, within the natural and co-inoculated fermentations four data clusters were identified. It could be speculated that more variation exists within the fermentation due to the longer fermentation time. The changing relationship of the FT-MIR spectra and chemical parameters between the clusters could be interpreted for insight, through the respective loading weights. A possible application will be that a local predictive model based on this insight will be more accurate, a lower standard error of calibration, than a global model. It was of interest to investigate if correlation exists between the BEM of the FT-MIR spectral fermentation data and the sensory descriptive and aroma compounds data of the corresponding finished wines. This was a three-data block comparison. Multiblock orthogonal component analysis (MOCA) finds the joint and unique variation between the data blocks. Joint variation and correlation were found between the three data blocks. No unique variation exists between the three data blocks. However, from the PCA score plot of the MOCA model observation C (small- scale co-inoculation) had larger variations between the three block models. Observation F (commercial-scale co-inoculation) had the least variation between the three block models. The other observations demonstrated similar variations. The reason for this could not be determined in this study and needs further research. The correlation between the three data blocks makes qualitative calibrations a possibility. The positive contribution is that qualitative outcomes could be possibly predicted during certain stages of alcoholic fermentation and encourages further research in this field. | en_ZA |
| dc.description.abstract | AFRIKAANSE OPSOMMING: Wyn is die resultaat van die impak van gesamentlike produksiebesluite wat geneem word in die wingerd en tydens die deurlopende stappe wat gevolg word in die wynkelder. Deesdae het verbruikers toegang tot wyne van dwarsoor die wĆŖreld. Dit het gelei tot ān verhoogde verbruikersaanvraag na wyne wat, in sensoriese terme, individualiteit, uitstaande kwaliteit en oorspong demonstreer. Wynprodusente interpreteer hierdie uiters kompeterende markaanvraag deur alternatiewe wynmaakstrategieĆ«, soos natuurlike of spontane fermentasies. Die mening word gehuldig dat wyn wat op hierdie manier geproduseer is, uniekheid en intrinsieke egtheid weerspieĆ«l. Die sensoriese profiel van ān wyn word grootliks beĆÆnvloed deur die fermentasieregime wat die wynmaker gebruik. Alkoholiese fermentasie (AF) verwys na die proses waartydens die bioverwerking van druiwesuikers, wat hoofsaaklik bestaan uit glukose en fruktose, omgeskakel word deur giste tot etanol, CO2 en sekondĆŖre metaboliete. Die sensoriese profiele van sekere witwynstyle word hoofsaaklik bepaal deur die geurverbindings wat in die druif self teenwoordig is, asook diĆ© verbindings wat tydens AF gevorm word. Argumentsonthalwe, kan AF beskou word as die belangrikste stap in wynbereiding. Derhalwe is die beheer en monitering van hierdie bioproses noodsaaklik vir ān voorspelbare duurte en uitkoms van die fermentasies, asook vir herhaalbare uitkomste van die wynstyl van een oesjaar na die volgende. Fourier transformasie mid-infrarooi (FT-MIR) spektroskopie is goed gevestig in wynlaboratoriums vir die roetine analise van chemiese parameters wat ān aanduiding gee van die verloop van AF. Uitgebreide navorsing is reeds gedoen op die ontwikkeling van kwantitatiewe FT-MIR spektroskopie-gebaseerde kalibrasiemodelle, terwyl slegs enkele studies FT-MIR spektra gebruik het vir kwalitatiewe kalibrasies. Geen studies wat verkennende data-ontginning van die inligtingryke AF FT-MIR spektra kon in die literatuur gevind word nie. Die visualisering van groot datastelle met behulp van multi-veranderlike data analitiese (MVDA) tegnieke kry tans baie aandag. Die resultate wat met hierdie tegnieke verkry word, gee ān goeie aanduiding van die waarde van die inligting vervat in die data, deurdat tendense, patrone en ekstreme datapunte beklemtoon word. Die benadering wat in hierdie studie gebruik is om die AF proses te visualiseer, is ān eerste in hierdie vakdisipline. Met die gebruik van Chardonnay druiwe (Datastel 1), gefermenteer by 'n konstante temperatuur en met ān drie-by-drie eksperimentele ontwerp, was dit moontlik om die variansie van die progressietrajekte in die onderskeie fermentasies te visualiseer. Die data, bestaande uit FT- MIR spektra en chemiese parameters, het bygedra om die progressietrajekte te interpreteer. Statistiese data-analise van die chemiese parameters het gekorreleer met die visualisering van die FT-MIR spektrale data en chemiese parameters, met die gebruik van veelvoudige gedeeltelike minste kwadrate regressie (GMK) en Batch Evolution Modeling (BEM). Hoofkomponente-analise (HKA) van die GMK en BEM data is gebruik om die fermentasies visueel te vergelyk. Diskriminant analise deur middel van ortogonale GMK is ook geruik in die visualisering van die fermentasies. Datastel 2 was ān historiese AF datastel van Chenin Blanc, Colombard en Chardonnay, wat by ān kommersiĆ«le wynkelder gegenereer is. Al die fermentasies was droog gegis (oorblywende suikers <5g/L). Geen inligting oor watter giste gebruik is vir AF was beskikbaar nie; slegs dat die beste fermentasiepraktyke toegepas is. Omdat die tydsduurte van die onderskeie AF grootliks van mekaar verskil het, was dit nie moontlik om die biologiese status van die fermentasies op ān bepaalde tydstip met mekaar te vergelyk nie. Derhalwe is ān relatiewe tydskaal in hierdie studie bekendgestel om die data te belyn, en om dit op ān generiese tydbasis te plaas. ān GMK model met die FT-MIR spektra as X- veranderlikes en relatiewe tyd die Y-veranderlike, het beduidende statistiese verskille tussen die tydsduurtes van die onderskeie fermentasies getoon. Met die gebruik van relatiewe tyd, kon die korrekte interpretasie van die progressietrajekte op ān sekere tydstip gemaak word en was dit moonlik om die vordering van die fermentasies korrek te voorspel. Natuurlike of ongeĆÆnokuleerde fermentasies het meer variansie in die FT-MIR spektrale data tot gevolg gehad. Die rede hiervoor was dat hierdie fermentasies nie geĆÆnokuleerd was met die Saccharomyces cerevisiae wyngis nie. In natuurlike fermentasies, word die druiwemos gegis deur inheemse S. cerevisiae en nie-Saccharomyces (NS) giste wat natuurlik teenwoordig is in die druiwemos. Interaksie tussen die inheemse giste vind plaas wat meer variasie in die verloop van die fermentasie tot gevolg gehad het, as wat die geval was met die fermentasies wat met kommersiĆ«le giste geĆÆnokuleer was. Datastel 3 het bestaan uit Sauvignon Blanc fermentasies. Hierdie wĆŖreldwye gewilde kultivar is gekies omdat heelwat navorsing reeds op die kultivar gedoen is, en die effekte van verskillende inokulasiestrategieĆ« teen so ān bekende agtergrond, duideliker ondersoek kon word. GeĆÆnokuleerde en natuurlike fermentasies is uitgevoer op klein- en kommersiĆ«le skaal. S. cerevisiae en Torulaspora delbrueckii (T. delbrueckii) is elkeen onderskeidelik geĆÆnokuleer in een klein-skaalse fermentasie. Twee ko-inokulasie strategiĆ«, naamlik opeenvolgende inokulasie van T.delbrueckii en S.cerevisiae, sowel as twee natuurlike fermentasies is uitgevoer op klein- en kommersiĆ«le skale. FT-MIR spektra is gegenereer tydens elk van die fermentasies. Chemiese parameters is bepaal deur die GMK kalibrasies van die FT120 Winescan spektrometer. Die chemiese aromaverbindings is gemeet deur gaschromatografie, terwyl die sensoriese profiele van voltooide wyne gegenereer is deur gebruik te maak van die sensoriese beskrywende analise metode. Natuurlike fermentasies het stadiger fermentasiekinetika getoon en het dus ān langer fermentastie-tyd gehad as geĆÆnokuleerde fermentasies. BEM, met relatiewe tyd (Y-veranderlike) en die FT-MIR spektra (X-veranderlikes), het die variansie tussen die fermentasies gevisualiseer. Die S. cerevisiae geĆÆnokuleerde fermentasie was meer verskillend, in vergelyking met die ander fermentasies. Dit is waargeneem in die Multi-veranderlike statistiese beheergrafieke (MSPC) en bundel beheergrafieke. Die PLS-treesĀ® hierargiese klassifikasiemetode is gebruik om die interne variansie van die fermentasies met behulp van die spektrale data te ondersoek. Drie datagroepe is geĆÆdentifiseer binne die geĆÆnokuleerde fermentasies, terwyl vier datagroepe in die natuurlike en ko-geĆÆnokuleerde fermentasies geĆÆdentifiseer is. Dit kan bespiegel word dat meer variansie bestaan binne natuurlike en ko-geĆÆnokuleerde fermentasies weens die langer tyd wat laasgenoemde fermentasie neem om te voltooi. Die veranderende verhouding tussen die FT- MIR spektra en chemiese parameters van die onderskeie groepe kon geinterpreteer word met behulp van MVDA tegnieke. ān Moontlike uitvloeisel van die bevinding is dat ān plaaslike prediksiemodel wat gebaseer is op gemete data, meer akkuraat kan wees, met ān laer standaard voorspellingsfout, as ān generiese model. Dit was van belang om vas te stel of daar ān korrelasie bestaan tussen die FT-MIR spektra, sensoriese data en chemiese aromakomponente van die ooreenstemmende voltooide wyne met behulp van ān drie-data blok vergelyking. Die Multiblock orthogonal component analysis (MOCA) tegniek is gebruik om die gesamentlike en unieke variansie tussen die datablokke te bereken. Gesamentlike variansie en korrelasie is gevind tussen die drie datablokke. Geen unieke variansie was egter in enige van die datablokke teenwoordig nie. Die kleinskaalse ko-geĆÆnokuleerde fermentasie het die grootste variansie getoon tussen die drie datablokke, terwyl die kommersiĆ«le- skaal ko-inokulasie fermentasie die minste variansie tussen die drie datablokke getoon het. Die rede hiervoor kon nie in hierdie studie vasgestel word nie en verdere navorsing is nodig. Korrelasies wat gevind is tussen die drie datablokke maak kwalitatiewe kalibrasies ān moontlike opsie om te ondersoek. Die bydrae van die studie is dat kwalitatiewe uitkomste moontlik voorspel kan word gedurende sekere stadiums van alkoholiese fermentasie en dit moedig verdere navorsing in hierdie veld aan. | af_ZA |
| dc.description.version | Doctoral | en_ZA |
| dc.format.extent | (14 unnumbered pages) 115 pages : illustrations (some color) | en_ZA |
| dc.identifier.uri | http://hdl.handle.net/10019.1/110503 | |
| dc.language.iso | en_ZA | en_ZA |
| dc.publisher | Stellenbosch : Stellenbosch University | en_ZA |
| dc.rights.holder | Stellenbosch University | en_ZA |
| dc.subject | Natural white wine -- Composition | en_ZA |
| dc.subject | Natural fermentation | en_ZA |
| dc.subject | Wine -- Sensory evaluation | en_ZA |
| dc.subject | Relative time scale | en_ZA |
| dc.subject | Alcoholic fermentation | en_ZA |
| dc.subject | Chardonnay grape must -- Effect of temperature on | en_ZA |
| dc.subject | Wine and wine making -- Analysis | en_ZA |
| dc.subject | UCTD | en_ZA |
| dc.title | Natural white wine alcoholic fermentation: a focus on progression trajectories and sensory outcomes | en_ZA |
| dc.type | Thesis | en_ZA |