Assessment of relationships between grape chemical composition and grape allocation grade
A wide range of compounds that contribute to wine’s appearance, aroma, flavour and texture are derived from grapes. Many of these compounds are known and measurable, and can be manipulated through viticultural and/or winemaking practices. Around the world, many wine companies use grape chemical measures to assess grape value or style; however, this approach has not been implemented widely in Australia.
By measuring a range of chemical compounds in multiple grape batches of different grades, this project aims to determine which compounds, independently or in combination, can differentiate between grape grades. The objectives are to determine how variable the chemical measures are across a wide range of fruit grades, if there is a relationship with fruit grade and if fruit can be clustered based on similarity of chemical composition. A primary aim of the project is to assess the practical application of grape grading measurements and to support wine producers who intend to apply these measures in their systems. Potential impacts for industry include the ability for grapegrowers to more efficiently produce grapes to defined specifications, and for winemakers to select fruit with greater confidence that it will be appropriate for a targeted wine style.
Can existing commercial grading allocations be predicted using previously identified and new chemical measurements?
Samples were sourced from regions across Australia and included 56 Cabernet Sauvignon, 62 Shiraz and 64 Chardonnay grape samples from up to nine commercially defined quality grades. Chemical analyses were selected based on a literature review and survey of the way local and international producers measure quality. Basic chemical measures included berry weight, titratable acidity (TA), pH, malic acid, °Brix and nitrogen content; potential negative quality markers included laccase, chloride and methoxypyrazine; targeted compositional measures (dependent on variety) included ‘grassy, green’ C-6 compounds, volatile thiol precursors, β-damascenone, the broad flavour measure phenol-free glycosyl-glucose, amino acid profile, phenolics and non-targeted spectral fingerprinting in mid- and near-infrared regions. Data were analysed with the multivariate statistical methods of partial least squares regression (PLS) and quadratic discriminant analysis (QDA).
PLS models developed for grade had R2 values of validation between 0.55 and 0.84, showing that this was not always the best modelling approach. Due to the fact that grade is not numerically defined and the relationships within the datasets were found to be non-linear, QDA (which predicts categories not values) gave better accuracy of grade prediction. The QDA prediction for Shiraz was 85% and indicated that higher values for °Brix, alpha-amino nitrogen, total yeast assimilable nitrogen, certain amino acids, tannin, absorbances at 280 and 520 nm, C6 compounds and chloride concentrations were associated with higher grade fruit. The QDA prediction for Cabernet Sauvignon was 90% and indicated that higher values for °Brix, certain amino acids, tannin, absorbances at 280 and 520 nm, glycosyl glucose and chloride concentrations were associated with higher grade fruit. The QDA prediction for Chardonnay was 95% and indicated that higher levels of TA, ammonia, malic acid, volatile thiol precursors, C6 compounds, glycosyl glucose, chloride and β-damascenone were present in higher grade fruit.
Spectroscopic fingerprints from MIR and UV-Vis scans best predicted Cabernet Sauvignon and Shiraz grade, while targeted chemical analyses best predicted Chardonnay grade using QDA. However, for Chardonnay using a reduced analysis dataset of only basic berry measurements, including pH, TA, ammonia, alpha-amino nitrogen, °Brix, berry weight and targeted UV spectra, prediction accuracy was reduced only slightly, to 93%. This was a promising observation, since most of these measurements are readily accessible to producers.
There is clearly potential for objective chemical measures to be defined by both targeted and non-targeted methods. A relatively simple analytical grading tool is achievable for grapegrowers and winemakers willing to commit to implementation but will require some ongoing refinements by grape variety, and calibration across multiple vintages.
In 2015 fruit batches for four style categories of Shiraz and Chardonnay were sourced from multiple vineyards and the trial was extended beyond the assessment of grape chemical composition to also include winemaking. The aim was to establish the relevance and predictive power of the identified grape compounds to forecast grape quality and wine style categories. Data analysis is being finalised for a final report due in December 2016.