The identification and use of objective measures for the accurate determination of red grape quality is of great interest for many Australian grape and wine producers. To achieve this objective, collaboration between wine producers and researchers is essential to identify practical grape quality measures that are associated with specific wine quality grades and styles. Building on previous work on grape and wine objective methods across multiple Shiraz grades, the study presented in this report aimed to understand the compositional differences between a narrow range of premium Barossa Shiraz categories. In both seasons of the study, the specified quality grades (1 and 2) could not be defined by targeted grape objective measures alone. However, certain premium vineyards did not produce wines in the expected premium wine category within a season, and could be designated as ‘marginal’ vineyards. Using two different approaches over two seasons of the study, it was found that very high levels of grape amino acids and ammonia were associated with vineyards categorised as ‘marginal’ and could be associated with losses in wine quality. The results suggest that grape nitrogen should be carefully managed in both the vineyard and at the winery as a first step to optimise wine quality.
An earlier Wine Australia-funded project (AWR1202) had aimed to identify the key objective measures in grapes which might be used to define commercial quality grading across a wide range of grades from the poorest quality through to premium for a single wine producer. This was particularly successful for Shiraz grapes, where the most important variables for defining quality grade were phenolics (colour and tannin) and nitrogen (amino acids and ammonia). Based on this successful proof-of-principle study, there was interest from another producer (Pernod Ricard Winemakers) to use the same objective tools to understand the difference between premium and ultra-premium Barossa Shiraz. This led to the initiation in 2016 of a new project on premium Shiraz quality, which is presented in this report. This project looked more specifically at grape phenolics using new techniques such as extractable tannin and colour, as well as non-targeted phenolic profiling, which enables multiple unknown compounds to be identified. Since total must nitrogen and individual amino acids are also known to influence the fermentation-derived volatile profile of wines, these measures were also included in the new project on premium Shiraz. Other techniques investigated were the use of mid-infrared (MIR), near-infrared (NIR) and UV-visible spectra of juices or homogenates (or their extracts).
In the 2017 and 2018 seasons, the producer identified 20 and 10 Barossa Valley Shiraz vineyards that were within the premium ‘intended use’ classifications 1 and 2 respectively. In the first season, five samples were taken across each vineyard, and in the second season ten samples were taken to gain greater understanding of within-vineyard variability. Grapes were analysed for multiple targeted and non-targeted objective measures. In the first season, the grape batches were also put through a micro-fermentation process. Finally, for commercial winemaking, each vineyard was harvested and kept separate in the winery for fermentation and barrel ageing in old oak over six to eight months. Commercial wines were then evaluated by a panel of winemakers to assign an ‘intended use’ classification and analysed for their chemical and sensory properties at the AWRI. The assigned ‘intended use’ grades spanned four quality categories: 1 (highest quality), 1.5, 2 and 3. Data generated from the study were analysed using multivariate statistical techniques in order to predict the respective grape and wine grades.
In the 2017 season, using the results of grape analysis alone, it was not possible to accurately predict the ‘intended use’ grade of more than 46% of the grape samples using multivariate statistics. Nevertheless, some grape-based measurements could be identified as more significant than others in distinguishing grape grades 1 and 2. Prediction models were also built using data from the micro-ferment wines associated with each grape batch. The wines were assessed immediately at the completion of fermentation, and were therefore not aged, and reflected only the basic extraction and conversion of compounds during fermentation. The inclusion of the wine data gave stronger models for grape grade prediction using multivariate statistics, at 70% accuracy. When data from non-targeted grape and wine phenolic analysis were added to the prediction models, their accuracy was improved up to 79%, and certain grape and wine compounds were further identified by mass spectrometry. The results of the grape grade study showed that although some grape-based measures could be identified as relevant in the prediction of grape grade, strong models of prediction could not be developed without the inclusion of wine data. This indicated that the conversion of grape metabolites during fermentation is related to, but not strictly correlated with, their concentration or extractability from the grape source. Further research into the identification of important wine metabolites from the non-targeted phenolic analysis may potentially shed light on other grape precursors which were not detected in the grape-based assays but which may be relevant to the definition of quality grade in the premium category.
Although the vineyards selected for the study were only from premium grape grades 1 and 2, it was evident after the commercial winemaking was completed that the vineyard grading system did not closely reflect the final wine grade assigned by the winemaker panel. In light of this finding, it was deemed relevant to the goals of the project to determine whether grape or wine chemical measures might be useful in predicting the ‘intended use’ wine grade, since this was of commercial importance. The same grape and wine chemical data used in the prediction of grape grade were therefore applied in models to predict the final wine grade. From this process, it was found that wine grades 1, 1.5 and 2 could not be distinguished by the targeted chemical data, but the models could strongly predict wine grade 3, at between 71 and 75% accuracy depending on the type of data used. From the grape data, it was found that grape batches that produced grade 3 wines had higher levels of grape nitrogen, including ammonia and total YAN and multiple amino acids. It was also observed that while total grape colour was relevant to the model, the anthocyanin to tannin ratio was more important as a predictor of grade 3 wines, and was generally lower than in grapes which resulted in wines graded to categories 1, 1.5 and 2. The sensory properties of the commercial wines from the different grade classes were also compared. It was found that large increases in the negative aroma attribute ‘tinned vegetable’ was associated with the grade 3 category relative to the higher wine quality grades, with losses in ‘dark fruit’ aroma and flavour also seen for grade 3 wines, accompanied by increases in ‘red fruit’ aroma and flavour. Grade 3 wines also had a browner colour, and had less opacity and purple colour than wines from the better quality grades. The relationship of grape nitrogen to the development of negative attributes or losses of positive attributes will require further study.
Based on observations from the 2017 season which showed that some vineyards had a large degree of variability in certain grape objective measures, the second year of the study aimed to look more closely at within-vineyard variation. From the analysis of the grape data in 2018/19, it was clear that significant within- and between-vineyard variability existed in the vineyards studied. The inclusion of variability data into the study was an important step forward, since it was found that high variability in phenolics, and high absolute levels of nitrogen (not necessarily nitrogen variability) were associated with ‘marginal’ vineyards (grade 2 vineyards not consistently performing to the expected wine quality outcome). This finding was not completely validated through the multivariate modelling approach and will require further research. Nevertheless, although grape nitrogen was generally highest in the ‘marginal’ vineyards, it was also found to be one of the grape-based measures which was generally more variable across all vineyards.
An important conclusion was that due to the importance of grape nitrogen to quality found in both seasons of the study, managing total vineyard nitrogen levels, as well as potentially reducing within-vineyard nitrogen variability, may be a factor of importance to improve quality across the board. Nitrogen management is also potentially important for quality control since DAP addition is a critical element of winemaking. Nitrogen fertiliser application in the vineyard and DAP addition in the winery should therefore be carefully considered in combination. To this end, growers and winemakers could apply different management strategies to heterogeneous vineyards to reduce variation and achieve quality goals. For example, a vineyard could be managed to reduce variability, or sub-sections of a vineyard could be harvested separately and subjected to different DAP addition protocols.
A final important result was that when using non-targeted grape-based measures in the 2018 season, prediction of both grape grades 1 and 2, as well as ‘marginal’ (grade 2/3) vineyards was possible with homogenate MIR spectra. The strongest prediction allowed the separation of ‘marginal’ vineyards from the other grape grade categories (R2 = 0.8, for 2/3 ‘marginal’ vineyards). Although not realised in the first season of the study, this was a promising result, with a more reliable and stronger model being produced using a readily available and easy-to-use method. Using MIR appears at this stage a most promising tool for detection of lower quality grapes early and quickly in the production chain. Future directions for the Australian wine industry would be to support the increased use of MIR for scanning grape homogenates early in the production process, not only for the prediction of grade but also for the development of rapid objective measures of grape composition for which calibrations due not yet exist (e.g. extractable phenolics) or which require further validation (e.g. YAN). A further key step would be to increase the skills base within the industry to use and apply non-targeted approaches such as MIR for the ongoing validation and prediction of vineyard and wine grade, and targeted improvements to management practices, according to individual producer needs.