Understanding the drivers of terroir in the Barossa Valley
Abstract
Despite being instrumental in selling a wine's story and commanding a price premium; terroir remains an enigmatic concept. This project aimed to scientifically investigate Barossa Shiraz terroir at regional, sub-regional and management scales using climatic, soil, vine, chemistry and sensory data. Binary models generally achieved good separation between Barossa Valley and Eden Valley within the Barossa Zone. Global models showed that samples from Northern Ground, Western Ridge and Eden Valley were relatively easy to classify compared to other sub-regions. This study improved our understanding of wine provenance and established novel approaches that can be applied to other wine growing regions.
Summary
Despite being instrumental in selling a wine’s story and commanding a price premium; terroir remains an enigmatic concept. This project aligned with the two main priorities areas in the 2015-2020 AGWA strategic plan – ‘Increasing demand and the premium paid for all Australian wine’ and ‘Increasing competitiveness’. This project aimed to understand how the vineyard drivers of wine uniqueness and hence terroir work, to determine the marker compounds or chemical profiles for regionality/unique Australian Barossa Shiraz wine and using this information to investigate strategies for manipulating wine quality in the vineyard to better express terroir. This included determining the key environmental drivers for fruit style/quality and how they can be better managed to best exhibit the typical style of the region in which they are grown.
The main findings from the study are outlined below.
Vineyard-only analysis provided a basis for considering sub-regionalisation of the Barossa GI. The analysis was confined to the land under vine and also made use of an updated climate dataset. The vineyard-only analysis clearly separated the Eden and Barossa Valleys, but within the Barossa Valley offers a basis for separation into three zones rather than a single one.
Yield and its components were measured in 24 Shiraz vineyards during three vintages in the Barossa zone. The frequency distribution of yield was L-shaped. The seasonal ratio of actual crop evapotranspiration and reference evapotranspiration was below 0.48 in most cases, highlighting the prevalence of a substantial water deficit in these vineyards. Yield gaps increased with increasing vine water deficit, quantified with carbon isotope composition in must. The yield gap was smaller with higher rainfall before budburst, favouring early-season vegetative growth and reproduction. Higher rainfall between flowering and veraison, favouring fruit set and early berry growth. The gap was larger with higher rainfall and lower radiation between budburst and flowering. The yield gap declined with altitude however the Eden Valley departed from this trend, with a yield gap that was higher than expected from its elevation. The yield gap increased linearly with vine age.
Vine phenology was assessed in three zones of 24 vineyards over three vintages using the E-L scale before veraison, and total soluble sugars (TSS) in berries during ripening. The spatial structure of the timing of phenological events followed gradients in topography and soils across the landscape, and were maintained despite the effect of the season (vintage). Phenology responded mainly to temperature until flowering, and to temperature and canopy size from flowering to maturity. The spatial variation in development was maintained despite vintage effects and management practices. Variation in temperature due to topography and elevation were the major drivers of vine phenological development until flowering. During ripening, development was driven by temperature and carbon capture and partitioning.
Of the grape berry chemistry measures conducted, grape volatile profiles gave the best evidence of some compositional attributes being aligned to the sub-regions. It appears that grape volatile compounds may be more affected variables at the geospatial scale rather than variables at the vineyard or intra-vineyard scale. Overlaying this data with the re-analysis of the geospatial clustering would be an interesting test of the hypothesis that this can drive the bound volatile compound profiles in Shiraz berries.
Analysis of the composition of grape anthocyanins, together with extractable measures of tannin and anthocyanin in grapes, statistical models were developed to attempt to distinguish the phenolic markers for each sub-region. Partial least squares-discriminant analysis (PLS-DA) models were only able to discriminate the Eden Valley samples from the other sub-regions over multiple seasons. Grape phenolics were found to be an important discriminating factor in the comprehensive multivariate analyses conducted using the entire dataset for the project.
For each season of the study, the wines from made from each site were analysed for anthocyanins, flavonols, tannin concentration and composition, polysaccharides, metals and multiple classes of volatile compounds. Phenolic compounds were important to all of the multivariate models developed, and it was of interest to note that for the Eden Valley, the key predictors were similar to the models developed for grapes on a seasonal basis. The results of the preliminary PLS-DA results have shown that wine composition could be used to discriminate the sub-regions on a seasonal basis, with greater consistency observed for the Eden Valley, and a lesser extent, the Northern Grounds.
Sensory analysis revealed that sensory profiles of Shiraz wines from the six Barossa sub-regions show several consistent sub-regional characteristics for both standardised small lot wines and industry scale wine samples across the four investigated vintages. Across the four vintages, Eden Valley, irrespective of winemaking, was separated most consistently compared to the other subregions. Western Ridge samples were consistently described as more textural wines for the small lot wines but this effect was not observed for the commercial samples. Additionally, wines from Northern Grounds were not associated with woody and intense fruit aroma and flavours in any of the vintages for the small lot set but were positively associated with the commercially produced ones. Indicating a consistent effect of the winemaking scale and practices on the sensory profiles.
Unsupervised clustering of samples was undertaken and the most consistent and stable k-means models were obtained with a two-cluster solution. The cluster solutions were consistent with cluster assignment and increasing vineyard elevation. A correlation to cluster assignment is also evident for soil cation exchange capacity, available water holding capacity, growing degree days, and mean January temperature with these measures of vineyard decreasing. The contributions of grape amino acids, grape volatiles, and wine volatiles to sample cluster assignments demonstrates the importance of grape and volatile composition to the two clusters in the data. Importantly, grape berry amino acid profiles can be modified through vineyard management such as nitrogen and water applications and these aspects of site management may account for some of the variance within the observed sample cluster assignment.
Interestingly, diethyl succinate, a volatile compound measured in grapes, was the only predictor ranked in the top 25% of importance in all regions as well as in the overall model. Other important predictors include, plant area index, sodium and cobalt in wine, as well as ethyl octanoate measured in grapes.
This study improved our understanding of wine provenance, and established novel approaches to measures of quality that can be applied to other wine growing regions. Grapegrowers, winemakers and wine brands can use this information to further inform their understanding of the expression of terroir in wine to help secure a sustainable future for the Australian wine industry.