GiESCO 2019 banner
IVES 9 IVES Conference Series 9 GiESCO 9 Machines and fire: developing a rapid detection system for grapevine smoke contamination using NIR spectroscopy and machine learning modelling

Machines and fire: developing a rapid detection system for grapevine smoke contamination using NIR spectroscopy and machine learning modelling

Abstract

Context and purpose of the study – Bushfires are a common occurrence throughout Australia and their incidence is predicted to both rise and increase in severity due to climate change. Many of these bushfires occur in areas close to wine regions, which receive different levels of exposure to smoke. Wine produced from smoke-affected grapes are characterised by unpalatable smoky aromas such as “burning rubber”, “smoked meats” and “burnt wood”. These smoke tainted wines are unprofitable and result in significant financial losses for winegrowers. This study investigated the use of near-infrared (NIR) spectroscopy and machine learning (ML) modelling for the rapid and non-destructive detection of grapevine smoke exposure by analysing grapevine leaves and/or grape berries.

Materials and methods – The trial was conducted during the 2018/2019 season at the University of Adelaide’s Waite campus in Adelaide, South Australia (34° 58’ S, 138° 38’ E) and involved the application of five different smoke and water misting treatments to Cabernet Sauvignon grapevines at approximately seven days post-veraison. Treatment vines were exposed to straw-based smoke for one hour under experimental conditions described previously by Kennison et al. (2008) and Ristic et al. (2011). Near-infrared (NIR) measurements were then taken from berries and leaves a day after smoking using the microPHAZIR TM RX NIR Analyser (Thermo Fisher Scientific, Waltham, USA) which has a spectral range of 1600-2396 nm. The NIR spectra were then used as inputs to train different ML algorithms, which resulted in two artificial neural networks (ANNs) with the best classification performance for either berry or leaf readings according to the different smoke treatments.

Results – Both ANN models found were able to correctly classify the leaf and berry spectral readings with high accuracy. The leaf model had an overall accuracy of 95.2%, 97.7% accuracy during training with a mean square error (MSE) 0.0082, 90.9% during validation with a MSE of 0.0353 and 88.1% during the testing stage with a MSE of 0.0386, while the berry model had an overall accuracy of 91.7%, 95.2% accuracy during training with a MSE of 0.0173, 86.4% during validation with a MSE of 0.0560 and 80.2% during the testing stage with a MSE of 0.0560. These results showed the potential of developing a rapid, non-destructive, in-field detection system for assessing grapevine smoke contamination following a bushfire using NIR spectroscopy and artificial neural network modelling.

DOI:

Publication date: September 28, 2023

Issue: GiESCO 2019

Type: Poster

Authors

Vasiliki SUMMERSON, Claudia GONZALEZ VIEJO, Damir TORRICO, Sigfredo FUENTES*

The University of Melbourne, School of Agriculture and Food, Faculty of Veterinary and Agricultural Sciences, Parkville 3010, Victoria, Australia

Contact the author

Keywords

bushfires, machine learning, smoke taint, climate change, non-destructive

Tags

GiESCO | GiESCO 2019 | IVES Conference Series

Citation

Related articles…

Bioclimatic shifts and land use options for Viticulture in Portugal

Land use, plays a relevant role in the climatic system. It endows means for agriculture practices thus contributing to the food supply. Since climate and land are closely intertwined through multiple interface processes, climate change may lead to significant impacts in land use. In this study, 1-km observational gridded datasets are used to assess changes in the Köppen–Geiger and Worldwide Bioclimatic (WBCS)

Variety and climatic effects on quality scores in the Western US winegrowing regions

Wine quality is strongly linked to climate. Quality scores are often driven by climate variation across different winegrowing regions and years, but also influenced by other aspects of terroir, including variety. While recent work has looked at the relationship between quality scores and climate across many European regions, less work has examined New World winegrowing regions. Here we used scores from three major rating systems (Wine Advocate, Wine Enthusiast and Wine Spectator) combined with daily climate and phenology data to understand what drives variation across wine quality scores in major regions of the Western US, including regions in California, Oregon and Washington. We examined effects of variety, region, and in what phenological period climate was most predictive of quality. As in other studies, we found climate, based mainly on growing degree day (GDD) models, was generally associated with quality—with higher GDD associated with higher scores—but variety and region also had strong effects. Effects of region were generally stronger than variety. Certain varieties received the highest scores in only some areas, while other varieties (e.g., Merlot) generally scored lower across regions. Across phenological stages, GDD during budbreak was often most strongly associated with quality. Our results support other studies that warmer periods generally drive high quality wines, but highlight how much region and variety drive variation in scores outside of climate.

Climate and the evolving mix of grape varieties in Australia’s wine regions

The purpose of this study is to examine the changing mix of winegrape varieties in Australia so as to address the question: In the light of key climate indicators and predictions of further climate change, how appropriate are the grape varieties currently planted in Australia’s wine regions? To achieve this, regions are classified into zones according to each region’s climate variables, particularly average growing season temperature (GST), leaving aside within-region variations in climates. Five different climatic classifications are reported. Using projections of GSTs for the mid- and late 21st century, the extent to which each region is projected to move from its current zone classification to a warmer one is reported. Also shown is the changing proportion of each of 21 key varieties grown in a GST zone considered to be optimal for premium winegrape production. Together these indicators strengthen earlier suggestions that the mix of varieties may be currently less than ideal in many Australian wine regions, and would become even less so in coming decades if that mix was not altered in the anticipation of climate change. That is, grape varieties in many (especially the warmest) regions will have to keep changing, or wineries will have to seek fruit from higher latitudes or elevations if they wish to retain their current mix of varieties and wine styles.

Exploring resilience and competitiveness of wine estates in Languedoc-Roussillon in the recent past: a multi-level perspective

The Languedoc-Roussillon wineries are facing a decline in wine yields particularly PGI yields due to many factors. Climate change is just ones, but is expected to increase in the future. There is also structurally a large heterogeneity of yield profiles among terroirs, varieties and strategies. This work investigates the link between yield, competitiveness and resilience to explore how resilient winegrowers have been in the recent past. To this end two approaches have been combined; (i) an accountancy database analysis at estate scale and (ii) municipality level competitiveness analysis. A new resilience indicator that characterizes the capacity of an estate to absorb yield variation is also defined. The FADN database between 2000 and 2018 of ex-Languedoc-Roussillon (France) and other data are used to analyse the current situation and the past evolution of competitiveness and resilience by type of estate (type of farm: PGI and/or PDO & type of commercialization: bulk and/or bottles). The net margin, which defines competitiveness, is not correlated to yield for all types but depends on the type of commercialization and the level of specialisation. The resilience indicator shows that the net margin of estates specialized in PGI is particularly sensitive to yield declines. We also show that price evolutions seem to compensate the effect of yield losses for the majority of types. Municipality scale analysis shows the links between local pedoclimate, yield, commercialization strategies and price. Overlapping a PDO with a PGI does not always increase a municipality’s PGI competitiveness. It is difficult to make links between causes and effects due to the complexity of the wine production system. Production diversification may be a solution. Resorting to the two level of analysis helps resolving the data gap that is necessary to explore the links between yield and economic performance of the wine estates in the long term.

Elucidating vineyard site contributions to key sensory molecules: Identification of correlations between elemental composition and volatile aroma profile of site-specific Pinot noir wines

The reproducibility of elemental profile in wines produced across multiple vintages has been previously reported using grapes from a single scion clone of Vitis vinifera L. cv. Pinot noir. The grapevines were grown on fourteen different vineyard sites, from Oregon to southern California in the U.S.A., which span distances from approximately hundreds of meters to 1450 km, while elevations range from near sea level to nearly 500 m. In addition, sensorial (i.e. aroma, taste, and mouthfeel) and chemical (i.e. polyphenolic and volatile) differences across the different vineyard sites have also been observed among these wines at two aging time points. While strong evidence exists to support that grapes grown in different regions can produce wines with unique chemical and sensorial profiles, even when a single clone is used, the understanding of growing site characteristics that result in this reproducible differentiation continues to emerge. One hypothesis is that the elemental profile that a vineyard site imparts to the grape berries and the resulting wine is an important contributor to this differentiation in chemistry and sensory of wines. For example, various classes of enzymes that catalyze the formation of key aroma compounds or their precursors require specific metals. In this work, we begin to report correlations between elemental and volatile aroma profiles of site-specific Pinot noir wines, made under standardized winemaking conditions, that have been previously shown to be distinguished separately by these chemical analyses.