Amid the recent bushfires in Victoria, a winemaker watched as his winery in Tonimbuck – where he has lived in and worked at for 40 years – was destroyed by the blaze. Meanwhile, in northern NSW, the entire 2019 vintage of Topper’s Mountain Wines was lost just 5 hours before picking was scheduled to begin. 70% of the crop was destroyed in the fire whilst the remaining crop was ruined by smoke taint.
Considering the frequent nature of bushfires which are predicted to increase in frequency and severity, smoke taint is becoming an area of concern. When they happen around veraison, smoke contamination of grapevines and grapes can occur. This can result in unsaleable wines due to high levels of taint with associated financial losses for wine producers and on-flowing costs for future wine sales.
At the moment, there are no practical tools to determine which plants, vineyard sections or specific grape bunches have been affected by smoke contamination after a bushfire. This is where novel research involving remote sensing and machine learning comes in.
A trial to investigate the physiological effects of smoke contamination in four grapevine cultivars was established at a commercial vineyard in South Australia’s Adelaide Hills and at a research vineyard at the Waite campus of The University of Adelaide in the 2009-2010 season. The measurements which were taken into account included leaf conductance (gL), infrared thermography imaging (IRTI) and infrared index (Ig)2,3. Chemical analysis of the levels of the levels of smoke-related compounds from berries and resulting wine was also conducted. Results from the canopy level measurements showed that the changes in the pattern of gL within the canopies due to smoke contamination can be recognised using IRTI and Artificial Neural Networks (ANN). Furthermore, smoke-related compounds in berries and wine can be modelled using machine learning algorithms based on non-invasive near infrared spectroscopy (NIR) in berries. It took a few years to analyse the data using new and emerging tools for data analysis, such as new algorithms available in machine learning modelling. Now we are able to revisit data and generate useful models for growers (A. Prof. Sigfredo Fuentes from The University of Melbourne).
Hence, Machine learning algorithms and non-destructive measurements of grapevine canopies and berries alter bushfires for suspected smoke contamination offer powerful tools for growers to assess levels of contamination and to potentially map out affected areas. This information can be used to manage contamination in the field. By mapping contaminated grapevines, growers can perform differential harvests from non-affected areas to produce wines without smoke taint and avoid spoiling the whole production.
Research aside, it seems that amidst all the tragedy of bushfires, there is another light at the end of the tunnel for winemakers. Late last year, a Californian winery hit by a bushfire set out to make the best out of the tragedy. Smoke-tainted grapes were picked quickly and turned into raisins instead of being thrown away. The raisins ended up being “quite good, in fact, with no smoky notes — and contained the antioxidants of a glass of wine, but in concentrated, edible form.” Amazing!
At the end of the day, it is continual research and the dedication to absolute quality by winemakers that will ensure the minimisation of likelihood of a wine that you buy being contaminated by smoke.
Stay tuned on the blog for Parts 2 and 3 of the Tackling Smoke Taint series!
Bedford, M. (2019). Bushfire rips through Topper’s Mountain vineyard devastating rare wines on harvest day. ABC News.
Graham, B., Smith, R. and Loomes, P. (2019). Township ‘wiped off map’ as winemaker watches life’s work destroyed on live television. news.com.au.
Ogletree, K. (2018). Father-Son Duo Turns Ruined Grapes Into Tasty Aid For Napa Fire Victims. NPR.