News

Early and accurate predictions of grape yields

AusCover’s Michael Schaefer (left) and Alex Held (right) conducting scans at a vineyard using the DWEL ‘echidna’ laser scanner in August 2014.  Repeat scans will generate data that can be used to monitor growth and accurately predict grape yields.

Two exciting new TERN projects show signs of bearing fruit for the Australian viticulture industry by assisting with yield and quality prediction.

FAST FACTS

TITLE:  Accurate and early yield predictions through advance statistical modeling

PARTICIPANTS:
Macquarie University (Van Sluyter, Ghan, Medlyn), Treasury Wine Estates (Petrie), NSW Department of Primary Industries (Dunn)

GRANT:
Australian Grape and Wine Authority

TIMELINE:
Oct 2014 – Sep 2017

OUTPUTS:
Web based computational tool to estimate grape yield: Vineyard data database; Peer reviewed publications

USED eMAST INFRASTRUCTURE:
eMAST bioclimatic Variables; NCI computational space

TERN is pleased to announce its assistance with the development of a grape yield prediction tool that will provide growers with accurate and early yield predictions. Macquarie University, in partnership with industry collaborators, is developing the tool with funding from the Australian Grape and Wine Authority. TERN’s Ecosystem Modelling and Scaling Infrastructure (eMAST) facility is assisting with pairing current and pre-existing weather data to selected grape regions.

Yields are influenced by weather up to 18 months prior to harvest and by crop load and vegetative growth of the previous season. As a result, yields vary substantially year-to-year because of compounding weather effects, with costly unpredictability. The general ways in which weather affects individual components of yield, such as the initiation of flowers, are known. However, a model incorporating the multiple factors affecting yield does not exist.

The premise of this new tool is to model the interactions of multiple variables that affect grape yield with advanced statistical methods not previously utilised in grape yield prediction. This project will use machine learning to create an accurate and easy to use yield prediction tool for use by Australian grape growers.

In a separate project, researchers associated with TERN’s AusCover facility are using TERN’s newest piece of research infrastructure, the DWEL ‘echidna’ laser scanner, to explore the possibilities of how this advanced technology can assist grape growers.

In August this year, the AusCover team completed the first round of scans with the echidna at a vineyard in south-eastern Australia.  The team will return to the site and conduct regular scans over the upcoming growing season to compile detailed imagery data on the growth of leaves and grapes over time.  Such data can then be used to accurately predict the yield.

The team hopes that in the not too distant future robotic laser scanners will be able to travel up and down the rows of a vineyard scanning and tracking not only grape growth, but also volume and chemistry.  Such a plan would no doubt help fully realise the potential of such advanced laser scanning equipment in an Australian agricultural context, and the AusCover team hopes to be able to carry out their plans in the near future, resources permitting.For more information on the DWEL laser scanner please contact Michael Schaefer of AusCover and CSIRO at michael.schaefer@csiro.au. For more information on the grape yield prediction tool contact Steve Van Sluyter of Macquarie University at steve.vansluyter@mq.edu.au. 

Two new TERN projects aim to accurately predict the yields of grapes like these Pinot Grigio wine grapes prior to harvest during vintage 2012 at Granton Vineyard in southern Tasmania (Photo courtesy of Flickr stefano_lubiana_wines CC BY 2.0)

Published in TERN newsletter November 2014

Share Article

Share on facebook
Share on twitter
Share on linkedin