Research trials taking place in Australian vineyards aim to find ways to use image-sensing technology to accurately estimate annual grape crop yields and replace the laborious and often unreliable manual system currently in use.

Being able to predict the annual grape yield from vineyards accurately and earlier in the growing season has long been high on the wish-list of wine producers.

Presently farmers or viticulturists have to manually sample grapes at different stages of growth and use past knowledge in the field to forecast eventual yield. It’s onerous and time-consuming and estimates early in the season can be out by more than 30 per cent, according to researchers.

But new technology being trialled by the University of New South Wales (UNSW) could provide the basis for a more reliable and inexpensive way to gather information needed for forecasting.

Three-tiered camera trial

The project by researchers from UNSW’s School of Mechanical and Manufacturing Engineering revolves around vehicles equipped with cameras. These are driven along vineyard rows collecting the video information to estimate yields earlier in the season, potentially before vines even begin to flower.

The project’s third and final season’s trials are now underway in vineyards in the NSW central west at Jarretts of Orange and in South Australia’s Clare Valley in vineyards owned by Australian wine giant, Treasury Wine Estates (TWE).

Funded by Wine Australia and being led by associate professor Gregory Dunn from the NSW Department of Primary Industries, who pioneered the development of objective measurement-based yield forecasting systems for the Australian wine industry, the team’s brief was to compare three different image-sensing systems: one based on a high precision camera, one on a GoPro camera and one on a smart phone.

Happily, the trials are showing that the more-low cost options – the GoPro and phone – are the most effective and practical.

Just a vineyard drive

Chief investigator Dr Mark Whitty says a big appeal of the method being trialled is its cost-effectiveness and ease of use, with data gathered while the vehicle driver is doing regular vineyard management.

“Whatever vehicle they might be using to drive through the vineyard can be used to collect the data as it moves along,” says Dr Whitty.

“It could be a tractor to do some slashing or applying fertilisers, or just a quad bike driving through for an inspection. The idea is you can just have the camera on the vehicle and do this without requiring any additional passes through the block – collecting the data during normal management processes. It makes it very simple for people to collect it.”

The information is then analysed by the research team using their software to create various yield maps then fed back to the farmers. The maps can help farmers identify ways to adjust management practices to boost yield and crop quality, as well as provide an overall block tonnage forecast.

“Based on the maps, [farmers] can do things like adjusting the water supply, bunch thinning if the crop load is too high, leaf plucking or shoot thinning,” says Dr Whitty.

Early results a boon

Early indicators are very positive, according to Dr Whitty.

“We’ve been able to do yield predictions just a few weeks into the season. Nobody else in the world that I’m aware of is able to provide such an early measure of shoots in the block and the mapping of the variation across the block. It means there is more time for the farmer to take intervention and that means it’s more cost effective.”

With industry estimates that $100 million is wasted each year on inaccurate yield forecasting, according to Dr Whitty, the payback for grape growers and wine producers could be substantial.

Certainly Gioia Small, Treasury Wine Estates (TWE)’ Regional Manager Sustainability, says the technology could reap big rewards by boosting the industry’s ability to better manage and plan for a season.

“Accurate crop forecasting is one of the wine industry’s high priorities,” says Small. “TWE invests significantly each year in understanding what the yields are likely to be.

“This knowledge drives agronomic decisions around vineyard management such as setting the right pruning levels as well as providing information that allows us to make the right grape purchasing decisions.

“Traditionally, crop forecasting involves destructive sampling of grape bunches to understand berry numbers and weights – this approach hasn’t changed much in the last 15 to 20 years. It is a laborious and time-consuming process. A faster, more accurate crop forecasting approach is something the industry has been looking at for some time.”

Looking to the future

Dr Whitty says the trials show that the combination of the smart phone – that can also be used to access an app being developed by UNSW’s Dr Scarlett Liu and Dr Stanley Lam for on-the-spot analysis of bunch numbers in the field – and the GoPro looks set to be the most effective approach. Both deliver strengths in the different data type each can gather, and together are a dynamic duo.

“The mobile phone gives us very small points of data in more detail-specific things like flower count and berry numbers on individual bunches, but doesn’t give us an overall picture of what’s happening everywhere in the block,” says Dr Whitty.

“That’s where the GoPro comes into it. You can use it to get wider coverage; the images can be used to map things like non-bearing sections of canopy or identifying missing vines.

“So they are complementary technologies in fact, and both very easy to use.”

“We’ve been able to do yield predictions just a few weeks into the season. Nobody else in the world that I’m aware of is able to provide such an early measure of shoots in the block and the mapping of the variation across the block. It means there is more time for the farmer to take intervention and that means it’s more cost effective.”

Dr Mark Whitty, Chief Investigator, UNSW School of Mechanical and Manufacturing Engineering