AI provides sweet results for Australian berry farms
Harvesting is one of the most expensive operations in commercial berry production, and one of the more complex to manage. Now, the power of data science and artificial intelligence (AI) is being applied to the process, with benefits for growers, the food supply chain – and berry lovers.
At the start of 2020 one of Australia’s leading fresh food growers, the Costa Group, started rolling out a technology innovation: an in-crop sensor and data analytics system called Sensing+ and Sensing+ Enterprise Analytics, a new yield prediction module.
Sensing+, developed by agriculture technology startup The Yield Technology Solutions, combines sensors and analytics to produce information that helps large commercial growers of irrigated perennials make important on-farm decisions such as when to plant, feed, irrigate, protect and harvest.
The Yield Prediction module – the result of a collaboration involving The Yield, Costa Group, the Food Agility Cooperative Research Centre and the University Technology Sydney (UTS) – then helps growers more accurately predict how much crop they’ll be able to send to buyers like the major supermarket chains
“We want to better understand and manage the specific growing conditions that improve the quantity and quality of our yields,” says Costa Group CEO Harry Debney.
Costa’s berries are grown in protective tunnel structures. The Sensing+ system measures the growing conditions inside those microclimates, using AI to provide in-tunnel weather predictions.
“We’re also rolling out the Sensing+ Yield Prediction module, which predicts our berry yields using AI and data from both our harvest management systems and the microclimate weather,” Debney says. “We’ve been impressed with the accuracy achieved to date compared with our manual approach.”
Costa is rolling out Sensing+ and the new yield prediction module at eight berry farms in NSW, Queensland and Tasmania.
Sensing+ measures microclimate at the farm, block, row and even what’s known as the “hot spot” level every 15 minutes, the founder and Managing Director of The Yield, Ros Harvey, explains. AI and machine intelligence then predict growing conditions at the points that have been measured.
AI will one day be absolutely standard in the agricultural industry. Using AI we can reduce waste and produce more food.
This is more precise than information based on averages available via traditional weather services and a digital leap forward from traditional hands-on observations conducted periodically in the field.
The yield prediction module then combines the microclimate data with the grower’s historical and current crop data to forecast output.
“That helps the grower to better plan their labour, equipment and logistics to maximise efficiency and control costs,” says Associate Professor Daniel Ramp, who leads a team of researchers embedded in the Sydney office of The Yield for this project.
In turn, this means growers have more certainty and can negotiate better prices with big buyers. Even a small improvement in the accuracy of yield forecasts can mean significant financial benefits to a large producer, he says.
Harvey says AI will one day be “absolutely standard” in the agricultural industry. “Probably the biggest change this industry will have to make is that we’ll go from biophysically modelling the world to using AI to model the world,” she says. “If you look at the Yield Prediction project, it’s precisely that leap that has happened.”
Previously, people believed it was impossible to predict yield on berries because every single growth stage was on the plant at the same time, she says. “Can you imagine trying to create a biophysical model simulating that? It was impossible. But you can do it using AI.”
Harvey says optimising production is important amid predictions that the world will need 60 per cent more food by 2050. “If that’s the case, how are we going to feed the world without wrecking the planet? The answer is that by using AI we can reduce waste and produce more food.”
Associate Professor Ramp also sees wide and long-term impacts of the work. “Growers are getting a much better understanding of how weather patterns and longer-term climate trends are influencing their operations,” he says. “That not only helps their forward planning, it could also help the banking sector direct finance on those farms that are able to show sustainability and resilience to climate change.”
A team of three from UTS – experts in data analytics and AI – continue to work with The Yield to further develop the models for the yield prediction module, which has been commercialised just six months into a two-year research program.
“This is a great example of how, using data and agile research methods, we can get research results out of the lab and into the field faster,” Food Agility’s Chief Scientist, Professor David Lamb, says.
“The incorporation of the research team with an industry partner, the co-design of the fundamental questions, and the agile approach to project delivery ensured a direct pathway to impact,” he says. “It’s a great example of true collaboration – with our industry partners learning from our researchers, and our researchers learning from our industry partners.”
so many different agricultural companies struggle to forecast the amount of produce that they'll make things like climate weather disease the vagaries of production trade a whole lot of uncertainty in what they're going to produce so our goal in this project is to be able to reduce that uncertainty by using artificial intelligence and data analytic methods that enable us to quantify that yield in a much more certain way than they do in the past the IOT sensors are extracting really fine scale information about what's going on so soil moisture or weather variables like temperature and rainfall and relative humidity and so forth because there's a much better handle on what those plants are actually experiencing all of this goes into this melting pot of data analytics using Mount official intelligence and enables us to forecast or predict what whatever it is we want really and in this case it's yield or the timing event you we have an iterative framework whereby we regularly update the growers that enables us to test assumptions that we've made about the way in which their businesses work but it also enables us to get clarity around expectations of the end-user and to make sure that the outcomes of the work that we're doing is readily able to be adopted within their organizations to lead to success the exposure that customers really fine-tunes the relevance of what you're doing we've got customers who are super excited about what we're trying to solve its problems that they're struggling with and they're spending a lot of money on and that actually there's an opportunity cost in their business that we're trying to solve for them so they're very motivated to help us and I think that's fantastic because we've actually changed the way we think about the problem we've actually learned so much from them about what really matters so one of the really interesting things as part of the project has been sort of showing data back to the growers and different with different views and it was interesting because when we came to them we had ideas of what we thought might be important in the views that they would like to see in the day they'd like they'd be interested in seeing but then they came back with us with what they thought was important what and what might be driving some of the changes in some of the sort of responses were seeing in the day based on the last meeting with the farmers we find our definition of the harvesting event is somehow different from their definition and we adjust that for our data aggregation and then we get improve our model performance the most favorite part of this project is touching with the true industry of data science previously my experience is mainly for the research is totally it's not totally different but I say 80% different here we're focusing more for the true farmers what the their needs and how can we improve our data science for their requirements you know very rapidly I mean the project isn't it hasn't even been run in a year yeah and we've already seen lots of progress and lots of outcomes already coming from that from our work and that is really that is a really nice change of pace this project exemplifies the wonderful outcomes when you have alignment of values and practice the incorporation of a research team with an industry partner the co.design of the fundamental questions the agile approach to project delivery is guaranteed to ensure a direct pathway to impact and fast impact you