“Accurately predicting harvest time and yield is the holy grail of agriculture. It allows clarity of availability to the entire food chain from the grower to the marketer and on to the consumer. Getting 90 percent accuracy rate up to six weeks out in our first three farms has been exceptional,” says CEO Darryn Keiller. Indeed: after three years of research, development, lab and farm trials; Autogrow’s FarmRoad solution has set a global benchmark for crop yield prediction with an initial 90 percent accuracy rate.
Under-production and over-production
“Under-production and over-production can financially impact a farm. Under supply brings both less revenue, potential financial penalties from purchasers e.g. supermarket chains and the need to buy off a competitor to meet contractual requirements. Over-production creates a surplus, which is then sold on the open market, usually at a price less than market value. It’s an unpalatable and expensive roller coaster ride.”
“When you consider the numbers, the ROI (return on investment) of increasing yield prediction of tomatoes by 10 percent - from 80 to 90 percent - based on a 30ha grower producing 60kg/sqm could be up to USD$1.3million. Savings can also be made with regards to labor by automating manual forecasting and through increased efficiency of farming practices.”
Darryn Keiller, CEO with Autogrow, launched the FarmRoad platform last year.
Large scale greenhouse tomato producers
FarmRoad’s Yield Prediction model has initially been created to service large scale greenhouse tomato producers combining the biophysical understanding of crop varieties, with crop and environmental data and proprietary A.I. based models and engines. The service is built and hosted on AWS cloud, and can be delivered to any enterprise farm operator, anywhere in the world.
“Not only do we have on average 90 percent accuracy but we are achieving that working with three different growers in Canada, Australia and New Zealand, using a mix of hydroponic substrate and soil and utilizing three different tomato cultivars - Marnax by Axia Seeds, Maxeza by Enza Zaden and Merlice by De Ruiter Seeds; showing the flexibility of our AI based prediction,” says Darryn.
90 percent accurate
Accurate yield prediction is dependent on available data and variables include weather, pest and pathogen events and management practices. The industry baseline for large scale greenhouse production ranges from no prediction, to 80 percent certainty up to two weeks in advance for the more experienced growers. FarmRoad is 90 percent accurate from one to six weeks in advance and anticipates achieving 95 percent accuracy within six months.
“Tomatoes are one of the most complex plants to apply yield prediction, but there is also a substantial amount of data available due to the crop registration techniques growers utilize. The key to prediction is the availability of data and we have been incredibly lucky to work with some fantastic growers with over 40+ years of experience who have shared their knowledge and data,” says Chief Technology Officer Jonathan Morgan.
“The first step is getting the data, but the biggest challenge has been turning the data into a form that works. When you look at environmental data, it is great for controlling systems but it’s not easy to go from the real time data to a prediction of how plants are going to grow. When you also add in unexpected variables like a grower changing from loose pick to truss tomatoes half-way through the growing cycle, then your accuracy rate can decrease.”
With yield prediction achieved, Autogrow is currently developing crop registration and crop planning services.
Personalized yield model
“Yield Prediction by FarmRoad, is one of our many industry leading services designed to enhance crop productivity and make farm operators more profitable. With their personalized yield model, growers can utilize their prediction and trends to materially improve farm profitability. The future is A.I. and digital farming, and we look forward to working with large scale growers who are looking to utilize their own data and gain a competitive and financial edge,” says Mr. Keiller.