Harvesting data to predict yield

CSIRO's Graincast app democratises data

BEST GUESS: CSIRO's Graincast app is harvesting publicly available data as an easy way to estimate soil moisture and yields - but there are trade-offs.

BEST GUESS: CSIRO's Graincast app is harvesting publicly available data as an easy way to estimate soil moisture and yields - but there are trade-offs.


The Graincast app harnesses CSIRO super computers to predict moisture and yield


A free app which hooks up to CSIRO super computers is being used by grain growers to estimate their cropping yields in real-time, however there are trade-offs.

Branded as Graincast, developers say the app is a test-case in the democratisation of agricultural data.

CSIRO group leader for integrated agricultural systems Dr Roger Lawes said Graincast had two major components, an app designed to give farmers information about their own paddocks, along with a broader technology which allowed the monitoring of Australian crop yields in near real-time.

"Graincast is technology designed to monitor the Australian agricultural landscape," he said.

"The app was designed after a number of farmers were interviewed to discover what sort of digital information they would like access to and how they would like to use it."

Dr Lawes said a significant number of the farmers interviewed indicated they wanted to have access to an app which would provide information on the levels of moisture in their paddock in real-time, without direct monitoring.

"In effect they wanted us to estimate it," he said.

"So we created the app so it could literally be operated for any paddock."

By utilising the national soil characterisation grid, spatial climate data from the Bureau of Meteorology along with the APSIM crop yield prediction model, Dr Lawes said Graincast supplied farmers with estimates of not only current soil moisture levels, but also real-time yield predictions.

"It showcased how we could democratise data and give it to farmers," he said.

"It is an app that hooks up to CSIRO super computers, so it is giving you in-paddock access to some of the most powerful computer facilities in the country.

"What we are trying to do is make the technology available to as many people as possible.

"But we are at the start of that journey, we are really at proof of concept stage."

As with any modelling system, there are limitations depending on the quality of the information available, junk in, junk out. Dr Lawes said there was an obvious trade-off between ease of accessing information and how specific and accurate the result was for an individual paddock.

"It is important to remember what we are using is publicly available information, and the best available information we can get today," he said.

"All data has errors and uncertainty attached, we acknowledge that.

"If for example growers wanted specific information about their soil and they were going to characterise the soil themselves with their agronomist, there are other apps out there that can take that detailed information to drive them."

Dr Lawes said Graincast app could be viewed as a tool to help make decisions easier.

"It's good for getting a sense of how the season is going and what your likely yield potential is," he said.

"However it is only good for winter cropping at this time, so if you were in a double cropping system it is probably going to be less successful."

Dr Lawes said Graincast had seen reasonable uptake through the proof of concept stage.

"We've had a reasonable uptake, as of last years over 200 growers from across Australia had accessed the app, we executed some 5000 individual workflows, and we limited the number of paddocks each farmer could access," he said.

"To make computing infrastructure available to farmers you have to think how you will deliver it, not only keeping the technology at the forefront, but also how you bring this computing power in a cost effective manner to industry.

"That's a challenge we have now embarked on."

The story Harvesting data to predict yield first appeared on Farm Online.


From the front page

Sponsored by