Plant breeders can now make predictions about the crops they are growing using computer modelling based on images from space.

It sounds futuristic, but this is a reality for some canola breeders in Australia and Mexico following work done by researchers at The Australian National University (ANU).

ANU Research School of Biology Professor Robert Furbank says with multiple satellites now providing coverage of agricultural land there’s an opportunity to use the images to support significant improvements to crop breeding.

“The satellites can generate images at 30cm resolution, which is high res enough to clearly see individual breeding plots of 10m² and, in some crops, such as cotton or larger horticultural species, you could even see single plants,” Professor Furbank says.

In a project supported by AFII’s Strategic Investment Program (SIP), Professor Furbank and his team worked with machine learning experts at the ANU School of Computing, as well as partners at La Trobe University, Australian Grain Technologies, CIMMYT Mexico and SmartSat CRC to match the satellite images up with on-the-ground observations made by humans.

“We have partnered with Australian grains breeding companies to trial the use of machine learning to analyse high quality satellite images enabling breeders to identify which crops perform best in the field without needing to have people on the ground collecting data,” he says.

“As part of the project, ANU researchers developed deep learning algorithms to predict important crop attributes such as flowering time and biomass in canola and wheat which could otherwise only be obtained through expensive daily field trips.

“This allows breeders to obtain key data on growth, flowering and performance of their best germplasm in a way which is cheaper and less time intensive.”

This project has now expanded to include canola breeders at NSW Department of Primary Industries and Regional Development.

Professor Furbank, who led the project at ANU, said SIP funding for the project was critical to bridge the gap between fundamental research and industry application.

“Without this investment from AFII to purchase satellite images and re-task satellites to image our crop trials, this industry collaboration would not have been possible,” he says.

“With more than one million of these crop trial plots grown every year across Australia by grain breeders, the potential value to industry of adopting this use of satellite technology is enormous.

“Ready access to data of the quality we obtained in our project would enable breeders to make a step change in their breeding while also saving millions of dollars.”

The results from the first stage of this project are available here.