When Karsten Nielsen was exploring a topic for his master’s thesis, he discovered no shortage of notions for using unmanned aerial vehicles (UAVs) in agriculture. But one idea stood out.
“I often found myself going down wormholes when reviewing literature,” he said. “There is an enormous amount of information that we know we could collect in research programs that just simply take such a large amount of time and effort that often they simply are not considered. Biomass is an example.”
Nielsen recently completed his thesis project for his master’s degree in the University of Saskatchewan (USask) College of Agriculture and Bioresources, using UAVs and associated computer image manipulation to tackle this hard-to-measure metric in crop development. The work supported his successful thesis defence and is the basis for an article he is preparing for a research journal.
“Biomass is an excellent indicator of plant growth rate and a tremendously important factor in determining yield,” he said. “It’s not the only factor, though. If a crop puts too much effort into biomass, it may put less energy into seed yield. There is a fine line.”
For farmers, how big and lush a plant is growing is a good indicator of crop health. For plant breeders, biomass can be used to answer important questions about plant growth and seed development.
To measure biomass, researchers must grow enough plants so some can be sacrificed to be regularly cut, dried, and weighed. It’s work that requires a lot of time and skilled hands.
“Biomass is almost never conventionally measured because it is a destructive measurement,” Nielsen said.
But what if biomass could be accurately measured, throughout the season, without destroying any plants? Nielsen wondered if UAVs and high-resolution digital photography could offer an answer.
He drew on the expertise of his co-supervisor Dr. Steve Shirtliffe (PhD), a field crop agronomist with the Department of Plant Sciences and senior researcher with the Plant Phenotyping and Imaging Research Centre (P2IRC). Shirtliffe is an expert in phenotyping using UAVs (phenotype refers to plants’ physical characteristics). For the plants themselves, Nielsen was guided by his co-supervisor, pulse crop geneticist Dr. Kirstin Bett (PhD) from the Department of Plant Sciences, who is also embracing UAV-generated data to further her own work.
Nielsen’s research meant becoming personally acquainted with some of the challenges that make researchers reluctant to measure biomass.
“By the end of the season, I needed a lot of space in the vehicle to move all of the material back to the lab to be weighed,” he said. “Keeping everything labelled appropriately became very important and the chance of losing data seemed high.”
Nielsen grew several varieties of lentil in five different Saskatchewan locations in 2017 and 2018, flying a UAV over the plots every two weeks, from when the lentil seedlings emerged to crop maturity. The advantages were immediately obvious.
“Flying the drone only took a small proportion of the time I allotted to data collection and took minimal effort,” he said. “If the field was muddy, I did not even need to enter it to fly the drone.”
The resulting images were combined into orthomosaics, that is, aerial photographs digitally stitched together and geometrically corrected so their scale is uniform, like a map.
Since the UAV images were taken from many different angles, it was also possible to create “3D point clouds.” For example, a single plant may show up in 20 different images, but from different angles. Using specialized software, these images can be combined and processed into a kind of virtual three-dimensional computer model for future analysis.
This is another powerful advantage of UAV-based imaging. Data can be gathered both for a specific project and to be set aside for future work. It also remains easily accessible.
“For example, while the purpose of one of my flights may have been to collect plot volume, I would have also been very easily able to get normalized difference vegetation index (NDVI), groundcover, and height information,” he said. “Even better, those images now serve as a record of the crop trial at that moment in time. If novel data extraction methods are developed in the future, they may be applied to the image set from an experiment that is no longer even in the field.”
Since it’s non-destructive, UAV-based imaging also allows researchers to look at the same plants as they develop over the growing season.
“Rather than just measuring the biomass at a key moment in the crop’s lifecycle, we could evaluate the rate of growth throughout the entire season,” Nielsen said.
UAV-based imaging does have limitations. Clouds, smoke from forest fires, and anything that throws a shadow can impair image quality. Plants themselves can block each other from the camera eye, distorting the data. On the operational side, high winds or rain can ground UAVs, and areas with high air traffic may need special permission from Transport Canada to operate. UAV operators in Canada also need to be licensed.
Back in the lab, data need to be analyzed, which requires highly skilled staff and specialized software, some of which is being developed at USask.
“The group at P2IRC has already made fantastic headway on that, so for a large number of traits, it is literally as simple as uploading and pushing ‘start,’” Nielsen said. “I expect the list of traits that this applies to will steadily continue to increase.”
Beyond adding a powerful tool to plant breeders’ toolkits, Nielsen speculates his project may even be immediately useful to producers for crops that are harvested for biomass, such as silage, forage, fibre for textiles, and bioenergy.
“I expect the indirect benefit to farmers through improved breeding efficiency will also be significant,” he said. “UAVs may allow varieties to be developed more rapidly producing higher, easier to grow crops. Desirable traits may be identified earlier, allowing them to be advanced more quickly.
“From a very large perspective, this will help to feed a growing global population.”
Agknowledge, December 2020