The development of improved varieties relies on identifying the best performing entries from a breeding population. However, high-throughput phenotyping tools beyond the plot harvester are still rare (White et al., 2012) and not used by Australian breeders. Whilst automated, high-throughput phenotyping platforms in greenhouses for the collection of precise phenotypic data are increasingly available, they operate in fully controlled environments, and QTL and candidate genes identified may not show a positive effect under actual field conditions. Another limitation of such phenotyping platforms is that they cannot address the need of breeders to screen thousands of entries in multiple environments. Unmanned Aerial Vehicles (UAVs) are a promising alternative: they can be easily transported, carry multiple sensors, and are becoming increasingly affordable (Zaman-Allah et al., 2015). In addition, they can characterize large numbers of plots within minutes which is important to avoid the effect of environmental changes and diurnal physiological processes on observed characteristics.
Given the potential offered by UAV’s, we are currently assembling an aerial imaging platform for field phenotyping. The UAV will carry several sensors, including an RGB camera, a multi-spectral camera and a thermal IR camera. This will allow us to phenotype a variety of plant and growth characteristics in actual field conditions, and to correlate these observations with, for example, plant biomass development, stress tolerance, and grain yield. The final product is targeted at breeders, enabling them to characterize their breeding materials faster, better, and cheaper, but the tool will also enable researchers to better understand germplasm performance in the target environment.
The first step consist of finalizing the UAV platform and establishing automated/semi-automated data extraction routines. Also, optimum operational parameters of the UAV need to be determined initially.
Then, the UAV platform will be used for extensive phenotyping of a limited number of field trials, covering breeders’ trials (and their specific objectives) as well as researchers’ trials (e.g., NUE mapping populations, trials for genome wide association studies (GWAS) and near association mapping (NAM)). Aerial imaging during the growing season will be accompanied by ground-based phenotyping of selected plant characteristics.
After acquiring the field image data, extraction of quantitative information from these images is the next critical step. Specific image analysis methods for UAV applications will be developed and implemented with the help of the team at the Phenomics and Bioinformatics Research Centre at the University of South Australia. Targeted is also the development of specific algorithms, focusing on the detection and characterisation of e.g., individual plant organs or of specific features related to abiotic stresses or diseases. Application of statistical and bioinformatics tools will then further improve the image analysis and the automation procedures to correlate field image findings with known information of direct relevance to plant biology.
Trevor Garnett (program 3 leader, UoA)
Stan Miklavcic (CI, UniSA)
Marie Appelbee (PI, LongReach)
Zohaib Khan (researcher, UniSA)
Stephan Haefele (CI, Rothamsted Research)
Vahid Rahimi-Eichi (PhD student, UoA)
Yuriy Onyskiv (technician, UoA)
Sanjiv Satjia (technician, UoA)
White JW, Anrade-Sanchez P, Gore MA, Bronson KF, Coffelt TA, Conley MM, Feldman KA, French AN, Heun JT, Hunsaker DJ, Jenks MA, Kimball BA, Roth RL, Strand RJ, Thorpe KR. (2012). Field-based phenomics for plant genetics research. Field Crops Res 133: 101-112;