ABSTRACT
Global food production needs to increase by more than 60% from 2015 to 2050 to meet the projected demand. At the same time, yield advances have slowed at both a production and a genetic improvement level. A better understanding of the traits that lead to greater yield potential and better adaptation to challenging climatic conditions is needed. While recent advances in DNA sequencing techniques have brought about high-throughput genotyping and allowed many breeding programs to assemble comprehensive genetic mapping resources, the tools to accurately phenotype a large number of genotypes in field trials has been lagging behind. With the advent of sophisticated spectral imaging, but also accurate geo-location devices, the development of proximal and remote-sensing high-throughput field phenotyping platforms has become feasible. Such platforms, combined with new innovative image-analysis and machine-learning tools, will likely become the benchmark in high-throughput plant phenotyping (HTPP) frameworks for determining plant responses at the leaf, plant, and canopy level. They promise to fill the phenotype-to-genotype gap that currently impedes the step change in plant breeding needed to meet the predicted increase in demand for food and fiber products. At the University of Queensland, we have developed a cost-effective HTPP platform that harvests sensing data from plants utilizing proximal point and spatial sensors. Here, we describe the set-up of a software pipeline to capture, manipulate, and analyze “BIG DATA” from proximal sensors onboard a tractor-based phenotyping platform. The development of these technologies not only allows the scaling up of cross-sectional phenotyping screens to the level required to keep pace with recent genotyping advances, but also increases the temporal resolution of phenotyping via the characterization of dynamic growth processes. We discuss the use of high-resolution characterization of cross-sectional and time-sequence data, obtained from a narrow-band hyperspectral camera, to derive estimates for crop cover and dynamic growth parameters in wheat and sorghum breeding trials. Application of these technologies across breeding plots will enhance phenotyping capabilities and hence the ability to discriminate among responses of genotypes. Furthermore, our research framework will enable the linking of “islands of knowledge” including Genomics, Phenomics, and Crop Modelling to deliver targeted outcomes across scales. This will enhance crop breeders’ understanding and knowledge of the relationships between traits, crop physiological processes, and genotypes across different environments.
