Novel field-based phenotyping methods for trait evaluation

Characterization of genetically modified organisms (GMOs) and gene-edited (GE) plants requires a significant amount of phenotyping and genotyping to determine whether the genetic modifications to the plant have a consistent and robust effect on plant phenotype. Costs to deregulate a GMO or to conduct whole-genome sequencing to demonstrate that an SDN-1 GE plant is GMO-free are high and are typically performed on the best-performing line(s) only. Therefore, researchers and breeders require extensive characterization of GMO and GE plant material to make the right selection. A current bottleneck is the phenotyping of GM/GE plant material in the field. Digital sensors and automation are being adopted more widely, allowing non-destructive and high-throughput measurements of plant performance. Facilities such as the NCRIS-funded Australian Plant Phenomics Facility (APPF) provide a range of phenotyping platforms for characterizing a large range of plant parameters. While this initially focused on greenhouse-based systems, field phenotyping has become a new focus.

Innovations in robotics and sensors have led to the development of mobile phenotyping platforms to characterize plant performance in the field. However, the number of parameters these systems can measure is still limited, requires validation, and as yet is not easy to deploy in a GMO field site.

An ARC LIEF grant awarded to the University of Adelaide, Australian National University, and the University of Melbourne has allowed the purchase of a modular robotic system (farm-ng Amiga) to be fitted out with a range of hyperspectral, LiDAR, and RGB sensors for characterizing plants in the field. These sensors replicate the types of sensors researchers and breeders use in the greenhouse at the APPF.

Unlike passive remote sensing, field-based LiDAR technology has the potential to capture unprecedented detail of plant 3D structure from individual plant to plot scale. The structural information plays a pivotal role in plant development. The candidate will be required to collect high-density point cloud and ground truth data using LiDAR from different agricultural experiments. This PhD project will focus on generating a voxel-based three-dimensional model from the LiDAR point cloud data, which will be meticulously processed through finely tuned mathematical functions to produce accurate canopy structural parameters such as leaf area index, leaf angle distribution, the proportion of botanical components, biomass, and plant height. This thesis also develops a framework to assess the associated uncertainty of above traits. Finally, the results will be evaluated under different conditions.