The Application of Next Generation Phenotyping Tools to a Wheat Breeding Programme

Abstract

With the advent of high-throughput genotyping modern plant breeding has reached a new frontier of high-volume, high-density, yet low-cost, genomic data. Previously the acquisition of this data has been a logistical bottleneck within breeding programmes, yet with genomic data now abundantly available to breeding programmes, it has been speculated that the collection of phenotype data will become the next operational bottleneck. That being the inability to phenotype all material for all desired traits within a programme . The journey to improve the collection of phenotypic data is well underway, with focus being placed upon next generation phenotyping (NGP) technologies, such as high-throughput field phenotyping systems, to aid in the pairing of genotype to phenotype. Numerous sensors and methods of deployment have been investigated for application within small-plot field trials and suggested as tools for wheat and other field-crop breeding programmes, though few have explored how these can be deployed at scale or the suitability of collected data for use by breeders. This thesis investigates the deployment of commercially available digital cameras and LiDAR sensors within large-scale wheat breeding field trials, assessing the suitability of collected data for its application within the analytical pipelines of breeding programmes. Digital cameras were deployed opportunistically within large-scale wheat breeding trials, and through basic open-source image analysis methods, were capable of objectively assessing colour-based traits traditionally scored with visual assessment, producing levels of heritability similar to or greater than traditional methods. As part of this process a tractor-based high-throughput phenotyping platform was developed for the deployment of digital cameras, leveraging upon infrastructure present within the breeding programme and enabling images to be captured at a speed of 7,400 plots per hour. Given the success of digital cameras to measure colour-based traits, digital cameras were also deployed manually at a small scale to measure above ground biomass, plant height and harvest index, using photogrammetric techniques. Though data capture and processing methods were low-throughput, correlations between digital and manually collected measurements were strong (up to r = 0.94), highlighting the potential of the three-dimensional point cloud data type. To further this investigation LiDAR sensors were deployed on the high-throughput phenotyping platform to collect point cloud data of wheat plots from multiple field sites and collection dates. Processed point cloud data correlated strongly to traditional measurements of above ground biomass and canopy height and was shown to be highly repeatable and suitable for integration in routine breeding analyses. The findings of this work demonstrate that commercially available digital cameras and Li- DAR sensors can be deployed within large-scale wheat breeding trials, in a high-throughput, non-destructive and non-disruptive manner, for the accurate and repeatable measurement of traits which are traditionally subjective, laborious and/or destructive. Investigation of these measurements showed their suitability for inclusion within routine breeding analyses, giving breeders confidence in the data collected by next generation phenotyping technologies. The findings of this work are not only relevant to wheat breeders, but also to breeders of other field-crops and scientists conducting field research at a large scale.