Development and Evaluation of Unmanned Aerial Vehicles for High Throughput Phenotyping of Field-based Wheat Trials.

Growing demands for increased global yields are driving researchers to develop improved crops, capable of securing higher yields in the face of significant challenges including climate change and competition for resources. However, abilities to measure favourable physical characteristics (phenotypes) of key crops in response to these challenges is limited. For crop breeders and researchers, current abilities to phenotype field-based experiments with sufficient precision, resolution and throughput is restricting any meaningful advances in crop development. This PhD thesis presents work focused on the development and evaluation of Unmanned Aerial Vehicles (UAVs) in combination with remote sensing technologies as a solution for improved phenotyping of field-based crop experiments. Chapter 2 presents first, a review of specific target phenotypic traits within the categories of crop morphology and spectral reflectance, together with critical review of current standard measurement protocols. After reviewing phenotypic traits, focus turns to UAVs and UAV specific technologies suitable for the application of crop phenotyping, including critical evaluation of both the strengths and current limitations associated with UAV methods and technologies, highlighting specific areas for improvement. Chapter 3 presents a published paper successfully developing and evaluating Structure from Motion photogrammetry for accurate (R2 ≥ 0.93, RMSE ≤ 0.077m, and Bias ≤ -0.064m) and temporally consistent 3D reconstructions of wheat plot heights. The superior throughput achieved further facilitated measures of crop growth rate through the season; whilst very high spatial resolutions highlighted both the inter- and intra-plot variability in crop heights, something unachievable with the traditional manual ruler methods. Chapter 4 presents published work developing and evaluating modified Commercial ‘Off the Shelf’ (COTS) cameras for obtaining radiometrically calibrated imagery of canopy spectral reflectance. Specifically, development focussed on improving application of these cameras under variable illumination conditions, via application of camera exposure, vignetting, and irradiance corrections. Validation of UAV derived Normalised Difference Vegetation Index (NDVI) against a ground spectrometer from the COTS cameras (0.94 ≤ R2 ≥ 0.88) indicated successful calibration and correction of the cameras. The higher spatial resolution obtained from the COTS cameras, facilitated the assessment of the impact of background soil reflectance on derived mean Normalised Difference Vegetation Index (NDVI) measures of experimental plots, highlighting the impact of incomplete canopy on derived indices. Chapter 5 utilises the developed methods and cameras from Chapter 4 to assess the impact of nitrogen fertiliser application on the formation and senescence dynamics of canopy traits over multiple growing seasons. Quantification of changes in canopy reflectance, via NDVI, through three select trends in the wheat growth cycle were used to assess any impact of nitrogen on these periods of growth. Results showed consistent impact of zero nitrogen application on crop canopies within all three development phases. Additional results found statistically significant positive correlations between quantified phases and harvest metrics (e.g. final yield), with greatest correlations occurring within the second (Full Canopy) and third (Senescence) phases. Chapter 6 focuses on evaluation of the financial costs and throughput associated with UAVs; with specific focus on comparison to conventional methods in a real-world phenotyping scenario. A ‘cost throughput’ analysis based on real-world experiments at Rothamsted Research, provided quantitative assessment demonstrating both the financial savings (£4.11 per plot savings) and superior throughput obtained (229% faster) from implementing a UAV based phenotyping strategy to long term phenotyping of field-based experiments. Overall the methods and tools developed in this PhD thesis demonstrate UAVs combined with appropriate remote sensing tools can replicate and even surpass the precision, accuracy, cost and throughput of current strategies.