Agriculture and Food Systems - Theses

Permanent URI for this collection

http://hdl.handle.net/11343/433

Filters

2020 1
1
1
Reset filters

Settings
Statistics
Citations
Start date: 2020 × Subject: High-throughput phenotyping × Subject: Biomass yield × Subject: Sensor technologies ×

Search Results

Now showing 1 - 1 of 1
  • Item
    Thumbnail Image
    Development of Advanced Phenomics Tools for Molecular Breeding of Biomass Yield Improvement in Perennial Ryegrass (Lolium perenne. L)
    Gebre, Alem Gebremedhin ( 2020)
    Increasing the biomass yield of perennial forage crops remains a crucial factor underpinning the profitability of grazing industries, and therefore is a priority for breeding programs. The rate of genetic gain for dry matter yield (DMY) in forage crops is likely to be increased with the development and application of genomic selection (GS) strategies. However, realising the full potential of GS will require an increase in the amount of phenotypic data and the rate at which it is collected. Phenotyping remains a critical bottleneck in the implementation of GS in forage species. Current assessment of DMY in forage crop breeding includes visual scores, sample clipping and mowing of plots, which are often costly and time-consuming. New ground and aerial-based platforms equipped with advanced sensors offer opportunities for fast, non-destructive and low-cost, high-throughput phenotyping (HTP) of plant growth, development and yield in a field environment. This thesis aimed to develop, validate and deploy sensor-based non-destructive aerial, and ground-based HTP platforms to estimate DMY of perennial ryegrass plants in-field. Firstly, calibration and validation of aerial and ground based HTP platforms were developed to provide a non-destructive method to accurately estimate perennial ryegrass height and DMY. Secondly, the thesis compared a range of traits alone and in combination to demonstrate sensor based DMY estimation. Thirdly, this study demonstrated the application of combining normalised difference vegetative index (NDVI) from multispectral imaging and ultrasonic sonar estimates of plant height to estimate the seasonal distribution of DMY of 48,000 perennial ryegrass plants. Fourthly, the study discussed the application of developed platforms and data processing workflow which allows for an increase in the accuracy of genomic breeding values prediction in GS. Calibration and validation results indicated that plant height measurements from ultrasonic sonar and light detection and ranging (LiDAR) sensors showed the potential to measure plant height with consistent repeatability under controlled conditions and in field trials. NDVI demonstrated the capability for non-destructive DMY estimation of individual ryegrass spaced planted and sward plots. However, DMY estimation from NDVI saturates as the biomass increases. The combination of plant height and NDVI was found to improve the DMY prediction accuracy by up to 10%. This was achieved by combining NDVI and plant height with a simple multiplication combination (NDVI multiply by plant height), with the possibility to further improve the accuracy by combining the parameters in different models. This thesis assesses further the feasibility of combining NDVI from multispectral imaging and ultrasonic sonar estimates of plant height to estimate DMY of single plants in a large perennial ryegrass breeding program. Results demonstrated that the best prediction models of DMY of spaced-planted perennial ryegrass plants come from the multiplicative combination of NDVI and plant height (NDVIsq_PH). The K-fold and random split cross-validation findings imply that the combination of NDVI and plant height improved prediction accuracy over the use of NDVI and plant height alone. This yielded an accuracy of the coefficient determination of DMY estimation of more than 0.63 and root mean square error (RMSE) for fresh herbage yield, and dry herbage yield was less than 33 gram per plant and 8 gram per plant, respectively across multiple growing seasons. Therefore, to assess the application of combining NDVI and plant height for accurate DMY, a computational workflow was developed for image acquisition, data processing and analysis of spaced-planted perennial ryegrass plants. Fifty advanced breeding lines and commercial cultivars represented by a total of 48,000 individual plants were used to develop and validate the computation workflow for DMY prediction across three growing seasons. Combining NDVI and plant height of individual plants was a robust method to enable HTP of DMY estimation in a large population of ryegrass breeding. Similarly, the plot-level model indicated good to high-correlation between the predicted and measured DMY across three seasons with coefficient determination between 0.19- 0.81 and root RMSE values ranging from 0.09-0.21 kilogram per plot. The model was further validated using a combined regression of the three seasonal harvests. This study will have a significant contribution to the wide application of sensor technologies in forage research and plant breeding.