Phenotyping of barley resistance to plant diseases on the leaf and greenhouse scale through hyperspectral imaging and data analysis

Abstract

Optical sensors become increasingly interesting as means to perform automated, non-invasive and objective monitoring in agriculture, but also have immense potential as tools for research and development. Especially hyperspectral sensors have proven to extract a multitude of relevant parameters about the interaction of plants with biotic and abiotic stresses alike through the large amount of precise information about changes in the plant spectral signature they can detect. However, despite multiple research studies, there are still areas within the field of hyperspectral imaging which have barely been investigated. <br /> This study focusses on three such areas, the potential of transmission based hyperspectral imaging for early disease detection and quantification, the early detection and quantification of disease symptoms on canopy scale in high-throughput, and the question of data compatibility of hyperspectral data from different scales. <br /> In order to achieve those goals three barley pathogens – Blumeria graminis f. sp. hordei, Puccinia hordei and Pyrenophora teres f. teres – were investigated in leaf scale time-series measurements, during which reflection and transmission data was gathered simultaneously through the HyperART setup. The effect of the distinct plant-pathogen interactions of the examined pathogens over the course of the inoculation were then used to evaluate the possibilities of transmittance hyperspectral images for early disease detection and quantification in direct comparison to the performance of reflectance data, giving valuable insights not only into the performance of transmission measurements but also about the underlying mechanisms through the interpretation of the results from the different pathogens. <br /> A high-throughput hyperspectral measurement system, which is able to perform time-series measurements in the greenhouse under field-like conditions on canopy level while being independent from environmental factors, was developed during the study. With this system it was possible to accurately detect powdery mildew symptoms on barley canopies and quantify the disease development automatically over the course of 30 days long time-series measurement, showing the possible advantages of non-invasive and objective measurements for phenotyping applications. Additionally, it was possible to detect and quantify necrotic lesions on the most resistant barley cultivar. These necrotic lesions could be linked via microscopic studies to resistance reactions of the cultivar upon powdery mildew inoculation, showing the possibilities to precisely track the progression of resistant plant cultivars on canopy level via automated hyperspectral measurement setups. <br /> Through the comparison of the hyperspectral data of time-series measurements investigating the plant-pathogen interactions of barley-powdery mildew, generated on both leaf and canopy level, through manual examination of the changes within the plant spectral profiles and the results of modern data analysis methods it was possible to confirm high similarities within both datasets. These results – and taking into account other recent studies with similar scope – support that hyperspectral data is comparable across multiple scales. These findings allow the usage of the wealth of results from scientific studies with hyperspectral imaging for more practical applications on higher scales.