Inhalt und Ziel des Forschungsprojektes
Unser übergeordnetes Ziel ist es, einen Arbeitsablauf zu gestalten, der es erlaubt, die Dynamik der Entwicklung von Winterweizen im Feld zu charakterisieren. Dies soll anhand von Messungen der vergangenen Jahre geschehen, die an unserer weltweit einzigartigen Feld-Phänotypisierungs-Anlage durchgeführt wurden. Der Arbeitsablauf soll so allgemein gestaltet werden, dass er von anderen Arbeitsgruppen für ähnliche Zwecke übernommen werden kann. Zu diesem Zweck wird eine Skala der Pflanzenentwicklung so definiert, dass mit verschiedenen Kamera-Konfigurationen feststellbar ist, wann die Pflanzen in bestimmten Entwicklungsstadien sind. Eine spezielle Kamera-Konfiguration wird hierfür neu etabliert; die Erfassung verschiedener Umweltfaktoren wird es erlauben, die Reaktion von vielen Sorten auf Umweltveränderungen zu charakterisieren.
Wissenschaftlicher und gesellschaftlicher Kontext des Forschungsprojekts
Unsere Arbeit wird wichtige Erkenntnisse darüber liefern, wie sich Weizen an verschiedene Klimaszenarien anpassen kann. Dies ist relevant für die Verbesserung der Weizenzüchtung und damit für die Welternährung. Wissenschaftlich wird das Projekt vor allem einem verbesserten Verständnis der Ökophysiologie von Pflanzen dienen.
Plant development is a dynamic process. Morphology, coloration and physiological processes change throughout development and are affected by environmental factors. Annual crops such as wheat need to develop under strongly varying environmental conditions. Assessing the genotype by environment interactions in an automated manner throughout crop development would be highly beneficial to improve approaches in crop modeling and breeding. The overall objective of this project is to establish a workflow of robust, automated image processing procedures capturing the performance of winter wheat genotypes. Individual procedures exist for certain traits but are applied mostly on data of few seasons, lacking generic applicability. The specific aims of the project are to i) elaborate robust workflows capturing ‘simple’ and ‘dynamic’ crop traits in high-throughput field phenotyping campaigns mostly based on RGB images taken from close distance ii) determine the response of a high number of wheat genotypes towards several environmental factors at different developmental stages iii) utilize this knowledge to predict the performance of previously analyzed genotypes under current and future climatic scenarios. To achieve this, the project will analyze existing and newly acquired images of wheat throughout multiple years at very high spatial resolution (pixel resolution < 1 mm), temporal resolution (at least two imaging campaigns per week) and for many genotypes (dozens to hundreds). This will be done on more than 230’000 images acquired in the ETH field phenotyping platform FIP since 2015 as well as on new data to be acquired in the FIP with a new multi-camera rig rendering an annual data volume of up to 23 TB. Traits such as germination rate and time point, duration of stem elongation or time point of heading will be extracted from these images, leading to an image-based characterization of the developmental progress of all genotypes. Extraction of these traits will be performed with image processing approaches such as color segmentation and with methods of artificial intelligence such as machine learning algorithms and convolutional neural networks. The coordinate system, in which those traits will be quantitatively captured is an adaptation of existing developmental schemes termed DYNAPHEN scale. It allows efficient establishment of environmental response functions useful for crop simulation models. Since annotated images will be made publicly available, this image dataset will form an important training set for data challenges dealing with even more genotypes and environments. The immediate impact of PHENOFLOW is that it will provide a robust workflow for automated trait characterization from single or multiple images, taking environmental co-variables into account, which is required by high throughput field phenotyping approaches. Thereby the project will be of value for the entire field phenotyping community and PHENOFLOW will stimulate future research in crop modeling, breeding, agronomy and physiology. It will facilitate optimized crop management and increase the resilience of wheat to climate change by allowing for more targeted crop breeding.
Last updated:18.06.2022
