- phenomaize@mrizp.rs
- Address: Slobodana Bajića 1,11185 Belgrade
About Project
Project Objectives (high-throughput field phenotyping)
The overall objective of the project is to explore the use of high-throughput field phenotyping (HTFP), specifically, Unoccupied Aerial Systems (UAS) derived data, in temperate hybrid maize breeding programs. Our aim is to expand knowledge and provide new skills regarding the use of UAS-based imaging in breeding and to introduce novel tools that would enable breeders to make more accurate and faster decisions compared to conventional selection approaches. Simply put, we want to build our capacities to conduct breeding more efficiently.
Specific PHENO_MaizE objectives (high-throughput field phenotyping):
- Collect UAS-derived RGB images of selected maize breeding populations at multiple time points throughout the season.
- Extract image variables for each flight (time point) in each environment.
- Use image variables to support decision-making in maize breeding.
- Acquire, share, and exchange new knowledge and skills to enhance the team and SRO excellence and visibility.
The main objective is to ensure undisturbed and efficient project implementation in a timely manner and within the budget constraints.
Description: The activities of Work Package 1 (WP1) include project management, financial management, and risk management. Specifically, WP1 involves initiating, planning, monitoring, executing, controlling, and completing the work of the project team to achieve the project objectives (MRIZP). Monitoring the project's compliance with the ethics code, data and privacy protection, and environmental protection is continuous. This will be accomplished through kick-off and regular meetings, data management and protection activities, timely purchase of equipment, adaptive resource and activity allocation, analysis of success for each project phase, and progress reports within the team and to the funding body. The Principal Investigator (PI) will organize kick-off and regular quarterly project meetings, with additional meetings scheduled as needed for adaptive planning and problem-solving during project implementation. The data management plan will be provided within the first three months of the project. Effective means of communication will be established to ensure the timely distribution of relevant information and documents.
The main objective is to collect Unoccupied Aerial Systems (UAS)-based visual images throughout the growing season and other single time-point measurements for 400 maize genotypes in three different environments.
WP2 activities are carried out in two stages. In the first year of the project, seeds for the multi-environmental field trials will be produced, and a successful protocol for field mapping will be established and tested. In the second and third years of the project, UAS-based field images will be collected from the field trials throughout the growing season, and field orthomosaics and DSMs (Digital Surface Models) will be generated for each time point and environment.
In all environments, terminal plant and ear height and stand count in each trial will be measured, while flowering times will be noted only in TC (test cross) trials. TC trials will be machine-harvested, and grain yield data, including grain moisture at harvest, will be collected in each location and year. TCs will be visually evaluated for overall appearance as well.
WP coordinator: Jovan Pavlov
The main objective is to extract plot-level image variables from each orthomosaic and DSM to tabular form.
The WP3 includes several steps: identification of individual plots, separation of plants from the ground, and finally extraction of plot-level information to tables.
The main objectives are to determine if digital traits can replace manual measurements, create prediction models for the traits of interest, and select the most informative image variables and flights during the season.
Initially, we will investigate whether accurately digitally derived plant number and terminal plant height can replace their manual measurements in the field. Additionally, we will define the optimal flight time relative to the maize growth stage to collect this image data. Subsequently, predictive models for traits of interest using image-derived variables will be developed. The variables demonstrating the most predictive power will be selected, and we will identify the most informative flights. Finally, orthomosaics will be tested in open-source machine learning frameworks to study deep-learning modeling.
Dissemination of the project results to both the scientific and non-scientific community and improvement of the team and scientific organization visibility are key goals.
Specific activities under Work Package 5 (WP5) include the establishment of the project webpage and promotion through social media channels. Additionally, dissemination to the scientific community will be achieved through publications and participation in conferences. Communication of the project results to the target audience will also be a priority.
IMPORTANCE AND RELEVANCE OF HIGH-THROUGHPUT FIELD PHENOTYPING
Plant breeding is a long-term and resource-demanding process, thus most research aims to explore different approaches that will improve accuracy and speed in plant selection. Phenomic selection is on its way to becoming a routine tool for use in plant breeding, but much more study needs to be conducted to enable wider adoption of the methodology.
The PHENO_MaizE will significantly contribute to the research field by studying different training populations and prediction scenarios using RGB HIGH-THROUGHPUT FIELD PHENOTYPING (HTFP) data. The obtained results are directly applicable to maize breeding. The first means of use use high-throughput field phenotyping, that will be tested is whether HTFP-derived data can replace manually measured data. Even this simple approach could greatly improve the breeding program.
Furthermore, PHENO_MaizE will contribute to the whole maize breeding community by providing answers to the next important questions, among others: how accurate is RGB UAS-based phenomic prediction, and which are the most important image-derived features and flights in temperate maize hybrid breeding?
The obtained image data will be kept for an unlimited amount of time and re-used at any point in the future for any purpose; thus, the potential for future research extension is unlimited.