Abstract
Stalk lodging is a major limitation to yield in maize and is estimated to cause over $6 billion annually in global yield losses. It arises from the interaction of structural, geometric, and material properties within a hierarchical system. Accurate characterization of lodging resistance requires phenotyping approaches that capture this behavior across large, multi-environment datasets while maintaining consistency across experimental conditions. Existing methods are often limited by lack of standardization, sensitivity to experimental conditions, and incomplete data. This dissertation addresses these challenges through the development of an integrated framework for biomechanical phenotyping, experimental validation, and missing data imputation.A high-throughput phenotyping pipeline was developed to measure internode-level geometry and whole-stalk mechanical traits while preserving plant-level identity across extensive datasets. The influence of experimental conditions was quantified, showing that displacement rate produces measurable differences in flexural stiffness (4–7%) and rind puncture measurements, while moisture content introduces larger changes in flexural stiffness on the order of 10–15%.
A unified imputation framework was developed that accounts for hierarchical structure and mechanical relationships within the data. Continuity-based methods provided the most accurate estimates for internode-level geometric traits, while regression models using mechanically related predictors produced the highest accuracy for stalk-level traits. Machine learning methods provided moderate improvements but did not consistently outperform mechanistically informed approaches.
Collectively, this work establishes an approach for generating, validating, and completing phenotyping datasets with substantial sample size and coverage, enabling more accurate and reproducible characterization of maize stalk biomechanics and stalk lodging resistance across environments. These advances support improved crop breeding strategies, more effective phenotyping technologies, and enhanced food security in cereal production systems.