Application of unoccupied aerial vehicles to assess the genetics and environmental responsiveness of maize growth and utility in predicting performance

Abstract

Climate change is a major threat to global food security as current crop varieties may not be capable of thriving in new growing conditions. Developing more resilient crops requires advanced tools to speed up the process. Recent advancements in unoccupied vehicles, available sensors, and computing infrastructure are facilitating increased utility of high throughput field phenotyping for crop improvement. This dissertation focuses specifically on opportunities and challenges in high resolution phenotyping using unoccupied aerial vehicles mounted with red, green, blue sensors to study maize growth patterns and predict end of season performance based on those growth patterns. I demonstrate that temporally resolved growth patterns can reveal novel information about the polygenic nature of complex traits and that utilizing different traits to derive growth curves (e.g. plant height versus canopy cover) can provide complementary knowledge about both the genetics and environmental responsiveness of maize growth patterns. I also show how temporal plant height and growth rate measurements throughout a growing season can be utilized with machine learning models to predict end of season grain yield in both commercial fields with a single hybrid grown over a variable landscape and in research and development settings with highly diverse germplasm. This research highlights both the potential and current limitations of temporal data for understanding the complex interactions between maize and the environment and application in crop improvement and commercial management practices.