Characterization of Soybean Grain Yield and Drought Tolerance Through Precision Phenotyping Techniques.

To feed the ever-growing population, new technologies that focus on increasing productivity with less input resources are crucial. With the introduction and wide implementation of high-throughput genotyping techniques and the significant cost reduction of these technologies, phenotyping has become the time and economic constraint in breeding and genetic programs. The use of precision phenotyping technology such as hyperspectral canopy reflectance measurements can alleviate some of this bottleneck and allow researchers to explore the genotype to phenotype relationship of complicated quantitative traits such as yield and drought/heat tolerance. The current experiment focuses on establishing a relationship between hyperspectral canopy reflectance measurements, canopy temperature, and chlorophyll content to grain yield across a wide range of genotypes, using prediction models created through partial least squares (PLS) analysis and multiple linear regression (MLR).  The experiment was conducted in rain-fed and irrigated environments in Manhattan, Kansas with Maturity Group III (MG III) and Maturity Group IV (MG IV) soybeans. PLS models were used as a data reduction technique to identify specific band regions contributing significantly to yield estimates. Candidate band regions were then subjected to MLR with backwards selection and selected based on coefficient of determination values, root mean square error, and the Akaike Information Criterion (model fit). Models exhibited coefficient of determination values ranging from 0.84 to 0.99. Canopy temperature, near-blue, green, red edge/shoulder, and the near infrared were selected most consistently within the models. Results indicate that separating irrigated and dryland environments created more precise predictive models than maturity group separation, suggesting stress indicators were sample independent. Also, results indicate using the drought tolerance index (ratio between irrigated and dryland environments), increases our ability to distinguish promising drought stress models compared to predicting irrigated and dryland yields separately.