Predicting Grain Yield based on Predictor Traits from High-Throughput Phenotyping, Genome-Wide Markers, and Pedigree to Increase Rates of Genetic Gain per Unit Time and Cost

Genomic selection (GS) and high-throughput phenotyping (HTP) promise to revolutionize crop improvement. Physiological traits measured using field-based HTP platforms could be useful as yield predictors. We tested if using aerial measurements of canopy temperature (CT) and green and red normalized difference vegetation index (GNDVI, RNDVI) as predictor traits in pedigree and GS models could increase prediction accuracy for grain yield (GY). Predictor traits on training and test sets, and GY on the training set were modeled as multivariate and compared to univariate models with GY on the training set only. Predictor traits increased prediction accuracy for GY by 11% and 15% for genomic and pedigree models, respectively. Pedigree models with predictor traits were 17% more accurate than GS models without predictor traits. This suggests that predictor traits measured by HTP can be used to enhance GS or as an alternative to GS when pedigree relationships are known.