Evaluation of ridge regression for country-wide prediction of genotype-specific grain yields of wheat

Abstract Large scale prediction of the performance of genotypes is fundamental for understanding genotype by environmental interactions (G × E), predicting accurately genotypic performance in specific environments, and increasing our knowledge to develop future crop varieties. We derived environmental limiting factors from daily weather data for critical crop growth phases of wheat ( Triticum aestivum ), using a winter wheat suitability model. The limiting factors that account for the effects of environmental variables on wheat productivity and phenology were then integrated into a matrix of environmental factors that related environments with observations through a ridge regression-best linear unbiased prediction (RR-BLUP) model. Prediction accuracy following a leave-one-site-out validation scheme was evaluated through correlations between predicted and observed yield for six winter wheat genotypes grown at 8 to 10 sites during three years. Accuracy (r = 0.01–0.75) was within the range of values reported in other studies. High prediction accuracies for certain sites and genotypes showed that the use of environmental limiting factors derived from gridded weather data into a RR-BLUP framework is a promising approach to predict genotypic performance in large areas. In contrast, the environmental and crop data collected in the variety trials and how these trials covered the territory limited the accuracy of predictions.