Reference Sets of Genotypes and Selection for Yield in Unpredictable Environments¹

Abstract

The concept of using a reference set of wheat ( Triticum aestivum L.) cultivars to monitor genotype × environment interaction and to position each environment relative to a breeding target was introduced. Techniques were examined for weighting trial information based on the similarity of trials to the target. These techniques were compared with selection based on mean performance before and after discarding trials with high coefficients of variation. Two weighting procedures based on distance measures were studied, namely canonical correlation and pattern analysis. Pattern analysis seemed superior but improvement compared to selection based on unweighted mean performance was marginal. The techniques may warrant further investigation for larger data sets and with a more restricted range of genetic diversity than used in this study.