Computer simulation of a selection strategy to accommodate genotype environment interactions in a wheat recurrent selection programme

Abstract Multi‐environment trials (METs) are used in plant breeding programmes to evaluate genotypes (lines/families) as a basis for selection on expected performance (yield and/or quality) in a target population of environments (TPE). When a large component of the genotype environment (G × E) interactions results from crossover interactions, samples of environments in METs that deviate from the TPE provide a suboptimal basis for selection of genotypes on performance expected in the TPE. To adjust for the negative effects of these deviations, a selection strategy that weights the data from the MET according to their expected frequency of occurrence in the TPE (i.e. a weighted selection strategy) was investigated. Computer simulation methodology was used to obtain preliminary information on the weighted selection strategy and compare it to the traditional unweighted selection strategy for a range of MET scenarios and G × E interaction models. The evaluation of the weighted selection strategy was conducted in context with the germplasm enhancement programme (GEP) of the Northern Wheat Improvement Programme in Australia. The results indicated that when the environments sampled in the MET matched those expected in the TPE, the unweighted and weighted selection strategies achieved a similar response to selection in the TPE. However, when the environments sampled in the MET did not match the expectations in the TPE and a large component of the G × E interactions resulted from crossover interactions, the weighted selection strategy achieved a greater response to selection in the TPE. The advantage of the weighted strategy increased as the amount of crossover G × E interaction increased or fewer environments were sampled in the METs.