Understanding and utilization of genotype-by-environment interaction in maize breeding

Abstract

Due to the interaction and noise in the experiments, yield trails for studying varieties are carried out in numerous locations and in the course of several years. Data of such trials have three principle tasks: to evaluate precisely and to predict the yield on the basis of limited experimental data; to determine stability and explain variability in the response of genotypes across locations; and to be a good guide for the selection of the best genotype for sowing under new agroecological conditions. The yield prediction without the inclusion of the interaction with the environments is incomplete and imprecise. Therefore, a great deal of breeding and agronomic studies are devoted to observing of the interaction via multilocation trials with replicates with the aim to use the interaction to obtain the maximum yield in any environment. Fifteen maize hybrids were analyzed in 24 environments. As the interaction participates in the total sum of squares with 6%, and genotypes with 2%, the interaction deserves observations more detailed than the classical analysis of variance (ANOVA) provides it. With a view to observe the interaction effect in detail in order to prove better understanding of genotypes, environments and their interactions AMMI (Additive Main Effect and Multiplicative Interaction) and the cluster analysis were applied. The partition of the interaction into the principal components by the PCA analysis (Principal Components Analysis) revealed a part of systematic variations in the interaction. These variations are attributed to the length of the growing season in genotypes and to the precipitation sum during the growing season in environments. Results of grouping by the cluster analysis are in high accordance with grouping observed in the biplot of the AMMI1 model.