Grupos-chave de maturidade relativa e formação de mega-ambientes para cultivo e melhoramento de soja no Brasil

Abstract

Multi-environment trials (MET) are performed by soybean breeding companies to evaluate and select genotypes in a target region (TR). Data from MET are large and need be evaluated by accurate statistical methods. A widely used method to evaluate data from MET is called GGE biplot. GGE analyses allows the evaluation of genotypes, test locations and mega-environment (ME) formation. The aims of this work were: I) Evaluate the use of GGE analyses for plant breeding, indicating which parameters, in terms of singular value partitioning (SVP), data centering and data scaling methods, are appropriate for each analysis; II) Evaluate the formation of MEs for the edaphoclimatic region (ECR) 402 on Mato Grosso (MT) state; and, III) Evaluate and indicate the best maturity groups (MG) for each ECR on macroregions 1, 2, 3 and 4 of soybean cultivation in Brazil. All analyses were performed in GGEBiplot software and R platform, with trials performed between the 2012/13 to 2016/17 crop seasons. When the objective of GGE biplot analyses is to evaluate ME formation or test locations, the SVP with a f value = 0 should be used. When the focus is the evaluation of genotypes, f = 1 should be used. The data centering with focus in G + GE is appropriate to all GGE analyses from phenotypic data, independent if the focus of the analyses is to evaluate test locations or genotypes. When the objective of the analysis is to evaluate only the association between vectors (test location) or ME formation, the scalings 1 and 2 are the best options. When the objective is to evaluate only the discrimination power, scalings 0 and 4 are appropriate. Together analyses can be performed with scaling 2. To evaluate genotypes, the scalings 1 and 2 are more appropriate, especially when the dataset are unbalanced. All parameters should be present together with the analyses (biplots) for the correct interpretation of the biplots. The ME formation analysis were performed summarizing the results between years, with data from three years. Two MEs were identified on ECR 402, where the main factors that delimit ME are: presence of cist nematodes (Heterodera glycines), altitude and management level of farms. Thus, the selection and recommendation of genotypes should be realized within each ME and not in the whole ECR. The selection and recommendation in ME improve the selection efficiency of genotypes with widely and specific adaptation, improving the mean yield in all TR. Grain yield data from 133 genotypes with MG between 4.8 to 9.1 were used to define the best MG for cultivation in the macroregions of soybean production in Brazil. Trials were performed at 83 locations in the four main macroregions for soybean cultivation in Brazil. The best MG were identified for each ECR. With low latitude, higher MG need be used to improve grain yield. In the same way, in similar latitudes, but in different altitudes, the MG should be suitable, according the variations. For each macroregion (MR), the best MG were: MR1, genotypes with MG between 5.3 to 5.9; MR2, genotypes with MG 6.0 to 7.0; MR3, genotypes with MG 7.1 to 7.9, and MR4, genotypes with MG 7.7 to 8.4. The best MG for each ECR was similar of macroregions, varying according altitude and other biological factors.