Proposta de estratificação de ambientes-alvo e definição do número ótimo de locais para ensaios multiambientes de soja no Brasil

Abstract

New challenges for crop production have intensified due to climate change. Soybean (Glycine max (L.) Merrill) is one of the crops most affected by these climatic uncertainties. Because of the strong genotype × environment interaction (G×E), breeders are expected to face increasing difficulty in selecting superior genotypes with annual genetic gains above historical values. One way to mitigate this difficulty is to reduce G×E effects. Reclassifying the target population of environments (TPE) into more homogeneous groups can be an effective alternative, since greater homogeneity in the target region reduces G×E effects and leads to more accurate genotype placement and selection. The objective of this study was to identify the formation of mega- environments (MEs) and to redefine the TPEs for each Brazilian soybean macroregion, as well as to determine the optimal number of locations and select key sites for conducting multi-environment trials (METs). We used 20 years of historical environmental data comprising 10 environmental covariates from 150 soybean growing locations across Brazil to assess ME formation and to redefine the TPEs. Phenotypic grain-yield data from eight years (2012–2020) nationwide were used to calculate the optimal number of locations (NOL) and to identify key MET sites for each edaphoclimatic region (REC). In this study, Brazil was divided into five macroregions and twenty RECs, differing from the currently used division. Each REC exhibited a distinct environmental profile. According to environmental typing, drier conditions were observed in ME2, making it the most challenging ME for soybean production in Brazil, while ME1 and ME3 showed milder temperatures. ME4 was characterized by good water availability, and ME5 by warmer climates. Additionally, ME1 and ME2 require attention to sowing windows due to variation in photoperiod duration during the soybean growing cycle. A redistribution of NOL is necessary: some RECs require an increase in locations, while others allow reductions. Overall, an 18% reduction in the number of locations is possible without compromising genotype selection or selection precision. Key sites for each REC were identified based on parameters that ensure high-quality, reliability, and TPE-representative data, supporting the selection of promising genotypes.