Adaptability and yield stability of soybean genotypes by mean Eberhart and Russell methods, artificial neural networks and centroid

Abstract

The soybean crop is prominent in national and international scenarios. A large part of the world production of soybean is cultivated in Brazil and this has been possible due to the performance of different technological areas, among them genetics and plant breeding. Soybean breeding has acted in the development and launch of new cultivars and for this it is required the studies of interaction genotypes x environments and those of adaptability and stability. Thus, the objective was to evaluate the adaptability and phenotypic stability of the grain yield of late-cycle soybean genotypes. Five experiments were conducted in the state of Minas Gerais, each of which was considered as an environment. In each, 17 soybean genotypes were evaluated in a randomized block design with three repetitions, for grain yield, in kg ha-1. The data were analyzed by means of individual (each environment) and joint analysis of variance. Subsequently, analyses of adaptability and phenotypic stability were performed using the methods of Eberhart and Russell (1966), Artificial Neural Networks (Nascimento et al., 2013) and Centroid (Rocha, Muro‑Abad, Araujo, & Cruz, 2005). The results indicated the classification of the analyzed genotypes for unfavorable, general or favorable adaptability, with high or low stability. DM-339 is indicated for favorable environments and UFV-18 (Patos de Minas), UFV91-651226, UFV99-8552093, UFV01-871375B, UFV01-66322813 and UFV99-8552099 are indicated as general adaptability, considering the three methods of adaptability and stability analysis.