Seleção, adaptabilidade e estabilidade genotípica na seleção de clones de cana- de-açúcar utilizando modelos mistos

• Ano de defesa: 2005
• Autor(a) principal: Bastos, Irlane Toledo
• Tipo de documento: Tese
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Universidade Federal de Viçosa
• Palavras-chave em Português: Cana-de-açúcar; Modelos mistos; Ciências Agrárias
• Link de acesso: http://www.locus.ufv.br/handle/123456789/10526

The augmented block design is commonly used in sugar-cane genetic improvement. The main reasons that made plant breeders adopt this design were the possibility of evaluating a large number of treatments without repetition in the initial stages of the programs and practicality of installing the trials in the field. However, the statistical analyses of this design have some limitations that compromise the accuracy of genetic parameter estimates. An alternative to evaluate more efficiently augmented block designs is the theory of models, since these are by definition incomplete and consequently unbalanced and non-orthogonal. Based on the theory of mixed models and using the Best Linear Unbiased Prediction- (BLUP), it was possible to predict the genotypic values of sugar-cane genotypes and from these to estimate the gains with the direct and indirect selection and the correlation among the parameters that evaluate the adaptability and genotypic stability according to Wricke (1956), Eberhart and Russell (1966) adapted by Carneiro (1998). A total of 72 sugar-cane genotypes were arranged in the augmented block design in seven environments, two experiments in each environment, with two repetitions. The traits appraised were: Brix, number of stalks per plot (NS), mean weight (kg) of stalks (MW), tons of Brix per hectare (TBH), and tons xof cane per hectare (TCH). The theory of mixed models applied to the augmented block design made it possible the selection of several clones superior to the controls RB72454 and RB835486. The direct gains were greater than the indirect ones in all the circumstances. The methodologies differences in relation to the bisegmented line pondered by the coefficient of residual variation (DRB-VC) and quadratic trapezium pondered by the coefficient of residual variation (QT-VC) were shown highly correlated to each other and with the genotypic values, classifying the most productive clones as giving the best performance. Unlike Wricke’s methodology (Wricke, 1956) that did not show significant correlation with the clones, but it was shown strongly correlated with the determination coefficient of the regression analysis. The parameter â ˆ 1i , which evaluates adaptability according to Eberhart and Russell (1966), had significant correlation with the genotypic values only in relation to the traits TBH and TCH. Whereas in all the appraised traits this parameter correlated significantly with DRRB- CV and TQ-CV methodologies, but in a more moderate way.