Eficiência da seleção visual para produtividade de biomassa verde de Brachiaria ruziziensis

Abstract

Visual selection is a practical alternative for plant breeding. However, its efficiency for visually discriminating higher productivity genotypes has been questioned. The aim of this work were to measure the efficiency of visual selection in relation to phenotypic mass selection, BLUP (Best Linear Unbiased Predictor) mean and selection by chance for the green matter yield of Brachiaria ruziziensis genotypes; and evaluate, for each mass selection strategy, the correspondence between the performance of the selected plants in unreplicated trial based on two cuts, and their clones evaluated in replicated trial and four cuts. We considered four mass selection strategies: vigor rating scale (visual), phenotypic means, BLUP means and by chance (no selection). The work was divided into two stages: 1) Evaluation of the visual selection efficiency in relation to the remaining selection strategies, in which were evaluated 2.309 genotypes, in augmented block design experiment, with two cuts; 2) Evaluation of the correspondence of the performance of individual plants and their clones, in which the clones of the plants selected were evaluated in triple lattice design experiment, with four cuts. In each stage, the traits evaluated were: vigor rating scale of the plant and green matter yield. The realized heritability for each selection strategy was estimated. Visual mass selection resulted in moderate efficiency in relation to phenotypic mass and BLUP for higher selection intensities, but presented good number of coincidence in low selection intensities. The realized heritability was higher for selection based on BLUP, followed by visual selection. The mass selection strategies by the phenotypic means, BLUP and visual presented similar gains. The effect of the genotype by environment interaction on the green matter yield was a limiting to the correspondence of the performance of the selected plants and their clones. Thus, selection must be performed with mild selection intensity. Visual selection may be a useful tool at initial stages of breeding programs for discarding genotypes.