Repeatability of quantitative characteristics in sweet orange through mixed-model methodology

Abstract

The objective was to estimate the repeatability coefficient of quantitative traits in multiple harvests of sweet orange, in order to infer the minimum number of evaluations necessary to identify superior genotypes of orange trees through the methodology of mixed models. The experiment was conducted in randomized blocks containing 55 sweet orange genotypes and three replications. The repeatability coefficients were estimated using the maximum residual likelihood method (REML) and the prediction of genotypic values using the best unbiased linear predictor (BLUP). The Selegen software was used to perform the statistical analysis. The average heritability of genotypes in eight seasons, individual and eight seasons repeatability, selection accuracy in one and eight seasons, repeatability determination coefficient, accuracy of permanent phenotypic values based on M years of assessment and efficiency of M assessments compared to situation where only one assessment is carried out. The predictive accuracy of the selection revealed a significant degree of certainty in the inferences made. Evaluation over six seasons can increase accuracy to 70% in selecting sweet orange genotypes for yield-related traits. For the prediction of average fruit mass, seven harvests are enough to obtain 80% accuracy.