Broad-sense heritability in mixed models for grapevine initial selection trials

This work addresses to the genetic analysis and selection in populations where the whole genotypic value is transmitted through generations, using grapevine yield data as a case study. Several models were applied to different types of data sets. The individual and mean year yield and the balanced and unbalanced data resulting from various experimental designs (completely randomized, randomized complete block and row-column) were used. The aims of present work were to study: (1) the applicability of a generalised measure of broad-sense heritability to evaluate the success of the genotypic selection and compare it with the classical approach; and (2) the effect of different models on the accuracy and precision of the genotypic variance component and on the generalised broad-sense heritability estimates. The results showed that the computation of a measure of generalised broad-sense heritability is very feasible and useful for evaluating the efficiency of genotypic selection. In this study, 88 % of the fitted models did not comply with the standards for applying the classical concept of heritability. The differences between both the classical and generalised broad-sense heritability estimates increased with the complexity of the model. Higher broad-sense heritability estimates were consistently obtained with the mean years. The most accurate and precise estimates of the genetic parameters were obtained with the spatial models. Finally, the genotypic variance component of yield and the generalised broad-sense heritability were consistently significant for all grapevine varieties.