ABSTRACT.- In forest tree breeding programs, open-pollinated families are frequently used to estimate genetic parameters and evaluate genetic merit of individuals. However, the presence of selfing events not documented in the pedigree affects the estimation of these parameters. In this study, 194 open-pollinated families of Eucalyptus globulus Labill. trees were used to compare the precision of estimated genetic parameters and accuracies of predicted breeding values with the conventional pedigree-based model (ABLUP) and the pedigree-genomic single-step model (ssGBLUP). The available genetic information for pairwise parent-offspring allows us to estimate an actual populational selfing rate of 5.4%. For all the growth and disease resistance traits evaluated, the inclusion of selfing rate was effective in reducing the upward bias, between 7 and 30%, in heritability estimates. The predictive abilities for ssGBLUP models were always higher than those for ABLUP models. In both cases, a considerable reduction of predictive abilities was observed when relatedness between training and validation populations was removed. We proposed a straightforward approach for the estimation of the actual selfing rate in a breeding population. The incorporation of this parameter allows for more reliable estimation of genetic parameters. Furthermore, our results proved that ssGBLUP was effective for the accurate estimation of genetic parameters and to improve the prediction of breeding values in presence of selfing events, thus a valuable tool for genomic evaluations in Eucalyptus breeding programs. © 2022, The Author(s).