Main Article Content
Genotype x environment interactions of stem volume were investigated by assessing the variation in 46 open-pollinated loblolly pine (Pinus taeda L.) families from first-generation cloned seed orchard in four genetic trials in the south and south-east Brazil. They were used to obtain least squares and restricted maximum likelihood (REML) estimates of variance components. Familie-by-trial interaction effects were evaluated by adjusting the mixed univariate model that contained data of two and four places tested by the likelihood ratio test. Breeding values from local data (univarate procedure) and predicted to the others sites (multivariate procedure) were obtained from best linear unbiased prediction (BLUP). The adjusted and average (obtained from local and predicted to other sites) breeding values were used to select parents and trees. The interaction effects and the adjusting of the mixed models were statistically significant, respectively, by F test and by likelihood ratio test. The loss of potential gain, sustained by not selecting the best families by site was 3.2%. For parents, the loss in mean productivity by indirect selection was respectively 2.3%. In the individual tree selection for seedling seed orchard, no loss of potential gain was observed when selection was carried by average genetic breeding values. For clonal seed orchard with the selection of ten more greater genetic breeding values trees, the interaction cause a inflation of 2% in the average productivity. In this case, the selection by average breeding values was the best procedure and may prove to be a useful tool, in selection stem volume, when genotype x environment interaction is significant.