Retrospective Evaluation of Parental Selection in Nursery Tests of Juglans regia L. Using a Mixed Model Analysis

Results of early testing in Juglans regia performed over the period 1993-2002 are presented. A total of 37 progenies were evaluated for establishment, growth traits, and phenology at ages one and two during two or more sowing years in the nursery. Independent culling selection was applied on parental trees to the family-mean values using specific truncation points for establishment, diameter and total height. In addition, parental selection was performed on unbiased predictions (BLUPs) of parental breeding values using a multivariate mixed model analysis in order to account for the unbalanced nature of the entire data set. Genetic parameters (heritabilities, correlations) of juvenile traits were also estimated. Except for second year growth traits, the genetic control of most characters was quite high, with heritability estimates ranging from 0.51 (establishment) to 0.93 (flushing date). Growth performance and establishment did not bear a common genetic control according to non-significant genetic correlations, but a higher growth was typical of early flushing families. Changes in parental ranking for growth after comparison of familymean and BLUP-based estimates of parental performance were considerable. This observation, together with the larger proportion of culled progenies in the former method (about 70%), suggests a sizeable loss of genetic gain by using unadjusted parental means. It is concluded that current evaluation and selection schemes using family-mean values should be reconsidered by i) relaxing truncation points for selection on establishment and growth traits and ii) re-evaluating progeny test data using a mixed model framework to unveil valuable material otherwise neglected due to unaccounted environmental influence on family performance.