Eleven Populus × generosa populations were developed in the Pacific Northwest by annual controlled hybridization of P. deltoides and P. trichocarpa between 1991 and 2001. Mass selection for Melampsora leaf rust resistance was observed in the field as a threshold character in identifying seedling phenotypes for clonally replicated evaluation. The effectiveness of the approach was assessed for each annual population by comparing the distribution of phenotypes in unselected seedling populations with the distribution of selected genotypes in the clonal field tests established in successive years and evaluated at the approximate same level of disease severity using two selection thresholds corresponding to chlorotic and healthy tissue. Bi-directional selection was used as an initial check on the efficacy of the procedure and resulted in a wide separation in liability between the positive (0.06 threshold units (T.U.)) and negative (−2.45 T.U.) selection groups when tested as clones. The other 10 seedling populations that were subjected solely to directional selection exhibited a mean increase in incidence above the first selection threshold at the clonal stage (47 versus 81%) that was accompanied by an improvement in population liability (−0.06 versus 0.50 T.U.) and a reduction in population standard deviation (0.83 versus 0.54 T.U.). The change in liability was strongly related by polynomial regression to selection intensity and a grouping of populations based on infection-season precipitation (r2=0.98). The mean liability of four of the 10 seedling populations observed during years of high infection-season rainfall was six-fold lower than the mean liability of those populations observed during the other six years of lower infection-season rainfall (−0.12 T.U. versus −0.02 T.U., respectively), indicating that populations undergoing evaluation during years of heavy precipitation experienced more intense rust exposure. Moreover, quadratic functions showed that populations undergoing rust evaluation during years of high rainfall were more responsive to increases in selection intensity above the vertex of the function (i.e. 13.20 versus 3.43 T.U.). Realized heritability averaged 0.63 for all ten populations subjected solely to directional selection.
