2004 IUFRO Forest Genetics Meeting Proceedings Efficiency of Using First-generation Information during Second-generation Selection: Results of Computer Simulation

BLUP (Best linear unbiased prediction) method has been widely used in forest tree improvement programs. Since one of the properties of BLUP is that related individuals contribute to the predictions of each other, it seems logical that integrating data from all generations and from all populations would improve both the precision and accuracy in predicting genetic values by increasing the effective number of observations on each genotype (White and Hodge 1989; Kerr et al. 2004). However, some studies based on computer simulation (e.g. Johnson 1998) and field data (e.g. Panter and Allen 1995) showed that including historical parental information actually did little to increase the efficiency of estimating breeding values under some circumstances. The objective of this study was to determine whether the inclusion of first-generation information sufficiently enhance the accuracy and precision of second-generation selection under different selection strategies and combinations of genetic parameters using stochastic data sets generated by computer simulation. Simulation Scenarios We assumed that the 1''-generation population consisted of 300 open-pollinated families with 100 trees per family. The individual-tree heritability h2(1) was set in the range of-0.05 to 0.35. Two different selection strategies were adopted for the enerati era ti on selection. (1) Strategy 1 (backward selection)- select top 48 parents; (2) Strategy 2 (forward selection)- select the best progeny from each of the top 48 families. All selections were based on estimated breeding values (EB Vs). The 48 selections were then used as 2nd-generation parents and crossed in a disconnected 2 x 2 factorial mating design which resulted in 48 crosses in total. The 2nd-generation progeny trials had six test sites with 20 trees per cross per site. Different levels of heritability (h2(2) = 0.05, 0.15, 0.25, 0.35), ratio of dominance to additive genetic variance (Vo / Va = 0, 1,2, 3), and genotype-by-environment interaction (V& / VA = 0, 1,2, 3) were assigned