SELECTION AGE AND RISK - FINDING THE COMPROMISE

Selection ages for height growth in coastal Douglas-fir on four sites on Vancouver Island were determined by stochastic simulation of age-to -age correlations. Stochastic perturbations of expected age-to-age cor relations at both family and within-family levels were derived from bo ot-strapping more than 4000 records of 23 years of observed height gro wth. Stochastic dominance principles were used to isolate stochastic e fficient (risk averse) selection ages in both family and forward selec tion schemes. The optimum age for family selection depended strongly o n the number of progenies tested in each family and on the test site. Early family selection (age <15) required at least 20 trees per family ; very early family selection (age <10) is feasible at a low risk at f amily sizes above 40. Selection ages in forward selection schemes depe nd on gain expectations from the within-family selection and on the nu mber of trees selected per family. Reliance on gain from within-family selection pushed the stochastic efficient selection age upwards. Sele cting fewer trees per family, but more families, increased the recomme nded selection ages. Age 17 appears to be minimum 'safe' age for conve ntional forward selection schemes.