SELECTION FOR AN OPTIMUM TREE GROWTH CURVE

Height growth trajectories of 100 half-sib families of Pinus pinaster AIT. were subject to selection that aimed at a constant improvement(he re 20%) far all ages or a gradual improvement of p log(age) 100% (here p=0.05). Height growth was under strong additive genetic control and prospects of genetic gain were promising. A 2nd degree polynomial on a log-log scale was successfully used to describe the height-age relati onship in all trees. Classical SMITH-HAZEL selection indices, growth c urve parameter indices, and continuous selection indices were also com puted for the 2 improvement objectives. The continuous selection index was a second degree polynomial on a log transformed time scale derive d from stochastic integration of covariance kernels Results suggest th at selection on the growth curve continuum is more efficient for both mass selection and family selection because it utilized more informati on than selection at discrete ages. The latter had to be restricted to 3 distinct ages in order to avoid singularity of covariance matrices. Selection on growth curve parameters was intermediate in efficiency.