This is a general question, you may check a standard deviation plot against the number of replicates N for the trait you want in a pilot survey, and take the optimal value for N. Tree parcels should not use border trees. I presume you got normal variables and that an Anova makes sense. You may refere to R-project agricolae or to http://www.wormbook.org/chapters/www_statisticalanalysis/statisticalanalysis.html of Dr Fay and Gerow.

One should see the homogeneity of the trees through a pilot survey, as suggested by other researchers, and find out the number. If the trees are heterogenous then ANCOVA would be appropriate wherein due consideration is given for variation among test trees.

The genetic structure of tree population is also important, i.e. if there is high heterogenity or are they rather homogenous. In case of highly heterogenous species/populations it is recommended to take much more individuals than in homogenous popualtions.

I add a powerpoint file with rigorous explanations from Oregon state university crop unit. The plot I meant is the one at slide 15  http://www.google.it/url?sa=t&rct=j&q=&esrc=s&source=web&cd=9&ved=0CFUQFjAI&url=http%3A%2F%2Fcropandsoil.oregonstate.edu%2Fsystem%2Ffiles%2Fclasses%2Fcss590%2Flectures%2FM10_plot_size_reps.ppt&ei=hPDmVK3rPIviapjjgKgM&usg=AFQjCNH8lLRyiPuxD8n5AuR_jhGPEJaGRQ&sig2=-6bp70kDjnqFGsYjQJjJdg&bvm=bv.86475890,d.d2s