Comparison of Experimental Designs for Clonal Forestry Using Simulated Data

Abstract

Abstract Various alternatives for the design of clonal field trials in forestry were studied using simulated data to identify “optimal𠇍 or “near-optimal𠇍 scenarios for the estimation of genetic parameters. The simulated field site consisted of a rectangular grid on which 256 clones with 8 ramets each were installed. Estimates of genetic parameters were compared for (1) single-tree and four-tree row plots; (2) several experimental designs (completely randomized, randomized complete block, incomplete blocks of various sizes, and row–column); (3) no mortality versus 257% mortality; and (4) different patterns of environmental variability (only patches, only gradients, and both patches and gradients). Use of single-tree plots, on average, increased the correlations between true and predicted clonal values by 57% over four-tree row plots and increased genetic gain from selection. Starting with a parametric broad-sense heritability (H B 2) of 0.25 for a completely randomized design, the experimental designs resulting in the highest H B 2 were row–column designs for single-tree plots and incomplete blocks with 32 blocks per replication when four-tree row plots were used. These designs increased average heritability 107% and 147% over a randomized complete block design, respectively. The only effect of 257% mortality was an increase in the variability of some variance component estimates.