Achieving higher heritabilities through improved design and analysis of clonal trials
Clonal testing was studied under different environmental patterns and experimental designs through simulation with the criteria of maximizing broad-sense heritability estimates and genetic gain from clonal selection. Several experimental designs were studied together with three patterns of environmental variability. In addition, empirical 95% confidence intervals for heritability estimates were compared with Dickerson's approximate method. Other elements studied included (i) conditions under which different environmental patterns yield high or low heritabilities and (ii) effects of varying the number of ramets per clone. Row-column designs produced the highest mean individual broad- sense heritability, but these designs were only slightly more efficient than incomplete block designs with small block sizes. For all experimental designs, Dickerson's approximate method for estimating the variance of heritability estimates produced reasonable 95% confidence intervals but overestimated the upper confidence limit of complex designs. Larger heritabilities were found with higher tree-to-tree spatial correlations and lower amounts of microsite residual variance, and varying gradients had negligible influence. The effect of implementing Latinization was significant on increasing heritability, but small in practical terms, and was more important for patchy surfaces. Experiments with more ramets per clone yielded higher clonal mean heritabilities, and using between four and six ramets per clone per site is recommended.