IUFRO Breeding theory and progeny testing

Clonal forestry is becoming a very important activity in breeding and deployment of many commercial forestry species. Because it is a relatively new practice, several aspects need to be understood and solved. Genetic testing of clones is one area for which guidelines are lacking for adequate characteristics of clonal field trials. The overall goal of this study was to identify ‘optimal’ or ‘near optimal’ experimental designs for the prediction of clonal values and estimation of genetic parameters in order to achieve maximum genetic gains from clonal testing. In this study, simulations of single-site trials with sets of unrelated clones ‘planted’ in environments with different patterns of variability were generated and studied.

Methods

The simulations were based on a single trial of 2,048 trees with 8 ramets for each of the 256 clones tested. The trees were “planted� in either single-tree plots (STP) or four-tree row plot (4-tree). For this study several sets of simulation were generated incorporating different site error patterns: only patchy (PATCH), only gradient (GRAD), and both together (ALL). These simulations were generated based on a field trial including the following experimental design: completely randomized (CR), randomize complete block (RCB), a variety of incomplete block designs (IB), and row-column design (R-C). Also, two levels of mortality were considered 0% and 25%; and additionally subsets of the simulated datasets where analyzed to study the effect of varying the number of ramets per clone.

Main findings

The main result was that considerable improvement can be obtained through selection of experimental design and statistical analysis. In particular, the use of STP increased the correlations between true and predicted clonal values by 5% over 4-tree plots and yielded greater genetic gain from selection.

Starting with a parametric broad-sense heritability of 0.25 for a completely randomized design, the experimental designs resulting in the highest heritability increase were: R-C designs for STP, and incomplete blocks with 32 blocks per replication (IB 32) for 4-tree plots. These designs increased average heritability 10% and 14% over a randomized complete block design, respectively. Unfortunately, the best designs always yielded more variable parameter estimates than simpler designs. The use of incomplete blocks (in one or two directions) provided explanation of a larger portion of total phenotypic variability and produced unbiased estimates of genetic variance components and clonal values. For STP, the smallest incomplete block under study had 8 trees per block and it is possible that smaller blocks could produce even greater improvements.

Average heritabilities and 95% confidence intervals for different plot types, designs and surface patterns.

For the different patterns simulated, the ranking from best to worst in terms of easy of accounting for spatial variability was ALL, GRAD and PATCH. The latter had the disadvantage that some of the small patches were confounded with the random error; hence, for this surface pattern lower heritability values should be expected frequently. Also, differences between plot types were smaller for the GRAD surface pattern.

Twenty five percent mortality, when compared to no mortality produced only slight changes in the statistics studied. The consequences were mostly reflected in an increase of the variability of some variance components and therefore, an increase in the variability of individual heritability.

As expected, experiments with more ramets per clone produced more precise variance component estimates and larger clonal mean heritabilities. Using 4 to 6 ramets per clone per site is recommended. More than 6 ramets produced only marginal improvements in precision of clonal means.

Author: Salvador A. Gezan, School of Forest Resources and Conservation, University of Florida, P.O. Box 110410. Gainesville, FL 32611-0410, USA.