Investigation of the application of best linear prediction for breeding and clonal production purposes in a Eucalyptus grandis population.
The genus Eucalyptus has been planted extensively throughout the world in tropical and subtropical regions, primarily because of its economic importance and use in wood and pulp production. Due to the growing demands for timber, forestry companies need to increase the productivity of available forest land. The genetic improvement of forest trees through selection and breeding involves a lengthy process of scientifically controlled trials focused on short-term and long-term goals using breeding and production populations. This investigation focused on the use of Best Linear Prediction (BLP) and its application to: (1) the prediction of genetic gains for a breeding population and, (2) the selection of superior individuals for clonal production of E. grandis. A CSIR dataset for a 20-year-old progeny trial involving 90 open-pollinated families was obtained. Four traits, namely, diameter at breast height (DBH), stem form, splitting and density were identified for use in this investigation. Relevant data were extracted and a file termed, Dataset created. Dataset was edited, standardized and corrected for the fixed effect of replication using SAS® procedures. Precise and accurate population parameter estimates are fundamental in determining breeding strategies and thus, heritabilities of each trait and phenotypic correlations between traits in Dataset were estimated using SAS® procedures. DBH was found to have the highest heritability (0.600), followed by density (0.492). The estimated heritability for stem form was 0.401 and splitting had the lowest heritability at 0.214. A high positive phenotypic correlation of 0.83 was estimated between DBH and stem form. The phenotypic correlations between other traits were close to zero. An index provides a weighted score for individuals, which takes all relevant information into account and allows individuals or families to be chosen for breeding and production purposes. Consequently, Best Linear Prediction (BLP) of individual breeding values were calculated using MATGEN® (2003). Thereafter, BLP values were used to determine the rankings of individual trees for 15 different selection indices. In order to determine the effect of selection on the change in the population mean of a trait, the breeding population's response to selection was predicted and compared across three selection strategies, namely: (1) individual selection, (2) single-trait index selection, and (3) multiple-trait index selection. The top 8% of individuals in the breeding population were selected for and the genetic gains were predicted. It was found that the response to selection was greatest when using individual selection. Furthermore, DBH had the best selection response for all three strategies as compared to the other traits under investigation. Fifteen indices, considering different numbers and choice of traits, were compared for commonality among rankings of the top 30 individuals. Two methods, namely, (1) a rank-correlation matrix and (2) a manual assessment, were used. The commonality between indices showed that a simple index, considering two traits (DBH and density) was equally effective (93%) in identifying genetically-superior individuals as the more complex index that considered four traits. Furthermore, it was possible to select for only three traits (DBH, splitting, density) and identify the same top 30 individuals as using the index that considered four traits. The researcher's goal was to find the most desirable individuals in the population to be used for production purposes, such as clonal forestry. Consequently, various selection options, specifying certain trait requirements, were used to select superior individuals for use in production and deployment. The "commercial selection" option was the only option successful in obtaining an individual that met the required criteria for the four traits in the population of 475 individuals. The results suggested that breeders should consider large populations and only a few important traits in order to obtain a greater number of individuals suitable for mass propagation in clonal forestry. In order to further investigate the effect of population size on the number of individuals suitable for clonal forestry, a hypothetical population was generated. This was accomplished using between family and within family standard deviation values obtained from Dataset. The large hypothetical population of 1000 individuals produced twelve individuals suitable for production purposes, as opposed to only one in the real population of 475 individuals. This result further indicates that a larger population provides a greater number of individuals appropriate for use in production and deployment. This investigation successfully addressed the aims by: (1) calculating individual breeding values (BLP) and ranking individuals, (2) predicting the breeding population's response to selection, according to three strategies, for the four traits under investigation, and (3) identifying superior individuals for use in commercial clonal forestry. As the work of tree breeders is aimed at improving the growth and quality of trees by increasing the frequency of desirable genotypes in the population, further research could focus on (1) the effect of different sets of economic weightings on index rankings in a population and (2) the influence that population structure has on the optimal genetic gains obtained.