In the development of an improved population of trees for a large scale reforestation program, different questions arise as to how the process should proceed most efficiently. Initial work would be undertaken to understand the environmental or abiotic limitations of the species of interest so that alternative species can be evaluated when climatic or edaphic variables present a significant risk. Once suitable environmental limits for key species are understood, developing an understanding of how to manage biotic risks is of great interest given the significant impact pathogens have on planted forest productivity. Once these abiotic and biotic risks are manageable, an emphasis is often placed on evaluating potential end-uses that would provide the greatest return on the investment in the planted forests. After the drivers of the abiotic, biotic and product risk factors are understood, decisions regulating the intensity of the supporting tree improvement program would need to be made so that systems to deliver genetically improved material may be put into place. This thesis examines the following four questions of relevance to the genetic improvement process of forest trees in general and uses spotted gum (Corymbia citriodora) populations to examine specific aspects of these questions:
1. How does the productivity of species change across environmental gradients?
2. Are assessments of damage caused by pathogens useful in identifying populations with
greater disease resistance or tolerance?
3. What importance should be placed on selection traits used to identify genetically improved
4. How may molecular markers be integrated into the selection process?
The common link between the exploration of these questions in the four research chapters of this thesis is the use of an empirical method to evaluate the performance of different cohorts in a variety of experimental designs. Statistical methods studied include: pattern analyses with landscape-level data from taxa comparison trials; comparisons of general and generalized linear models to assess disease resistance; mixed models for genetic parameter estimation; and multivariate models for the assessment of associations between DNA markers and phenotypic traits.
Following the initial introduction chapter, the second chapter of the thesis focuses on a novel method for classifying changes in species performance across environmental gradients and presents results in a new form of the classic biplot. The third chapter finds very high genetic correlations between growth and disease resistance and indicates there is little to gain from moving from general to generalised linear models. The fourth chapter details a comprehensive methodology for genetic evaluation including among-site and among-trait multivariate analyses for genetic parameter estimation and breeding value prediction with a key finding highlighting the importance of volume production in multi-trait selection indices. The fifth chapter presents and alternative marker-trait association method and demonstrates that testing the significance of the genetic correlation between markers and traits provides different results than the standard model that accounts for variation through the inclusion of markers in general linear models. Taken together, this thesis provides methodologies to guide the various decisions involved in the genetic improvement process of undomesticated forest tree species with a specific focus on the application of mixed models to guide decision making processes.