Improving the efficiency of selection in a plant breeding program using information on correlated traits, ancestry and environments

[Truncated abstract] This thesis presents how information on correlated traits, ancestry and environments can be used within a mixed model framework to improve selection in plant breeding. The motivating example is canola (Brassica napus L.). Plant survival data in blackleg disease of canola are often composed of multiple measures used to form a derived variable, such as percent survival values, which is then subject to analysis. Instead, a bivariate linear mixed model approach is proposed in which the two variables are the initial and final plant counts. This approach is demonstrated using data from blackleg disease nurseries in the 2009 growing season in Australia. The counts were considered as two 'traits', which are a ected by different biological, genetic and environmental influences. The bivariate mixed model approach for the analysis of plant survival data not only provided a more detailed picture but also a more accurate assessment of the impact of disease resistance compared with the univariate analysis of percentage survival data. The release of new cultivars onto the market is preceded by extensive testing of varieties across target environments and growing seasons in multienvironment trials (METs), which is a core process in plant breeding. Another related objective is the selection of parents for the next cycle of breeding. The inclusion of pedigree information in the MET analysis satis es both objectives. Using the 2011 subset of data from a canola breeding program, this thesis demonstrates the use of spatial analysis of individual trials and then extends this to an across site analysis using a MET and factor analytic (FA) mixed model framework. The efficiency of this process is demonstrated in the spatial analysis of individual trials to control within trial environmental impacts when pedigree information is included.