Thesis (MScFor (Forest and Wood Science))--University of Stellenbosch, 2007. / This study evaluates the variation that exists between six sites and between more than 200 Pinus patula families established across the sites for various traits utilised in a tree improvement programme. The traits utilised were growth at ages five and eight years, gravimetric densitometry of a sub-sample of the top 100 families at age eight, and microdensitometry and fibre morphological characteristics as determined by image analysis of increment cores, extracted from a sub-sample of the top 30 families. Significant differences were found between locations for growth at ages five and eight, density (both gravimetric and micro densitometry) and fibre properties. For each of the site combinations, utilising Type B- genetic correlations, the interaction between families and locations were evaluated in order to determine which locations could be grouped together in order to determine the effect various sites will have on the deployment of material. Significant differences were also found between families for the various traits investigated, which would indicate that desired trait or trait combinations can be selected for in a tree improvement programme. Heritability estimates for growth varied across sites, ranging from 0,32 to 0,57 at five years and 0,34 to 0,59 at eight years for family heritability. The individual tree estimates ranges from 0,08 to 0,27 at five years and from 0,09 to 0,26 at eight years. The standard errors associated with the heritability estimates for growth however indicate that the estimates, especially those of the individual trees should be used with caution. The heritability estimates for density and fibre morphological characteristics on the family and individual tree level are on a number of sites very high, although this is associated with large standard errors. Indications were that the traits can be combined effectively into a multi-trait selection index, since the phenotypic and genotypic correlations indicated mostly favorable or slight negative correlations between traits.