A stem-diameter data set of five combined trials of 'Pinus radiata' D. Don was used to identify and determine the nature of genetics by environment (GxE) interaction. The restricted maximum likelihood approach was applied to handle the main issues of the multi-environment trial analysis: (1) Testing sources of heterogeneity of variance and lack of between-sites genetic correlation; (2) Modelling the heterogeneity of error variance among trials and micro-environmental variation within each trial; and (3) Selecting the best model for prediction of breeding values. Model comparison was based on the criterion of log-likelihood. The significance of variance components was tested by the likelihood ratio test which showed that all sources of GxE interactions were highly significant, indicating that GxE interactions occurred in these five trials due to both the heterogeneity of variances and the lack of correlation. Estimates of Type B genetic correlations were increased slightly by correcting for the heterogeneity of variances. The full model, which accommodated heterogeneity of error variances between trials, spatial variation within trials, and fitting a separate GxE interaction variance for each trial, was superior to other models for this multi-environment trial.

In breeding forest trees, as for livestock, the goal is to capture as much genetic gain as possible for the breeding objective, while limiting longand short-term inbreeding. The Southern Tree Breeding Association STBA) is responsible for breeding Australia's two main commercial forest tree species and has adopted algorithms and methods commonly used in animal breeding to achieve this balance. Discrete generation breeding is the norm for most tree breeding programmes. However, the STBA uses an overlapping generation strategy, with a new stream of breeding initiated each year. A feature of the species bred by the STBA ('Pinus radiata' and 'Eucalyptus globulus') is the long interval (up to 7 years) between when an individual is mated and when its progeny is first assessed in field trials and performance data included in the national performance database. Mate selection methods must therefore recognize the large pool of unmeasured progeny generated over recent years of crossing. In addition, the substantial delay between when an individual is selected in a field trial and when it is clonally copied into a mating facility (breeding arboretum) means that selection and mating must occur as a two-step process. In this article, we describe modifications to preselection and mate selection algorithms that allow unmeasured progeny (juveniles) to be recognized. We also demonstrate that the addition of hypothetical new progeny to the juvenile pool is important for computing the increase in average co-ancestry in the population. Methods outlined in this article may have relevance to animal breeding programmes where between mating and progeny measurement, new rounds of mating are initiated.

TREEPLAN is an advanced analytical tool providing accurate and precise predictions of genetic values to operational tree breeders. This project aimed to exploit TREEPLAN's analytical power for the benefit of all sections of the forest growing industry; in particular, the deployment sector and programs not traditionally associated with the Southern Tree Breeding Association. The outputs include: the routine inclusion of reproductive and fitness traits into the assessment framework; better prediction of stand performance through competition models; the undertaking of pilot studies using data from large industry programs; and the better modelling of the genetic structures of hybrid populations.

The POTATOPLAN project aims to adapt comprehensive genetic evaluation systems used in forest tree and livestock breeding to potatoes. These systems use Best Linear Unbiased Prediction (BLUP) to incorporate all measurements from all relatives and correlated traits to best predict the additive and total genetic values of all genotypes for all traits. This enables optimal selection of genotypes for breeding and deployment as any genotype at any stage of testing can be compared with any other, including commercial varieties. The method can remove biases due to natural and artificial environmental variation while accounting for GxE, genetic trends over time due to selection, differences in data amount and quality, and varying sampling procedures. The models have been adapted to potatoes by estimating additive genetic effects using an additive relationship matrix that recognises the polyploid nature of most potato varieties. Family merit score counts (number of selected progeny) are converted into a binary selection trait for the genotypes in the family, with a value of 1 for selected named progeny and 0 for unselected progeny. This trait is used to predict the value of the selected progeny for all traits that contribute to the merit score through the genetic correlations determined on an experimental basis. Measurements of genotype performance using different plot or sample sizes is accounted for by using weighted analysis with the individual plant or standard sample size error variance as the reference point. These approaches should allow the more widespread adoption of BLUP methods in potato and other crop breeding. POTATOPLAN also includes a comprehensive database of pedigree, measurements, trial designs, genetic and other model parameters, and resultant genetic values to make them useable for breeders and growers.

Genetics, genetics x environment (GGE) biplot analysis is an effective method, based on principal component analysis, to fully explore multi-environment trial data. It allows visual examination of the relationships among the test environments, genotypes, and the genotype x environment (GxE) interactions. Data from multi-environment trials of 'P. radiata' D. Don containing 165 to 216 families in five environments were used to demonstrate the results and application of GGE biplot analysis. There were non-overlapping clusters of two and three sites, which indicated two distinct environments. The best family for both of the distinct environments was also identified. Genetic correlations among sites ranged from 0.98 to -0.50, indicating that there were large GxE interactions among the test environments.