Drivers of genotype by environmental interaction in radiata pine as indicated by multivariate regression trees

Productivity of forest tree plantations can be maximized by matching genetically adapted planting stock to environments where they perform best. We used multivariate regression tree (MRT) analysis with environmental predictors to quantify and characterize the nature of genotype by environment interactions (G x E) of radiata pine diameter at breast height (DBH) grown in New Zealand. The analysis was carried out for 21 provenance trials, and 48 progeny trials of second-generation selections that are widely used in plantation forestry today. To quantify the maximum variance explained by G x E, we used unconstrained clustering of genotypes based on their performance across all sites. Subsequently, the clustering was constrained by climate and soil variables, i.e. the putative causes for G x E. Unconstrained clustering explained 62% and 58% of the observed G x E variance in provenance and progeny trials, respectively. Constrained clustering explained approximately 50% and 25% of the G x E variance in provenance and progeny trials, respectively. Minimum temperature was identified as an important driver of G x E in both provenance and progeny trials. Environments can be grouped into warm humid sites, where most second-generation selected genotypes performed better, and cold sites, where specific genotypes performed best. Based on the progeny trials, only marginal (ca. 3%) gains can be made by targeted deployment to warm humid sites, but more substantial (approx. 20%) genetic gain can be made on cold sites, compared to current deployment strategies.