Comparison of canonical correlation and regression based focal point seed zones of white spruce

Journal Article

Development

The focal point seed zone methodology determines spatially explicit areas of adaptive similarity for any selected geographic point and is used to match seed sources and planting sites. A total of 127 seed sources (provenances) of white spruce (Picea glauca (Moench) Voss) from Ontario and western Quebec were established at a greenhouse and in six field trials throughout Ontario. Growth and phenological variables were measured over three growing seasons. Two focal point seed zone methodologies were employed: (i) using models derived from principal components analysis (PCA) of biological response variables followed by multiple linear regression against climate variables and (ii) using models derived from canonical correlation analysis (CANCOR). While both approaches use climate data to model adaptive variation, CANCOR reduces the number of steps in the analysis by simultaneously finding the relationships of biological and climatic variables that maximize the covariance between two data sets. Although more of the variation in adaptive biological traits was actually described by climate variables using the PCA-regression approach, this method produced intuitively less realistic patterns. Both methods showed similar overally geographic trends, but the CANCOR method had a finer resolution, especially in southern Ontario, persumably due to statistical efficiency; growth was modeled by all climate variables.