Back to Study Design
	A multi agency project funded by US EPA's STAR Program

Site selection, continued

Making use of all the environmental data (stressor and natural) to combine segments into clusters required both principle components analysis (PCA) and cluster analysis. Once segments were combined into clusters, at least one segment was selected for sampling from each cluster to ensure that the major environmental gradients were covered.

To reduce the 207 environmental variables to a more manageable number, a PCA was run for each of the seven categories.

The first seven PCs from each of the seven categories (total of 48 PCs*) were used as input variables into the cluster analysis. A separate cluster analysis was run for each of the ecoprovinces , northern (212) and southern (222) (Keys et al. 1995), covering the Great Lakes basin.

The groups sampling the largest number of sites evaluated segments in a random order within each cluster during site selection. The groups sampling fewer sites made as many of their selections as possible from the pool of sites already selected by the other groups. This ensured the maximum sampling overlap among groups.

* One stressor category (coastline attributes) had only 6 PCs.