Variance Partitioning
The remaining 38 explanatory variables were grouped into
three subsets to yield ecologically interpretable variance
components as follows: (1) variables describing the geographic
position (G) of the water body, (2) regional scale
(R), and (3) local scale (L) variables (Fig. 2, Table 2).
The variation partitioning technique used has been previously
described by Borcard and others (1992) and hence
we will not go into detail here. In brief, the procedure
allows for the variance in the explanatory data set to be
partitioned into different variable components through the
use of covariables (i.e., variables whose influence is partialled
out of the analysis). Initially, this technique was
used to partition variation in ecological data sets into
environmental and spatial components (e.g., Økland and
Eilertsen 1994) and has been extended by incorporating
three sets of explanatory variables (e.g., Anderson and
Gribble 1998).