In our second analysis we assessed the spatial autocorrelation
of the abiotic environment in our finescale
survey. To do this, we used a Mantel test to model
a trend between geographic distance and environmental
change. Environmental change was calculated as the
Euclidean distance of all the standardized environmental
variables that were significant in the fine-scale
CCA (McCune and Mefford 1995). Geographic distance
was calculated as the Euclidean distance between
sites to produce the shortest on-the-ground distance
between the centers of each pair of plots. A similar
Mantel test of the mesoscale survey was unnecessary
because of the sampling design. It is important to note
that the Mantel test is not meant to compare directly
to the ordination results, as it measures a different variance
and reports weaker correlations than tests on raw
data (Dutilleul et al. 2000). We use it instead to look
at the nature of the clustering of similar environments
over geographic distance and to complement the results
of the ordination.