using multi-dimensional scaling (NMDS) to evaluate the similarity
among samples based on their species composition and abundance
(Wilkinson, 1990). The fit of configuration distances to the original
data was evaluated by calculating stress using Kruskall’s stress
formula, with values near 0 indicating a better fit. We examined
Shepard diagrams (plots of the distance between points in the final
plot with observed dissimilarities in the original data) to verify that
they appeared as straight lines or smooth curves, also indicating
good fit with the original data.
Habitat data were combined into categories that reflected their
growth form and structure, and percent cover of these categories
was calculated as the proportion of the 20-point intercepts where
that substrate was encountered. These included percent cover of
conifers (trees and shrubs combineddue to lowcover), percent cover
of lowbroadleaved shrubs (5 m tall),
percent cover of ferns and forbs, grasses (including sedges and
rushes), percent cover of bare ground (including rock or slash), and
percent cover of exotic invasive plants. These values were averaged
for sample points, and log-transformed to improve normality
and equality of variances.We used canonical correlation analysis to
identify multivariate correspondence between bird communities
and habitat (McGarigal et al., 2000). Species values were standardized
by maxima and site values were standardized by totals to
emphasize relative differences among species rather than differences
due to abundance. This analysis was conducted using
the vegan package in R 2.3.0 (Oksanen et al., 2006). We compared
habitat characteristics between treatments using ANOVA nested
by site.
We calculated nest success for each site in each