Environmental Engineering Reference
In-Depth Information
be defined post-hoc using subjective methods or more objective measures which
assess the homogeneity or heterogeneity of groups (Sharma
1996
). In general, the
practitioner's knowledge of the study system is most important when defining
groups that are helpful to modeling the system (not too many or too few groups
for understanding). The process of non-hierarchical K-means cluster analysis is
becoming more popular (Carr et al.
2010
), in which the practitioner first determines
the number of groups and then a computer program optimizes a statistical parame-
ter within those groups (McCune and Grace
2002
).
The non-parametric multi-response permutation procedure (MRPP) can be used
to assess within group homogeneity and to test for significant differences between
groups based upon multivariate data (McCune and Grace
2002
). However, statisti-
cally significant differences are not always ecologically-meaningful.
Currently, classification for mapping purposes is more frequently accomplished
remotely using vegetation reflectance from the visual and near infrared spectra.
These remotely-detected pixel signals are frequently combined into groups using
supervised or unsupervised classification with K-means clustering to identify the
spectral signatures of different wetland plant communities (Zhang et al.
2011
).
Remotely-sensed and classified communities can be mapped very easily (Midwood
and Chow-Fraser
2010
), however, the level of detail in these classifications is
necessarily limited. Numerous statistical packages can perform cluster analyses,
including the freeware R package vegan (Oksanen et al.
2011
) and PC-ORD
(McCune and Mefford
2011
).
5.4.5 Ordination
Typically, a practitioner will have multiple sites or samples, with each sample
described in numerous ways (the abundances of multiple species, environmental
characteristics, etc
). Ordination is a method of discovering patterns and underly-
ing structure in this multivariate data (Kenkel
2006
). Because ordination diagrams
and the process of ordination itself can be confusing, ordination information is
typically not directly presented to lay people or political decision-makers. How-
ever, that does not mean that it has no role to play in wetland conservation and
management. Ordination has been used to assess the effects of management
practices on wetland plant communities (Hall et al.
2008
); compare damaged,
restored, and reference plant communities (Rooney and Bayley
2011
); assess the
community-level effects of exotic species invasion (Mills et al.
2009
); and gener-
ally understand how environmental degradation affects wetland systems (Carr
et al.
2010
). A complete discussion of ordination techniques, their assumptions,
and mathematical background can be found in McCune and Grace (
2002
), Kenkel
(
2006
), and Legendre and Legendre (
1998
).
There are several different types of ordination, but all involve reducing the
variability in a large data set down to a few axes that express the primary patterns
...
