Environmental Engineering Reference
In-Depth Information
Fig. 5.10 Ordering sites according to their multivariate plant species composition using indirect
gradient analysis (ordination).
Numbers
are wetland sites. (a) A data set consisting of only two
species and six sites. (b) A data set consisting of numerous species in six sites. The underlying
structure in the species information has been extracted by ordination methods into two axes of
variation. Site ordination scores are related to environmental measurements using correlation
in the data using the correlation between the multiple variables (McCune and Grace
2002
). Sometimes there is a particular environmental factor of interest. If this is the
case, then direct gradient analysis, or positioning samples along axes determined by
environmental measurements, is most appropriate (Kenkel
2006
). The practitioner
can then examine the relations between the species within those sites and the
environmental factor of interest (Fig.
5.8
).
Indirect gradient analysis does not assume the importance of any particular
environmental gradient
a priori
, but rather lets the plant species data order itself.
In multivariate speak: “arranging plots in species space.” There are several ways of
using ordination, but the most common is to plot sites along axes of species
composition, and then correlate these axes with environmental variables to deter-
mine which environmental variables most strongly influence the plant
communities. One can conceptualize this more easily by beginning with a system
of multiple sites with only two species. It is easy to plot samples or sites within
species space, with each axis representing the abundance of species A or B
(Fig.
5.10a
). In Fig.
5.10
sites 1, 2, and 3 are similar in their composition of species
A and B, and sites 4, 5, and 6 are similar to each other. In a cluster analysis, these
two groups would most likely cluster together. Ordination typically involves many
more than two species, but it is very difficult to create a graph that contains 100 axes
for each different species. In order to make meaning, ordination mathematically
sorts through the variation in these different species to draw out the strongest
patterns (based on correlation or similarity), and these patterns are reduced to
usually two or three axes (Fig.
5.10b
). Each site is assigned a score or position
along each axis, based upon its species composition. Because each site has
associated environmental measurements, the ordination score can be correlated
with the environmental factors to determine how these factors affect the community
as a whole rather than an individual species.
Various methods exist for indirect gradient analysis: PCA (principle com-
ponents analysis), Bray-Curtis, NMS (nonmetric multidimensional scaling), CA
