Environmental Engineering Reference
In-Depth Information
provide a more comprehensive data set. Important species located just outside of
the study plot can also be included.
In many ways, the relev
´
gives a richer picture of the wetland plant community
than systematically-placed plots. However, skilled field botanists are needed to
execute it properly. Relev
´
data are more difficult to statistically summarize at
smaller spatial scales due to the smaller number of plots. Over large spatial scales,
species presence and abundance estimates can be used to assemble solid commu-
nity descriptions. Practitioners also use permanently-established relev´s to detect
community change over time. Relev´s may not be the best solution in systems with
numerous discrete plant communities concentrated within small spatial scales,
because collecting data from a high number of relev´s can be time-consuming
(U.S. EPA
2002
).
5.3.5 Number of Samples
The number of samples is always a compromise between the resources available
and the desire to collect as much data as possible to adequately characterize the
system of interest. Ideally, a pilot study should be conducted prior to implementing
a sampling scheme. The pilot study reveals the amount of variability in the wetland
vegetation and can therefore give an idea of how many samples are needed to
adequately describe that variability. The practitioner will typically vary the number
and/or size of samples in the pilot study. The data are then used to create species
accumulation curves (for species richness: number of species) or performance
curves (for other measures). From these curves, the investigator can estimate the
point at which additional samples yield minimal additional information. This point
optimizes the efficiency and accuracy of sampling. The pilot study data should also
be used in statistical power calculations (see Chap.
1
).
A species accumulation curve is obtained by comparing the mean cumulative
number of species to the number of samples (or size of plot, Fig.
5.6a
). The
asymptote of this curve is the point at which an adequate number of samples has
been collected to characterize the richness of the system. In practice, this value is
tedious to compute (consider: calculating the mean number of species for sample
size of one is quite easy, but what about all pairs of samples for sample size of two?).
However, there are computer programs (such as EstimateS; Colwell
2009
) that will
calculate for you based on your data matrix of samples and species. Likewise, the
performance curve is obtained by plotting the mean and variability of some attribute
against the number of samples (or size of plot, Fig.
5.6b
). When there is no further
change in the mean (within acceptable limits), the number of samples is adequate.
If there are no resources available for a pilot study, the adequacy of sampling can
be assessed post-hoc using these methods. Statistical methods have been developed
to estimate actual species richness from inadequate samples (discussed later in this
chapter). In the absence of a pilot study, an important rule of thumb is to collect at
least 20 samples at each site to meet the demands of some statistical methods (such
as linear regression).
