ECOFACT Volume 1 Introduction and Approach - Centre for Ecology ...

Assigning new vegetation plots to classes within the CVS and exceptions:
In order to allocate vegetation plots (quadrats or relevèes) to the existing hierarchical
classification, a binary decision tree was constructed. At each node of the tree a decision method
was implemented, appropriate to the classification being emulated. The decisions were based on a
partition of multidimensional species-space. The resulting decision tree produced a deterministic
result, allocating each vegetation plot to a single vegetation class. It should be emphasised that
this procedure gives a precise allocation of each individual plot to the appropriate class of the
CVS, based on the entire information available on the species content of that plot.
This structure has been implemented as a software package running under Microsoft WindowsTM.
So far this package has been made available on request for testing, and has performed well. It is
currently available on the World Wide Web
(http://www.ceh.ac.uk/products/software/CEHSoftware-CVS.htm). This package enables a user
to classify sample plots for a variety of objectives.
The classification has also been incorporated into MAVIS (Modular Analysis of Vegetation and
Interpretation System), a package currently being tested which provides ready access to the
vegetation analysis procedures of CVS, NVC, CSR and Ellenberg values. This software allows the
user to enter species lists for vegetation units either interactively or in batch mode from a data
file. Once a vegetation unit or units have been allocated to a class or classes, the software allows
the user to determine their positions on the three main vegetation gradients in Great Britain, as
determined from the CVS.
The CVS allocation software provides a means of sorting the different assemblages, using the
whole list of species present. The major exception is saltmarsh, which was excluded from the
analysis as it was represented by only 38 plots and is not robustly described within the range of
CVS classes. The five most frequently recorded species were common saltmarsh-grass (Puccinellia
maritima), sea plantain (Plantago maritima), annual sea-blite (Suaeda maritima), sea aster (Aster
tripolium) and Spartina spp., but others such as glasswort (Salicornia spp.) may be locally
important. However, if exceptional plots are being observed, eg amongst scrub on sea cliffs,
and/or if plots are being examined out of season, then users should be aware that problems could
arise in the allocation.
Tests have been done, e.g. on limestone pavements, where no plots were placed in the random
survey, and the system assigned the list of species to an appropriate group. It is difficult to
provide statistical rules that are easy to apply in order to verify whether a class is appropriate,
and judgement must be involved in applying the above guidelines. In critical studies appropriate
statistical analyses should be carried out in order to confirm intuitive interpretation.