THE USE OX STATISTICS I N PRYToSOCIOLOOY
85
the coincident units were generated; and this-will give most weight to
those species and sites likely to differentiate most powerfully between
closely related noda. There is clearly more work to be done on determining the most efficient methods of both abstracting and characterizing noda by statistical methods: all that can be said at present is that
the concept seems valuable and might help to resolve existing phytosociological difficulties as to the nature and relationships of abstract
vegetation-units.
F. RELATIONSHIP WITH THE ENVIRONMENT
Although the erection of abstract units defined by both species and
site could be legitimately regarded as the end of the strictly phytosociological operations on the data from a given area, it must be
remembered that, for analytical purposes, the sites have so far been
regarded as purely spatial entities with no external reference except
their position on the surface of the earth. For ecological purposes,
however, some knowledge of the particular habitat conditions relating
to particular groups of species is required: we must therefore now
return to the original concept (p. 64) of vegetation as a threefold system
of plantlsitelhabitat relationships. The sites form the link between the
species and the general environment ; and, whereas at present we have
dealt exclusively with specieslsite relationships, we must now examine
sitelhabitat relationships and use the sites to establish connection
between the two. The question of what habitat features to record in a
given situation must always be to some extent subjective, based on the
investigator’s intuition as to the range of features most likely to be
involved; within this range, however, the question of which of the
recorded features show the best correlations with variations in the plant
cover needs to be objectively resoIved.
We have already indicated (p. 65) that there are three basic approaches
to the general problem. We may examine specieslsite relationships in
the first instance and use these as the reference system for site/habitat
relationships ; we may make sitelhabitat relationships the focus of our
interest and relegate specieslsite relationships to a subsidiary position ;
or we may examine both systems independently and attempt to
correlate the two sets of results. Although there are a priori reasons
for preferring the &st approach, some mention should nevertheless
be made of the other two.
Once site-groups have been obtained from specieslsite data by any
method of statistical analysis, these groups can actually be mapped
and this in itself may give some guide to habitat va3iations of ecological
importance. Agglomerative methods of classification will have concentrated on floristic similarities between the sites, while subdivisive
D
C.E.R.
85
the coincident units were generated; and this-will give most weight to
those species and sites likely to differentiate most powerfully between
closely related noda. There is clearly more work to be done on determining the most efficient methods of both abstracting and characterizing noda by statistical methods: all that can be said at present is that
the concept seems valuable and might help to resolve existing phytosociological difficulties as to the nature and relationships of abstract
vegetation-units.
F. RELATIONSHIP WITH THE ENVIRONMENT
Although the erection of abstract units defined by both species and
site could be legitimately regarded as the end of the strictly phytosociological operations on the data from a given area, it must be
remembered that, for analytical purposes, the sites have so far been
regarded as purely spatial entities with no external reference except
their position on the surface of the earth. For ecological purposes,
however, some knowledge of the particular habitat conditions relating
to particular groups of species is required: we must therefore now
return to the original concept (p. 64) of vegetation as a threefold system
of plantlsitelhabitat relationships. The sites form the link between the
species and the general environment ; and, whereas at present we have
dealt exclusively with specieslsite relationships, we must now examine
sitelhabitat relationships and use the sites to establish connection
between the two. The question of what habitat features to record in a
given situation must always be to some extent subjective, based on the
investigator’s intuition as to the range of features most likely to be
involved; within this range, however, the question of which of the
recorded features show the best correlations with variations in the plant
cover needs to be objectively resoIved.
We have already indicated (p. 65) that there are three basic approaches
to the general problem. We may examine specieslsite relationships in
the first instance and use these as the reference system for site/habitat
relationships ; we may make sitelhabitat relationships the focus of our
interest and relegate specieslsite relationships to a subsidiary position ;
or we may examine both systems independently and attempt to
correlate the two sets of results. Although there are a priori reasons
for preferring the &st approach, some mention should nevertheless
be made of the other two.
Once site-groups have been obtained from specieslsite data by any
method of statistical analysis, these groups can actually be mapped
and this in itself may give some guide to habitat va3iations of ecological
importance. Agglomerative methods of classification will have concentrated on floristic similarities between the sites, while subdivisive
D
C.E.R.