- Formalized reproduction of an expert-based phytosociological
classification -
607
Discussion
The Cocktail method, combined with the similarity-
based assignment of relevés to vegetation units, satis-
factorily reproduced the expert-based classification of
subalpine tall-forb vegetation of the Czech Republic
(Kočí 2001) at the level of associations. We formulated
unequivocal assignment criteria for 14 of 16 associa-
tions recognized by Kočí (2001) (Table 2). The remain-
ing two associations of the original classification could
also be formally defined, but we decided to abandon
them as they lacked positive differentiation and were
highly similar to some other associations. Tables 3 and 4
show generally good agreement between the expert-
based classification and the formalized classification
proposed in this paper. A poor agreement was found
between some expert-based associations composed
mainly of generalist species and their corresponding
Cocktail definitions; however, this was substantially
improved by the subsequent similarity-based assign-
ment of relevés to the associations. These results show
that the expert-based vegetation classification can be
successfully reproduced by the formalized methods and
converted to the computer expert systems for identifica-
tion of vegetation units (Noble 1987). Similar results
can also be achieved by other methods such as neural
networks (Ejrnaes et al. 2002). However, the absolute
agreement will rarely be achieved since the expert-
based classifications contain various inconsistencies and
consider also non-floristic classification criteria such as
vegetation structure, chorology, stand history, position
in successional seres, and abiotic site factors (Westhoff
1967; Feoli 1984; Pignatti et al. 1995).
Our ability to reproduce narrowly conceived vegeta-
tion units as associations was due to involving the
dominance of individual species as a classification crite-
rion, in addition to the presence of species groups.
Previous applications of the Cocktail method showed
that if only species groups without other classification
criteria were used (Bruelheide 1995; Jandt 1999, 2000;
Bruelheide & Chytrý 2000), it was mostly possible to
define alliances or broad groups of associations but not
the associations. On the other hand, studies employing
dominance or presence of individual species in addition
to the species groups (Pflume 1999; Täuber 2000) were
quite successful in reproducing the traditional associa-
tions. These results reflect the fact that most associa-
tions traditionally recognized in Central Europe are
defined not only by species presence/absence but also
by quantitative proportions among species.
Cocktail definitions of vegetation units may overlap
to some degree, which results in an inconvenient prop-
erty that some relevés may be assigned to more than one
vegetation unit. These overlaps are larger the broader
the definitions of vegetation units are. Therefore it is
often necessary to create narrow definitions including
only the most typical, core relevés of vegetation units.
Still, even with narrow definitions some relevés may be
assigned to more than one vegetation unit. Jandt (1999),
Pflume (1999) and Täuber (2000) fixed this issue by
introducing a hierarchy of vegetation units. They started
with the unit that was subjectively given the highest
Table 4. Relationships between the associations of the expert-based classification and of the formalized classification. Both the
relevés assigned directly by the Cocktail definitions and relevés assigned by subsequent similarity calculations are considered in the
formally defined associations. The associations of the two classifications are compared by the phi coefficient (multiplied by 100);
negative
Φ values are not shown and values higher than 50 are printed in bold. Numbers of associations are the same as in Table 2.
Formally defined
Expert-based associations recognized by Koćí (2001)
associations
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
No. of relevés 34
84
120
36
41
55
15
13
4
59
14
15
23
53
37
115
1
64
69
-
8
11
-
-
-
-
-
-
-
-
-
-
-
-
2-3
184
-
55
46
-
-
2
-
-
-
-
-
-
-
-
-
-
4
25
-
-
-
55
18
-
-
-
-
-
-
-
-
-
-
-
5
25
-
-
-
3
58
-
-
-
-
-
-
13
-
-
-
-
6
31
-
-
-
-
-
71
-
-
-
-
-
-
-
-
-
-
7
22
-
-
-
-
-
-
77
28
-
-
-
-
-
-
-
-
8
8
-
1
-
-
-
-
-
58
17
-
-
-
-
-
-
-
10
58
-
-
-
-
-
-
-
-
-
81
-
10
1
-
-
-
11
14
-
-
-
-
-
-
-
-
-
-
100
-
-
-
-
-
12
28
-
-
-
-
26
-
-
-
-
-
-
32
37
-
-
-
13
11
-
-
-
-
-
5
-
-
-
-
-
-
56
-
-
-
14
54
-
-
-
-
-
-
-
-
-
1
-
-
-
91
-
-
15
38
-
-
-
-
-
3
-
-
-
-
-
-
-
-
82
-
16
156
-
-
-
-
-
-
-
-
10
-
-
-
-
-
-
79