Formalized reproduction of an expert-based phytosociological classification

- 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

---