Development and evaluation of new approaches in phytosociology with special regard to vegetation classification
Phytosociology had its starting point in the beginning of the 20th century as one of several schools which deal with vegetation from a scientific point of view. Josias Braun-Blanquet, who-se textbook ‘Pflanzensoziologie’ had been published in its first edition in 1928, could be seen as the founder of this school. Basic ideas of the Braun-Blanquet approach of vegetation science are the following:
• Plant stands on plots are documented in a standardised manner, the so-called vegetation relevés, which form an adequate mean between being too superficial and being too lavish. They comprise information on location, site conditions and vegetation structure as well as a complete list of the occurring plant species each of those quantified by means of a combined cover-abundance scale.
• Concrete plant stands (phytocoenoses) are thought to be assignable to abstract plant communities (phytocoena), which are characterised and separated one from another with regard to their whole species combination (floristic-coenological method). Since each of its constituent species has a certain ecological optimum and a specific geographical range it should be expected that they as a whole result in a clearly limited ecological space and synareal in which the community is distributed. Therefore it is not necessary to use the accordance in site conditions and geographical range as additional classification criteria.
• Phytocoena are arranged into a hierarchical system according to their floristic similarity. The principal ranks of this system are bottom to top association, alliance, order and class. Those so-called syntaxa are given scientific names deriving from one or two species names.
• As means for recognition and distinction of syntaxa diagnostic species are derived from the phytosociological data: Species which are restricted to a large extent to one such syntaxon are called character species, whereas differential species only differentiate one side only.
Phytosociology has undergone an enormous impetus within the last century and has integrated most of the other former schools of vegetation science step by step. By means of the Braun-Blanquet approach, millions of relevés from all continents have been collected until now. For several countries comprehensive overviews of the plant communities occurring on their territories have been published. Since the 1960s numerous attempts have been made to automate the classification process by means of computer programs (numerical syntaxonomy). Phytosociology also found its way into legislation in the field of nature conservancy as protected habitats defined by the occurrence of certain syntaxa.
In spite of this quantitative success phytosociology was always exposed to intense criticism, going as far as to blame it for not being scientific. Indeed, it must be stated that consistent and generally accepted methods are missing in syntaxonomy as one of the core subjects of phytosociology – even though a huge amount of phytosociological papers have been published and also numerous textbooks. Therefore the major aim of this Ph.D thesis is to develop such methods and to back them up from a theoretical point of view.
As a first step, the three major purposes of vegetation classifications are determined: (1) Naming of the research object to enable communication about them. (2) Reduction of data and making them representable. This means that instead of determining and describing the distribution of numerous species doing the same only for several plant communities will be suitable generally. (3) Framework for the ‘inductive generalisation’: The floristic-coenological delimitation of phytocoena results into a far reaching accordance of the phytocoenoses belonging to them with regard to species combination and site conditions. Phytocoena therefore are suitable reference entities especially for ecological research and nature conservation. Moreover they can be used for bioindication.
These purposes lead to the requests to be made on vegetation classifications which preferably should be meat. Most important are both the inner coherence of the separated units with respect to their major properties and their simple and clear discernibility. As further criteria the following should be mentioned: (1) completeness of the system. (2) Stability of the classification. (3) Tolerance against heterogeneous data. (4) Supra-regional applicability. (5) Applicability for different purposes. (6) Hierarchical structure. (7) Equivalence of syntaxa of the same rank. (8) Adequate number of discerned entities.
As a whole these criteria are already fulfilled in the ‘classical’ form of the Braun-Blanquet approach. Nevertheless some weak points remain:
• Until now precise and operational definitions for major concepts are missing. Syntaxonomy therefore is often difficult to comprehend for laymen. Instead of clear and verifiable criteria for and against a certain classification authors of phytosociological papers often refer to their own ‘expert knowledge’ or quote the opinion of a renowned phytosociologist.
• Even though it has been well known for a long time, that numerous phytocoenoses and phytocoena are lacking character species, there is no generally accepted and methodically sound way to include them into the phytosociological system. As a result such plant stands frequently are not documented by relevés or at least not taken into account in the final syntaxonomic treatment.
• Among phytosociologists belief in minimal areas is widely distributed. These should be relevé areas, beginning with which the species number does not increase substantially any more with further enlargement of the area. But it is well known for a long time by numerous empirical data as well as via theoretical arguments that every increase in area will cause an increase in the mean species number. The only problem ist that this increase is hardly recognizable when plotting the two axes of a species-area-curve in a linear scaling. This is because the function more or less follows a power function. The assumption of the existence of minimal areas lead many phytosociologists not to draw their major attention to relevé areas in syntaxonomy as long as they approximately correspond to the assumed mimimal areas. Relevé sizes for different vegetation types suggested by various textbooks differ by 150,000. Due to the dependence of species numbers from area it is not permissible to compare relevés of different area sizes. Most of the synthetic properties of plant communities as well are influenced by area size. This especially holds true for constancy which is the percentage of occurrences of a certain species within a set of relevés. By means of a conceptual model and empirical data a function is derived which approximately describes this dependence: St (A) = 1 – (1 – St0)(A / A0)^0,42, where St is the constancy and A the area size.
The classification method put forward is formulated as an axiomatic system comprising twelve suggestions of definitions:
1. Phytocoenoses are defined in a pure operational manner as the plant individuals of different species growing in a time-space unit of a certain dimension. Neither discreteness nor integration are demanded a priori. This definition includes expressively all synusiae thriving in that time-space unit as for example epiphytes.
2. The so-called ‘basic syntaxonomic axiom’ states that every phytocoenosis belongs – within a syntaxonomic system – to exactly one syntaxon of a certain rank.
3. The classical definition of fidelity degrees by SZAFER & PAWŁOWSKI (1927) is both contradictory and impractical. Therefore a new differential species criterion is presented in which it says that a species can be called differential of one syntaxon versus another syntaxon of the same rank if its constancy is at least twice as high and this difference in commonness most probably is not due to chance. This formulation goes back to BERGMEIER & al. (1990) but uses constancy percentages instead of constancy classes as otherwise there would be unreasonable changes in the minimum requirements for differential species below and above the class borders.
4. Whereas constancy at association level and below is defined as percentage of relevés in which the species occurs, a constancy reference value (short: constancy) of a higher syntaxon should be calculated as a mean of the constancy values in all associations belonging to it. This me-thod of calculation is due to the fact that associations are considered the basic units of the system and prevents the results from being influenced from different examination intensities in different associations.
5. A differential species below the vegetation class must fulfil the differential species criterion against all other syntaxa of the same rank within the syntaxon of the next higher rank. To a-void both taxa from being named differential species which nearly do not contribute to the discernibility of the syntaxa and effects due to chance a restriction has to be appended: Only those taxa should be given the status of differential species which have at least 20 % constancy within the particular syntaxon and not more than 20 % in the compared syntaxa.
6. As common differential species of a class are those taxa named which are a nowhere seen character species within the particular structural type, but fulfil the differential criterion of two or three classes against all other classes of this type.
7. Character species of a syntaxon are those taxa which meet the differential species criterion compared with all other syntaxa of the same rank within a structural type. The function of character species is twofold: (1) In the classification process the principal demand of the existence of character taxa ascertains the approximate equivalence of syntaxa of one rank, which especially holds true for associations. (2) When a classification is done on the basis of the complete species combination, character species as those taxa which have a clear sociological optimum in a particular syntaxon can be used best as a means for recognition and discrimination of the entities. If the next higher syntaxa occur nowhere together, as an exception from the general rule, one taxon could be named character species in two independent syntaxa. The major reason for restricting the character species to structural types is the dependency of constancy from area size. Since the customary relevé sizes widely differ between different vegetation types, which seems to be sensible at least to some extent, it is not acceptable to classify all of them within one system. It seems to be appropriate to discern two (vegetation types with and without phanerophytes) or three (vegetation types with phanerophytes, without phanerophytes, but with other vascular plants and those built up solely by one layer of mosses, lichens and algae) floristically defined structural types a priori. Some problems related to the practical implementation of this rule are discussed and a pragmatic approach is recommended.
8. Species which meet the character species criterion within more than one syntaxon fitted into one another, which is often the case, are called transgressive character species.
9. With regard to syntaxa not possessing character species of their own it is recommended to apply the central syntaxon concept of DIERSCHKE (1981) at all syntaxonomic levels below the class. Accordingly, at a maximum one central syntaxon not or not sufficiently characterised by character species of its own can be distinguished within a syntaxon of the next higher rank. This is named in the same manner as all other syntaxa. As compared to the ‘deductive meth-ode’ developed by KOPECKÝ & HEJNÝ (1971) and other approaches to deal with negatively characterised syntaxa the suggested method has several theoretical and practical advantages. These are discussed in detail. Most important is probably the fact that the presented approach avoids the introduction of different ways of naming syntaxa, which unavoidably gives the impression that there would be an ecological difference – which in fact does not exist.
10. Syntaxa from association to class level either must be sufficiently characterised by character species of their own or be the central syntaxon of the next higher entity. ‘Sufficiently’ in this case means a constancy sum of all character species of at least 100%.
11. The association is the lowest syntaxon that could be characterised by character species of its own and not divided further in such syntaxa or otherwise it can be regarded as the central syntaxon of a (sub-) alliance.
12. The class is the highest syntaxon characterised well by character species within one structural type. To measure the ‘quality’ of a syntaxon the sum of the constancy values of its character species seems to be appropriate. An adequate classification then would be one in which the constancy sums of all classes are as high as possible. Both the minimum and the mean of those should be maximised.
Some further recommendations for syntaxonomic work are presented which do not form part of the above axiomatic system:
• ‘Graduated’ hierarchies are – from an information theoretical point of view – favourable to ‘flat’ ones. In many cases they reflect the structure of syntaxonomic data better.
• Below association level it seems inappropriate to prolong the linear-monohierarchical classification from above. Here a multidimensional-polyhierarchical approach in which different complexes of differentiating factors stand aside with equal importance has its advantages.
• Formal syntaxonomic classification above class level is to be rejected since the class is ac-cording to the suggested definition 12 the uppermost syntaxon.
The syntaxonomic concept leads to requests upon the drawing up of relevés, the most important of which are the following:
• Since – as has been shown – sound syntaxonomic classification is only possible on the basis of relevés of the same (or at least a similar) size, standard sizes for relevés in different structural types are proposed.
• Due to the above definition of phytocoenoses and to the often high diagnostic value of non-vascular plants it is requested to document in relevés of phytocoenoses (holocoenoses) in principle all macroscopically visual photoautotroph organisms, including all synusiae such as epiphytes.
The practical application of the presented syntaxonomic concept could be understood as a problem of optimisation which has to be treated in an iterative process: The most similar relevés on a floristic basis are to be combined together until the emerging units either possess character species of their own or can be regarded as the central association of a higher unit which evolves from the further joining of these basic units (associations).
The treatment continues with analyses on the question which quotients and differences of percentage constancy values accord to statistical significant differences in commonness. They lead to the recommendation to complete – in case of low numbers of relevés – the requested constancy quotient q > 2 from the differential species criterion by an minimal constancy difference of Δ = (2 ⋅ n1)-1/2, where n1 is the number of relevés in the particular syntaxon.
Additionally, some recommendations concerning the adequate presentation of syntaxonomic data in form of tables arise from the presented syntaxonomic concept:
• It is essential always to give the (mean) sizes of the relevé areas, also in the case of constancy tables.
• Constancy values should be presented as percentages and not in form of constancy classes to allow the application of the differential species criterion.
• To make the results transparent, it is advisable, to give columns for all syntaxonomic levels in constancy tables and not only for associations and their subdivisions.
Nomenclature is the second indispensable part of syntaxonomy besides the classification. It is ruled by the ‘International Code of Phytosociological Nomenclature’ (ICPN), which has been published in its third edition by WEBER & al. (2000). The basic function of the ICPN is to guarantee unambiguity and stability of the scientific names of syntaxa. To achieve this, two major principles are employed, that of priority and that of nomenclatural types. Although this construction and the Code as a whole have proven to be useful, some possibilities for improvement remain, of which the following should render prominent:
• It seems necessary to clearly separate the system of phytocoena and that of synusiae. It is a logical contradiction if one entity at the same time is regarded as syntaxon and synusia as it is common practice at the time being. It is therefore recommended to put the synusial system on a clear nomenclatural basis, by reserving separate endings for its ranks within the ICPN.
• The demands for original diagnoses of associations should be increased, to avoid the further accumulation of nomenclatural ‘ballast’ in literature. It should be pointed out in the ICPN that the requested full species list for a description of an association has to include bryophytes and lichens as well.
• The ruling of subassociations by the ICPN should be omitted, especially due to the fact that it is incompatible with multidimensional subdivisions.
How the presented ideas work when practically applied is shown in detail by several examples, most of which are taken from the comprehensive treatment of the plant communities in the German state Mecklenburg-Vorpommern by BERG & al. (2001b), where they have been use in a large scale survey for the first time. In particular, it is shown how more equivalent classes could be achieved in the system and how central syntaxa could solve classificatory problems about which has been debated at length but to no avail.
Distribution maps of syntaxa have been published rarely so far. One possibility to generate such synchorological maps is the superimposition of distribution maps of their diagnostic species. This results in maps of ‘potential synareals’. This method is both applicable to outline and lattice maps. How this could be done is described in detail. Examples of either case are shown and their interpretation is discussed.
Finally, the presented syntaxonomic concept is examined as a whole: Due to the axiomatic construction it is consistent. It meets the above given requests for vegetation classifications better than any other known approach. It has proven its practical suitability in BERG & al. (2001b), where it enabled the establishment of a syntaxonomic classification of the extant vegetation types within a region, based on a large databank. A major strength of the approach could be seen in the fact, that clear criteria are given to evaluate different classifications with respect to which of those is conform with the method at all and – if there being more – which is the best of them. Likewise high importance has the applicability of the presented method both in manual table work and as basis for an implementation in a numerical classification algorithm. The latter will be a necessary property of any syntaxonomic approach with which the data of one of the large national and supranational vegetation-plot databases arising can be analysed one day.