How the errors in the process of vegetation analysis in the field and the data processing affect the results of classification (with arctic communities as an example)

Abstract

A list of species with an access of their “amount” (number of individuals, true/projective cover, biomass) on a plot of a standard size is the information that is necessary for an objective classification of plant communities, no matter what principles it is based on. Information on species composition, the variation both in their “amount” and constancy in the pool of geobotanical relevés is the basis for their clustering and the delimitation of syntaxonomical units. The only possible documents recording this information are geobotanical relevés, both published in the open press and stored in databases/archives. The completeness of species list within these depends on such parameters as time spent working in the field and technique (standard eye assessment at the sample plot (25 or 100 m2), a series of smaller (less than 1 m2) plots as well the researcher’s professionalism. The statement about the need to obtain a complete list of species in each stand seems an axiom, which is not fulfilled in practice. In Taymyr, when describing zonal communities for more than 2 hours, were recorded about 75 % of species, found on a permanent, carefully studied, sample plot of the same association. It is not necessary to comment that eye assessment of both composition and quantitative parameters are far fr om perfect. The same “amount” of species (abundance, cover) can be reflected differently not only by various researchers, but even by one, and not only in different years and areas, but as well in one season depending on such factors as what reléve was before, at what time of day (evening lighting in the Arctic is a serious factor), in what weather, etc. The result is influenced by factors such as the size and shape of a sample plot. The size is obvious: it should be no smaller than minimal area i. e. an area that gives an adequate idea of the composition of the described plot (Barkman, 1958, 1993). For the Arctic, according to the results of special work (Matveyeva, 1998), an area of 25 m2 was recommended for the species richest communities with a complex horizontal structure and 9 m2 for all others. The most frequent, generally accepted shape is a square. The use of another one depends on the community configuration: in narrow, elongated, winding, it must be “adjusted” to the outline of the stand, or it is better to abandon a single large plot in favor of several smaller ones. Location of the sample plot in space: preferably the most central, equidistant from the community boundaries. The smaller size of the community in general or its narrowness is fraught by the effect of visinizm (Barkman, 1958, 1990): the plot and therefore the list will get species of neighboring communities. No less problematic is the eye assessment of the species”amount”. It is generally accepted to evaluate projective cover, since neither to count the number of individuals, nor the determine of the true cover, and even more no biomass, in numerous relevés, is unrealistic. Despite the enormous field experience of Western European phytosociologists during the first half of the last century, who become convinced that it was impossible to determine the projective cover by eye with an accuracy of 1 %, they came to a reasonable decision to use grades. In the practice of classification according to the Brown-Blanquet approach, the 7-grade scale (r, +, 1, 2, 3, 4, 5), periodically slightly modified, became the most widely used (see: Becking, 1957). To our great regret, we gradually began to use individual authors’ scales with larger number of grades — as a result the same numbers in different scales represent different cover in percent. This is more or less acceptable as long as we are talking about one paper. However when a lot of data are collected from different, often geographically remote, areas, and a large number of relevés are put into a single table, the possibility of an error in assessing the species cover is huge. Recently again, researches began to show the projective cover with an accuracy of 1 %, arguing that percent can be converted into grades, but not vice versa. The objection is the same: it is impossible to determine the projective cover by eye with an accuracy of 1 %. Not only that each researcher has his own mistake (before starting work in the field in Taymyr, we checked our estimates more than once: the difference in smaller values always differed by 1–3 %, in higher ones (after 25 %) by 5–10 %); the same person will give a different percentage depending on many reasons. Hence, if it is not specified how the projective cover was assessed (for example, on 100 m2 sample plot on each 1 m2, i.e. 100 squares), then the figures 8 %, 11 %, 19 %, 38 %, 41 %, etc. — are deliberate misrepresentation. An intermediate result of the discussion of field errors: the eye assessment of the composition, the total number of species and their amount depends on such a number of tricks that its accuracy leaves much to be desired. What is the alternative? It is available and repeatedly tested. From personal experience in Taimyr, this is the use of small (from 0.1×0.1 to 0.5×0.5 m) plots on which species identification and estimation of their numbers is incomparably more accurate than on a large test site. Another approaches were: to use a ­graduated piece of metal-wire — a field device for creating a virtual plot in the form of 3 circles (0.1, 0.01, 0.001 m2 in size) for obtaining data in heterogeneous Arctic communities, proposed by Danish phytosociologist T. Böcher (1975); to estimate separately the “amount” and number of species for each element of intra-community mosaic. Always the practice on identifying species on smaller plots gave results 1.5–2 times higher than in the standard relevé, i. e. the species richness of the Arctic communities is underestimated and often very much (up to 30 %). It should also be noted that it is always a great success if specialists in different plant groups work together in the field, for the Arctic in particular these are bryologists and lichenologists: the numbers of species in relevés are always many times higher. All the above facts are the arguments of the evident incomplete list of species under standard field practice. What follows from this? First position is that: 1) the number and set of species in different communities of the same association is always different (“... the number of species grows steadily as the number of relevés increases even in the most homotonous types of vegetation” (Barkman, 1990: p. 1217); 2) the species richness of the association (coenoflora) is many times higher than that of a particular stand. Few examples: in polar desert zone on Bolshevik Isl. in the zonal ass. Deschampsio-Aulacomnietum turgidi Matveyeva 2006 179 species were identified in 18 stands, with 49–84 in each one (70 in average), i.e. 2.5 times less than in the association; for other associations (same place) this ratio is 1.7–2.8 (Matveyeva, 2006); on Taymyr, this indicator varies from 1.0 to 3.4, and there is no connection with either the zonal position or the number of relevés (Matveyeva, 2009). Having in mind the significant differences in species richness of communities and syntaxa, it would be possible to see the unevenness of species distribution in the landscape, which is also indicated by the species spread in constancy classes with a high (up to 40– 50 %) proportion of species of the low (1–20 %) constancy. The question whether this reflects the reality is asked extremely rarely. Greatly possible is that a large number of species with low constancy in syntaxa is a consequence of their oversight in the field. In the data processing such field errors are not taken into account, and in the procedure of syntaxon differentiation one attaches importance to variances in the presence of low-constant (I and II classes) species. The difference in one step fits into the statistical error for any class (even assuming the unbelievable that every species has been recorded in each stand that does not happen in practice). Hence, it is incorrect to consider such weak differences in the species constancy (V–IV, IV–III, III–II, II–I) as essential for choosing selective character species (occur in several syntaxa, but more often in one). Against giving diagnostic value to low-constant species, in particular those with low (r, +) abundance, experts in phytosociology have been warning for a long time (see: Barkman, 1990, 1991). Outside the scope of analysis is also a topic of how many relevés are needed for definition a new syntaxon, as well as on which territory — in one or in several sites. However just this is what gives reason of describing new variants of already known associations in another area wh ere another researcher works with his own errors in obtaining data. An answer to the question of what to do with all this is as follows: no matter how tempting it is to attract indicators of species composition, richness and constancy, for any kinds of assessment and comparisons, they should be used with great care, having in mind the methodological errors in their obtaining. But even if forget about the relevé quality, their table treatment is also not at the highest level (three example are given). In order not to complete the given essay on the subject of how imperfect we are in our attempts not only in understanding, but even in describing nature on such a sad note, here, as an ironic defense, is a quote from the ancient Greek (485–410 BC) philosopher Protagoras: “... about every subject we can say both in two ways and in the opposite way”, which in our practice is often sounds as: “but the author sees it like that”. However if we consider ourselves as people of science, we have to operate with facts, to justify our position by evidence, not by assumptions, guesses, and emotions ... In any comparisons of the species diversity of communities and the use of these data for their classification, it should be keeping in memory th