Formal statistical techniques have been defined to take the subjectivity out of community description. These techniques allow the data from community studies to sort themselves, without the investigator putting in any preconceived ideas about which species tend to be associated with each other or which environmental variables correlate most strongly with the species distributions. One such technique is classification.
Classification begins with the assumption that communities consist of relatively discrete entities. It produces groups of related communities by a process conceptually similar to taxonomic classification. In taxonomy, similar individuals are grouped together in species, similar species in genera, and so on. In community classification, communities with similar species compositions are grouped together in subsets, and similar subsets may be further combined if desired (see Ter Braak & Prentice, 1988, for details of the procedure).
The rotifer communities of a number of lakes in the North Island of New Zealand (Figure 16.7a) were subjected to a classification technique called cluster analysis (Duggan et al., 2002). Eight clusters or classes were identified (Figure 16.7b), each based solely on the arrays of species present and their abundances. The spatial distribution of each class of rotifer community in the New Zealand lakes is shown in Figure 16.7a. Note that there is little consistent spatial relationship; communities in each class are dotted about the island. This illustrates one of the strengths of classification. Classification methods show the structure within a series of communities without the necessity of picking out some supposedly relevant environmental variable in advance, a procedure that is necessary for gradient analysis.
Ordination is a mathematical treatment that allows communities to be organized on a graph so that those that are most similar in both species composition and relative abundance will appear closest together, whilst communities that differ greatly in the relative importance of a similar set of species, or that possess quite different species, appear far apart. Figure 16.7c shows the application of an ordination technique called canonical correspondence analysis
(CCA) to the rotifer communities (Ter Braak & Smilauer 1998). CCA also allows the community patterns to be examined in terms of environmental variables. Obviously, the success of the method now depends on having sampled an appropriate variety of environmental variables. This is a major snag in the procedure - we may not have measured the qualities in the environment that are most relevant. The relationships between rotifer community composition and a variety of physicochemical factors are shown in Figure 16.7c. The link between classification and ordination can be gauged by noting that communities falling into classes A-H, derived from classification, are also fairly distinctly separated on the CCA ordination graph.
Community classes A and B tend to be associated with high water transparency ('Secchi depth'), whereas those in classes G and H are associated with high total phosphorus and chlorophyll concentrations; the other lake classes take up intermediate positions. Lakes that have been subject to a greater level of runoff of agricultural fertilizers or input of sewage are described as eutrophic. These tend to have high phosphorus concentrations, leading to higher chlorophyll levels and lower transparency (a greater abundance of phytoplankton cells). Evidently, the rotifer communities are strongly influenced by the level of eutrophica-tion to which the lakes are subject. Species of rotifer that are characteristic of particularly eutrophic conditions, such as Keratella tecta and K. tropica (Figure 16.7d), were strongly represented in classes G and H, while those associated with more pristine conditions, such as Conochilus unicornis and Ascomorpha ovalis, were common in classes A and B.
The level of eutrophication, however, is not the only significant factor in explaining rotifer community composition. Class C communities, for example, while characteristic of intermediate phosphorus concentrations, can be differentiated along axis 2 according to dissolved oxygen concentration and lake temperature (themselves negatively related because oxygen solubility declines with increasing temperature).
What do these results tell us? First, and most specifically, the correlations with environmental factors, revealed by the analysis, give us some specific hypotheses to test about the relationship between community composition and underlying environmental factors. (Remember that correlation does not necessarily imply classification involves grouping similar communities together in clusters in ordination, communities are displayed on a graph so that those most similar in composition are closest together subsequently, it is necessary to ask what varies along the axes of the graph ordination can generate hypotheses for subsequent testing
Figure 16.7 (opposite) (a) Thirty-one lakes in the North Island of New Zealand where rotifer communities (78 species in total) were sampled and described. (b) Results of cluster analysis (classification) on species composition data from the 31 lakes (based on the Bray-Curtis similarity measure); lake communities that are most similar cluster together and eight clusters are identified (A-H). (c) Results of canonical correspondence analysis (ordination). The positions in ordination space are shown for lake sites (shown as letters A-H corresponding to their classification), individual rotifer species (orange arrows in top panel) and environmental factors (orange arrows in lower panel). (d) Silhouettes of four of the rotifer species. (After Duggan et al., 2002.)
causation. For example, dissolved oxygen and community composition may vary together because of a common response to another environmental factor. A direct causal link can only be proved by controlled experimentation.)
A second, more general point is relevant to the discussion of the nature of the community. The results emphasize that under a particular set of environmental conditions, a predictable association of species is likely to occur. It shows that community ecologists have more than just a totally arbitrary and ill-defined set of species to study.
Was this article helpful?