One of the reasons for there being so many observations ofcommunities in nature seems to be intimately related to receive wisdom (paradigms) about units of study. If an ecologist strongly believes that plants and/or animals are in communities, with few species overlapping from one community to another and with fairly abrupt boundaries, there can be problems for designing objective sampling. Suppose it is thought, or it has been described, that there are, say, three communities along some environmental gradient. Each of these communities is associated with particular dominant or abundant species (A and B; C; D and E; respectively). A study in some new area may well involve searching along the gradient until A and B are numerous together. This purportedly identifies the presence of the first community. The second community is identified to be where C are numerous, and the third where D and E are numerous together. Now, sampling around the three areas which have been already designated to be containing different communities will, inevitably, reveal great differences in the species making up the communities and in the densities or relative abundances of species in each community. Similarly, if two areas are defined to contain the same community, because they each have an abundance of the species that dominate that community, sampling in the two areas will inevitably reveal quite a lot of similarity.
Instead, sampling along the gradient should be at random or regular intervals, depending on the nature of the study and the hypotheses being tested. Then, at least, there is no prior dogma defining where to sample. Data about the identities or abundances ofspecies at each point sampled can then be analyzed to test hypotheses.
Sampling without unquestioned acceptance of the existence of communities is, however, also fraught with difficulties, as illustrated in Figure 1. In the first case (Figure 1a), there are three very well-defined communities along an environmental gradient. Very few species transcend these structures. Sampling at the sites indicated will reveal different sets of species, with very little overlap. The data would reveal that communities exist because sample points contain quite different sets of species.
Unfortunately, if the species are entirely independently distributed along the gradient, this type of sampling will not be useful. Sampling at intervals along a gradient (as in Figure 1b), where species have independent distributions, would also reveal apparent communities. There would be different sets of species at each place sampled. The data would be very similar whether or not communities really exist.
Independence of distributions is essentially the null hypothesis against which to test the hypothesis that species are clumped as communities, with nonrandom commonality of boundaries. There is, of course, a third alternative that species have boundaries that are more regularly spaced than is the case for randomly distributed species. This is, however, irrelevant to any attempt to identify the existence of communities.
This has led to the realization that identification of nonrandom patterns in arrangements of species along environmental gradients requires data about the actual distributions, rather than sampling of organisms at intervals across the habitat. One methodology uses contiguous quadrats across the gradient. The numbers ofspecies that have a boundary in each quadrat are recorded (as in Figures 1c and 1d). So, transects of quadrats up a mountainside would be examined. As one progresses from the bottom upward, the first occurrence of a member of a particular species indicates its lower boundary. The number of these lower boundaries is recorded for each quadrat. Similarly, the quadrat in which the last member of a species is encountered denotes its upper boundary.
This is illustrated for the two distributions, each with 18 species, in Figure 1. For the species in Figure 1a, the numbers of lower and upper boundaries per quadrat are shown (Figure 1c). The same data are shown for the distribution in Figure 1b; there are clear differences. The clumped species have more boundaries in some quadrats (i.e., more 2's and 3's) than shown by the randomly scattered species. The former also have more and larger gaps between quadrats with boundaries.
Methods to analyze such data began to appear in ecological papers in the mid-1970s. Note that it is therefore a recent phenomenon to have statistical techniques to use in tests of hypotheses about the existence of communities where these are defined in terms of nonrandom coincidences of their boundaries. Prior to the last 30 years, it was very difficult to use the sorts of data collected by community ecologists to distinguish between communities (i.e., with clumped boundaries), random and regular patterns of distributions of species.
Was this article helpful?