Communities of Harvestmen
4: Continental comparisons
Widely separated communities.
Communities of Harvestmen4: Continental comparisonsWidely separated communities. |
|
|
To illustrate the use of these multivariate techniques the 261x122 (samples x species) data set, used in §2 for species diversity measures, can be analysed by ordination and classification. But first a word of caution: the data samples in this set are rather heterogeneous. We have to admit that the sampling intensity and duration varies greatly over the range of samples. As a result the number of individuals varies in a way that is not related directly to the actual population densities. By converting the species’ raw counts to proportions of each sample, this drawback can be ameliorated. Thus relative species’ abundances can form the basis of the analysis, retaining an important element of the quantitative information. The presence/absence samples are a very small proportion of the total data set and do not impair the overall pattern of results that is obtained.
Furthermore, the classification analysis by TWINSPAN can be carried out on just the presence/absence of species, which retains most of the information/variation across the samples anyway in a heterogeneous data set (Gauch, 1977) such as ours. The low values of occurrence frequency shown in Table 2 suggest a high rate of species change between samples and justify such an approach. Note that although the data matrix involves 122 species, the average species richness, as described above, is only five species per community. Additionally, the species’ low occurrence frequencies (Table 2) indicate that there are no “trashy cosmopolitan” species in the data set and that they all show discontinuities in their distribution.
For our complete data set, the sample scores for axes I and II are plotted in Figure 4. For axis I, λ = 1.0, indicating completely distinct species composition along axis I, where the Japanese and South American communities are clearly separated from the European (including British) ones [what's λ?]. The North American communities are split, those from Florida lying alongside the Japanese and South American points, those from Canada and Maine being in with the European samples. Further separation is evident against axis II (λ = 0.968), with some of the Canadian samples mingling with Europe and other Canadian and the Maine communities remaining separate. The sequence of species against the major axis I is used indirectly in Table 5. So distinct are the faunas of the different regions that a plot of the species’ scores against axes I and II yields just the same overall pattern of points as for the samples. Ordination is being used here simply to seek out patterns of variation in the data set. DCA and other ordination methods are examples of gradient analysis and if sufficient information is available the axes may be interpreted in relation to real environmental gradients, assuming that species turnover occurs as one moves along the gradient. The latter assumption is not invoked in a search simply for data patterning, but in fact could be true on a long time scale permitting evolutionary and distributive processes.
Thus, even though sample location was not used in the analysis, which was based purely on species composition, the ordination analysis gives a clear arrangement of the communities in relation to geographical location. Figure 4 illustrates how ordination analysis separates faunistically distinct communities into clearly separated clusters. Implicit in this approach is an underlying gradient in species’ distributions (and community composition), which is not fully sampled in our data set. Augmentation of these data by additional samples from the geographical regions covered and, especially from others, may well “fill in some of the gaps” and a repeat of the analysis on a more complete data set could reveal any underlying gradients.
Classification of the 261 samples by TWINSPAN, using just presence/absence of species, generated the dendrogram shown in Figure 5. This reflects the information shown in Table 5, in which the species are ordered and classified according to the TWINSPAN process. It is interesting to see the clear geographical patterns in Figure 5, shown by the regions included in each of the classes (end-groups) ranging from Japan in class A across to the Americas in classes F, G and H. The first three divisions all have λ = 1 [what's λ?], indicating completely different species complements for the sets of samples separated by these divisions: classes G (South America) and H (North America) split off at division 1 and then from each other at division 3, as well as class A (Japan) separated off at division 2. Even the splitting of the North American communities seems appropriate, with 14 of the Canadian samples in class D (along with many European and British ones) and the remainder in class F along with the Maine samples. The remaining North American samples, from Florida, are in class H closer to the South American communities in class G. The clear separation of the classes between geographical regions is verified by χ2 = 731.6 with 28 degrees of freedom, p < 0.001 for the contingency of the eight classes over the five regions.
These classes differ markedly in their characteristic species, as considered in terms of their constancy (Table 5).
Class A, the Japanese samples, is not dominated by any particular species, but has eight species with relatively low constancy values (range 13 – 38%).
Class B is entirely European and is more diverse with 34 species, though only four of these have constancy exceeding 50%: Trogulus nepaeiformis and Egaenus convexus (both with constancy of 56%), Trogulus tricarinatus (78%), and Rilaena triangularis (52%).
Class C, also European, is similar with 35 species and a maximum constancy of 71% for Amilenus auranticus; three other species have constancy >50%: Leiobunum rupstre at 59% and Platybunus bucephalus and Mitopus morio, both at 53%.
Class D includes three regions (Europe, Great Britain and North America) and has a correspondingly high number of species (47) of which only two have high constancy: Oligolophus tridens at 50% and Mitopus morio at 55%.
Class E is very special, comprising just five montane samples in Austria, with five species of which three have high constancy: Mitopus glacialis (100%), Dicranopalpus gasteinensis and Mitopus morio (both at 60%); the remaining two species both have constancy of 40% - Ischyropsalis kollari and Mitostoma chrysomelas.
Class F comprises 11 North American samples which share 12 species, the two most frequent being Leiobunum calcar and Odiellus pictus, both with constancy of 73%.
Class G involves many more species (31) which are present in the five South American samples in this class, but only one - Thrasychirus dentichelis - has a high constancy of 60%.
Finally, class H only has five species, three with high constancy: Vonones ornata and Hadrobunus sp., both at 83%, and Leiobunum aurugineum at 75%.
Separation of the communities in relation to their environments is a little less straightforward, but some patterns are suggested by the summary shown in Table 6; indeed, regarding this as a contingency table indicates a significant variation in classes’ frequencies between habitats, with χ2 = 128.6 with 63 degrees of freedom, p < 0.001. There is some confounding of habitat with region, for example class A, which is Japanese woodland but we have no data from other Japanese habitats. The scatter of samples across widely separated geographical locations complicates the analysis in relation to habitat, but this is explored below in relation to the European and British data.
|
Back to Opilionid Communities |
or |
 Back to Arachnologia |
 |
Back to Home page |