Previous Page Table of Contents Next Page


Mathematical analysis of surveys of vegetation in the Sahelian zone

J.C. BILLE *

(*) J.C. Bille, Chercheur ORSTOM, B.P. 1386, Dakar, Senegal.


A. processing of the surveys
B. Pluriannual variations


SUMMARY

In order to be able to study as objectively as possible the vegetation of a Sahelian dune zone and to classify the phytosociological samples, data were processed by computer, using the mathematical analysis method of Roux and Roux (1957). The results obtained made it possible to define ecological groupings; to assess their variability as well as the affinities between different species. Changes in the flora related to the climate are also perceptible, and it is proposed to choose the maximum individualisation of the groupings to characterise the vegetation.

A. processing of the surveys

Attempts to describe tropical plant associations come up against a dearth of data in the entire area and lack of knowledge of the relationships between the plant formations, with the result that classical phytosociology has not in this case yet become accepted. In general, action is limited to the preparation of lists of species linked with a given ecological feature (soil hydromorphology, presence of a cuirass or a gritty horizon, granulometry of the substratum, etc.). Surveys of vegetation in which each species is accompanied by an abundance/dominance figure remain rather subjective, although the numbering is codified, because of the heterogeneity of the formations observed, particularly in the driest areas: here only the presence or absence of species have been taken into consideration.

The technique for classifying surveys (Roux G. and Roux M., 1957: "Concerning some methods of classification in phytosociology", Rev. Strat. Appliq. 14: 59-72) uses factorial analysis of correspondences: the species and the survey are regarded as a set of realizations of chance variables, and the results are expressed using the smallest number of these variables by rotating the datum lines. It then becomes possible to project constellations of surveys (or of species) in planes defined by the datum lines taken two at a time, and in general this is limited to the first five datum lines, which express the maximum elongation of the "cloud" of points.

Figure 1 shows the image obtained for the case under study by means of projection of the surveys in the plane of the first two datum lines. The surveys are indicated by different symbols depending on their topographical location: white circles for dune vegetation; white squares at slope bottom; white triangles for sharp rises, and black circles for the centres of depressions; while black triangles and black squares represent surveys carried out in woody areas, outside and within the depressions respectively. The points are spread out in five constellations, which are more or less compact:

1. A fairly homogeneous dune area where the surveys all include Aristida funiculata, A. mutabilis, Schoenefeldia gracilis, Lepharis linariifolia and Polycarpaea linearifolia as dominant species;

2. A constellation close to the one above in which may be found, in addition to the species under (1), Cenchrus, Eragrostis tremula, Commelina, Dactyloctenium, and in cases Diheterpogon hagerupii or Ctenium elegans;

3. A group of sciophytic surveys which are much more variable, with Chloris prieurii, Brachiaria hagerupii, Panicum laetum or Digitaria velutina;

4. A homogeneous area under trees in depressions, with Pennisetum pedicellatum, Papilionaceae, Triumfetta, Cassia Marremia, etc.;

5. A fairly heterogeneous group of heliophilous surveys of hydromorphic soils, characterized by Panicum humile, Eragrostis spp., Zornia glochidiata or Echinochloa corona.

The advantages of the method are three-fold: the classification of measurements takes fully into consideration all the species present in the area, without granting any of them special importance compared with the others; the greater or lesser compactness of each "cloud" is an indication of the homogeneity of the corresponding ecological group; and the distances between constellations express the individualization of one group towards the others. In addition, the same principle may be used to project the representative points of the species, thus highlighting those which specialize in a particular area and those which differentiate less.

Fig 1.

Fig 2.

B. Pluriannual variations

The previous exercise was repeated over the same control points during 1970, when rainfall was reduced to 200 mm. over a shorter period; the new image obtained appears in figure 2, from which it may be noted that:

1. The dune grouping was hardly affected, since its projections shifted in relation to the datum lines; at most it was slightly more homogeneous because of the dwindling or disappearance of chance species (small Cyperaceae, Oldenlandia, Monsonia, and so on).

2. The slope bottom and sharp rise area was assimilated to (1) to the point that it could not really be dissociated from it, and that it disappeared for the year 1970; indeed, certain characteristic species did not appear (Ctenium) or covered less area (Diheteropogon, Dactyloctenium).

3. The sciophytic group outside depressions became specialized, and there was no longer any continuum towards the previous groups; indeed, it became assimilated to group (4), several of the species in which were, in 1970, common to the two types with poorer flora. The temporary surface water group was less variable (absence of Eragrostis, Panicum humile, and Andropogon pinguipes, and general spread of Zornia); the reason is that the ground was submerged only to a very slight extent.

The new aspect of the vegetation is clearly as objective as the previous one, and experience has shown that to each year there corresponded a special mathematical representation linked to the variations in a complex set of ecological conditions: in 1972, the existence of infrequent Borreria- and Blepharis covered areas of type (2), but located in depressions; in 1974, the almost homogeneous distribution of Chloris prieurii assimilating the first three types of vegetation, and so on. This implies two main consequences:

- Without lengthy previous experience of dry savannah it seems difficult for an observer to obtain an exact idea of the vegetation over a single control year. For example, for the region studied here descriptions have often been made of areas of dwarf plants attributed to over-poor soils, whereas it has since been possible to observe that the experience and location of such areas were a matter of chance, and that it was a simple case of small areas disadvantaged because of storms, the nanism being an adaptive reaction to the drought on the part of numerous Sahelian herbaceous plants.

- A choice must be made, for the cartographic representation of the vegetation, between its least varied appearance (which may well be most frequent) and its potential variability, which will be realized in optimal plant growth conditions. The ideal solution would be one expressing a synthesis of the various possibilities and, if the grouping of neighbouring formations is the solution which presents the least risk of error, taking into account the various micro-environments certainly provides more insight, since each unit most often has a particular level of production despite sharing flora with various facies.

One would then be led to characterize herbaceous vegetation on the supposition of optimal individualization conditions, and indicating jointly the greater or lesser constancy of the originality of each element depending on climate and, possibly, on other factors (for example, exploitation).


Previous Page Top of Page Next Page