Multivariate Methods in Plant Ecology: I. Association-Analysis in Plant Communities

Williams, W; Lambert, J

The valuable pioneer work by Goodall (1953) on the use of interspec. associations for sorting quadrats into groups is based on his definition of a homogeneous unit of vegetation as one in which all species-associations are indeterminate or nonsignificant. The methods he eventually recommends, from empirical studies on data from the Australian Mallee, is to sort on the most abundant species involved in a positive association, pooling the residuum at each stage. Since statistical methods of this kind, however, inevitably require much large-scale computation, it is necessary both to examine the stat. foundations of any method proposed and to assess whether the ecological information obtained justifies the time and labor involved. Furthermore, the analysis of complex communities can normally only be brought within the reach of the practicing ecologist if the method is programed for a digital computer, and this requirement must be borne continuously in mind if prohibitive expense in computing time is to be avoided. It is implicit in the use of any sorting method that it is expected to reveal an underlying structure simpler than the raw matrix of associations. This is equivalent to assuming that the associations are not all independent; but if this is the case, the system is multivariate, and must be studied as such. With this in mind, we shall show, on theoretical grounds and by reference to two selected heathland communities, that a method of subdivision more effective than Goodall's can be devised: and we shall then attempt to assess the type of ecological information thereby obtained.

Multivariate Methods in Plant Ecology: I. Association-Analysis in Plant Communities

Series