Objective methods for the classification of vegetation. I. The use of positive interspecific correlation

Abstract

It is argued that correlations between the presence or quantity of different species recorded in quadrats imply heterogeneity in the vegetation in which the quadrats were placed. If the quadrat data can be so divided that within each subdivision no interspecific correlations occur, these subdivisions represent elementary classification units of the vegetation, and an objective classification may be arrived at if methods can be found of satisfying this requirement.

Four procedures are described for dividing a set of quadrat data into groups satisfying the above criterion of homogeneity. The simplest and apparently the most satisfactory consists in finding the most frequent species showing significant correlations with others, and separating in the first instance all quadrats contain- ing this species. Within this group correlations are again tested, and the process is repeated until a group with no significant correlations has been extracted. All other quadrats are then lumped together and the process is begun again from the beginning. When the whole collection of quadrat data has been divided in this way into homogeneous groups, these groups are recombined where this can be done without the resulting larger group showing interspecific correlations.

The four procedures suggested are illustrated on data from a small area of the Victorian Mallee. Some of the results suggest that classification is an inap- propriate method of dealing with variation in this type of vegetation, and that a coordinate treatment might be more suitable.