A Rapid Multivariate Method for the Detection and Classification of Groups of Ecologically Related Species

Abstract

A method is suggested for the rapid analysis of large ecological surveys by computers with limited high speed storage. Coincidence of occurrence rather than homogeneity is taken as the fundamental property of ecological groupings. The species and the quadrats that contain them are classified according to their similarity with the population sample as a whole or with a sub-set of this population. Using the statistic μ'2 = Σ(oi - ei)2 a test is made of species interaction on the group properties of the population sample and on each sub-set. The quadrats are then divided depending on whether or not they contain the species with the highest interaction. As each species is tested for its interaction value the method is multivariate. The process is repeated on the sub-sets and a hierarchial division is made of the data. A statistic is used to define an ecological group and quadrat sets that fail to reach the desired level of significance are classified as ecotone or transitional units. As the method is based on the calculation of species and quadrat attributes of a continuous nature the method lends itself to ordination studies as well as to classification. The analysis system, written in Fortran II D, is available on application to the authors.