Multifactorial analysis of distance in studies of ecological community structure

A framework is described for partitioning the matrix of between-sample distances that forms a starting point for many ecological studies into contributions attributable to different factors in a structured data set. This partitioning enables a series of ordinations to be produced that better enable an insight to be gained into the effects of the factors. More detailed application of the same partitioning provides a decomposition of each factorial effect into single degree-of-freedom contrasts, which enables reasons for observed trends to be investigated. The methods are illustrated by application to data from a study of marine community structure.