An Application of Detrended Correspondence Analysis and Non-Metric Multidimensional Scaling to the Identification and Analysis of Environmental Factor Complexes and Vegetation Structures

Abstract

(1) A four-stage computational procedure is proposed for the analysis of complex vegetation-environment systems. This consists of: (i) an initial ordination by Detrended Correspondence Analysis (DCA), summarizing the structure in a vegetation data set; (ii) simple 2 x 2 correlation (using Kendall's rank-order correlation coefficient) of a matrix of environmental data and DCA sample scores; (iii) environmental plexus construction by means of the application of Non-metric Multidimensional Scaling (NMDS), using the environmental correlation matrix as data; and finally (iv) construction of an integrated plexus, incorporating both environmental variables and DCA ( vegetation) gradients. (2) The procedure was tested on data from the Storbreen glacier foreland, Jotunheimen, Norway, as part of a wider study of this successional system. On the basis of the NMDS plexus diagrams, two principal structures were identified: (i) a 'terrain-age factor complex', principally comprising measures of soil development and site disturbance and (ii) an 'exposure-moisture regime-snowmelt factor complex'. Clear links were suggested between these complexes and two DCA axes. These relationships were found to be compatible with the trends displayed in the species ordination. (3) The analytical sequence, which involves an extension of previous applications of both DCA and NMDS, provides a comprehensive framework for the examination of vegetation-environment interactions. The most novel feature of this sequence is the representation of the relationships in interpretable plexus diagrams. The approach is recommended for the generation of ecological hypotheses which interrelate complex gradients in environmental factors and vegetational composition.