Nodal analysis of vegetation data

Hale, William Harriskine Grant (1981) Nodal analysis of vegetation data. University of Southampton, Doctoral Thesis.

Abstract

The aim of this investigation has been to develop nodal analysis, a particular numerical technique for analysing vegetation data, in the light of the advances in computing ability that have become available since the technique was originally proposed. 'Node' were conceived as recurring central components or points of reference in the variation in species occurrences in site, the term nodum referring to any such abstract vegetational unit erected. The original method using qualitative data attempted to provide an objective way of defining node, which previously had been performed subjectively. It was hoped that further consideration of various methodological alternatives would allow development towards the best method of nodal analysis possible, and that therefore any remaining problems in interpretation would be due to the concept of nodal analysis and not to the system of computation. This is a prerequisite to a fuller consideration of the merit of nodal analysis as a method of analysing vegetational data. Initial attention was directed at refining the original method of defining nodal coincidences between separate analysis of the sites and of the species. Various changes to this method were made. The classificatory strategy was altered from a monothetic technique known as association analysis to use a polythetic alternative technique previously described known as cVCR. A stopping rule was developed to define the termination level for this hierarchical classification, since for nodal analysis the number of groups is critical. Lastly, the manner of defining nodal status was based on the proportion of positive entries in a cell in comparison to other cells in the two-way table formed. These modifications gave improved results over the original version of the method. Besides this modified original method, some consideration was given to alternative strategies based on existing methods. None proved wholly satisfactory, however. An entirely new approach was then pursued based on optimisation of the internal concentration of the entries in the data matrix, attempting to make the process of nodal definition more flexible than in the modified original method. Some success was achieved using a parameter derived from the 'contiguity' of positive entries to concentrate the entries. However, the method was very slow to opeiatr. and proved less satisfactory than had been hoped. After comparative testing the new method and the modified original method were considered to be the most promising of the alternative strategies. Brief consideration was then given to the extension of these to quantitative data. The 'information' used by these different methods was discussed. It was concluded that progress had been made towards defining the best strategy for nodal analysis, and that consideration of the value of the nodal approach was now required.

This record has no associated files available for download.

Contributors

Download statistics