Abstracts Biological Sciences

Numerical classification of range vegetation and statistical analysis of its ecology

by William Tucker Pyott

The purpose of the research was to computerize mathematical procedures for the analysis of range vegetation and environmental data. The specific objectives were as follows: 1. to develop and apply computer techniques to the classification of vegetation in order to provide a phytosociological framework within which to investigate the ecology of range vegetation, and 2. to adapt computer software for multivariate statistical techniques to the analysis of vegetation-environment relationships. In fulfilling these objectives a synthesis of classical phytosociological principles and methods with techniques of numerical taxonomy was achieved. A Euclidean mathematical model for numerical taxonomy was selected. Hierarchical, polythetic-agglomerative classification procedures, measures of interspecies and interstand relationships, and forms of data (species presence vs. species amount) were compared. Computer programs were written in FORTRAN for classification. Data used were from sagebrush-grass ranges of eastern Oregon as well as from the literature. Classification results were evaluated according to the principles of the Braun-Blanquet system, by the chaining coefficient, and by stepwise discriminant analysis. The latter analysis was also used to determine the degree of correlation of environmental variables with vegetation pattern (from classification). Ward's method, which minimizes the within groups sum of squares (error S. S.), was selected as the classification strategy for both species and stands. Squared Euclidean distance for presence only data was selected as the coefficient of interspecies relationships. It produced the best definition of species groups and had the lowest chaining coefficient. Squared Euclidean distance calculated from standardized vegetation data (such as % foliage cover transformed to zero mean and unit variance by species) gave as good or better results for stand classification as the squared distance calculated from presence only data (depending on the data set). Species were classified first to determine groups of potential differential species (internally cohesive, mutually exclusive species groups). Only these species were retained as the basis of stand classification. In the stepwise discriminant analysis of stand groups, differential species were invariably the first few entered as discriminating variables. Patterns of similarities and differences appeared in the environmental data which were highly similar to patterns in the results of the vegetation classification. In stepwise discriminant analysis of environmental data, site variables and soil physical and chemical variables were equally prominent as the first discriminants entered. The hierarchical nature of the classification procedure makes possible the determination of species groups which identify vegetation units (and thus environments) at different levels of synthesis. Thus the same data base can be used for broad land use planning purposes as well as for intensive management of…