An Integrated Multivariate Analysis of Forest Communities of the Central Great Smoky Mountains

Abstract

An integrated sequence of multivariate techniques was applied to vegetation and site data from 244 stands in the central Great Smoky Mountains National Park. The analyses included cluster analysis, stepwise multiple discriminant analysis with both species and site variables, reciprocal averaging, canonical analysis and canonical correlation. Taken together, they provided a system of successive refinement and cumulative evidence for analysis of vegetation and vegetation-site relationships. Nineteen forest community types were identified and examined. At the broad level, the vegetation pattern was most directly related to elevation, topography and stand history. However, some soil variables were important, specifically soil/?H and clay content of the ? horizon. The most useful measures of topographic effects were topographic position and total annual potential solar irradiation (a function of slope angle and aspect). Number of dead Castanea dentata remains provided a measure of past disturbance which was related to the vegetation pattern. Introduction The Great Smoky Mountains National Park, located in eastern Tennessee and western North Carolina (approximately 83? 30' W longitude, 35? 45' ? latitude), contains large areas of undisturbed and little-disturbed temperate forests. General descriptions of the physiography, climate, flora, general vegetation and land use history are available elsewhere (Cain, 1930, 1931, 1945; Whittaker, 1956, 1966; King and Stupka, 1950; Hoffman, 1964; Frome, 1966; Golden, 1974). Whittaker's (1956) monograph is the most comprehensive treatment of the general vegetation pattern. He provided a direct gradient analysis which related species populations and subjectively defined community types to elevation and to a qualitative moisture complex-gradient based on topographic characteristics. He did not obtain any data on soils. His scope was the entire national park and involved data from locations scattered throughout it. The study reported here was designed to supplement previous studies. It employs an integration of several multivariate techniques which have recently become widely used. The scope is the middle elevation forests of the central portion of the park. This was defined as forest stands at 750-1600 m elevation in an area ca. 20,000 ha, roughly centered on Mt. LeConte. Field observation indicated that the vegetation pattern of this area varied significantly in certain details from the general pattern reported by Whittaker (1956). For example, the high elevation Quercus alba communities he reported are absent here, and Tsuga canadensis is more prominent than in the remainder of the park. This paper has two primary purposes: (1) to more clearly define the forest vegetation pattern and forest vegetation-site relationships in the central Great Smoky Mountains, and (2) to illustrate the integrated use of a combination of multivariate techniques for defining and describing such vegetation attributes. Methods and Results data Data from 244 forest stands were used in the analyses and descriptions. Each stand was sampled using a 0.08 ha circular plot in which trees > 12.7 cm dbh were tallied by species in 5-cm dbh classes. Sample locations were selected primarily from topographic maps with the objective of obtaining a reasonable distribution among topographic situations in the elevation range 750-1600 m. Species importance percentages for each stand were computed as one-half the sum of relative density and relative basal area. 37 This content downloaded from 157.55.39.162 on Thu, 11 Aug 2016 04:52:52 UTC All use subject to http://about.jstor.org/terms 38 The American Midland Naturalist 106(1) Only tree species present in at least three stands (30 species) were used in the analyses. Taxonomic nomenclature follows Little (1953). Topographic measurements taken included elevation, slope angle, slope direction and slope position (as percent distance from ridge to cove or valley bottom). The potential annual solar irradiation (in langleys) was estimated for each site from the tables of Frank and Lee (1966). Soil profile descriptions and samples were taken from a soil pit near the center of each plot. Soil p?i and texture for each horizon were determined in the laboratory. numerical analyses An integrated sequence of multivariate procedures, involving both classification and ordination, was used to explore the vegetative and environmental relationships among the communities. The sequencing and integration of the procedures are illustrated in Figure 1. Cluster analysis and similarity sorting. ?Minimum within-group dispersion using standard distance (Orloci, 1967), a polythetic agglomerative clustering algorithm, was used to provide a basis for defining stand groups which represent community types. The data were species importance values. A dendrogram was constructed using output