Date of Award
Master of Science
Ecology and Evolutionary Biology
Herman H. Shugart, Jr.
J. S. Olsen, H. R. DeSelm
The purpose of this study was to evaluate the effectiveness of using discriminant analysis in assessing plant niches. As a preliminary to establishment of the Environmental Research Park Program at Oak Ridge, Tennessee, five sites were inventoried for herbaceous species. From this inventory, four sympatric species of Galium and seventeen co-occurring herbaceous species were selected for discriminant analysis. The four species of Galium were treated as two data sets: one was composed of information collected at one site (a mesic hardwood area) and the other contained data from two cedar sites of shallow soil over lime-stone bedrock. The seventeen herbaceous species all occurred in the mesic hardwood area.
Variables selected for the analysis included site characteristics such as aspect and percent slope, biotic variable such as total woody basal area and litter composition, and (for the seventeen species) soil characteristics including pH and texture. Biotic variables were included for possible allelopathic influence and as indicators of environmental variables.
Univariate and multivariate analysis of variance on the three data sets showed significant differences for most variables included in the analysis. All variables with zero value for one or more species and highly correlated variables were eliminated before the discriminant analysis was performed. F-ratios for overall discrimination among groups was highly significant (p < 0.001) in all three data sets.
The first discriminant function was found to be most highly correlated with variables related to canopy density in both Galium data sets, while the second is most highly correlated with variables related to pH. The third discriminant function (available only from the cedar set) was most highly correlated with cedar litter and cedar basal area. Species ordering on the discriminant functions satisfactorily approximated known general habitat descriptions.
In the seventeen species data set, percent beech litter, slope position, pH and steepness of slope, and soil texture were respectively most highly correlated with the first, second, third and fourth discriminant functions. These axes interpreted as moisture axes, the first reflecting soil and air moisture, the second average moisture, the third drainage and the fourth water holding capacity.
Univariate comparisons of presence versus absence of seven of the seventeen species found to have the closest means of occurrence showed significant differences in mean pH values for all but one species. Discriminant analysis of presence versus absence data added little information to the univariate comparisons.
Combining the results of the univariate presence versus absence data with the discriminant analysis results in a comparison of fundamental and realized niche characteristics. The significance of pH values in determining species absence versus species presence implies the fundamental niches of these species include pH as a limiting factor. However, these species have quite similar mean pH values and results of the discriminant analysis implicate the importance of moisture axes in circumscribing realized niches.
Mann, Linda K., "Discriminant Analysis of Some East Tennessee Forest Herb Niches. " Master's Thesis, University of Tennessee, 1977.