Date of Award
Master of Science (MS)
This study presents an analysis of the use of classification and ordination techniques for the detection of sub-communities at a reclaimed coal refuse dump and strip minded area in southeast Kansas. Cover values for 58 grass, forb, and woody species were estimated in 103 systematically positioned plots across the reclamation grassland using the modified Daubenmeyer cover scale. The data were submitted to an array of default and variant classification and ordination analysis programs in order to assess their ability to detect sub-communities. The classifications obtained from the analyses were mapped and analyzed for differences in species composition, diversity, and spatial contiguity. Of the two classifications produced by TWINSPAN, the variant analysis was better able to detect a sub-community characterized by high cover levels of Panicum virgatum, which appears to be exerting a negative impact on local diversity and evenness compared with area where the grass is present at lower cover levels. Both TWINSPAN analyses detected sub-communities characterized by a consociation of Bouteloua curtipendula, Andropogon gerardii, and Solidago candensis, with higher species diversity and evenness than sub-communities of a B. curtipendula and A. gerardii association; the variant TWINSPAN analysis was better able to detect the spatial parameters of this sub-community. The default Canonical Correspondence Analysis, a constrained ordination technique, indicated the effect of two environmental variables, frequency of standing water and level of mine refuse at the soil surface, account for only a fraction of variation in species composition within the reclamation area. Within the environmentally-focused ordination diagram produced by the default analysis, there occurred a separation in species composition between plots influenced by high levels of one or the other of the two variables. However, as the ordination diagram yielded no clearly-defined clusters of plots representing discrete, natural sub-communities, the classification derived from teh partitioning of ordination in which species planted during the reclamation were made passive. This may indicate that the systematic sowing of a homogeneous seed mix nine years earlier still confounds a more typical ecological separation of grassland species. However, this variant ordination, and two others, did not prove useful in creating more interpretable ordination diagrams.
Yates, Karen Frances, "The Evaluation of Two Types of Multivariate Analyses Applied to Grassland Vegetation Data from a Reclaimed Coal Mine Area in Southeast Kansas, USA" (1996). Electronic Theses & Dissertations. 264.