Masters Theses

Date of Award

12-1990

Degree Type

Thesis

Degree Name

Master of Science

Major

Forestry

Major Professor

Edward R. Buckner

Committee Members

Glenn Smalley, John Rennie, Esteben Walker

Abstract

A land classification system has been developed for the Interior Uplands of Tennessee, Kentucky, Alabama, Georgia and Virginia (Smalley 1984b). The system utilizes information from geology, soils, and physiography to divide the landscape into units with homogeneous forest site potential. No detailed vegetation studies were made during the development of the system, raising questions as to its applicability for predicting forest cover within each landtype. The scope of this study was to 1) determine if distinct forest cover occurs within and between landtypes, 2) identify and describe the forest cover types occurring within landtypes and 3) investigate underlying factors controlling spatial arrangement of forest cover on the landscape. If distinct forest spatial patterns can be correlated to landtype descriptions, Smalley's system would be a valuable tool to efficiently characterize and map forest communities.

This study was conducted in two compartments of Prentice Cooper State Forest located on the southern tip of Walden Ridge on the Cumberland Plateau west of Chattanooga, TN. Rectangular, 0.10-acre plots were used to describe sub-plots for sampling vegetation from the overstory, understory, sapling/shrub and seedling/herbaceous forest strata as well as measure selected physical site characteristics. Sample plots were allocated by landtype using a random start with subsequent systematic location. Final statistical analyses used the information from 139 plots situated on four major landtypes.

Variable distribution and variance structure was evaluated for each data set to determine appropriate multivariate techniques. Parametric discriminant analysis was used to investigate the distinctness of forest cover within landtypes. Farthest neighbor cluster analysis was used to document forest cover within landtypes. Principal factor analysis was used to investigate underlaying factors controlling forest community variation.

Discriminant analysis revealed sampled landtypes had relatively distinct forest cover. The overstory misclassification matrix had the highest percent of correctly classified sampling plots of all single forest stratum data sets. As the information from successively subordinate forest strata was included in the analysis, the percentage of correctly classified plots approached 100.

Farthest neighbor cluster analysis revealed common communities occurring within different landtypes: chestnut oak (Quercus prinus L.), white oak (Quercus alba L.) and shortleaf pine (Pinus echinata Miller) communities occurred on all four sampled landtypes. A scarlet oak (Quercus coccinia Muenchh.) community occurred on landtypes 5, 6, and 17. A black oak (Quercus velutina Lam.) community occurred on landtypes 3 and 5. Landtype 17 contained three communities not found on the other landtypes; yellow-poplar (Liriodendron tulipifera L. ), northern red oak (Quercus rubra L.)/ and eastern hemlock (Tsuga canadensis (L.) Carr.)

Principal factor analysis revealed that an ecosystem element or element complex varying along an elevational gradient was the dominant factor in controlling forest community variation. Microenvironmental ecosystem elements appear to be secondary controllers of forest community variation. Site variables characterized physical aspects of the landscape. At best, these variables were indirect measures of ecosystem elements responsible for forest community variation, resulting in speculation as to the actual factors contributing to the majority of forest community variation.

This study examined the utility of a landtype classification system for efficiently characterizing forest cover on the landscape. The statistical techniques involved proved adequate to describe the forest resources within specific landtypes. Parametric discriminant analysis revealed forest cover within landtypes was relatively distinct. Contradictory to these results, farthest neighbor cluster analysis revealed common communities occurring within different landtypes. The discrepancy between the two techniques arises from the function of discriminant and cluster analysis. Discriminant analysis pools vegetation information within landtypes to develope discriminating criteria while cluster analysis agglomerates vegetation information within landtypes into similar clusters. The differences between similar communities in conjunction with differences attributable to communities not common to all landtypes was significant enough to reveal relatively distinct forest cover within the sampled landtypes. Species Relative Importance Values of overstory species might be inadequate to determine if similar communities from different landtypes are distinct. Future research should focus on developing an appropriate variable(s) that incorporates bole height or quality to aid in distinguishing between similar communities that occur on different landtypes.

Download
Files over 3MB may be slow to open. For best results, right-click and select "save as..."

Share

COinS