Date of Graduation

2001

Document Type

Thesis

Degree Type

MA

Committee Chair

Robert Hanham

Abstract

The goal of this thesis research was to show that spatial data do affect the results of the analysis of vegetation-environment relationships. Vegetation-environment relationships were modeled within the Allegheny Mountain section of the Appalachian Plateaus physiographic province in western Maryland. Specifically, this research focused on two state forests, namely Garrett and Potomac State Forests, in Garrett County, Maryland. Four regression models were utilized in this research to model the basal area and diversity of the fifteen most prominent species relative to environmental factors within the two forests combined. A standard Ordinary Least Squares regression model with diagnostics for spatial effects was initially employed to model the basal area and diversity of the fifteen most prominent species. Secondly, a spatial lag model was utilized to account for autocorrelation among the dependent variables. Third, a spatial error model was employed to account for autocorrelation among the error terms. Finally, a heteroskedastic spatial regimes model was used to account for regional differences between the two state forests. Spatial lag and spatial error problems were found to be present and significant in six of the fifteen tree species. The heteroskedastic spatial regimes model showed spatial heterogeneity was a significant problem in thirteen of the fifteen species and the diversity index. The research results show that spatial heterogeneity appears to be the largest problem overall, followed by spatial lag, then spatial error. This research shows that spatial effects are present in relationships of this sort, and that they must be controlled for if the researcher is committed to an accurate assessment of the relationship between variation in vegetation and environment.

Download

Share

COinS