Predictive modeling of deciduous forest stand composition and distribution using GIS and ecological land classification : a regional-scale test in the Clinch-Powell watershed of Tennessee

Climatically-induced changes in the hydrologic cycle will likely affect the distribution and composition of forest stands in the Clinch-Powell watershed of Tennessee. To assess the potential impact and magnitude of these changes and their corresponding effect on water resources requires baseline information on the current location, extent, and composition of forest stands. Conventional field-based sampling methods are too expensive and time-consuming to be of use. Satellite image data have advantages over other data sources for large-area land cover classifications, in part because of their frequent repeat cycles, large-area samples, wide spectral range, and amenability to automated classification. However, Landsat Multispecteral Scanner data (MSS) and Thematic Mapper (TM) data, because of their coarse spatial resolution and limited spectral resolution, require the use of ancillary data to improve the accuracy of the classification. In this study I tested the feasibility of combining coarse-resolution remotely-sensed TM data with data on soils and topography within the structure of a hierarchical rule-based ecological land classification system to predict forest composition in the Clinch-Powell watershed of east Tennessee. The model predictions were compared with forest inventory data to judge the accuracy of the model and the general applicability of this method. Results show that given current databases and knowledge of soil/site relationships to trees, ecological land classification is not viable at a regional scale.