Wildland classification with multivariate analyses and remote sensing techniques

Abstract

Wildland classification is a prerequisite to many tasks in the land management planning process. Although a variety of general wildland classification frameworks have been proposed for the United States, much work remains to develop detailed-level classification. This study was conducted in three parts to examine three related aspects of wildland classification at the detailed level. The first part of the study was designed to examine the utility of multivariate analysis techniques in developing an ecologic land classification for a specific area. The second part of the study was designed to assess the similarity between two independently developed classifications for the same area. The third part of the study was designed to evaluate the use of remote sensing data for identifying detailed land classification units on the ground. Vegetation dominance data were collected in 102 sample stands in the Manitou Experimental Forest, Colorado. The data were analyzed by an iterative sequence of cluster analysis, canonical ordination, discriminant analysis, and subjective interpretation to identify important groupings of the sample stands. Digitized aerial photographic data (1:50,000 scale color, infrared) from the sample stands were analyzed to evaluate the ability to identify stands representing the habitat type classes from part one of the study. Adding three physical site variables - elevation, slope, and aspect - to the analysis increased identification accuracy to 97 percent.

Citation

Radloff, David L. 1983. Wildland classification with multivariate analyses and remote sensing techniques. Fort Collins, CO: Colorado State University. 106 p. Dissertation.