Mapping of Biological Communities in Northwestern Montana

Publication Date


Type of Culminating Activity


Degree Title

Master of Science in Biology



Major Advisor

Stephen J. Novak


The study and management of biological communities depends on systems of classification and mapping for the organization and utilization of resource information. This project examined the feasibility of mid-scale mapping of floristic units, such as alliances and associations, in northwestern Montana using advanced remote sensing technology and modeling techniques. The primary goal of this study was to identify a methodology for modeling forest alliances and associations of the region.

In this study, satellite imagery was used to develop a polygon configuration through segmentation and merging of spectral data, resulting in a base polygon coverage of object primitives from which all map feature space was classified in a step-wise modeling process. By applying multiple classifiers, image objects were classified hierarchically using a decision tree method, first by physiognomic categories and then by the floristic units themselves - alliances and associations. Reference data were synthesized from existing inventory plot records through a series of screens necessary to identify suitable plot data. All suitable plots were labeled by alliance and association to form the base reference data set. Fifteen percent of reference data were reserved for the validation of map products, and the remainder of the plots were designated as training data for map modeling. Key spectral and biophysical differentia were identified for the image classification by physiognomic and floristic classes. While fuzzy rule sets were generated for partitioning physiognomic classes, a nearest neighbor function was used to separate alliances and associations, where the contrasts between classes are less distinct.

Test maps were produced for nine forest alliances and 24 associations across the upper Kootenai River subbasin. Error matrices were constructed for the alliance and association maps based on stratified random selection of map validation samples for each map. Accuracy estimates varied between 60% and 61% for alliance and association maps, respectively. The map products provide an acceptable level of accuracy for landscape-level analyses at the geographic scales of 1:24,000 and larger. Alternative techniques, including extending the depth of the image classification decision tree, are identified to augment the modeling of alliances and associations for forest types while also discussing limitations of the reference data. A qualitative summary of map products is provided along with recommended uses and caveats.

Files over 30MB may be slow to open. For best results, right-click and select "save as..."