| This research explores the use of fuzzy classification to create cartographic representations of vegetation regions with graded boundaries, or graded area-class maps. The research is divided into four major sections.; First, the relationship between fuzzy classification and conceptual models of vegetation is examined by comparing fuzzy classification with traditional classification and ordination techniques used in ecology. Fuzzy classification is shown to implement a conceptual model of vegetation consisting of core regions and transition zones. As such, it imposes a minimal structure on observations, affording the flexibility to capture complex geographic patterns while still providing anchors for spatial cognition.; Next, a data reduction framework is used to develop numerical techniques for fuzzy classification which seek to maximize the information content of a classified representation. A fuzzy classification technique based on principal components analysis is shown to maximize information content under certain general conditions. The technique is demonstrated and compared to fuzzy k-means clustering, the most widely used technique for fuzzy classification of geographic data.; Third, the data reduction framework is expanded to include the criteria of cognitive simplicity in map interpretation. This leads to a Pareto-optimization problem with two competing goals, which is demonstrated by constructing a series of ecoregion maps of equal information content based on tree genus distribution data for New York, New Jersey and Pennsylvania. The series ranges from a 4-class graded area-class map to a 13-class crisp map. By holding information content constant, the goal of cognitive simplicity is isolated for further analysis.; Last, an experiment is designed to compare the relative communicative efficacy of crisp and graded area-class maps. It is hypothesized that crisp area-class maps will facilitate tasks involving lookup of values at specific locations, while graded area-class maps will facilitate global pattern recognition and memory tasks. Results of task performance by 42 subjects are consistent with the hypotheses but inconclusive.; In sum, the research presents a theoretical framework that can be used to guide cartographic representation of gradation in ecological landscapes. The research suggests that graded area-class maps facilitate ecological analysis, and are an effective means of communicating ecological information. |
