Development of an enhanced adaptive resonance theory mapping system for watershed classification

Li, Pu (2009) Development of an enhanced adaptive resonance theory mapping system for watershed classification. Masters thesis, Memorial University of Newfoundland.

[img] [English] PDF - Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission.

Download (7MB)


Watershed classification is a process that classifies watershed sub-basins into certain groups due to similarities and/or differences in their characteristics. Such a process is of necessity and importance to support the decision making and practice of watershed monitoring, modeling, and management and helps in reducing the set up and running cost and improving efficiency. A watershed system is usually characterized by a large variety of topographical, hydrological, and ecological features, which provides the basis for watershed classification and also makes it a challenging task. Furthermore, many of the features and their interrelationships are hardly measured or quantified accurately due to the complexity and uncertainty of the system. Numerous studies have been conducted on watershed classification but the comprehensive consideration of both systematic complexity and uncertainty in the classification process is lacking. There is a need of more efficient and reliable approaches of watershed classification to deal with complex and uncertain features. -- This research aims to fill the gap by developing a novel classification system based on the enhanced adaptive resonance theory (ART) mapping approaches to classify complex watershed features under uncertainty for supporting watershed modeling and management. The developed system is composed of: (1) a two-stage adaptive resonance theory mapping (TSAM) approach by integrating multitier ART into the system to form an unsupervised learning module for cluster centroid calculation and a supervised learning module for normalized original input classification; and (2) an integrated rule-based fuzzy adaptive resonance theory mapping (IRFAM) approach by incorporating fuzzy set theory and rule-based operation to the system to form an unsupervised learning module for cluster centroid calculation and two supervised learning modules for criteria combination and fuzzified input classification. -- To test the feasibility and efficiency, the developed system was applied to a real-world case study in the Deer River watershed, Canada. The results indicated that the watershed sub-basins were properly classified into preset target groups by both approaches in the given conditions (e.g., vigilance = 0.7). The TSAM approach could efficiently solve the problem of difficulties in criteria generation by using ART unsupervised classification and centriod determination in the first stage and feed the criteria to the ARTMap supervised classification in the second stage. In comparison with the TSAM, the IRFAM approach could take advantages of fuzzy set theory to generate full criteria combinations to match the input patterns and use the rule-based operation to screen the matched patterns into the target groups. This can efficiently handle the classification for the input patterns with a high degree of uncertainty and wide ranges of variations. In the case that there are not sufficient information for generating fuzzy membership functions, the TSAM could be a better choice than the IRFAM from a feasibility perspective; otherwise, the IRFAM could provide more accurate classification results than the TSAM.

Item Type: Thesis (Masters)
Item ID: 8756
Additional Information: Includes bibliographical references (leaves 127-138)
Department(s): Engineering and Applied Science, Faculty of
Date: 2009
Date Type: Submission
Library of Congress Subject Heading: Watershed management--Mathematical models; Watersheds--Classification

Actions (login required)

View Item View Item


Downloads per month over the past year

View more statistics