WATERSHED SIMILARITY ANALYSIS USING INTEGRATION OF GIS AND UNSUPERVISED-SUPERVISED ARTIFICIAL NEURAL NETWORKS

摘要

An integration of Geographic Information Systems (GIS) and Unsupervised-Supervised Artificial Neural Networks (ANNs) was applied to quantify the similarity of watershed characteristics, including geometric, meteorologic, and hydrologic parameters. While the Self-Organizing Features (SOFMs), a type of unsupervised ANNs, was used to perform the clustering process, a supervised ANNs such as Multilayer Perceptron, was used to examine the classification accuracy. The purpose of this study was to determine the best fitness "home" for an additional watershed from an existing knowledge base. The result can provide further "transplant" or "synthesis" of hydrological information between watershed systems, such as the streamflow hydrograph and meteorological forcing. An example using a knowledge base with 193 watersheds, which were described by 15 geometric characteristics and 3 hydrological parameters, was demonstrated. The classification and clustering sensitivity due to the number of involved parameters and clustering dimension are discussed.