Modeling and managing uncertainty in object-based geospatial information systems.

摘要

This thesis investigates uncertainty in object-based Geospatial Information Systems (GIS). This issue is of vital importance to users since uncertainty information aids users in determining the fitness of the data to their particular application.; The theories that are used to deal with uncertainty—information theory, probability theory, evidential theory, fuzzy sets theory and geostatistics—are briefly reviewed. Each of these theories can be used to analyze a specific aspect of uncertainty. A general theory to handle all aspects of geospatial uncertainty has yet to be developed.; The techniques for modeling uncertainty in GIS are classified into four categories: analytical, simulation, experimental and error descriptors. Analysis of the categories has shown that each method has associated advantages and disadvantages and that the method selection is directly related to functional approximation errors, computer efficiency, and economic constraints.; Current models of geospatial uncertainty are analyzed. It is found that they are oversimplified. Therefore, more complete uncertainty models that describe the positional variations of spatial objects are proposed and evaluated.; A data reduction strategy for representing curvilinear boundaries is proposed. This recursive spline approximation has shown that a 70% data reduction can be achieved in boundary delineation for the cases studied in this thesis. However, data reduction depends on the assumed approximation error.; Uncertainty management problems are viewed as closely related elements. The components of the framework are uncertainty identification, uncertainty modeling, uncertainty communication, uncertainty reduction, and uncertainty absorption. A framework is proposed to deal systematically with uncertainty management issues in object-based GISs.; Finally, a software package for uncertainty modeling and visualization has been developed which facilitates the implementation of the proposed uncertainty management techniques.; Further research may be directed towards studying the uncertainty implications due to scale changes, integrating temporal and topological uncertainty in the management strategies, searching for the most effective methods for uncertainty communication, and analyzing the impact of spatial database uncertainty upon decision making and risk analysis.