GEOAGENT-BASED KNOWLEDGE SYSTEMS

Author:

Yu, Chaoqing

Graduate Program:

Geography

Degree:

Doctor of Philosophy

Document Type:

Dissertation

Date of Defense:

November 11, 2005

Committee Members:

Donna Jean Peuquet, Committee Chair/Co-Chair
Alan Maceachren, Committee Member
Brenton Yarnal, Committee Member
John Yen, Committee Member

Keywords:

geographic process
representation
GeoAgents
GIS
knowledge

Abstract:

Modern geography focuses on studying processes. In addition to observed phenomena, the study of geographic processes must (and does) place emphasis on understanding how components interact within geographic systems. As a fundamental tool for geographic representation and spatial analysis, current GISystems (geographic information systems) are nevertheless still data centered. While they are good at representing “what” and “where” information, they have limited capabilities in representing higher-level knowledge. This is because in the current GISystems there is a lack of means of capturing and representing human understanding of geographic processes to address “how” and “why” questions. In addition, non-observational factors such as laws, policies, regulations, plans, and cultural elements (e.g. religions, customs) cannot be easily represented. Instead of the traditional data-centered approach, this dissertation presents a knowledge-oriented strategy for the representation of geographic processes. To reach that end, two major steps are adopted: (1) introducing the concept of GeoAgents as the spatiotemporally distributed knowledge-representation components, and (2) presenting an integrated approach to incorporate multiple knowledge-representation techniques with geospatial databases. GeoAgents are defined in this dissertation as spatial, dynamic, and scale-dependent agents within an explicitly geographic context. By incorporating GeoAgents with graph-based concept maps, rule-based expert systems, quantitative models, and geospatial databases, this research develops a Java-based prototype — GeoAgent-based Knowledge System (GeoAgentKS) — that allows the representations of diverse kinds of geographic knowledge and spatial data to be integrated in a single cohesive software system. To examine the knowledge-oriented strategy of geographic representation in real-world problems, GeoAgentKS are employed in a case study to represent the complex geographic processes relevant to community water systems (CWSs) in Central Pennsylvania. In this case study, geographic knowledge is captured via interpretation of the pre-existing documents and computer-based-concept-mapping interviews with domain experts. To evaluate the usability of GeoAgentKS, evaluation interviews with different experts and novices were conducted to assess the adequacy of the knowledge representation and the effectiveness in conveying knowledge. The experts in the evaluation interviews believed that it was possible to use the GeoAgentKS to represent the complex, dynamic and scale-dependent human-environment interactions. And the knowledge stored in the GeoAgentKS could be quickly learned by novices.