Knowledge-Based Visualization Design for Geographic Information Systems

This paper presents taxonomies for visualization design in Geographic Information Systems (GIS) and a methodology to formalize and implement GIS visualization knowledge. Relevant taxonomies include characterizations of spatially-referenced attribute data and of the user's visualization objectives. The visualization design methodology decomposes complex data sets into simple relations that are encodable by primitive visualization methods. Partial visualizations are composed using a set of specific composition operations until all complete visualizations of the data set have been found. The complete visualizations are evaluated and put in the correct order of effectiveness. The methodology has been implemented in the knowledge-based system Vizard. This system designs two-dimensional GIS visualizations and offers users the possibility to select and modify visualizations in a graphical user interface. Vizard is a working system that is useful for exploratory data visualization in GIS. It h elps novice and casual users of GIS to produce consistent and informative visual representations of data sets and to avoid ineffective or misleading visualizations.