Date of Award


Level of Access

Open-Access Dissertation

Degree Name

Doctor of Philosophy (PhD)


Spatial Information Science and Engineering


M. Kate Beard-Tisdale

Second Committee Member

Peggy Agouris

Third Committee Member

Max J. Egenhofer


Geographic information systems (GISs) allow users to analyze geographic phenomena within areas of interest that lead to an understanding of their relationships and thus provide a helpful tool in decision-making. Neglecting the inherent uncertainties in spatial representations may result in undesired misinterpretations. There are several sources of uncertainty contributing to the quality of spatial data within a GIS: imperfections (e.g., inaccuracy and imprecision) and effects of discretization. An example for discretization in the thematic domain is the chosen number of classes to represent a spatial phenomenon (e.g., air temperature). In order to improve the utility of a GIS an inclusion of a formal data quality model is essential. A data quality model stores, specifies, and handles the necessary data required to provide uncertainty information for GIS applications. This dissertation develops a data quality model that associates sources of uncertainty with units of information (e.g., measurement and coverage) in a GIS. The data quality model provides a basis to construct metrics dealing with different sources of uncertainty and to support tools for propagation and cross-propagation. Two specific metrics are developed that focus on two sources of uncertainty: inaccuracy and discretization. The first metric identifies a minimal?resolvable object size within a sampled field of a continuous variable. This metric, called detectability, is calculated as a spatially varying variable. The second metric, called reliability, investigates the effects of discretization on reliability. This metric estimates the variation of an underlying random variable and determines the reliability of a representation. It is also calculated as a spatially varying variable. Subsequently, this metric is used to assess the relationship between the influence of the number of sample points versus the influence of the degree of variation on the reliability of a representation. The results of this investigation show that the variation influences the reliability of a representation more than the number of sample points.