The stochastic simulation approach: Tools for representing spatial application uncertainty

This dissertation's goal is to establish some practical methodology allowing decision makers to quantitatively describe spatial application uncertainty. Most geographic applications provide a decision maker with single answer, with no supporting information about the reliability of the answer. Describing spatial application uncertainty allows a decision maker to know the range of potential application results. Knowing the range of application results provides a decision maker with an understanding of how much faith to put into the application results using a measure known to the decision maker: the variation of the results itself. This dissertation focuses on how the uncertainty of spatial data affects the uncertainty of spatial applications by applying the following approach:; Source maps should include methods for random field generation included in the metadata. Random fields, which provide both a representation of a map's uncertainty magnitude and spatial dependence, perturb probabilistic definitions of maps creating potential map realizations. Potential map realizations provide inputs to geographic information system applications (or any spatial heuristic) by repeatedly running Monte Carlo applications creating a series of potential results. Finally, the data and application's validity is known by analyzing the statistical representation of the potential results.; Although computationally expensive, this process is the only practical way of including all forms of analytical heuristics available to map users without explicitly declaring errors for each of the analytical heuristics.; This dissertation demonstrates a complete working model of this stochastic simulation process from describing source map uncertainty to analyzing application uncertainty. Due to the complexity of the problem, this research is still in an exploratory phase, with many parts of the model needing adjustment or better understanding. However, this dissertation provides a modular working spatial data uncertainty model to the research community, allowing incremental improvements to be easily incorporated.