A geographical approach for integrating belief networks and Geographic Information Sciences to probabilistically predict river depth

Nathan Lee Hopper, University of Southern Mississippi


Geography is, traditionally, a discipline dedicated to answering complex spatial questions. Although spatial statistical techniques, such as weighted regressions and weighted overlay analyses, are commonplace within geographical sciences, probabilistic reasoning, and uncertainty analyses are not typical. For example, belief networks are statistically robust and computationally powerful, but are not strongly integrated into geographic information systems. This is one of the reasons that belief networks have not been more widely utilized within the environmental sciences community. Geography's traditional method of delivering information through maps provides a mechanism for conveying probabilities and uncertainties to decision makers in a clear, concise manner. This study will couple probabilistic methods with Geographic Information Sciences (GISc), resulting in a practical decision system framework. While the methods for building the decision system in this study are focused on the identification of environmental navigation hazards, the decision system framework concept is not bound by this study and can be applied to other complex environmental questions.