Assessing similarity of dynamic geographic phenomena in spatiotemporal databases.

Abstract

The growing availability of routine observations from satellite imagery and other remote sensors holds great promise for improved understanding of processes that act in the landscape. However, geographers' ability to effectively use such spatiotemporal data is challenged by large data volume and limitations of conventional data models in geographic information systems (GIS), which provide limited support for querying and exploration of spatiotemporal data other than simple comparisons of temporally referenced snapshots. Current GIS representations allow measurement of change but do not address coherent patterns of change that reflects the working of geographic events and processes. This dissertation presents a representational and query framework to overcome the limitations and enable assessing similarity of dynamic phenomena. The research includes three self contained but related studies: (1) development of a representational framework that incorporates spatiotemporal properties of geographic phenomena, (2) development of a framework to characterize events and processes that can be inferred from GIS databases, and (3) development of a method to assess similarity of events and processes based on the temporal sequences of spatiotemporal properties. Collectively the studies contribute to scientific understanding of spatiotemporal components of geographic processes and technological advances in representation and analysis.