Methods for ad-hoc delineation and analysis of categories of spatio-temporal events

Analysts are faced with increasing volume and complexity of spatially and spatio-temporally referenced events to analyze. One means of taming this volume and complexity is to develop methods and tools that can identify patterns, including spatio-temporal structure like clusters, in event data. To understand these methods and tools, we first present some of the motivation for the work, and then we detail the software architecture that we will use to support categorical analysis of spatio-temporal events.

Methods for analyzing spatial events, and spatio-temporal events, have experienced a recent renaissance. This upsurge in interest has occurred in part because of novel high-quality event sources which can provide complex data with geographic and temporal referents. Examples of such event sources include geographically located Twitter postings linked to documents, or photographs and video taken with GPS-enabled mobile phones. The products of these event sources are a vast stream of events that are linked with heterogeneous and voluminous data, including textual, imagery data as well as numerical data. One of the ways of making numerical sense of such heterogeneous data is to consider the text or media as a set of tagged categories. We can then apply methods for detecting structure in spatio-temporal events, including recently developed methods for disease outbreak detection, to these categories.

We are developing methods and associated software that will allow users to tag or label events, analyze them, and interact with visual representations of the event structure detected by the analysis. An integrated software system, called STempo, will provide the user with a fixed set of analysis tools and coordination topology to work from. The category tagging and structure detection tools will also be worked into the larger set of tools available in the GeoViz Toolkit, an interactive system for geographic visualization and analysis.