A State-Space Model for Understanding Spatial Dynamics Represented by Areal Data

Most geographical phenomena, if not all of them, are essentially dynamic processes instead of static forms. The continuing advances of data acquisition and dissemination technologies have made available massive amounts of spatiotemporal data for the research of such dynamic processes. In the past decade, a surge of statistical models have been proposed in GIScience and related fields for modeling geo-referenced spatiotemporal data, with point support in particular, by accounting for complex spatiotemporal interactions. Much less attention, however, has been given for the spatiotemporal data collected and aggregated over geographical units (areal data), which are all-important and common data sources in many scientific disciplines. Many existing methods are developed particularly for spatiotemporal disease mapping (e.g., Schmid & Held 2004), and they tend to assume that areal units share a common and well-behaved temporal trend. These assumptions hinder the applications of such methods in dynamics with complex and heterogeneous spatiotemporal interactions, which (unfortunately) are typical in GIScience.