The Uniqueness of Spatio-Temporal Data Unique and the Critical Role of GIS
Spatial data is characterized by a set of longitude and latitude coordinates (or x and y), and usually modeled from an object-based or location-based approach (Peuquet, 2002). These modeling approaches are not contradictory, but rather complimentary. For instance point data is used when mapping crime events and disease occurrence (McElroy et al., 2003; Chainey & Ratcliffe, 2005) and can easily be overlaid with raster data, while linear features are used for network modeling. An interesting question is whether the spatial distribution of these events is clustered or not, leading to the identification of hot spots. Increasingly however, spatial data has also been augmented with attributes and temporal coordinates, for instance the time stamp associated with an event.
Researchers (Peuquet, 2002; Andrienko & Andrienko, 2006) have proposed two approaches to incorporating time in spatial data. In an object-based approach, temporal extent is attached to each entity as an attribute, while in a continuous approach individual objects are considered as attributes and attached to a given location in space and time. In the object based approach for instance, GIS provide a unique platform that facilitates the linking of temporal and non-spatial attributes to geospatial locations by means of a unique identifier (ID). By means of structured query language (SQL), events occurring within a certain time interval can be extracted, and statistical techniques applied to test whether they exhibit space-time patterns.
Due to the unique nature of space-time data, it is thus straightforward to combine temporal and spatial queries. Consequently GIS is undergoing a new phase where two critical issues are in (1) the development and applications of techniques for the identification of clusters of spatial association in space and time -or in the attribute space, and (2) the development of space-time visualization techniques. These issues can be very challenging for large datasets. Our paper fits directly into the first concern, which is the development and application of space-time methods to identify clusters in space and time.