Spatiotemporal Data Models

Building Blocks

To explain spatiotemporal data models, it is useful to first briefly describe characteristics of independently formed spatial models and temporal models. Geographic information systems (GIS) depend on spatial data models, of which there are two basic types: raster and vector. These spatial data models use different spatial primitives to represent spatial properties. The raster model uses a collection of grid cells or pixels to represent locations, and each pixel is associated with an attribute value. In this model, the spatial properties (location, size, and shape) of a real-world entity cannot be directly represented, as they are dependent on the size and shape of the underlying pixels, as seen in Figure 1. Because of its representational structure, this model is better suited to representing the spatial variation in essentially continuous attributes, such as elevation or temperature. The vector model uses points, lines, and regions defined by coordinates to represent the location, size, and shape of entities. This model explicitly represents objects, their location, extent, and spatial relationships among them.

Temporal data model primitives include instants and intervals. They also model different characteristics of time, including linear, cyclic, and branching time. Temporal data models also separate time into transaction time, or the time at which a fact is recorded in a database, and valid time, the time when an event, action, or change occurred in the real world.

The earliest spatial data models in geographic information systems were strictly static, and there were both logical and pragmatic reasons for the exclusion of time. Most geographic features modeled in early GIS, such as mountains, lakes, and rivers, have persistent identities and location, and from this perspective they could be considered static. Early spatial data collection technologies, such as photogrammetry, were also too expensive to repeat frequently, so typically only single versions of geographic entities existed. In these models, time effectively was assumed to be constant. Users could not query the database about any past states of geographic features, as only one state was represented.

...