Calibrating spatial interaction models from GPS tracking data: An example of retail behaviour

Global Positioning System (GPS) technology has changed the world. We now depend on it for navigating vehicles, for route finding and we use it in our everyday lives to extract information about our locations and to track our movements. The latter use offers a potential alternative to more traditional sources of movement data through the construction of trip trajectories and, ultimately, the construction of origin-destination flow matrices. The advantage of being able to use GPS-derived movement data is that such data are potentially much richer than traditional sources of movement data both temporally and spatially. GPS-derived movement data potentially allow the calibration of spatial interaction models specific to very short time intervals, such as daily or even hourly, and for user-specified origins and destinations. Ultimately, it should be possible to calibrate continuously updated models in near real-time. However, the processing of GPS data into trajectories and then origin-destination flow matrices is not straightforward and is not well understood. This paper describes the process of transferring GPS tracking data into matrices that can be used to calibrate spatial interaction models. An example is given using retail behaviour in two towns in Scotland with an origin-constrained spatial interaction model calibrated for each day of the week and under different weather conditions (normal, rainy, windy). Although the study is small in terms of individuals and spatial context, it serves to demonstrate a future for spatial interaction modelling free from the tyranny of temporally static and spatially predefined data sets.