Abstract
When studying geographical phenomena, different levels of spatial and temporal granularity often have to be considered. While various approaches have been proposed to analyse geographical data in a multi-scale perspective, they have all focused on either spatial or temporal attributes rather than on the integration of space and time over multiple scales. This study introduces the continuous spatio-temporal model (CSTM), a conceptual model that seeks to address this shortcoming. The presented model is based on (1) the continuous temporal model (CTM), a multi-scale model for temporal information, and (2) the continuous spatial model (CSM), an extension of CTM for multi-scale spatial raster data. At the core of the presented conceptual model is a spatio-temporal evolution element or, in short, stevel, which is described by four variables: (1) pixel location, (2) spatial resolution, (3) temporal interval, and (4) temporal resolution. By varying one or more of these variables, a CSTM-tree consisting of (sets of) stevel arrays is created, forming the basis of an exhaustive CSTM-typology. These arrays can then be used to systematically cluster spatio-temporal information. The value of our approach is illustrated by means of a simplified example of mean temperature evolution. Various suggestions are made for modifications to be developed in future research.
Funding
The authors gratefully acknowledge the financial support by the Research Foundation Flanders (FWO) and by the University Research Foundation (BOF-UGent).
Notes
1. In the tradition of the picture element or ‘pixel’ for 2D raster data and the volume element or ‘voxel’ for 3D raster data, here the atomic element is referred to as a ‘stevel’, or spatio-temporal evolution element.
2. This model was originally called the continuous triangular model, but for reasons of consistency with the newly presented models in this study, ‘triangular’ has been replaced by ‘temporal’.
3. In contrast to CTM and stevel, for the interpolated version, and are used.