Moving Beyond Snapshots

Marceau et al set out to develop temporal database and analysis functionality in a GIS environment.  They are motivated to do this by the inadequacies of current GIS software to deal with temporal elements, which results in a de facto spatiotemporal analysis approach known as the “snapshot model”: limited to knowledge about specific moments in time with no formal inferences about what happens in between or the process behind any observed changes.

The authors point out two major shortcomings with the snapshot model, the first being a lack of temporal topology.  In temporal analysis, the most important topological relationship is order, or the sequence of events along a timeline, which establishes whether an event happens before, concurrent with, or after another event. Other topological relationships exist as entities appear, disappear and change through time.  Evolving suburban land use is a good example of this, as residential entities may appear in agricultural land, growing and merging into each other until remaining pockets of agricultural use split and eventually disappear.

The other important shortcoming with snapshots is the choice of sampling interval, or temporal resolution, which can have important effects on what events are noted by researchers and thus the conclusions drawn from spatiotemporal analysis.  While even the input data to the article’s case study is basically in temporally coarse “snapshot” form, the authors attempt to address the issue of sampling interval by adding attributes to spatiotemporal land use entities such as “beginning min”, “end max” and “duration max”.  These values incorporate a degree of uncertainty into the model to more transparently deal with the limitations associated with temporal resolution, but I was disappointed to note that there is little here in the way of concrete solutions to the problems of temporal interpolation.

Marceau et al’s model provides a basic framework to harness temporal considerations within the existing vector GIS paradigm.  A key question is whether this approach will be robust enough to be extended into more types of spatiotemporal analysis (such as multi-linear or branching time), or whether we will yet require an entirely new GIS foundation to achieve this.


Comments are closed.