Modeling Approaches in GIS and Environmental Modeling

The three broad approaches represent how the existence or source of any hazard is modeled as risk in relation to vulnerable receptors. This implies that both the source and the receptors have to be identified and that there is some model of how that source impacts the receptors. A source may be identified as a specific object (a smoke-stack, effluent outfall), a zone (unstable slope from which specific landslides may occur, an area of seepage, a fault line), or diffuse over the whole area (strong wind event, heavy rainstorm).

Generally, the first of these would be classified as point sources and the latter two as nonpoint sources. Difficulties arise over such an inductive classification because a nonpoint source in one area may originate from a point source in another. Thus, general air pollution in one area may originate from specific factories elsewhere. Depending on the scale of study, it may not be possible to model all the individual point sources, but treat the aggregate effect as a nonpoint source. Point and nonpoint sources need to be treated differently in GIS (point, line, polygon, or field) and are likely to influence the form of analysis. Receptors can be people, the flora and fauna, properties and land, again with different ways of representing these in GIS with consequences for the modeling. It is, however, the modeling of the means by which the source has the ability to impact the receptors that is distinctive within the taxonomy:

1. Spatial coexistence: This approach assumes that there is a reasonably simple or obvious spatial link between sources and receptors. Thus, for example, by taking the floodplain and overlaying it on settlements one might infer that anywhere where polygon “floodplain” and polygon “settlement” spatially coincided, people were vulnerable to a flood hazard. This approach relies heavily on conceptual and empirical models.
2. Source/pathway characterization: Not all spatial relationships between source and receptor can be conceived of as simply as is the case in the previous category. Where there is a significant and well-defined transport process linking source and receptor, then that transport process is explicitly modeled. Thus, this approach relies heavily on process models (deterministic or stochastic) that are either lumped parameter or distributed parameter.
3. Cluster detection: Whereas the first two approaches in the taxonomy are ex ante, that is, are carried out to forward predict some event that can then be mitigated against in some way, this approach is ex post, that is, to detect that some event has or is happening to the receptors, to identify a source and then ensure that the hazard has ceased or needs to be mitigated. This is also empirical modeling using GIS, statistical techniques, and artificial neural networks (ANN) to search for space–time clusters in the event occurrences among the receptors that indicate that something has gone wrong.

Before coming to the case studies, we need to explore the difference between environmental modeling within GIS and environmental modeling with GIS. In the former, the entire environmental model is realized within GIS. In the latter case, GIS are linked with external models. Some would feel the distinction is unimportant; after all, it is possible to carry out the complete taxonomy of approaches within GIS. Well, yes and no. Both spatial coexistence modeling and cluster detection are heavily weighted in their approaches toward the spatial dimension of phenomena and are eminently suitable to be carried out in GIS or GIS-like modules. One problem here is that most commercial GIS software are poorly endowed, if at all, with cluster detection methods and that these need to be programmed either as externally linked programs or using the GIS internal macro language capability.

Source–pathway characterization is somewhat different. While fairly simple transport processes, for the most part reliant on routing over topography, can be programmed as internal macros, the implementation of efficient finite element method (FEM), for example, is beyond these macro tools despite their growing sophistication. For many source–pathway characterization approaches, it still remains far more satisfactory to use specific environmental simulation software with GIS as an important complementary tool.

This combined approach has grown in popularity due to the “dual recognition of environmental problems with compelling spatial properties, but also with a complexity that cannot be adequately explored through interrogation and recombination of geographic data alone” (Clarke et al., 2000). But there is another important reason. There are many instances, for example, in engineering orientated applications where a client mandates the use of certain classes of model for which formal validation is available. In some places, such models are used as de facto standards and one’s own macro implemented in GIS isn’t going to cut much ice. They may even be quite specific about the nature of GIS software being used … and the database. This is a form of quality assurance where specification of tools of known quality means that the tools are not going to be the weakest link and are capable of supporting good quality work. Under these circumstances, tool coupling has to take place.

Share this post