A comparison of two interaction matrix coding techniques used in a GIS-based tool for air quality assessment

The paper describes the development of a fast and easy-to-use qualitative tool for preliminary
assessments of urban air quality related to road traffic. The tool is particularly aimed at the
ability and budget of local government. It uses a novel interaction matrix-type methodology
combined with mapping overlay, performed via a GIS. More specifically, the interaction matrix
provides the weighting factors, which show the impact of each variable involved in a system
on the target variable, air quality, as well as on the system as a whole. These weighting
factors are used in the GIS to produce vulnerability maps. The maps visualise vulnerability to
air pollution due to the combined effect of a number of interacting factors, and thus indicate
areas susceptible to poor air quality. This results in a considerable reduction in computing
time and complexity compared to the use of a sophisticated numerical model, as the user of
the GIS tool only needs to perform mapping overlays in the GIS (using the previously derived
weighting factors).
The particular aim of this study was to compare two different methods for quantifying the
interactions between variables in the matrix. The first method used constant coefficients,
whose values are based on parametric studies performed using an advanced dispersion
model or on good engineering judgement. The second method used a more sophisticated and
versatile quantification of the interactions between variables, via analytical or semi-empirical
relationships. In the latter method, the matrix was formulated computationally, so that the
interaction weightings for different conditions can be obtained automatically.
The technique was applied to the case study of an urban area with a high traffic throughput, in
the UK. Two different interaction matrices were constructed for urban air quality linked to road
traffic, based on the above methods. The GIS results based on both matrix methodologies
were compared to the results of a more intensive dispersion numerical model in terms of
pollutant dispersion patterns and hot spots. Both sets of results were shown to compare
favourably with those of the numerical model. The results based on the more sophisticated
matrix coding were found to be in closer agreement with those of the numerical model.