Mapping traffic atmospheric emissions for epidemiological studies using atmospheric dispersion models and geostatistical methods: a case study

In some cases, epidemiological studies require the air pollutant concentrations at the exposure points. In these cases air dispersion models represent a very important tool. When additional points of exposure are inserted or when some exposure points must be relocated, spatial interpolators can be used in place of new runs of the air dispersion model. In this work the uncertainties and the problematic related to spatial interpolation methods are inspected. The case studied is based on an epidemiological study aimed to study the risk of childhood leukemia associated with benzene exposure due to traffic emissions. The concentration values of benzene computed by the atmospheric dispersion model ADMS are taken as reference and compared with the concentration values computed using several interpolation methods and additional data sets of concentrations computed by ADMS in the same area. The comparison is done following two approaches: the summary statistics of the differences and the correctness of the assignment of the exposure points to the concentration categories used in the epidemiological study. These comparisons show that the values computed by the interpolators are very problematic: important differences and categories assignment and categories uncertainties were found. The main conclusion of this work is that the use of interpolators must be done with extreme caution. Moreover, it is highlighted the importance and the potential pitfalls of exposure modelling methodologies when assessing the health effects of environmental pollutants