Air pollution modelling in the Ruhr Area: Land-Use Regression vs. Dispersion Chemistry Transport

F Hennig; D Sugiri; L Tzivian; K Fuks; S Moebus; KH Jöckel; D Vienneau; T Kuhlbusch; K de Hoogh; M Memmesheimer; H Jakobs; U Quass; B Hoffmann

doi:10.1055/s-0036-1586669

Air pollution modelling in the Ruhr Area: Land-Use Regression vs. Dispersion Chemistry Transport

F Hennig ¹, D Sugiri ², L Tzivian ², K Fuks ², S Moebus ³, KH Jöckel ⁴, D Vienneau ⁵, T Kuhlbusch ⁶, K de Hoogh ⁵, M Memmesheimer ⁷, H Jakobs ⁷, U Quass ⁸, B Hoffmann ⁹

¹Institut für Arbeitsmedizin und Sozialmedizin, Center for Health and Society, Medizinische Fakultät, Heinrich Heine Universität Düsseldorf,, Düsseldorf
²IUF-Leibniz Institut für Umweltmedizinische Forschung, Düsseldorf
³Universitätsklinikum Essen, Essen
⁴Institut für Medizinische Informatik, Biometrie & Epidemiologie Universitätsklinikum Essen, Essen
⁵Department of Epidemiology and Public Health, Swiss Tropical and Public Health Institute, Basel
⁶IUTA e.V., Bereich „Luftreinhaltung & Nachhaltige Nanotechnologie', Duisburg
⁷Rhenish Institute for Environmental Research (RIU), Köln
⁸IUTA e.V., Air Quality & Sustainable Nanotechnology Unit, Duisburg
⁹Institut für Allgemeinmedizin, Universitätsklinikum Düsseldorf, Düsseldorf

Abstract

Background: Two commonly used modelling approaches to assess air pollution (AP) concentration for investigating health effects in epidemiological studies are land use regression (LUR) and dispersion chemistry transport models (DCTM). Choice of model can influence the exposure estimation and consequently the health effect estimation as well. In the Ruhr Area in Germany, the location of multiple epidemiological studies, residential AP concentrations have been modelled with a LUR model as part of the European Study of Cohorts for Air Pollution Effects (ESCAPE-LUR), and with an European Air Quality, Dispersion and Chemistry Transport Model (EURAD-CTM) as part of several research projects investigating health effects of residential air pollution exposure.

Objectives: We aimed to compare ESCAPE-LUR and EURAD-CTM model specifications, and to evaluate the impact of model selection on exposure estimation.

Methods: We compared modelled residential exposure to the air pollutants PM_2.5, PM₁₀ and NO₂ for participants of the Heinz Nixdorf Recall Study, located in the Ruhr Area. Comparison of models included qualitative measures, such as input data, initial aim of modelling, modelling characteristics and the temporal and spatial resolution, as well as quantitative measures, such as goodness of fit and agreement (i.e. correlation coefficient (r) and mean ± SD difference). Since the EURAD-CTM approach provided source-specific AP estimates, comparison of qualitative measures included traffic-specific and industry-specific estimates apart from all sources.

Results: While the ESCAPE-LUR provides a point estimate specifically for traffic-related AP, it is a temporally stable long-term exposure estimate. Alternatively, the EURAD-CTM estimates a time-varying average AP concentration at the small area level (i.e. 1 km²). Furthermore, with different input data (monitoring and land use vs. emissions and meteorology), exposure estimates differ (ΔLUR-CTM: 3.7 ± 1.3 µg/m3 for PM_2.5, 9.8 ± 2.4 for PM₁₀ and -7.4 ± 4.9 for NO₂). Accordingly, overall agreement between ESCAPE-LUR and EURAD-CTM was weak to moderate on a residential basis (r < 0.4). However, restricting EURAD-CTM to sources of local traffic only, respective agreement was good (r > 0.8).

Conclusion and practical relevance: Combining the strengths of both applications is a necessary step to enhance exposure assessment and associated analysis of adverse health effects.