Spatial-temporal air pollution models for national-scale health studies

Weiyi Wang, Daniela Fecht, John Gulliver, Sean Beevers, Medical Research Council (Great Britain)
Air pollution exposure assessment is fundamental to establish the associations between air pollution and health outcomes. In the recent decade, exposure models have been widely used to assess the long-term effects of air pollution. Studies analysing the short-term effects, however, are still relying on measurements obtained from routine monitoring network. The aim of this thesis is to develop spatially resolved daily and annual air pollution models using land use regression (LUR) techniques
more » ... can be applied in air pollution epidemiological studies in Great Britain. The first part of the thesis developed national air pollution exposure models NO2, PM2.5, PM10 and O3. The spatial models followed a traditional land use regression technique, and the spatio-temporal models used a generalised additive mixed model framework (GAMM). The predictors include geographic information system (GIS)-derived land use variables and estimates from a chemical transport model (CTM). The second part demonstrated an application of daily estimates in a case-crossover design to assess the health effects of short-term exposure at the individual level. Overall, the hybrid approach incorporating land use information and CTM estimates improved model performance compared to traditional LUR, and produced more accurate long-term exposure estimates. The GAMM framework integrating land use variables and daily CTM estimates successfully captured local space-time variability over a large geographic domain. Using the modelled daily NO2 concentrations in a case-crossover design, NO2 was found to be associated with emergency hospital admissions for both asthma and other respiratory diseases. The short-term and long-term exposure models developed in this thesis have great potential in reducing exposure misclassification in epidemiological studies.
doi:10.25560/97230 fatcat:ei5tr3ixcndedan77jibjhgnoe