Fine Scale Spatio-Temporal Modelling of Urban Air Pollution

Urban air pollution is a leading environmental health concern. However, the association between urban air pollution and health outcomes is not consistently reported in the literature, likely because of inaccurate exposure assessment induced by spatial error. In this study, a spatio-temporal model is presented, which integrates harmonic regression and land use regression (LUR) to estimate urban air pollution at fine spatio-temporal scale. The space-time field is decomposed into space-time mean and space-time residuals. The mean is estimated by linear combinations of harmonic regression components, and the spatial field is modelled with LUR. The residuals account for spatio-temporal deviation from the mean model. Using data from a regulatory monitor network and geographic covariates from a LUR model, the study yields monthly nitrogen dioxide estimates at the postal code level for Calgary, Canada. The model yields a satisfactory fit (R2 = 0.78). The space-time residuals exhibit non-significant to moderate spatial and temporal autocorrelation.