57. Mixed Deterministic-Statistical Modelling of Regional Ozone Air Pollution

We develop a physically motivated statistical model for regional ozone air pollution by separating the ground-level pollutant concentration field into three components, namely: transport, local production, and large-scale mean trend mostly dominated by emission rates. The model is novel in the field of environmental spatial statistics in that it is a combined deterministic-statistical model, which gives novel perspectives on the modeling of air pollution. The model is presented in a Bayesian hierarchical formalism, and explicitly accounts for advection of pollutants, using the advection equation. We apply the model to a specific case of regional ozone pollution – the Lower Fraser Valley of British Columbia, Canada. As a predictive tool, we demonstrate that the model outperforms existing, simpler statistical modelling approaches. Our study highlights the importance of simultaneously considering different aspects of air pollution as well as taking into account the physical bases that govern the processes of interest.