A reactive state-space model for prediction of urban air pollution

Abstract

The generic reaction set (GRS) model offers a convenient framework for studying photochemical smog production. Its highly condensed seven equations are deduced from the principal reactions that produce photochemical smog (such as photolysis of reactive organic species, oxidation of NO to NO 2, photolysis of NO 2, etc.), and have been validated with the CSIRO outdoor smog chamber data. The performance of the model has been found comparable to more detailed photochemical mechanisms such as the CBM-IV. This paper expands the GRS model to include spatial advection and diffusion in the airshed. Via an appropriate numerical scheme, this extended dynamic model is transformed into the state space form, from which interpolation and prediction can be performed using the Kalman algorithm. The model is implemented on a simple grid of seven stations in the Sydney monitoring network. One-step ahead forecasts are derived for observed as well as unobserved locations. Comparison with observed data indicate that the model performs quite well, in particular, it traces the ozone episodes accurately.