Global and local sensitivity analysis of urban background ozone modelled with a simplified photochemical scheme

Abstract

The Generic Reaction Set (GRS) is a simplified photochemical scheme that allows estimation of ozone (O3) concentrations resulting from nitrogen oxides (NOx) and reactive organic compounds (ROC) emissions in an urban area. Due to its simplicity and relative good performance, it has been included in the algorithms of many atmospheric dispersion models. However, few works assess quantitatively the sensitivity of simulated O3 concentrations to the uncertainty in the GRS input variables. In order to do so, in this work we carry out two sensitivity analyses: global (applying an atmospheric dispersion model to a representative real case) and local (using a box-GRS model under a wide range of hypothetical conditions). In the global study, a Monte Carlo analysis is performed to estimate the uncertainty of maximum O3 concentrations that is caused by possible errors in the GRS input variables, applying the DAUMOD-GRS model in the Metropolitan Area of Buenos Aires. Results show that the initial concentration of O3 dominates at all analysed receptors while that of NOx can make a non-negligible contribution if its error is relatively large. In order to further analyse the effect of each parameter individually, a local sensitivity analysis is performed using a box-GRS model under a wide range of conditions. Except for the small NOx-limited region, sensitivity indexes computed for +-1 ppb changes in the initial concentrations of ROC, NOx and O3 are in the ranges 0.00-0.35, 0.00-0.81 and 0.05-0.96, respectively, with that of ozone dominating over most of the isopleth diagram space. In turn, its relative impact increases with decreasing ROC initial concentrations. Reaction rate coefficients have also different effects on O3 peak concentrations depending on the initial conditions of the system. Our results show quantitatively the change of dominant variables under different environments.