Uncertainty analysis of ozone formation and response to emission controls using higher-order sensitivities

Understanding ozone response to its precursor emissions is crucial for effective air quality management practices. This nonlinear response is usually simulated using chemical transport models, and the modeling results are affected by uncertainties in emissions inputs. In this study, a high ozone episode in the southeastern United States is simulated using the Community Multiscale Air Quality (CMAQ) model. Uncertainties in ozone formation and response to emissions controls due to uncertainties in emission rates are quantified using the Monte Carlo method. Instead of propagating emissions uncertainties through the original CMAQ a reduced form of CMAQ is formulated using directly calculated first- and second-order sensitivities that capture the nonlinear ozone concentration-emission responses. This modification greatly reduces the associated computational cost. Quantified uncertainties in modeled ozone concentrations and responses to various emissions controls are much less than the uncertainties in emissions inputs. Average uncertainties in modeled ozone concentrations for the Atlanta area are less than 10% (as measured by the inferred coefficient of variance [ICOV]) even when emissions uncertainties are assumed to vary between a factor of 1.5 and 2. Uncertainties in the ozone responses generally decrease with increased emission controls. Average uncertainties (ICOV) in emission-normalized ozone responses range from 4 to 22%, with the smaller being associated with controlling of the relatively certain point nitrogen oxide (NOx) emissions and the larger resulting from controlling of the less certain mobile NOx emissions. These small uncertainties provide confidence in the model applications, such as in performance evaluation, attainment demonstration, and control strategy development.