Sensitivity analysis of chemical mechanisms based on field data.

摘要

Ozone air pollution damages the health of humans and ecosystems; it has been the focus of decades of research and regulatory action. The atmospheric processes causing ozone are complex and contain hundreds of chemical reactions, including reactions of reactive gases hydroxyl (OH) and hydroperoxyl (HO 2). Chemical mechanisms describe these complex chemical reactions mathematically and hence are essential for air quality modeling.;In this study, as a first step toward improving model performance, sensitivity analysis was applied to study the effect of the uncertainties in all model constraints and inputs on the uncertainties in OH, HO2, and ozone production predictions. Uncertainties in hundreds of model parameters were assigned in a reasonable manner for measured amounts of model constraints, photolysis rates and kinetic rates, and the product yields, typically ranging from about 10% to a factor of two. 28,500 Monte Carlo runs were made with given sets of different initial conditions based on the measurements of this field campaign.;The model-measurement discrepancies were examined based on over 100 base cases from 16 days of field data. The relative uncertainty (+/-1sigma) exhibits a persistent diurnal pattern: high uncertainty at morning rush hour (about 35%), low uncertainty in the afternoon (about 20%), and intermediate uncertainty at night (about 25%). The sources of model uncertainty are dominated by the uncertainties in chemical schemes (30-60% from kinetic rates and 10-40% from product yields), while the uncertainties in measurements are less influential. The most important model parameters are generally associated with the amounts of monoterpenes and acetaldehyde, the photolysis of HONO and HCHO (&;The effect of uncertainties of model parameters on the prediction of whether ozone production is limited by nitrogen oxides (NOx) or by volatile organic chemicals (VOCs) was examined for over 30 transition cases between NOx-sensitive and VOCsensitive regimes. The impacts of the most important parameters on ozone production limitations by NO x or VOCs were quantified. The greater values of the NO amount, the reaction rates of NO2 + OH, NO + HO2 and ISOP (isoprene peroxy radicals) + NO rather more VOC-sensitivity, while the higher values of higher aldehydes and nonoterpenes, and kinetic rates associated with reactions of OH with aldehydes and xylenes, internal alkenes with O3, and ISOP with HO2 rather more NOx-sensitivity of ozone under the studied conditions.;The mechanism-mechanism differences in model predictions were examined based on about 30 cases from four typical days of data. The difference in kinetic rates is the major source (averaged 53-91%) for nighttime when the mechanism-mechanism discrepancy is larger, while the difference in product yields are more important (averaged 49%) during daytime when the discrepancy is smaller. The major contributors were revealed to be associated with the reactions of internal alkenes with O3, alkenes/xylenes/ aldehydes with OH, and organic peroxy radicals with HO2 or NO.;Model sensitivity of OH and HO2 was also examined by varying each model parameter by a factor of three one-at-a-time to test the model response if the uncertainties in some model parameters were greatly underestimated. Thirty-one model parameters were considered influential with the corresponding percentage of HOx variation greater than 10%. For most of these important model parameters, both small and large perturbations of their constrained values lead to significant variations of OH and HO2. However, some model parameters were found influential only when larger variations were assumed, but not influential within their assigned uncertainties. Such model parameters included the amounts of ozone and several internal alkenes, and product yields of peroxy radicals generated from reactions of HC3 with OH and RO2 with NO.;The sensitivity analysis applied in this study helps evaluate model uncertainty comprehensively and improve model performance by quantifying the uncertainty sources from the model parameters associated with chemical mechanisms. These results can then be used to improve chemical schemes to better represent the complex atmospheric processes and to achieve better model-measurement agreement eventually. With a better understanding of these complex atmospheric processes, more effective regulatory action can then be taken to reduce the ozone air pollution and improve the air quality.