Sensitivity of Regional Oxidant Model Predictions to Prognostic and Diagnostic Meteorological Fields

Abstract

Objective analysis and diagnostic methods are used to provide hourly meteorological fields to many air quality simulation models. The viability of using predictions from the Pennsylvania State University?National Center for Atmospheric Research Mesoscale Model version 4 (MM4) together with four-dimensional data assimilation, technique to provide meteorological information to the U.S. EPA Regional Oxidant Model (ROM) was studied. Two numerical simulations were performed for eight days using the ROM for a domain covering the eastern United States. In the first case, diagnostically analyzed data were used to provide meteorological conditions, while in the second case the MM4's prognostic data were used. Comparisons of processed diagnostic and prognostic meteorological data indicated differences in dynamical, thermodynamical, and other derived variables. Uncertainties and forecast errors present in the predicted vertical temperature profiles led to estimation of lower mixed-layer heights (? 30%?50%) and a smaller diurnal range of atmospheric temperatures (? 2 K) compared with those obtained from the diagnostic data. Comparison of area-averaged horizontal winds for four subdomains indicated minor differences (? 1?2 m s?1). These differences systematically affected the estimation of other derived meteorological parameters, such as friction velocity and sensible heat flux. Processed emission data also showed some differences (? 1?5 ppb h?1) that resulted from the differing characteristics of the diagnostic and prognostic meteorological data. Comparison of predicted concentrations of primary (emitted) chemical species such as NOx and reactive organic gases in the two numerical simulations indicated higher values (1?5 and 1?25 ppb, respectively) when the prognostic meteorological data were used. This result was consistent with the lower estimated values of the ROM's layer 1 and layer 2 heights (planetary boundary layer) with the prognostic meteorology. However, comparison of predicted ozone concentrations did not indicate similar features. Area averages of predicted concentrations of ozone for four subdomains indicated both increases and decreases (+1 5 to ?10 ppb) over the area averages predicted by the ROM using diagnostic meteorological data. These results indicate that the prediction of trace gas concentrations and the nonlinearity in the model's chemistry are sensitive to the type of meteorological input used.