| In this study, the concept of scale analysis is introduced to evaluate the performance of regional-scale air quality models. To this end, seasonal time series of observations and predictions from the RAMS/UAM-V modeling system for ozone and meteorological variables are spectrally decomposed into fluctuations on the intra-day, diurnal, synoptic and longer-term time scales. The results indicate that the modeling system underestimates the total variance of the ozone time series (energy) when compared with observations at most locations, but shows a higher mean value than the observations. Furthermore, for all components, the model predicts that the average spatial correlation for ozone extends over longer distances than that for the observations. It is shown that correlations between the model predictions and observations for ozone are insignificant for the intra-day component, high for the diurnal component because of the inherent diurnal cycle but low for the amplitude of the diurnal component, and highest for the synoptic and baseline components. Furthermore, analyses of ozone observations and model predictions reveal that only those components that contain fluctuations with periods equal to or greater than one day carry the information that distinguishes ozone episode days from non-episodic days. The magnitude of the intra-day fluctuations is nearly invariant in time. Which of the longer-term fluctuations is dominant in a particular episode varies from episode to episode. These results suggest that for regulatory purposes, modeling periods must be longer than one single episode and the modeling domain must cover the eastern United States to increase the confidence in model predictions. In addition, the effect of simulation length on the predicted ozone improvements stemming from emission reductions is examined. The results demonstrate that for regulatory purposes, model simulations need to cover for longer time periods than just the duration of a single ozone episode; this is necessary not only to perform a meaningful model performance evaluation, but also to quantify the variability in the efficacy of an emissions reduction strategy. |
