Evaluation of the performance of RAM with the Regional Air Pollution Study data base

The RAM air quality simulation model's performance is examined using the Regional Air Pollution Study (RAPS) Level-7 data base. Time series analyses were performed to test the adequacy of RAM in simulating the dynamics of pollutants in the atmosphere. Power spectrum and auto-correlation analyses show that the predicted concentrations do not have the same temporal characteristics as the observations during the winter period.Both paired and unpaired analyses are included to critically examine the model performance. The paired comparisons, including those performance measures suggested by the AMS Woods Hole workshop, indicate a better agreement between predicted and observed data at the rural sites than at the urban sites. When the data are segregated according to wind speed and atmospheric stability class, it is found that there is very little agreement between the predicted and measured concentrations under extreme stability (either very unstable or very stable) and low wind speed (less than 2 ms⁻¹ conditions.The measured and predicted daily maximum concentrations are subjected to the ‘bootstrap’ sampling procedure to develop the distribution of the differences between observed and predicted concentrations. These distributions suggest that the errors are random, and, therefore, RAM cannot be calibrated to improve model performance within the urban area. Further, the analyses suggest that improvement of RAM's performance may be realized through a better characterization of area source emissions within the urban area and the inclusion of other physical processes such as a fumigation algorithm in the model.