Evaluation of the AERMOD Dispersion Model as aFunction of Atmospheric Stability for an Urban Area

摘要

The AERMOD dispersion model was used to compute ambient air concentrations of SO2for 1-hr, 3-hr, and 24-hr using the emission inventory data for Lucas County, Ohio for theyear 1990. The estimated concentrations were classified based on the stability parameter,Monin-Obukhov length (L), for the two monitoring stations located in the area. The datawere divided into two atmospheric stability classes (stable and convective cases) as usedin the AERMOD model. These categories were further grouped into five sub categoriesbased on the value of L to learn about fine details of model performance. The modelevaluation was done using several statistical parameters used in air quality studies.AERMOD did not yield a satisfactory performance in predicting 1-hr and 3-hr averageconcentrations for the multi-source region but showed a slightly better performance inpredicting the 24-hr concentrations using urban option for the land use parameters. Themodel had a tendency to underpredict in both the stable and convective cases. In the 24-hr averaging period Fa2 showed a better performance than FB. The model seemed toperform better for the Main Street station than the Collins Park station. Limited analysisusing different land use parameters reported in the paper indicates that modelperformance may improve for certain cases. Other sources of error include theunavoidable scatter due to differences between wind direction and actual transport for anaveraging period, formulation of the model, and the emission inventory. The results ofthe study should be used cautiously because of the limited scope of the evaluation. Futurework should focus on the role of land use parameters in predicting concentrations at themonitors and finding ways to quantify errors due to other factors. However, it is clear thatmore guidance is needed in order to apply the AERMOD model for multi-source regions.Alternative schemes to divide the data should also be considered for analyzing modelperformance.