The performance of four widely used dispersion models (AERMOD, a single-equation Gaussian formulation, and two versions of CALPUFF) for predicting ambient hydrocarbon concentrations at a regional air quality monitor in the Eagle Ford Shale oil and gas production region was assessed. Model performance was found to vary considerably based on the performance objective, meteorological conditions, and temporal resolution. Among the models evaluated in this work, the methods used to estimate the dispersion coefficients and whether the model was plume- or puff-based strongly influenced model performance. Uncertainties in meteorological and emissions inputs also played an important role in model performance, but the significance of their impact varied depending on the performance objective. Techniques to identify and address model uncertainties and for selecting the best performing model for a given application are suggested.
Graves et al. (Mon,) studied this question.