A comparison of the WIND System atmospheric models and MATS data

The results produced by two of the WIND System atmospheric models, PUFF/PLUME and 2DPUF, were compared with a select group of eight MATS experiments to determine the performance of the models. Three of the MATS experiments employed TRAC vehicle sampling and the remaining five used a line of fixed samplers. The performance of the models was based on certain dispersion characteristics that are important in emergency response situations. Both PUFF/PLUME and 2DPUF were executed with the same source term and meteorological data. When the numerical results from the models were compared to the observed values from the MATS experiments, it was found that 2DPUF produced concentrations and plume widths that were closer to the observed values than PUFF/PLUME. Both models did not produce any bias in the values of the concentration when individual data points were examined; however, PUFF/PLUME consistently overpredicted the peak and total concentrations. 2DPUF did not exibit any bias in the peak and total concentrations. When wind direction errors were removed, 80--84% of the concentrations from PUFF/PLUME and 88% of the concentrations from 2DPUF where within a factor of 10 of the observed values. In some instances, both models were able to predict concentration values that were comparible to a more complex, three-dimensional model called MATHEW-ADPIC. Considering all of the possible uncertainties associated with dispersion modeling, PUFF/PLUME and 2DPUF performed reasonably well. The differences between the dispersion forecasts made by PUFF/PLUME and 2DPUF and the observed surface tracer concentration are very similar to many other emergency response models based on the Gaussian assumption