Comparison of monitored air quality data with the predictions of ADMS-3

Version 3 of CERC's Air Dispersion Modelling System has been used to construct a model of air quality in the lower Trent Valley, UK, in order to test the predictions against field measurements. The area is predominantly agricultural, however there are also three large coal fired power stations. There are no other significant sources of SO2 close enough to affect the area significantly. Four continuous air quality monitoring stations are operated around the lower Trent Valley, and a further six SO2 bubbler sites measure daily averages. The predictions of a model of the air quality in the lower Trent Valley for 1998 are compared with the measured values of SO2 at all the specific monitoring sites, and some analysis has been made of the sensitivity of the model to individual input meteorological parameters. A brief comparison was also made with the USEPA model AERMOD. The results of ADMS-3 for long term statistics, in particular the 99.9(th) percentile, are good when the complexity of the model is considered. The predictions are also stable to small perturbations in the input data set, suggesting a generally robust model formulation. By contrast, when individual hours are considered the prediction accuracy is significantly poorer, and a large variation can be induced from small perturbations to the input meteorology. The predictions of ADMS-3 and AERMOD agree quite well for concentrations up to about the 99(th) percentile. Above this the behaviour of the two models diverges rapidly, with large over-predictions from AERMOD, and moderate under-predictions from ADMS-3 above about the 99.9(th) percentile. It is concluded that the predictions of ADMS-3 are broadly representative for long-term statistics, but the model is less successful when predicting individual hours.