Keywords: atmospheric pollutants, dispersion, Gaussian models, atmospheric dispersion models, environmental impact assessment, Portugal, sulphur dioxide, modelling, air pollution, environmental pollution
Sulphur dioxide dispersion from a main point source in Portugal: comparison between models
A simulation of sulphur dioxide dispersion from a power plant was carried out using CRSTER, MPTER and POLUX models. The first two models are from the Environmental Protection Agency and the third one is from the Department of Environment and Planning of the University of Aveiro. Both episodic and long-term simulations were made, including model performance comparison and analysis of model accuracy. For all models considered, mean hourly values, paired in space and time, exceeded the measured values by more than a factor of two, showing low correlation coefficients and NMSE (normalised mean square error) significantly different from zero. Long-term simulation for 1992 was made by comparing model performances for four receptors. Taking into account Portuguese air quality standards, the 50th and 98th percentiles of measured and calculated daily concentration data were used for comparison, along with maximum and average daily concentrations for that year. Results indicate that CRSTER has a tendency to overpredict daily measured concentrations, and the other two underpredict in all cases, except for one receptor where all three models overpredict. The correlation coefficients and NMSE found are low, of the order of 0.1 to 0.26, and 4 to 8, respectively. However, the results in terms of 98-percentile and maximum of daily concentrations show good agreement between calculated and measured values. This might mean that the models can be used for regulatory purpose with reliable results, even if the comparison indicates low statistical performances in terms of bias, NMSE and correlation coefficient, most probably due to an impairing in time of measured and calculated concentrations.