ESTABLISHMENT OF A MODEL NETWORK AND ITS APPLICATION FOR THE PREDICTION OF THE AIR POLLUTION IN A MESOSCALE AREA

摘要

First applications of the model network to predict the air pollution on the European scale and in a subregion have been successfully carried out. In general, the average diurnal cycle of the ozone concentration is quite well predicted. Only the increase of the mean ozone concentration in the morning begins in the simulation earlier than in the measurements. The statistical evaluations show that the model simulations in the subregion Southwest Germany predict the ozone concentrations rather well. In the time period between 3 UTC and 21 UTC the deviation between the simulated and measured ozone values was less than 10 ppb in 50% of the cases. This percentage increases to 60% for the time period between 11 UTC and 16 UTC, where the highest ozone concentrations occur. But the slope of the regression line decreases from 0.71 to 0.44 and the correlation coefficient from 0.80 to 0.57, respectively. These data prove that the lower ozone concentrations are overestimated and the higher values are underestimated by the simulations. Mainly, the spatial variability of the ozone concentration for the individual days in the period between 11 UTC and 16 UTC is not sufficiently simulated. On the contrary, the temporal variability of the simulated ozone concentrations at individual stations agrees better with the measured one. In order to improve the model simulations, the reasons for the observed discrepancies have to be analysed by corresponding sensitivity studies.