The uncertainty in modelled air concentrations of NOx due to choice of emission inventory

Emission data, together with meteorology, is the most important input to air quality models. In this work we compare two simulations, performed with the Fine Resolution Atmospheric Multi-pollutant Exchange (FRAME) model. In the first model run, we use an emission inventory prepared for Poland by downscaling the national total emissions to a 5km x 5km resolution grid. For the second model run, we replace the NOx road transport emission with the TNO-MACC II emission inventory available for the entire Europe. The grid mesh of both emission inventories was unified for this work to a 5km x 5km resolution, as required by the FRAME model. The FRAME model is run for the year 2008 and compared with the NOx measurements. Both inventories are in close spatial agreement, with a grid to grid correlation coefficient of 0.75. The FRAME model run with TNO-MACC II emission inventory shows better agreement with NOx measurements gathered at 106 stations if the mean bias is considered. For the model run with TNO-MACC II inventory, bias is -0.7 μg m-3, and the RMSE is 8.19 μg m-3. The bias for the FRAME model run with the national road transport emission inventory is -3.10 μg m-3 and the RMSE 7.97 μg m-3.