LAUSR

Abstract We apply the regional chemistry-transport model WRF-Chem to investigate the air pollution characteristics in three Chinese metropolitan areas (Beijing, Shanghai and Guangzhou) and assess the model performance in wintertime (January 2010) and summertime (July 2010) conditions. Three simulations utilizing the HTAPv2, REASv2 and MACCity anthropogenic emission inventories are conducted and compared with satellite and ground-based measurements to assess the sensitivity of the model predictions to anthropogenic emissions. Model-predicted surface meteorological fields are overall in good agreement with surface observations. However, the model tends to overestimate maximum surface wind speed, a feature also found in previous WRF-Chem modelling studies. Large discrepancies between the three emission inventories are found, specifically in the spatial distribution and the magnitude of the emissions (e.g. over 98%, 86% and 85% differences in some regions for carbon monoxide (CO), nitrogen oxide (NO) and organic carbon (OC) emissions, respectively). As a result, we find large differences between the simulations. When compared to satellite and surface observations, the HTAPv2-and REASv2-based simulations better reproduce the magnitude and spatial patterns of CO total columns and NO2 tropospheric columns derived from MOPITTv6 and GOME2 satellite observations, respectively. The simulations satisfactorily reproduce the air pollution characteristics in January 2010 as observed at surface monitoring stations in Beijing and Shanghai. The HTAPv2-based simulation better captures the pollution episode that occurred in Beijing in mid-January 2010, which is found to be related to a combination of different meteorological factors that led to the build-up of pollution. The high photochemical activity and convective turbulence in July 2010, in addition to differences in the emissions used, led to more pronounced differences between the simulations and lower statistical skills in particular for trace gases. This study shows that a considerable part of uncertainties in air quality model predictions for China can be attributed to uncertainties in current anthropogenic emission inventories. It highlights the need for more accurate and highly resolved emissions in order to improve the accuracy of model predictions. However, this study reveals that in some cases, model-observation discrepancies cannot be attributed only to inaccuracies in the emissions. For example, we found that regardless of the adopted emissions, all of the simulations fail to capture the temporal variability of particulate matter in Guangzhou during January 2010, indicating that factors other than emissions have a considerable influence on the atmospheric composition in this region. This work shows that inaccuracies in the predicted meteorological parameters (e.g. wind speed/direction, relative humidity, etc.) significantly affect the model performance, as found for example in the model predictions for Shanghai for July 2010, which show high and abrupt increases of NOx and particulates due to rapid decreases in wind speed in the model.