Evaluation of modeling NO₂ concentrations driven by satellite-derived and bottom-up emission inventories using in situ measurements over China

Abstract. Chemical transport models together with emission inventories are widely used to simulate NO₂ concentrations over China, but validation of the simulations with in situ measurements has been extremely limited. Here we use ground measurements obtained from the air quality monitoring network recently developed by the Ministry of Environmental Protection of China to validate modeling surface NO₂ concentrations from the CHIMERE regional chemical transport model driven by the satellite-derived DECSO and the bottom-up MIX emission inventories. We applied a correction factor to the observations to account for the interferences of other oxidized nitrogen compounds (NO_z), based on the modeled ratio of NO₂ to NO_z. The model accurately reproduces the spatial variability in NO₂ from in situ measurements, with a spatial correlation coefficient of over 0.7 for simulations based on both inventories. A negative and positive bias is found for the simulation with the DECSO (slope  =  0.74 and 0.64 for the daily mean and daytime only) and the MIX (slope  =  1.3 and 1.1) inventories, respectively, suggesting an underestimation and overestimation of NO_x emissions from corresponding inventories. The bias between observed and modeled concentrations is reduced, with the slope dropping from 1.3 to 1.0 when the spatial distribution of NO_x emissions in the DECSO inventory is applied as the spatial proxy for the MIX inventory, which suggests an improvement of the distribution of emissions between urban and suburban or rural areas in the DECSO inventory compared to that used in the bottom-up inventory. A rough estimate indicates that the observed concentrations, from sites predominantly placed in the populated urban areas, may be 10–40 % higher than the corresponding model grid cell mean. This reduces the estimate of the negative bias of the DECSO-based simulation to the range of −30 to 0 % on average and more firmly establishes that the MIX inventory is biased high over major cities. The performance of the model is comparable over seasons, with a slightly worse spatial correlation in summer due to the difficulties in resolving the more active NO_x photochemistry and larger concentration gradients in summer by the model. In addition, the model well captures the daytime diurnal cycle but shows more significant disagreement between simulations and measurements during nighttime, which likely produces a positive model bias of about 15 % in the daily mean concentrations. This is most likely related to the uncertainty in vertical mixing in the model at night.