Abstract. Satellite retrievals of vertical column densities (VCDs) of tropospheric nitrogen dioxide (NO₂) normally do not explicitly account for aerosol optical effects and surface reflectance anisotropy that vary with space and time. Here, we conduct an improved retrieval of NO₂ VCDs over China, called the POMINO algorithm, based on measurements from the Ozone Monitoring Instrument (OMI), and we test the importance of a number of aerosol and surface reflectance treatments in this algorithm. POMINO uses a parallelized LIDORT-driven AMFv6 package to derive tropospheric air mass factors via pixel-specific radiative transfer calculations with no look-up tables, taking slant column densities from DOMINO v2. Prerequisite cloud optical properties are derived from a dedicated cloud retrieval process that is fully consistent with the main NO₂ retrieval. Aerosol optical properties are taken from GEOS-Chem simulations constrained by MODIS aerosol optical depth (AOD) data. MODIS bi-directional reflectance distribution function (BRDF) data are used for surface reflectance over land. For the present analysis, POMINO level-2 data for 2012 are aggregated into monthly means on a 0.25° long. × 0.25° lat. grid.

POMINO-retrieved annual mean NO₂ VCDs vary from 15–25 × 10¹⁵ cm⁻² over the polluted North China Plain (NCP) to below 10¹⁵ cm⁻² over much of western China. Using POMINO to infer Chinese emissions of nitrogen oxides leads to annual anthropogenic emissions of 9.05 TgN yr⁻¹, an increase from 2006 (Lin, 2012) by about 19 %. Replacing the MODIS BRDF data with the OMLER v1 monthly climatological albedo data affects NO₂ VCDs by up to 40 % for certain locations and seasons. The effect on constrained NO_x emissions is small. Excluding aerosol information from the retrieval process (this is the traditional "implicit" treatment) enhances annual mean NO₂ VCDs by 15–40 % over much of eastern China. Seasonally, NO₂ VCDs are reduced by 10–20 % over parts of the NCP in spring and over northern China in winter, despite the general enhancements in summer and fall. The effect on subsequently constrained annual emissions is between −5 and +30 % with large seasonal and spatial dependence. The implicit aerosol treatment also tends to exclude days with high pollution, since aerosols are interpreted as effective clouds and the respective OMI pixels are often excluded by cloud screening; this is a potentially important sampling bias. Therefore an explicit treatment of aerosols is important for space-based NO₂ retrievals and emission constraints. A comprehensive independent measurement network with sufficient spatial and temporal representativeness is needed to further evaluate the different satellite retrieval approaches.