Space-based retrieval of NO<sub>2</sub> over biomass burning regions: quantifying and reducing uncertainties
Abstract. The quality of space-based nitrogen dioxide (NO2) retrievals from solar backscatter depends on a priori knowledge of the vertical profiles of NO2 and aerosol optical properties. This information is contained in an air mass factor (AMF), which accounts for atmospheric scattering and is used to convert the measured line-of-sight "slant" columns into vertical columns. In this study we investigate the impact of biomass burning emissions on the AMF in order to quantify NO2 retrieval errors in the Ozone Monitoring Instrument (OMI) products over these sources. Sensitivity analyses are conducted using the Linearized Discrete Ordinate Radiative Transfer (LIDORT) model and the GEOS-Chem chemistry-transport model with an improved daily biomass burning emission inventory. Aircraft in situ data collected during two field campaigns, Arctic Research of the Composition of the Troposphere from Aircraft and Satellites (ARCTAS) and Dust and Biomass-burning Experiment (DABEX), are used to evaluate the modeled aerosol optical properties and NO2 profiles over Canadian boreal fires and western Africa savanna fires respectively. Biomass burning aerosols increase the AMF by 3 to 15% over boreal fires, while they decrease the AMF by −10 to −30% over savanna fires. The presence of an elevated aerosol layer over west Africa due to the Harmattan front explains the negative aerosol effect over this area. The impact of fires on the AMF is driven by the NO2 shape profile perturbations, which decrease the AMF by −10 to −60% over both regions. Aerosol and shape factor effects are most sensitive to surface reflectance and clouds. In particular, retrieval errors associated with shape factor uncertainties can increase by a factor of 2 due to the presence of clouds. In contrast with conclusions from previous studies, we demonstrate that in the presence of pre-existing clouds, the effect of aerosols on the AMF cannot be fully accounted for through the modified retrieved cloud parameters. Finally, a new method that uses slant column information to correct for shape factor error in the retrieval is proposed and tested over west African fires.