A high spatial and temporal coverage global gridded tropospheric NO2 dataset (2007-2021) based on OMI and GOME-2

NO_x interacts with both air pollution and short lived climate forcers, including ozone, nitrate aerosol, CO, and VOCs, which in turn cause damage to the atmosphere, degrade the environment and threaten human health. Based on the DOAS algorithm, satellite measurements can provide decadal and long-term and grid-by-grid coverage over the entire globe for NO₂ and a few other trace gasses. This work merges two different satellites using NO2 retrieved by DOAS. The overpass times, wavebands used for the retrieval, and the uncertainties of the sensors are different. We take advantage of this variability to produce a merged product which relies on the local strengths of each sensor. Applying machine learning and the DINEOF method, this work generates a 15-year dataset with more overall data, fewer cloud-covered pixels, and data which is of higher quality and lower error. The spatial coverage of the reconstructed dataset is improved by 60% compared with the original datasets. A few specific and scientifically important results of this are explained in detail including: first, higher data coverage and quality over mountain basin regions which are even more clear than using the TROPOMI product, and second higher precision as compared with surface based remotely sensed profiles in highly polluted regions.