Reconstructing Missing Information And Increasing Spatial Resolution For OMI Tropospheric NO₂ Data

Nitrogen dioxide（NO₂）,a pollutant in the Earth’s atmosphere,which seriously affects human health and ecosystems.Satellite remote sensing measurements have played an important role in quantifying NO₂ concentrations at large scales,among which the Ozone Monitoring Instrument（OMI）has provided continuous measurements for up to 17 years and its tropospheric NO₂ vertical column densities are widely used.However,due to the data product with insufficient spatial coverage and low spatial resolution,it has been difficult to meet the needs of monitoring atmospheric changes dynamically and detecting the distribution of NO₂ concentrations in urban scale.In response,this study fills the missing data of OMI due to“Row Anomaly”or cloud contamination,and improves the spatial resolution.Firstly,a framework for reconstructing missing data from OMI sensor in collaboration with NO₂ concentrations from morning and afternoon measurements is proposed to generate a set of daily tropospheric NO₂ products with high spatial coverage in mainland China.Secondly,a machine learning algorithm is used to establish the spatial correlation of remote sensing products with different resolutions,thus increasing the resolution of historical OMI measurements and generate a set of tropospheric NO₂ data with high spatial resolution in Jiangsu.The results in this study show that:1)there is improved spatial and temporal coherence on a day-to-day basis,and the amount of data doubled,with 40%more data available from 2015 to 2018;2)the spatial resolution is improved to 0.05°to clearly observe the high-value emission regions within each city in Jiangsu province,and the NO₂ concentration near the emission source spatial gradient is enhanced;3)the results are reliable overall,with a good agreement with Multi-Axis Differential Optical Absorption Spectroscopy measurements（correlation coefficient:0.75-0.85 for the reconstructed product and 0.76 for the high resolution one）;4)the mean of reconstructed NO₂vertical columns during 2015 and 2018 is consistent with the original data in the spatial distribution,while the standard deviation decreases in most places over Mainland China（-10.4%）,and the data product after increasing the spatial resolution reduces the overestimation of NO₂ concentrations in suburban and rural areas,and the normalized difference can be less than-0.5.The data-driven algorithm framework proposed in the study can adapt to the future development,and the reconstructed and enhanced spatial resolution results are expected to assist in the assessment of health exposures to air pollution and the construction of“top-down”emission inventories,contributing to air quality and climate research.The paper has 45 figures,7 tables,and 158 references.