Version 2 Ozone Monitoring Instrument SO2 product (OMSO2 V2): new anthropogenic SO2 vertical column density dataset

The Ozone Monitoring Instrument (OMI) has been providing global observations of SO2 pollution since 2004. Here we introduce the new anthropogenic SO2 vertical column density (VCD) dataset in the version 2 OMI SO2 product (OMSO2 V2). As with the previous version (OMSO2 V1.3), the new dataset is generated with an algorithm based on principal component analysis of OMI radiances but features several updates. The most important among those is the use of expanded lookup tables and model a priori profiles to estimate SO2 Jacobians for individual OMI pixels, in order to better characterize pixel-to-pixel variations in SO2 sensitivity including over snow and ice. Additionally, new data screening and spectral fitting schemes have been implemented to improve the quality of the spectral fit. As compared with the planetary boundary layer SO2 dataset in OMSO2 V1.3, the new dataset has substantially better data quality, especially over areas that are relatively clean or affected by the South Atlantic Anomaly. The updated retrievals over snow/ice yield more realistic seasonal changes in SO2 at high latitudes and offer enhanced sensitivity to sources during wintertime. An error analysis has been conducted to assess uncertainties in SO2 VCDs from both the spectral fit and Jacobian calculations. The uncertainties from spectral fitting are reflected in SO2 slant column densities (SCDs) and largely depend on the signal-to-noise ratio of the measured radiances, as implied by the generally smaller SCD uncertainties over clouds or for smaller solar zenith angles. The SCD uncertainties for individual pixels are estimated to be similar to 0.15-0.3 DU (Dobson units) between similar to 40 degrees S and similar to 40 degrees N and to be similar to 0.2-0.5 DU at higher latitudes. The uncertainties from the Jacobians are approximately similar to 50 %-100 % over polluted areas and are primarily attributed to errors in SO2 a priori profiles and cloud pressures, as well as the lack of explicit treatment for aerosols. Finally, the daily mean and median SCDs over the presumably SO2-free equatorial east Pacific have increased by only similar to 0.0035 DU and similar to 0.003 DU respectively over the entire 15-year OMI record, while the standard deviation of SCDs has grown by only similar to 0.02 DU or similar to 10%. Such remarkable long-term stability makes the new dataset particularly suitable for detecting regional changes in SO2 pollution.