Estimating Surface SO2 Concentrations over Eastern China Using OMI Satellite Data and GEOS-Chem

Sulfur dioxide (SO₂) is one of the most commonly monitored air pollutants as it contributes to aerosol formation and growth, as well as acid rain. Using satellite data to monitor surface SO₂ concentrations can fill in the gaps of ground-based air quality networks in rural and remote areas. In this study, we employ planetary boundary layer SO₂ vertical column densities (VCDs) retrieved from the Ozone Monitoring Instrument (OMI) to estimate seasonal and annual surface concentrations over eastern China, a region well-known for abundant SO₂ emissions, during 2015-2018. We first use OMI-constrained Community Emissions Data System SO₂ emissions to run GEOS-Chem model simulations. We then use the simulated SO₂ vertical profiles to convert OMI SO₂ VCDs to surface concentrations. The OMI-constrained SO₂ concentrations have a good agreement with the spatial distribution of in-situ measurements, but the magnitudes are biased low by about 60-80%. The significant underestimation may be due to errors in the simulated profile shape, which are partially attributed to the coarse model resolution. The GEOS-Chem concentrations are underestimated by about 80% as compared with in-situ measurements, whereas the VCDs are overestimated by just over 250% compared to OMI, resulting in surface-to-VCD ratios that are too small. Unlike the in-situ measurements, the OMI derived concentrations do not show a consistent seasonality. Both the in-situ and OMI data show decreases in seasonal and annual average SO₂ concentrations from 2015 to 2018, but the decreases observed by the in-situ measurements are 2-4 times larger than OMI. OMI-derived concentrations are slightly more accurate in regions where SO₂ loading is high, like the North China Plain, compared to rural areas with low SO₂ loading. We also notice that as SO₂ concentrations decrease over time, the statistical relationship between OMI-derived concentrations and in-situ measurements weakens. As global SO₂ emissions continue to decrease, it becomes increasingly challenging to derive surface SO₂ concentrations using satellite data and/or global models that have relatively coarse resolution. Alternative methods (such as machine learning) and/or higher resolution geostationary satellites like TEMPO may be more useful for estimating surface SO₂ concentrations.