Evaluation of the accuracy of the Sentinel-5 Precursor operational SO2 products over China

The Sentinel-5 Precursor (S-5P) was launched into low earth orbit with a single payload TROPOspheric Monitoring Instrument (TROPOMI) on 13 October, 2017. TROPOMI is able to provide atmospheric composition observations including sulfur dioxide (SO2) daily and globally. It has a higher spatial resolution, specifically with a ground resolution of 7 kmx3.5 km in the nadir, than previous satellite instruments such as Ozone Monitoring Instrument (OMI, 13 kmx24 km) and Ozone Mapping and Profiler Suite (OMPS, 50 kmx50 km). The three types of S-5P operational SO2 products (offline (OFFL), Reprocessing (RPRO) and Near Real Time (NRTI)) are on basis of the differential optical absorption spectroscopy (DOAS) algorithm, and available from the S-5P Pre-Operations Data Hub. To evaluate these data, the results of OFFL are compared to the measurements obtained by ground-based Multi-Axis DOAS (MAX-DOAS) and China National Environmental Monitoring Center (CNEMC). We find that the S-5P operational SO2 products significantly overestimate SO2 concentrations over Northern China. Compared with four ground-based MAX-DOAS measurements in terms of the weekly average SO2 vertical column densities (VCDs), the S-5P operational SO2 products exhibit an overestimation of 61%-140%, with the correlation coefficient less than 0.5. The a priori SO2 profiles in S-5P operational products are from daily Tracer Model (TM5) simulations. SO2 surface concentrations are converted by the S-5P operational SO2 products and a priori SO2 profiles used in the retrievals. The correlation coefficient between the SO2 surface concentrations measured by CNEMC over Northern China and those from the S-5P operational products is higher than 0.5, and the S-5P operational products at these CNEMC sites show an overestimation of about 54.6% on average. In this study, the SO2 products (TROPOMI USTC SO2 products) are retrieved by the University of Science and Technology of China (USTC) with the optimal estimation algorithm (OE) from TROPOMI measurements, with the a priori SO2 profiles from monthly average GEOS-Chem simulations. The mean biases between the TROPOMI USTC SO2 products and the ground-based MAX-DOAS measurements are reduced significantly (-4.8%-22%), together with an improvement in the correlation coefficients (0.72-0.89). TROPOMI USTC SO2 surface concentrations are transformed by the TROPOMI USTC SO2 products and a priori SO2 profiles in the retrievals. Based on the evaluation statistics of the S-5P operational SO2 products, a smaller mean bias of -25.8% is found between the TROPOMI USTC SO2 surface concentrations and CNEMC measurements over Northern China. Compared to the TROPOMI USTC SO2 products in the heavily polluted area (33 degrees-45 degrees N, 110 degrees-122 degrees E), the S-5P operational SO2 products have an overestimation of around 41.3%, with a correlation coefficient of 0.73. In addition, S-5P operational SO2 products show a larger area of heavy pollution than TROPOMI USTC SO2 products. Additionally, the sensitivity analysis indicates many influential factors of the SO2 retrievals using OE algorithm, such as radiative calibration, fitting window and a priori SO2 profiles. In the fitting window of 310.5-326. 0 nm, the SO2 columns retrieved by the a priori profiles from monthly GEOS-Chem simulations show better agreement with the MAX-DOAS measurements than those from daily TM5 simulations. A possible reason is that the emission inventory used in GEOS-Chem is based on a Chinese power sector database for 31 provinces supported by CMEP (Chinese Ministry of Environmental Protection), resulting in a better performance of GEOS-Chem than that of TM5 over China in simulations. And the overestimation of the S-5P operational SO2 products over Northern China may be attributed to the fitting window used.