Evaluation of aerosol optical depth and aerosol models from MODIS and VIIRS retrieval algorithms over North China Plain

After the launch of Suomi National Polar-orbiting Partnership (S-NPP) equipped with the Visible Infrared Imaging Radiometer Suit (VIIRS) instrument in late 2011, the aerosol products of VIIRS have received much attention. Currently there are two aerosol products of VIIRS by using different algorithms: VIIRS Environment Data Record data (VIIRS_EDR) and aerosol products by applying MODIS-like algorithm to VIIRS (VIIRS_ML). In this study, the aerosol optical depth (AOD) at 550nm and properties of aerosol models used in the two VIIRS algorithms (VIIRS_EDR and VIIRS_ML) are compared respectively with their corresponding quantities retrieved from the ground-based Sunphotometer measurements (CE318) during May 2012-March 2014 at three sites over North China Plain (NCP): metropolis-Beijing, suburban-XiangHe and regional background site-Xinglong. The results show that the VIIRS_EDR AOD has a positive mean bias (MB) of 0.04-0.06 and the root mean square error (RMSE) of 0.22-0.24 in NCP region. Among three sites, the largest MB (0.10-0.15) and RMSE (0.27-0.30) are observed in Beijing. The results of evaluation of VIIRS_ML for each site and quality flags analysis are similar to VIIRS_EDR, but in general the VIIRS_ML AOD shows better than VIIRS_EDR except for the MB (0.13-0.14). The model comparisons show that the occurrence percentages of both dust and clean urban aerosol in VIIRS_EDR (82% for Beijing, 73% for XiangHe and 50% for Xinglong) are significantly larger than that for CE318, the latter shows the polluted urban aerosol is the dominant aerosol especially for Beijing (67%) and XiangHe (59%) sites. The values of Single Scattering albedo (SSA) from VIIRS_EDR are higher than from CE318 in all aerosol modes, with a positive bias of 0.03-0.06 for fine mode, 0.18-0.22 for coarse model and 0.03-0.08 for total modes and the aerosol microphysical properties used in the VIIRS_EDR algorithm for AOD retrieval show a large difference with the counterparts from CE318 inversion results. The differences of SSA between VIIRS_ML and CE318 are mostly less than 0.05, indicating the model used in VIIRS_ML seems more reasonable than VIIRS_EDR in NCP region. Therefore, the predefined aerosol models and the corresponding microphysical properties in the VIIRS_EDR algorithm are likely to be repressive and should be refined in NCP region.