Credibility and statistical characteristics of CAMSRA and MERRA-2 AOD reanalysis products over the Sichuan Basin during 2003-2018

Owing to its large population, unique topography, and rapid industrialization and urbanization, the Sichuan Basin (SCB) is one of the most polluted regions in China, as reflected by the large aerosol optical depth (AOD). In this study, the credibility of two long-term atmospheric composition reanalysis datasets-the Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) and Copernicus Atmosphere Monitoring Service Reanalysis (CAMSRA)-are evaluated using Moderate Resolution Imaging Spectroradiometer (MODIS) AOD products over the SCB during 2003-2018. The time series of the three datasets at seasonal, monthly and annual scales exhibit some interesting similarities and differences, but with all of them demon-strating a prominent slowing down of the severe air pollution status over the SCB since 2012 (-0.25/7yrs for MODIS and CAMSRA and-0.12/7yrs for MERRA-2). CAMSRA shows good consistency with MODIS at all time scales (correlation coefficients varying from 0.89 to 0.97 at four seasons) despite its systematic overestimation (similar to 0.2) throughout the 16-year study period. MERRA-2 has a negative deviation during 2003-2006 and then maintains an increasing positive deviation in the following years (0.05-0.2). This positive bias increases as MODIS AOD declines prominently from 2014. The good temporal consistency of CAMSRA is partly attributable to the bias correction of the assimilated remotely sensed AOD from MODIS and AATSR (Advanced Along-Track Scanning Radiometer). The differences between the reanalysis and MODIS are obvious during October-December, which may be related to the cloud contamination effects as well as the uncertainties of emissions in each model system. The uniform decreasing trend of the three datasets since 2012 supports the effects of the tough emissions reduction policy, which is also evident in AOD changes of anthropogenic species. It is concluded that atmospheric composition reanalysis data with good spatial and temporal coverage can be considered in future air quality analyses over the SCB. However, the bias and regional dependence should be handled carefully.