Monitoring multiple satellite aerosol optical depth (AOD) products within the Copernicus Atmosphere Monitoring Service (CAMS) data assimilation system

The Copernicus Atmosphere Monitoring Service (CAMS)provides near-real-time forecast and reanalysis of aerosols using the ECMWFIntegrated Forecasting System with atmospheric composition extension,constrained by the assimilation of MODIS and the Polar Multi-Sensor AerosolOptical Properties (PMAp) aerosol optical depth (AOD). The objective of thiswork is to evaluate two new near-real-time AOD products to prepare for theirassimilation into CAMS, namely the Copernicus AOD (collection 1) from the Seaand Land Surface Temperature Radiometer (SLSTR) on board Sentinel 3-A/Bover ocean and the NOAA EPS AOD (v2.r1) from VIIRS on board S-NPP andNOAA20 over both land and ocean. The differences between MODIS (C6.1), PMAp(v2.1), VIIRS (v2.r1), and SLSTR (C1) AOD as well as their departure from themodeled AOD were assessed at the model grid resolution (i.e., level-3)using the 3-month AOD average (December 2019-February 2020 and March-May 2020). VIIRS and MODIS show the best consistency across the products, which isexplained by instrument and retrieval algorithm similarities. VIIRS AOD isfrequently lower over the ocean background and higher over biomass burningand dust source land regions compared to MODIS. VIIRS shows larger spatialcoverage over land and resolves finer spatial structures such as thetransport of Australian biomass burning smoke over the Pacific, which can beexplained by the use of a heavy aerosol detection test in the retrievalalgorithm. Our results confirm the positive offset over ocean (i) betweenTerra/MODIS and Aqua/MODIS due to the non-corrected radiometric calibrationdegradation of Terra/MODIS in the Dark Target algorithm and (ii) betweenSNPP/VIIRS and NOAA20/VIIRS due to the positive bias in the solar reflectivebands of SNPP/VIIRS. SLSTR AOD shows much smaller level-3 values than therest of the products, which is mainly related to differences in spatialrepresentativity at the IFS grid spatial resolution due to the stringentcloud filtering applied to the SLSTR radiances. Finally, the geometrycharacteristics of the instrument, which drive the range of scatteringangles sampled by the instrument, can explain a large part of thedifferences between retrievals such as the positive offset between PMAp datasets from MetOp-B and MetOp-A.