Spectral Deconvolution Algorithm for Global Fine-Mode Aerosol Retrieval in the 1990s Using Dual-Angle Satellite Aerosol Data

Fine-mode aerosol optical depth (fAOD) is related to anthropogenic aerosols, which play an essential role in global climate change and the atmospheric environment. However, knowledge of the global fAOD in the 1990s is very limited. Compared with general satellite-based fAOD retrieval methods, the spectral deconvolution algorithm (SDA) provides an entirely different way to calculate fAOD. In this study, we derived three new SDA-based global daily fAOD ( $0.55\mu \text{m}$ ) datasets from 1995 to 2003 with 1 $^{\circ }\,\,\times1^{\circ }$ spatial resolution using the second Along-Track Scanning Radiometer (ATSR-2) aerosol product data. In global-scale validation, the application of SDA to Swansea University (SU)–ATSR aerosol products for fAOD retrievals showed the best agreement with AErosol RObotic NETwork (AERONET) fAOD, with a correlation coefficient of 0.72, a root-mean-square error (RMSE) of 0.12, and $\sim $ 65.36% of retrievals inside of the estimated error envelope (EE) (±[0.05 + 15%]). Regionally, SDA improves retrieval performance in Southeast Asia and USA when using the SU and ATSR-dual view (ADV) aerosol products. However, the application of SDA to Oxford-RAL Aerosol and Cloud (ORAC)-based aerosol data considerably worsened its accuracy compared with the official ORAC fAOD. With the improved SU–SDA fAOD, the long-term trend showed a considerable decrease in fAOD in Central South America from 1995 to 2003. However, significantly increasing trends were observed in India and China. As SDA uses spectral AOD only, this study demonstrates that it is a promising method for optical satellite spatial fAOD retrieval.