An improved algorithm for retrieving the fine-mode fraction of aerosol optical thickness. Part 2 : application and validation in Asia

Abstract

Since small aerosol particles are mostly anthropogenic, the fine-mode aerosol optical thickness (fAOT) can be used to infer PM2.5 amounts. However, satellite-based fAOT products such as those from the Moderate Resolution Imaging Spectroradiometer (MODIS) are highly uncertain over land. An improved fAOT retrieval method called the look-up table–spectral deconvolution algorithm (LUT-SDA) was tested and improved using data from Asia. The improvement is achieved by accounting for seasonal changes instead of using constant annual mean values of the aerosol parameters used in the LUT-SDA derived from the Aerosol Robotic Network (AERONET) data from 2010 to 2014. Compared with the previous version of the LUT-SDA developed for Beijing, Hong Kong, and Osaka, the updated LUT-SDA generates more accurate fine-mode fractions (FMFs) with the total mean root-mean-square error (RMSE) decreasing from 0.24 to 0.18. The updated LUT-SDA was then applied to retrieve fAOT and was validated by retrievals from 45 AERONET sites over the period 2015 to 2016. A good accuracy was achieved by this method with 31% of the validation sites having > 50% of retrievals falling within the estimated error (EE) envelope ± (0.05 + 0.15 × AERONET fAOT) and 42% of the validation sites having 40–50% of retrievals falling within the EE envelope. In the total validation and comparison with the MODIS Collection 6 fAOT, the fAOT retrievals from the LUT-SDA agreed more closely with AERONET retrievals, showing a low bias. About 48% of the LUT-SDA-based fAOT retrievals fell within the EE envelope (RMSE = 0.29), while ~22% of the MODIS-based fAOT retrievals fell within the EE envelope (RMSE = 0.42). The fAOT was significantly underestimated by the MODIS algorithm in most areas of Asia with many values of zero. This study demonstrates that the refined LUT-SDA method is valid for the large-scale estimation of fAOT from satellite images.