LookUp Table-based Spectral Deconvolution Algorithm (LUT-SDA) for Fine Mode Fraction (FMF)
The fine-mode fraction (FMF) can be a useful tool to separate natural aerosols from man-made aerosols and to assist in estimating surface concentrations of particulate matter with a diameter < 2.5 um. A LookUp Table-based Spectral Deconvolution Algorithm (LUT-SDA) was developed here for satellite-based applications using data such as MODerate resolution Imaging Spectroradiometer (MODIS) measurements. This method was validated against ground-based FMF retrievals from the Aerosol Robotic Network (AERONET). In comparison with the MODIS C6 FMF product in three study areas (Beijing, Hong Kong, and Osaka), FMFs estimated by the LUT-SDA agreed more closely with those retrieved fromthe AERONET with a very lowbias. Eighty percent of the FMF values fell within the expected error range of ±0.4. The root mean square error (RMSE) was 0.168 with few anomalous values, whereas the RMSE for the MODIS FMF was 0.340 with more anomalous values.
Satellite-based PM2.5 estimation using fine-mode aerosol optical thickness
Estimation of ground-level PM2.5 from satellite-derived aerosol optical thickness (AOT) is fraught with uncertainties, a major one being inability to retrieve AOT for fine-mode aerosols (fAOT) that is more closely related to PM2.5. Except over dark oceans, fAOT are either unavailable or unreliable over land. As such, the vast majority of studies have employed total AOT instead. In this study, we compared the total AOT and fAOT for surface PM2.5 estimation using ground-based measurements
collected in Xingtai, China from May to June 2016. The correlation between PM2.5 and fAOT was higher (r =0.74) than that between PM2.5 and total AOT (r =0.49). Then, applying the LUT-SDA to MODIS, we generated a regional PM2.5 product in the NCP from December 2013 to June 2015. The retrieval results are compared with PM2.5 measurements at 30 stations. The two agree well with a correlation coefficient R=0.8 and a root mean square error (RMSE)=18.9 mg/m3 for a total of 921 samples.
