Characterization of atmospheric aerosols using MFRSR measurements

Measurements by multifilter rotating shadowband radiometers (MFRSRs) constitute a valuable global data set with contributions from hundreds of instruments deployed worldwide. The geographical coverage of MFRSR networks is complementary to that of AERONET and often provides better spatial density of measurement sites, especially in the United States. We describe our recently updated analysis algorithm for MFRSR data that allows partitioning of the spectral aerosol optical depth (AOD) into fine and coarse mode AOD and retrieval of the fine mode effective radius. Our recent sensitivity study demonstrated that for a typical measurement accuracy 0.01 of AOD, the trade-offs between the spectral aerosol extinction and NO2 absorption in the visible range effectively prevent unambiguous retrieval of NO2 column from MFRSR data and may also bias aerosol size distribution retrievals. This has prompted us to adopt a new retrieval approach, which utilizes climatological NO2 (based on SCIAMACHY satellite retrievals) and uses column ozone from TOMS measurements. The performance of this new approach was evaluated using the long-term data set from the Southern Great Plains (SGP) site operated by the U.S. Department of Energy Atmospheric Radiation Measurement (ARM) Program. We present a detailed intercomparison of total, fine, and coarse mode AOD and fine mode effective radius between two MFRSRs located at the SGP’s Central Facility and with the correlative AERONET Sun-sky inversion results (Version 2) derived from a collocated CIMEL Sun photometer. The comparison between two MFRSRs demonstrated good consistency of both the measurements and the analysis. Agreement with AERONET inversions is remarkably good, in that differences in AOD components do not exceed the expected measurement accuracy of 0.01, while the retrieved values of fine mode effective radius show no relative bias and only 0.03 mm random error (standard deviation of the differences). We show that if only data with large enough AOD (more than 0.06 at 870 nm) are selected, this error is reduced by a factor of two, becoming about 10% of a typical fine mode effective radius value.