Development of a cloud-screening algorithm for direct and diffuse AODs from the Skyradiometer Network

Abstract

We developed a cloud-screening algorithm for direct and diffuse aerosol optical depths (AODs) from the Skyradiometer Network (CSDD). Variabilities of direct AODs were checked not only to screen cloud-contaminated (cloudy) data effectively but also to make direct AODs available for aerosol study, as in the Aerosol Robotic Network (AERONET). Skyradiometer data were used from the Seoul National University (SNU) site for three years from 2012 to 2014. CSDD tested the spectral and temporal variabilities according to the Ångström exponent at the first stage, and the temporal smoothness in the form of the coefficient of variation at the second stage. The algorithm for CSDD was constructed to minimize the differences with optical properties of cloudy data (and clear-sky data as well) based on the cloud cover from the synoptic station. A number of cloudy data that was not screened by the previous algorithms was removed, and the absolute value of the bias total could be substantially lowered. The performances of algorithms for cloud detection were also examined using lidar observations at the study site in terms of accuracy, probability of detection (POD), and false detection rate. The statistics of cloud detection for CSDD were generally comparable to those of AERONET for direct data, and the POD for diffuse data was improved to the level of direct data.