Development and Validation of Cloud Screening Algorithm for Gosat/cai

A new and versatile cloud screening algorithm (CLoud and Aerosol Unbiased Decision Intellectual Algorithm, CLAUIDA) bas been developed for multi-channel imagers on board satellites, and is applied to the operational procedure for GOSAT/CAI (a 4-channel imager) cloud screening. CLAUDIA consists of many threshold tests to deal with various cloud types on various surfaces, and is designed to aim at neutral, no-biased cloud screening. Considering that the cloud optical thickness in nature continuously changes and the border between clouds and clear sky is often indistinct, clear confidence level, which is a value to represent certainty of clear or cloud discrimination, is introduced to avoid distinct classification. In addition, threshold tests in the algorithm are categorized into two groups according to their characteristics. A threshold test in group 1 tends to mistake clear sky for cloud, whereas a threshold test in group 2 tends to incorrectly classify cloudy areas as clear. The representative clear confidence level for group 1 is calculated to be cloud conservative, and that for group 2 is to be clear conservative. Consequently, the overall clear confidence level results in neutral. Visual inspection for several examples of cloud screening for CAI L1B data proves that the results over ocean without sunglint and desert are roughly correct. For validation, the results of clear confidence level derived from GOSAT/CAI data are compared to that from Aqua/MODIS data, to which the same cloud screening algorithm is applied. The validation suggests that discrimination over ocean without sunglint is almost in good agreement with each other. Over land, there are some types of the surface where GOSAT/CAI tends to result in cloudy tendency compared to MODIS, because of the lack of effective threshold tests. * Corresponding author. [email protected].