CLOUDET: A Cloud Detection and Estimation Algorithm for Passive Microwave Imagers and Sounders Aided by Naïve Bayes Classifier and Multilayer Perceptron

CLOUDET, a cloud detection and estimation algorithm for passive microwave imagers and sounders is presented. CLOUDET is based on a naïve Bayes classifier and multilayer perceptron. It is applied to the special sensor microwave imager/sounder (SSMIS), and the ECMWF integrated forecast system (IFS) cloud liquid water information has been used to train the algorithm. CLOUDET is applicable to both ocean and land-surface types. CLOUDET has been developed and evaluated by employing two different groups of radiometric information, namely, the humidity channels near 183 GHz (humidity algorithm) and the window channels between 19 and 91 GHz (window algorithm). It has been revealed that both humidity and window algorithms can detect cloudy scenes over ocean at a confidence level of more than 90% (80% over land). The analysis indicates that the humidity algorithm has a better ability in detecting cloudy scenes over ocean than the window algorithm (CSI = 0.98 vs. CSI = 0.93). The opposite is true over land-surface type, revealing a CSI of 0.85 by humidity algorithm as opposed to CSI of 0.88 by window algorithm. The estimation of cloud by the CLOUDET algorithm has also been very promising during the validation effort. In particular, the correlation coefficient obtained over ocean through the use of the window algorithm is 0.70 (MAE 0.04 mm and RMSE 0.09 mm). The presented algorithm CLOUDET can be served as a stand-alone tool to reject and identify the cloudy scenes as well as to estimate the cloud liquid water path amount prior to assimilating the radiances into numerical weather prediction model.