Improvement of CALIOP Cloud Masking Algorithms for Better Estimation of Dust Extinction Profiles

Abstract

Mineral dust suspended in the atmosphere affects the Earth’s radiation budget. To accurately predict the effect of dust on the climate system, information regarding its extinction profiles is needed. The Cloud-Aerosol Lidar with Orthogonal Polarization (CALIOP) instrument on the Cloud-Aerosol Lidar and Infrared Pathfinder Satellite Observation (CALIPSO) satellite has enabled the global observation of the vertical distributions of aerosols and clouds since June 2006. To correctly retrieve extinction coefficients from CALIOP signals, the lidar-observed layers must be classified into aerosols or clouds. The cloud masking algorithms of CALIOP should be improved since the cloud mask products occasionally misclassify dense dust as clouds. This study attempts to discriminate misclassified clouds from the CALIOP cloud mask with a discriminant analysis. The training data are collected by tests with the CloudSat cloud mask, the Moderate Resolution Imaging Spectroradiometer (MODIS) cloud mask, and relative humidity. Discrimination of dust from clouds is successful in cases over land and water surfaces during the daytime and nighttime. In contrast, the discrimination model of previous studies was inadequate during the nighttime since training data were not collected during the nighttime. The accuracy rate of the linear discriminant function classification is 91.7 % for misclassified clouds. The cloud mask is most frequently misclassified in the Taklimakan Desert. The proportion of misclassified clouds to the observed dust is ∼34.6 % (below 2 km) in the desert. Comparison of our results with CALIOP level 3 products indicates that the extinction profile using the improved cloud mask is at most twice larger than that of CALIOP level 3 products. This study suggests that the smaller extinction coefficients of CALIOP level 3 products are mainly caused by misclassification of dust as clouds in the vertical feature mask.