Bayesian Cloud Top Phase Determination for Meteosat Second Generation

Abstract. A comprehensive understanding of cloud thermodynamic phase is crucial for assessing the cloud radiative effect and is a prerequisite for remote sensing retrievals of microphysical cloud properties. While previous algorithms mainly distinguished between ice and liquid phases, there is now a growing awareness for the need to further distinguish between warm liquid, supercooled and mixed phase clouds. To address this need, we introduce a novel method named ProPS, which enables cloud detection and determination of cloud top phase using SEVIRI, the geostationary passive imager aboard Meteosat Second Generation. ProPS discriminates between clear sky, optically thin ice (TI), optically thick ice (IC), mixed phase (MP), supercooled liquid (SC), and warm liquid (LQ) clouds. Our method uses a Bayesian approach based on the cloud mask and cloud phase from the lidar-radar cloud product DARDAR. Validation of ProPS using six months of independent DARDAR data shows promising results: The daytime algorithm successfully detects 93 % of clouds and 86 % of clear sky pixels. In addition, for phase determination, ProPS accurately classifies 91 % of IC, 78 % of TI, 52 % of MP, 58 % of SC and 86 % of LQ, providing a significant improvement in accurate cloud top phase discrimination compared to traditional retrieval methods.