Probabilistic Retrieval of All-Day Overlapping Cloud Microphysical Properties

Globally, approximately 25% of clouds are considered overlapping, which are critical to the Earth’s radiation budget and the evolution of weather systems. However, traditional physical methods fail to retrieve all-day overlapping cloud microphysical properties from passive remote-sensing satellites due to their complex vertical structure, which remains an ongoing challenge. To address this, we propose a probabilistic deep learning model to retrieve overlapping cloud microphysical properties from the Aqua satellite’s thermal infrared channels and integrate this algorithm into DaYu CLoud Analysis System (DaYu-CLAS), with the model referred to as Overlap-CloudDiff. The results show that DaYu-CLAS excels in cloud-phase classification with an overall accuracy of 88.18% and a multi-layer cloud precision rate of 76.08% during the daytime, while the retrieval results for upper-layer ice clouds yield RMSEs of 6.66 µm for cloud effective radius (CER) and 2.78 for cloud optical thickness (COT), and lower-layer water clouds with RMSEs of 19.60 µm (CER) and 11.76 (COT). DaYu-CLAS outperforms the deterministic model with the same input during the daytime, particularly in capturing probabilistic distributions. Additionally, generating diverse ensemble members helps the model estimate uncertainty, enhancing retrieval reliability.