Multiscale Aerosol–Cloud Classification Algorithm of the DQ-1/ACDL

Abstract

Spaceborne LiDAR is a unique tool for profiling clouds and aerosols globally. Following the retirement of the Cloud–Aerosol LiDAR and Infrared Pathfinder Satellite Observations (CALIPSO) mission, DQ-1, carrying the aerosol and carbon dioxide detection LiDAR (ACDL), now fills this observational gap. However, traditional single-profile aerosol–cloud classification algorithms often misclassify overlapping layers and generate “striping artifacts,” characterized by unrealistic horizontal discontinuities between adjacent profiles, which significantly degrade data quality and usability. To address these issues, we propose a multiscale classification algorithm that partitions LiDAR scenes using dynamic windows (ranging from $1\times 1$ to $8\times 8$ ) based on signal-to-noise ratio thresholds and classifies layers using particle backscatter coefficients rather than total attenuated backscatter (TAB). This approach effectively reduces striping artifacts and improves aerosol–cloud layer identification, achieving a 74.6% reduction in horizontal discontinuity compared with the cloud–aerosol LiDAR with orthogonal polarization (CALIOP) scene classification algorithm (SCA). Furthermore, comparisons with Moderate Resolution Imaging Spectroradiometer (MODIS), CALIOP, and the Modern-Era Retrospective Analysis for Research and Applications, Version 2 (MERRA-2) for dust distribution, cloud-top height (CTH), and cloud phase show general consistency and physically reasonable behavior across diverse regions, supporting the applicability of the ACDL products. The proposed framework ensures the quality of ACDL datasets and offers a transferable approach for optimizing layer classification in other spaceborne LiDAR missions.