Key points are not available for this paper at this time.
Accurate cloud detection is critical for spaceborne photon-counting light detection and ranging (LiDAR) systems like ICESat-2. This study presents the Normalized Difference Cloud Index (NDCI), a new radiometric algorithm for high-resolution daytime cloud identification. By conceptualizing the LiDAR system as an ultra-narrowband radiometer, the NDCI synergistically integrates active laser backscatter and passive solar background radiance. This approach amplifies atmospheric scattering signatures while suppressing interference from heterogeneous surface reflectance, providing a physically consistent framework that requires only standard ATL03 photon data. Validation across cryospheric, arid, and coastal regions demonstrates robust performance, with F 1 scores consistently exceeding 0.81. Through regional threshold optimization via receiver operating characteristic (ROC) analysis, the algorithm achieves an F 1 score of 0.97 and an area under the curve (AUC) of 0.99 over the Greenland ice sheet. A key advantage of the NDCI is its applicability to ICESat-2 weak beams, which currently lack dedicated operational cloud flags, thereby addressing spatial inconsistencies in multi-beam atmospheric screening. The NDCI provides a potential approach for cloud filtering in existing and future spaceborne photon-counting LiDAR data processing. Nevertheless, the NDCI algorithm has certain limitations. Due to its foundation on the Lambertian reflection model, its detection sensitivity to optically thin clouds is diminished over calm water surfaces under specular reflection. Additionally, the algorithm relies on accurate surface signal photon identification, and signal extraction errors in complex, rugged terrain may give rise to false positives in cloud detection.
Zhang et al. (Thu,) studied this question.