4:45 PM - 5:00 PM
[ACG41-30] Optical and microphysical properties of water cloud retrieved based on the three-dimensional radiative transfer
Keywords:cloud, radiative transfer, GCOM-C
Cloud properties like cloud optical thickness (COT) and cloud-droplet effective radius (CER) use look-up tables based on one-dimensional radiative transfer (1DRT) calculations, assuming full cloud coverage and homogeneity within individual pixels. However, the satellite image pixel is often partly covered by cloud, and natural clouds exhibit complex heterogeneity across spatial scales. Subpixel cloud cover and three-dimensional radiative transfer (3DRT) effects largely influence cloud property retrieval. The 3D radiative effects depend on many parameters including sun-cloud-satellite geometry, subpixel cloud fraction (SCF), in-cloud inhomogeneity, cloud geometrical thickness, COT, CER, aerosols, and surface reflection, and one of the challenges is to represent such complicated relationships.
For cloud retrieval based on 3DRT, multi-pixel methods look promising. The multi-pixel method retrieves a multi-pixel cloud property array from a multi-pixel observed-radiance array. Convolutional neural networks capture spectral and spatial features and naturally trace the complicated 3D radiative effects and cloud characteristics. We have developed the 3D Radiative Effect Correction (3REC) method, in which the 3D radiative effect (defined as the 3D–1D difference in radiance) is estimated from observed radiance, simultaneously estimating the SCF. COT and CER are retrieved from 1DRT radiance after the 3REC.
We have applied the 3REC method to the Second-generation Global Imager (SGLI) onboard the GCOM-C satellite, analyzing seasonal and global distributions. It revealed systematic biases in conventional 1DRT retrieval of water cloud properties over ocean. 3DRT-based global mean COT and CER tend to be larger and smaller, respectively, than their 1DRT counterparts. This is mainly because of partial cloud cover. Water cloud fraction largely decreases, particularly over subtropical oceanic regions primarily covered by small cumulus clouds. The 3DRT–1DRT differences depend on the average SCF and cloud inhomogeneity. 1DRT-based retrieval substantially underestimates cloud inhomogeneity due to radiative smoothing. These results suggest that actual clouds exhibit greater complexity and heterogeneity than previously recognized in existing global cloud products. This enhanced understanding of cloud properties contributes to more reliable atmospheric modeling capabilities. Ongoing work includes validation of the SCF by high-resolution satellite measurement, inter-product comparison of cloud water path, and testing the radiance closure.
For cloud retrieval based on 3DRT, multi-pixel methods look promising. The multi-pixel method retrieves a multi-pixel cloud property array from a multi-pixel observed-radiance array. Convolutional neural networks capture spectral and spatial features and naturally trace the complicated 3D radiative effects and cloud characteristics. We have developed the 3D Radiative Effect Correction (3REC) method, in which the 3D radiative effect (defined as the 3D–1D difference in radiance) is estimated from observed radiance, simultaneously estimating the SCF. COT and CER are retrieved from 1DRT radiance after the 3REC.
We have applied the 3REC method to the Second-generation Global Imager (SGLI) onboard the GCOM-C satellite, analyzing seasonal and global distributions. It revealed systematic biases in conventional 1DRT retrieval of water cloud properties over ocean. 3DRT-based global mean COT and CER tend to be larger and smaller, respectively, than their 1DRT counterparts. This is mainly because of partial cloud cover. Water cloud fraction largely decreases, particularly over subtropical oceanic regions primarily covered by small cumulus clouds. The 3DRT–1DRT differences depend on the average SCF and cloud inhomogeneity. 1DRT-based retrieval substantially underestimates cloud inhomogeneity due to radiative smoothing. These results suggest that actual clouds exhibit greater complexity and heterogeneity than previously recognized in existing global cloud products. This enhanced understanding of cloud properties contributes to more reliable atmospheric modeling capabilities. Ongoing work includes validation of the SCF by high-resolution satellite measurement, inter-product comparison of cloud water path, and testing the radiance closure.