Optimizing Ice Cloud Representations with a Mixed Super-ellipsoidal Scheme using Polarized Radiance Observations

Accurate representation of ice crystal shapes is critical for simulating polarized radiance and improving radiative transfer in climate and weather models. In this study, we systematically developed a database of 1071 super-ellipsoidal ice crystals covering a broad range of aspect ratios, roundness, and surface roughness, and used POLDER-3/PARASOL polarized radiance observations from 2009 to identify optimal shapes and roughness parameters and obtain a mixed super-ellipsoidal scheme across latitude bands. Based on these results, we evaluated a mixed super-ellipsoidal scheme against traditional single-habit particles from the TAMUice2016 database (e.g., Plate, 5-Plate, 8-Column) using 2012 observations. Conventional single-habit models exhibit varying performance across latitude bands, often underestimating or overestimating polarized radiance in specific regions. In contrast, the mixed super-ellipsoidal models demonstrate consistently higher correlations and lower RMSE, with robust performance across wavelengths and latitudes. These results indicate that observationally constrained mixed super-ellipsoidal scheme provides a flexible framework for ice cloud representation in radiative transfer simulations, paving the way for improved studies in microphysics and polarization-based retrievals.