Modeling Optical Properties of Non‐Cubical Sea‐Salt Particles.

Dry sodium chloride forms cubic crystals, while marine aerosol particles often display more or less irregular deviations from this ideal form. In this study, three non‐ideal cuboidal and octahedral model geometries are investigated. Superellipsoids are tested as a model to simulate the linear backscatter depolarization ratio and the extinction‐to‐backscatter ratio. Gaussian random cubes as well as convex polyhedra are investigated as possible model candidates to quantify the error introduced by simplified model geometries, such as superellipsoids. Uncertainties in the real and imaginary part of the refractive index are studied, and their effect on the optical properties is compared to that caused by morphological variations. Optical calculations were performed at a wavelength of 532 nm using the discrete dipole approximation and the T‐matrix method. The considered size range is representative for marine aerosol generated at low to moderate wind speeds. It is found that cuboidal superellipsoids predict depolarization and extinction‐to‐backscatter ratios that are consistent with observations. On the other hand, octahedral superellipsoids strongly overestimate the depolarization ratio. Gaussian random surface perturbations result in a positive shift of the depolarization ratio compared to cuboidal superellipsoids. By contrast, convex polyhedra yield results that more or less randomly scatter about those of regular cubes. Thus, convex polyhedra are a promising candidate for modeling random errors, while Gaussian random cubes are not. Uncertainties in the refractive index result in perturbations of the depolarization and extinction‐to‐backscatter ratio that are of comparable magnitude as those caused by perturbations of the geometry. Key Points: The depolarization and lidar ratio of dried sea‐salt aerosol are well represented by cuboidal superellipsoids over a representative size rangeConvex polyhedra provide a randomization model suitable for simulating random model errorsModel geometries strongly deviating from cubes, such as octahedra and Gaussian random cubes, overestimate the depolarization ratio [ABSTRACT FROM AUTHOR]