In the treatment of ice cloud radiative properties, two approaches have been taken: (1) parameterization of exact results from electrodynamic scattering theory in terms of bulk microphysical properties (e.g. IWC and effective size) and (2) parameterization of electrodynamic scattering theory itself in terms of explicit microphysical properties (e.g. size distribution parameters, ice crystal mass and area relationships). The strength of the first approach lies in the single particle calculations, while the strength of the second approach lies in the explicit coupling of microphysical and radiative properties. This study tests the weakest aspect of the second approach: the accuracy of ice crystal single scattering properties.

The radiation scheme tested here was developed by Mitchell and colleagues. Extinction and absorption efficiencies (Qext and Qabs) are determined from a modified version of the anomalous diffraction approximation (ADA), which includes the processes of internal reflection and refraction, and photon tunneling. Although it is used in GCMs and satellite retrieval schemes, it has never been rigorously tested against electrodynamic scattering theory applied to ice crystals. This study first tests the schemes ability to predict the Qext of a laboratory ice cloud over a wavelength range of 2-18 microns, relative to Qext measured via FTIR. The size distribution and aspect ratios of the hexagonal columns comprising the ice cloud were then used in T-matrix calculations to predict Qext and Qabs (of the size distribution) from 2-18 microns wavelength. The modified ADA predictions of Qext and Qabs were compared with the T-matrix solutions for Qext and Qabs, as well as the measured values of Qext. This new implementation of T-matrix incorporates the exact geometry of hexagonal columns, without approximating them as spheroids or circular cylinders.

Using a photon tunneling factor of 0.60, the mean difference between the Qext predicted from modified ADA and the measured Qext was 3.5%, with the same agreement (3.5%) obtained for the T-matrix calculations. The excellent agreement between T-matrix and measurements provides confidence that Qabs predicted by T-matrix can be used to test the accuracy of Qabs predicted by modified ADA. Over the 2-18 micron wavelength range, the mean error was 5.0% and the maximum error was 15%. Using the Baran-Havemann tunneling parameterization with modified ADA resulted in a maximum error in Qabs of 13%. Comparing modified ADA Qext with T-matrix, the mean error was 4.1% and the maximum error was 8%.