Investigation of the sensitivity of ice cloud radiative properties to the parameterization scheme of ice crystal habit distributions in NCAR Community Atmosphere Model 3.0

The bulk radiative properties of clouds directly modulate the cloud-radiative forcing and therefore directly influence the role of clouds in the climate. In most current climate models, the radiative properties are parameterized as a function of the microphysical and single scattering properties of the cloud particles. As many recent studies reveal, current parameterization schemes result in a large variation of the cloud radiative properties and feedbacks in different climate models. In turn, clouds and their interactions with radiation remain as one of the major sources of uncertainty in climate studies. In this research, we investigate the influence of the ice crystal habit distribution on the parameterization of the ice cloud radiative properties in the NCAR Community Atmosphere Model 3.0 (CAM3.0). We compare cloud radiative properties between a parameterization scheme in which all ice crystals are assumed to be solely hexagonal columns, and several modified versions using different ice particle habit distributions that include different particle habits such as aggregates and bullet rosettes. The comparison effort focuses on two steps in the parameterization process. In the first step, the cloud radiative properties are parameterized as functions of the effective particle size. For this step, we find that the impact of changing the habit distribution is minimal and can be neglected. For the second step, the effective particle size is parameterized as a function of cloud temperature. We find that the impact of the assumed habit distribution cannot be neglected. For a given cloud temperature, our new parameterization scheme tends to generate smaller effective particle sizes than those predicted from the current scheme. For a given effective particle size, the shortwave reflectivity from our parameterization is found to be about 20% larger when ice water path is smaller than 50 (g/m^-2). However, the difference in the cloud temperature domain becomes statistically unimportant in comparison with the variance caused by the uncertainty associated with the particle size distribution. While the difference in the longwave emissivity due to the difference of the habit distribution rarely exceeds 10% for a given effective particle size and ice water path, differences as functions of cloud temperature become considerably larger and comparable to the variance caused by the uncertainty of the particle size distributions.