Frick, C. (Corresponding author)FZJ*

2012
Zürich, ETH

Report No.: 20624

Abstract: The prediction of snowfall is particularly challenging for numerical weather prediction models. At the beginning of this dissertation, an event of wet and heavy snowfall in NW Germany is investigated. Simulations reveal a low predictability of the event more than one day in advance. An appropriate simulation of the synoptic scale processes is as essential for an accurate snowfall prediction as an adequate microphysical representation of precipitation. Even for a perfect simulation of the dynamics and the atmospheric background conditions, the phase of surface precipitation is difficult to predict, especially under near-surface melting conditions. Additionally, a direct simulation of partially melted snowfall is not possible for the standard microphysical parameterization of the COSMO model. Therefore, the development and implementation of a new melting scheme including a new prognostic variable, the meltwater of snow, is presented. The introduced bulk microphysical parameterization allows an internal mixing of water and ice in the snow category of the COSMO model. Liquid water fraction is assumed to vary with the size in the represented snowflake ensemble, leading to a faster melting of smaller snowflakes compared to larger ones. For a first validation of the effects of the new melting scheme, two wet snowfall events are simulated using the standard and the new parameterization. Approximately one third of the surface precipitation is predicted as snowfall for both schemes with 1 to 2 percent higher snow fractions for the new melting scheme. A categorization of surface snowfall in dry and wet snow for the new parameterization reveals that approximately one third of the surface snowfall is predicted to be wet snow. The modified melting process of the new parameterization leads to an onset of rain at lower altitudes and a deeper vertical penetration of snow into the potential melting layer. The evolution of the precipitation phase from snow to rain is decelerated especially for snow fractions below 40 percent and liquid to ice ratios larger than 1. Overall, the new melting scheme slows down the melting process, slightly enhances surface snow fraction, and allows a direct prediction of wet snowfall at the surface. The new melting scheme also allows a realistic simulation of a melting layer of wet snow, which is related to the “bright band” in radar imagery.