ALBERT

Description: The evolution of a hydrometeor ensemble („cloud“ ) can be described using a balance equation for its size spectrum. In numerical weather prediction or climate models, however, this approach is too time consuming. It is therefore necessary to capture, if only approximately, the on-going microphysical processes in a cloud using a parame- terised form of modelling. The parameterisation of sedimentation alone is already a demanding task. If its standard form is used in a two-moment scheme, the mean mass of the hydrometeors will be too large for a cloud physics context. Existing approaches try to avoid excessively large mean masses by altering dynamically the parameterisation assumptions in the running model. In this work, a new fundamental approach is presented: the assumption of a spectrum containing particles of all sizes is replaced by its truncation at a particle size realistic in cloud physics. The calculation of integrals over the truncated spectrum requires a new technique, which demands higher computational effort. In this work, comparisons of the new parameterisation with already existing parameterisations are made while changing the initial conditions and the advection scheme. With the new method, the parameterisation error is reduced by up to 50 %. The new paramerisation also constitutes a great improvement compared to the standard form when transferred to modelling an ensemble of solid hydrometeors. This is illustrated using a particle type commonly found in polar regions. Furthermore, for the first time a B-distribution is used as a basis for a cloud microphysics parameterisation. Its domain of definition is bounded by construction. This distribution, however, appears not suitable for use in two-moment schemes, because one free parameter has to be derived from model data. Extending the sedimentation model with drop collisions and starting with a cloud droplet spectrum, it is not possible to judge the quality of the sedimentation paramete- risations, because the coagulation rates are dominated by the choice of the sedimetation velocity for small droplets. If the initial spectrum already contains a sufficient number of raindrops, however, application of the new method again reduces the parameterisation error by up to 50 %.