Why can't models simulate mixed-phase clouds correctly?

Motivated by the importance of clouds for weather and climate due to their radiative impact, this thesis addresses the current poor representation of thin, stratiform mixed-phase clouds by state-of-the-art numerical weather prediction (NWP) models. Due to the supercooled liquid water present at cold temperatures near the cloud top, mixed-phase clouds strongly influence the amount of solar radiation reaching the surface and have a net cooling effect on the climate. Supercooled liquid water content is underestimated, by a factor of 2 or more, in all 5 NWP models tested and ERA-Interim when compared with ground-based remote-sensing observations of mixed-phase clouds. The ice water content is better predicted, but ice cloud fraction is underestimated. A new ice cloud fraction parameterization is developed to correct this bias, based on radar observations. EMPIRE, a new high-resolution single column model is developed and used to determine the most important processes for maintaining mixed-phase clouds. It is found that altering the model specification of ice particles (size, fall speed, concentration or habit) affected the liquid water content and most also affect the ice water content. A key reason why models underestimate liquid is an overestimate of ice growth rate but parameterizing N0 as a function of ice water content based on aircraft measurements leads to a significant improvement. A strong sensitivity to the model vertical resolution is identified. At coarse resolutions EMPIRE produces less than 2% of the liquid water content of high resolution simulations. This is because the coarse resolution model does not resolve the vertical profile of temperature, liquid and ice near the cloud top. By adding a parameterization of the vertical structure of the upper part of the cloud, the resolution sensitivity is largely removed suggesting that the implementation of such a parameterization in NWP models could improve their simulation of mixed-phase clouds.