Improving mixed-phase cloud representation in weather and climate models

Stratiform mixed-phase clouds are radiatively important and have the potential to be a large negative feedback in a warming climate. Yet, their existence is poorly understood and their representation in numerical weather and climate models is very crude. In many forecast and climate models the ratio of liquid and ice in mixed-phase clouds is solely dependent on temperature, with entirely liquid at temperatures above 0°C and entirely ice below a critical temperature, usually around -15°C. In contrast to this, aircraft and lidar observations have shown that supercooled liquid water exists at temperatures as cold as -35°C, often in thin layers 200-300 m deep at the top of ice clouds and can in some cases persist in excess of 12 hours. Around 20% of clouds between -10°C and -15°C are reported to contain layers of liquid water whilst comparison of models with lidar and radar observations show most models underestimate mid-level cloud occurrence.

A single column model has been developed to understand how interaction of complex microphysics leads to the formation of supercooled water and the relative importance of radiative cooling, ice nucleation mechanisms and turbulent mixing in maintaining liquid. Furthermore, we try to understand why current models are missing these clouds, whether due to poor resolution or missing processes. The model is run with forcing from ERA-interim over a single site where radar and lidar forward models are used to compare the model output with observations. The model is run over a large number of cases to statistically assess the sensitivity to all of the model settings.

Results demonstrate a number of interesting sensitivities, for example increasing ice fall speed results in less coexistence of cloud liquid and ice and therefore the Bergeron process is less active which results in a more persistent liquid layer. Details of this and other sensitivities of mixed-phase clouds will be presented together with some discussion on how models might better represent these clouds.