Concept of a one-Moment Ice-Cloud Scheme avoiding Saturation Adjustment

Abstract

Persistent contrails and contrail induced cirrus are responsible for a large fraction of global aviation’s effective radiative forcing. The persistence of contrails could be avoided by rerouteing aircraft around ice supersaturated regions in the upper troposphere. Unfortunately, the forecasting of these regions is challenging, where one of the problems are inaccurate ice cloud parameterisations in numerical weather prediction models, many of them making use of saturation adjustment. In this approximation, the air inside ice clouds is assumed to be exactly saturated, even though supersaturated water vapour is frequently observed inside cirrus. In this thesis, the concept of a new one-moment ice-cloud scheme avoiding the use of saturation adjustment is proposed. A simple humidity distribution for newly generated cloud parts is derived with the help of a stochastic box model that treats one model grid box as an ensemble of a large number of individual air parcels. This distribution is used to model the decay of in-cloud supersaturation explicitly in two alternative parameterisations. One introduces the in-cloud humidity as an additional prognostic variable in order to evolve it in time. The other one adjusts the in-cloud humidity to its equilibrium value in some cases, but in turn does not need an additional prognostic variable. The new parameterisations are then tested against a parameterisation that uses saturation adjustment, where the stochastic box model serves as a benchmark. It is shown that saturation adjustment underestimates humidity both shortly after nucleation, when the actual cloud is still highly supersaturated, and also in aged cirrus if cooling proceeds, as the actual cloud remains in a slightly supersaturated state in this case. In contrast, the new parameterisations show great improvement compared to saturation adjustment, as they stay closer to the stochastic box model in any considered case. It is concluded that the new parameterisations are promising, but need further testing in more realistic frameworks.