An investigation into the performance of four cloud droplet activation parameterisations
Cloud droplet number concentration prediction is central to large-scale weather and climate modelling. The benchmark cloud parcel model calculation of aerosol particle growth and activation, by diffusion of vapour to aerosol particles in a rising parcel of air experiencing adiabatic expansion, is too computationally expensive for use in large-scale global models. Therefore the process of activation of aerosol particles into cloud droplets is parameterised with an aim to strike the optimum balance between numerical expense and accuracy. We present a detailed systematic evaluation of three cloud droplet activation parameterisations that are widely used in large-scale models and one recent update. In all cases, it is found that there is a tendency to overestimate the fraction of activated aerosol particles when the aerosol particle "median diameter" is large (between 250 and 2000 nm) in a single lognormal mode simulation. This is due to an infinite "effective simulation time" of the parameterisations compared to a prescribed simulation time in the parcel model. This problem arises in the parameterisations because it is assumed that a parcel of air rises to the altitude where maximum supersaturation occurs, regardless of whether this altitude is above the cloud top. Such behaviour is problematic because, in some cases, large aerosol can completely suppress the activation of drops. In some cases when the "median diameter" is small (between 5 and 250 nm) in a single lognormal mode the fraction of activated drops is underestimated by the parameterisations. Secondly, it is found that in dual-mode cases there is a systematic tendency towards underestimation of the fraction of activated drops, which is due to the methods used by the parameterisations to approximate the sink of water vapour.