This paper describes the basis of an improved ice nucleation scheme for use in a cloud-resolving model (CRM) with bulk-water microphysics. The aim of this scheme is to capture essential characteristics of the evolution of mixed-phase and cirrus clouds which are due to the limited availability of ice nuclei (IN) whilst retaining the relative computational simplicity of bulk-water microphysics.

Current deposition ice nucleation schemes sometimes assume the ice crystal concentration does not fall below the number of active IN given by Meyer's (or Fletcher's) equation. This effectively leads to an infinite supply of ice crystals, because when ice crystal fall-out removes activated IN from the grid-box, during a subsequent time step the effective IN concentration may be restored by changes to the temperature or ice saturation ratio. In such schemes, the maintenance of a stable supercooled cloud is highly dependent on parameter values of the IN scheme. Prognostic deposition IN enables ice crystals to be initiated once the grid-box goes above a specific ice supersaturation, and only replenished by entrainment from surrounding unactivated regions. Similarly, prognostic contact aerosol can be used to give a limited supply of contact-freezing initiated ice crystals. The aerosol collection rate includes Browniain motion and thermophoretic and diffusiophoretic forcing. The thermophoretic forcing significantly enhances collection during droplet evaporation. It is commonly observed that ice initiation tends to occur in downdrafts of wave clouds. This can be modelled by prognostic evaporation IN, which are a special subset of CCN that activate when the droplet evaporates.

Prognostic CCN aerosol concentration enables the modelling of homogeneous freezing of supercooled haze particles in a water-subsaturated environment. Parcel model studies show that only some of the haze particles can freeze before their ice depositional growth depletes the environmental water vapour and the remaining haze particles become too concentrated to freeze. The homogeneous freezing rate used in the parcel model calculates an effective droplet temperature (which depends on droplet molality), and for highly concentrated haze particles this temperature is high and freezing is inhibited. The actual fraction of haze particles that freeze is found to be a function of ambient temperature, relative humidity and updraft velocity.

The simulation is the Met Office large eddy model (LEM) version 2.3. The LEM is a cloud resolving model with three-phase microphysics parametrizations and interactive radiation code. Moist processes are represented by prognostic variables for water vapour, liquid water and ice water mixing ratios. The thermodynamic variable is the potential temperature. The model is initialised with water vapour, liquid water and temperature profiles to give a 500m thick cloud layer. IN and cloud condensation nuclei (CCN) aerosol profiles are also initially defined and are subsequently depleted by the ice initiation processes. Further ice cannot be initiated until radiatively driven turbulence at the cloud top and bottom entrains unused aerosols into the cloud layer.

Results are presented from simulations of an idealised altostratus cloud layer with various warm heterogeneous ice initiation processes active, and a cirrus cloud layer with only homogeneous freezing active.