Evaluating a Stochastic Ice Microphysics Parameterization Scheme in Simulations of Deep Convection

Description

Parameterizing ice microphysics in cloud-scale models impacts important features of mesoscale convective systems such as cloud and precipitation structure, cloud radiative forcing, and the vertical redistribution of heat, momentum, and aerosols. However, properly simulating these features is challenging given the major simplifications of microphysical parameters and processes that are implemented in schemes. For example, most microphysics schemes employ relationships with empirically-derived constant parameters, despite observations suggesting considerable variability of such parameters. An intuitive method to account for the natural variability of ice particle properties is to employ a stochastic approach. Applying guidance from aircraft in situ observational retrievals of ice particle properties during the DOE ARM Midlatitude Continental Convective Clouds Experiment (MC3E), a new stochastic ice microphysics parameterization has been developed for the Weather Research and Forecasting (WRF) model. This approach stochastically varies the parameters of the mass-size (m-D) relationship (m = aD^b) of unrimed and partially-rimed ice in the Predicted Particle Properties (P3) scheme, allowing for the prefactor (a) and exponent (b) to be sampled stochastically from observationally-based distributions with a prescribed spatiotemporal autocorrelation scale. Additional simulations are analyzed in which the prefactor of the ice fallspeed-size relationship (V = cD^d) is stochastically varied. Results from two MC3E cases indicate that the stochastic m-D scheme alters distributions of anvil cloud optical depth, even for the same ice water path (IWP). The spread in optical depth as a function of IWP is also able to envelop the observed relationship, while this variability is virtually nonexistent in ensembles employing large-scale initial and boundary condition perturbations or in the stochastic fallspeed ensembles. It is subsequently shown that this spread impacts distributions of simulated cloud radiative effect. Rain rate distributions vary among ensemble members using the stochastic fallspeed scheme, which is seen only minimally using the stochastic m-D scheme. This variability in rain rate distributions is more evident in a simulated bow echo case compared to a synoptically-forced squall line case, suggesting that the stochastic fallspeed scheme more strongly impacts convective systems driven by mesoscale dynamics.