Sensitivity of simulated deep convection to a stochastic ice microphysics framework



Stanford, McKenna — University of Utah
Morrison, Hugh Clifton — UCAR

Area of research

Cloud Processes

Journal Reference

Stanford M, H Morrison, A Varble, J Berner, W Wu, G McFarquhar, and J Milbrandt. 2019. "Sensitivity of Simulated Deep Convection to a Stochastic Ice Microphysics Framework." Journal of Advances in Modeling Earth Systems, 11(11), 10.1029/2019MS001730.


Observations show that atmospheric ice particle properties vary at different times and locations even for similar cloud conditions and thermodynamic environments. Yet, the way ice particles are represented in weather and climate models typically does not account for this variability. Herein, we develop a stochastic ice microphysics parameterization that includes spatiotemporal variability of parameters controlling relationships between ice particle mass, fallspeed, and size, and investigate the impact of this variability on simulated convective systems.


The impact of ice particle property natural variability on simulated cloud systems has not yet been thoroughly investigated. We introduce a novel framework to account for this variability and show that cloud radiative forcing is sensitive to variability in the ice mass-size relationship while precipitation rate is sensitive to variability in the ice fallspeed-size relationship. These sensitivities exceed those from applying small random perturbations to the initial temperature field. Considering this variability may therefore have important applications to properly representing parameter uncertainty in numerical weather forecasts and climate prediction. 


A stochastic framework for ice single-particle properties is applied to the Predicted Particle Properties (P3) microphysics scheme within the Weather Research and Forecasting (WRF) model. In separate suites of simulations, we allow for variability in (1) parameters of the mass-dimension (m-D) relationship (m=aDb) of unrimed and partially rimed ice and (2) the fallspeed-dimension (V-D) relationship of all forms of ice. This scheme allows these parameters to vary in time and space according to a prescribed spatial and temporal autocorrelation scale. The m-D scheme samples from an observed distribution of m-D parameters retrieved from aircraft during ARM’s Midlatitude Continental Convective Clouds Experiment (MC3E).

In simulations of two MC3E cases, the relationship between cloud optical depth and ice water path (IWP) through anvil regions of mesoscale convective systems (MCSs) is shown to be highly sensitive to different fixed a-b parameter pairs (i.e. a-b parameters are fixed for a given simulation but varied among ensemble members, Figure 1a). The stochastic scheme, tested at short, mid-range, and long spatiotemporal autocorrelation scales, is shown to also produce a variable optical depth-IWP relationship (Figure 1b), with the largest variability produced by the longest spatiotemporal autocorrelation scale. Thus, the stochastic scheme produces variable cloud radiative forcing for anvils with the same integrated condensate mass content. This optical depth-IWP variability is not produced by an ensemble with perturbed initial and lateral boundary conditions (ICBC, Figure 1c) nor by an ensemble with white noise added to the initial temperature field (White Noise, Figure 1d). However, spread in purely optical depth distributions is larger for the ICBC ensemble compared to the stochastic simulations due to the ICBC ensemble producing substantial variability in MCS structure and evolution. The stochastic V-D scheme produces variable rain rate distributions and accumulated precipitation that exceeds the spread produced by the White Noise ensemble but is smaller than that produced by the ICBC ensemble or by an ensemble employing constant, fixed-parameter changes to the V-D relationship.