Waves to Weather

Breadcrumb Navigation


B1 - Microphysical uncertainties in hailstorms using statistical emulation and stochastic cloud physics

Project leaders: Prof. Dr. Corinna Hoose, Dr. Annette Miltenberger, PD. Dr. Michael Kunz

Other researchers: Lena Frey (PostDoc), Patrick Kuntze (PhD student)

Many European high impact weather events, such as hailstorms, are associated with low local predictability, as is common for deep moist convection. This is partly caused by forecast errors on the synoptic scale, which control the predictability of convective initiation, but also by various sources of uncertainty on small scales. Both effects determine the location and trajectories of convective cells, precipitation formation pathways, and, finally, the critical forecast variables precipitation and hail, which are the focus of the project. In Phase 1, we have used statistical emulation of idealized simulations to quantify the contributions of uncertainty in environmental and microphysical parameters to the variance of the output. In addition to cloud condensation nuclei and ice nucleating particle concentrations, the type and size of warm bubble and cold pools used as artificial convection triggers was shown to be particularly important for the simulated clouds. In addition, it was discovered that the numerical treatment of process splitting in the COSMO model leads to an undesired time step dependence, adding a significant error source for precipitation and hail also in real-case simulations. With a different treatment of process splitting, the timestep dependence was removed.

In Phase 2, we will focus on real-case simulations of organized convection associated with large hail that have a high potential of error growth to larger scale meteorological conditions. We will use the model ICON-ART on a cloud resolving scale with an interactive treatment of aerosol and cloud microphysics. A simplified version of ICON-ART will be developed to meet the computational requirements of large ensemble simulations. From an ensemble of sensitivity experiments with varying important microphysical parameters (e.g., those related to the parameterization of ice nucleation, fall speed, saturation adjustment), we will develop an emulator setup for selected cases of organized convection in order to systematically quantify the relative contributions of these factors to the uncertainty of the forecast variables of interest. Furthermore, the uncertainty in cloud microphysics and the related latent heat release can feed back onto the cloud field structure and the larger scale flow evolution. We will complement the sensitivity ensemble by simulations using small random perturbations to the boundary layer temperature, moisture, and wind profiles and simulations with stochastically selected cloud microphysical parameters in each timestep. By analyzing the error growth in the various simulations, the systematic impacts of cloud microphysical uncertainties above random perturbations of ambient conditions will be assessed, and thus their role for quantitative forecasting.