Influence of the turbulence and convection parameterization on convective initiation in kilometer-scale simulations

Thursday, 18 December 2014
Steven Böing1,2, Jürg Schmidli1,2, Oliver Fuhrer3 and Christoph Schär1, (1)ETH Swiss Federal Institute of Technology Zurich, Zurich, Switzerland, (2)Center for Climate Systems Modeling (C2SM), Zurich, Switzerland, (3)Federal Office of Meteorology and Climatology MeteoSwiss, Zurich, Switzerland
The next generation of numerical weather prediction models will run with a grid spacing of about 1 kilometer. Deep convection is coarsely resolved at this grid spacing, but shallow clouds and boundary layer turbulence are too small to be resolved. Previous experience at MeteoSwiss and other institutions has shown significant biases at this resolution: shallow clouds tend to be underrepresented and a too sudden onset of deep convection occurs.

We use idealized simulations to investigate the ability of a kilometer-scale NWP model (COSMO) to accurately represent the initiation and development of moist convection. In particular, we look into the role of boundary layer turbulence, horizontal mixing and the convection scheme. We do this by systematically exploring a number of case studies (GCSS cases as well as two previously documented cases of convection over topography), using both the kilometer-scale resolution setup and an LES setup with higher resolution. Using the latter setup as a reference, we aim to identify weaknesses and suggest improvements in the formulation of moist convection and turbulence. We look into the spatial distribution of turbulence, clouds and precipitation and use conditionally sampled statistics to investigate the properties of clouds and flow patterns.

In order to represent non-precipitating shallow convection, the use of a subgrid-scale convection scheme is necessary in the kilometer-scale simulations. The standard convection scheme in COSMO captures the onset of clouds, but the liquid water content and mass-flux near cloud top are overestimated. This behavior can be improved, however, with modifications in the closure, convective triggering and the entrainment/detrainment formulation.

The kilometer-scale model performs surprisingly similar to LES simulations for phenomena which it only coarsely resolves, such as convective cold pools, slope flows and marginally resolved updrafts. We further discuss the effect of the horizontal mixing formulation and turbulence parameterization on e.g. the mass-flux, the strength of updrafts and the precipitation rate.