Evaluating the Turbulence Representation in a Numerical Weather Prediction Model over Mountainous Terrain

The horizontal resolution of numerical weather prediction models (NWP) keeps decreasing towards the hectometric range. However, although topography, land-use, and other static parameters might be better resolved, the performance of physical parameterizations also has to be evaluated and assessed.
One of the most challenging environments for modern NWP models is mountainous topography - namely because, among other issues, turbulence parameterizations were developed for horizontally homogeneous and flat terrain. This assumption is clearly violated in complex terrain, leading to the underestimation of the turbulence kinetic energy and an unrealistic representation of the mountain boundary layer.
In this study, we perform limited-area simulations with the state-of-the-art NWP model ICON (Icosahedral Nonhydrostatic Model) across resolutions (1km, 500m, 250m, 125m) in the Inn Valley, Austria for a day where boundary-layer processes dominate. A thermally-induced valley wind circulation forms on this day, typical for wide valleys. The model is evaluated with observations from the CROSSINN measurement campaign, providing besides the usual meteorological parameters also measurements from LIDAR systems, radiosondes, turbulence eddy-covariance towers, and scintillometers. This data pool of observations allows us to evaluate the current turbulence parameterizations of ICON, starting with the (pre-)operational NWP resolution (1km) across the turbulence grey zone towards large-eddy simulation resolutions (125m). Both the one-dimensional (1D, Mellor-Yamamda type) and the three-dimensional (3D, Smagorinsky closure) turbulence schemes of the model compared with the observations across the grid resolutions to identify the resolution at which the 3D scheme starts to add value over the 1D scheme. Finally, we can check whether further improvements in the turbulence schemes are necessary for the turbulence grey zone.