Introducing TABLES: First insights from the Terrain-induced Atmospheric Boundary-Layer flow: Large-Eddy Simulation model intercomparison study

Large-eddy simulations (LES) have become one of the most important tools to study boundary layer turbulence due to the increase in computing power in recent years. Research including LES spans from idealized simulations of flow over horizontally homogeneous and flat terrain to realistic hindcasts nested in mesoscale simulations. Due to the wide range of potential applications of LES models, many different LES codes were developed. Differences between these codes include the spatial and temporal discretization, the dynamical cores, the sub-filter-scale turbulence models, and the representation of terrain. Benchmark studies comparing the outcomes of LES from different models have been performed for a variety of scenarios and weather regimes (stable boundary layer, very stable arctic boundary layer, stratocumulus, idealized convective boundary layer flows exist for flat terrain), but no such study exists for complex-terrain flows yet.

The Terrain-Flow Atmospheric Boundary Layer Large-Eddy Simulation (TABLES) Model Intercomparison Study was designed to fill this research gap and provide a benchmark for convective boundary layer flows above idealized bell-shaped hills. In this simple scenario, baroclinicity due to differential heating along sloping surfaces generates a thermally-driven circulation that exerts a profound impact on the spatial heterogeneity of turbulence in the boundary layer. The project consists of 12 research groups bringing 12 different LES codes. While all models use the same initial and boundary conditions, the models differ in the dynamical core and spatial discretization, as well as in the implementation of terrain, sub-filter-scale model and surface model. Each research group was asked to provide simulations of free convection and forced convection, both over flat terrain and over a 100-m-high ridge. In addition, research groups working with compressible models provided also one free convection and one forced convection simulation over a 1000-m-high ridge.

Preliminary analysis suggests that mean variables and turbulence statistics from the flat simulations agree remarkably well between the different models. Also, general flow structures of convective boundary layer flows above idealized ridges agree well, but differences in both the strength of these flow structures and their location across the ridge are more pronounced for terrain-induced flow cases.