A Pan-NCAR Large Eddy Simulation and Single Column Model Case Study

Abstract:

As part of an effort to leverage the extensive observational and modeling capabilities across the National Center for Atmospheric Research (NCAR), results from a case study from the South Atmosphere Studies (SAS) during Summer 2013 are described. Initial conditions and surface flux forcings are applied quasi-analytically to reflect an aggregate daytime period in a dry stable environment. The case data are applied consistently to a single-column version of the Community Earth System Model (CESM) and to three NCAR LES models, nominally the NCAR, WRF and FastEddy formulations. The simulations have both chemistry and meteorology foci. Here we focus on comparing the turbulent meteorology in the planetary boundary layer.

In a configuration closely resembling that of the full 3D atmosphere in CESM, the vertical and time-step resolutions are clearly insufficient to simulate the observed boundary layer evolution. Compared to LES, vertical mixing of humidity, PBL growth rate and the PBL-top potential temperature gradient are too weak. Divergence from the LES simulation accelerates after midday, when the vertical mixing from the high-order turbulence scheme in CESM, the Cloud Layers Unified By Binormals (CLUBB) parameterization, underestimates the vertical extent of eddy humidity fluxes. Although the evolution of the second-order turbulent moment is well simulated, the third-order moment is too weak by an order of magnitude compared to LES. The weak penetration by asymmetric turbulence may explain the poor PBL growth in CESM later in the day. When increasing vertical-resolution by a factor of 2 or 3, we then see convergence of the solution in the column model, but yet it is still not able to reproduce the deeper mixing and PBL top characteristics, seen across all the LES companion simulations.

A more coherent observational, LES and large-scale modeling activity, such as presented here, provides a consistent workflow for improving large-scale models, and developing parameterizations. This is enabled with comparison to LES benchmarks, and more appropriate validation against campaign-based observations.