An evaluation of LES turbulence models for scalar mixing in the stratocumulus-capped boundary layer

The stratocumulus cloud-capped boundary layer under a sharp inversion is a challenging regime for largeeddy simulation (LES). Here, data from the first research flight of the Second Dynamics and Chemistry of the Marine Stratocumulus field study are used to evaluate the effect of different LES turbulence closures. Six different turbulence models, including traditional TKE and Smagorinsky models and more advanced models that employ explicit filtering and reconstruction, are tested. The traditional models produce unrealistically thin clouds and a decoupled boundary layer as compared with other more advanced models. Traditional models rely on specified subfilter-scale (SFS) Prandtl and Schmidt numbers to obtain SFS eddy diffusivity from eddy viscosity, whereas dynamic models can compute SFS eddy diffusivity independently through dynamic procedures. The effective SFS Prandtl number in dynamic models is found to be ~0.5 below the cloud and ~10 inside the cloud layer, implying minimized mixing in the cloud. In contrast, the SFS Prandtl number in traditional models is about 1/3 throughout the entire boundary layer, suggesting spuriously strong mixing in the cloud. The SFS Schmidt number in the dynamic models also changes independently from the SFS Prandtl number, whereas in traditional models they are identical, meaning that the efficiency of the turbulent mixing of water content is forced to be the same as that of heat. Since it is very difficult to know in advance the SFS Prandtl and Schmidt numbers in a given flow, dynamic models may provide a more realistic representation of scalar mixing in LES.