Surface Flux Parameterization in Atmospheric Large Eddy Simulation: Necessity to Filter out Turbulent Fluctuations

Atmospheric large eddy simulation (LES) require the use of a surface flux parameterization of turbulent transports near the ground. Conventional parameterizations used for Reynolds- Averaged Navier–Stokes simulation may not be accurate for LES: the grid values of wind speed in LES contain turbulence fluctuation, so that they may be inappropriate to substitute in the surface flux parameterization. A sensitivity experiment with an atmospheric LES show that the issue in the surface flux parameterization is relatively unimportant to domain-averaged wind speeds, but significant for their variations.

The results of direct numerical simulation (DNS) and wind tunnel experiments are examined to how surface momentum flux relates to wind speeds in turbulent boundary layer. The surface flux is hardly correlate with wind speed fluctuations due to turbulence. Both DNS and wind tunnel experiments suggest temporal (spatial) filtering over the correlation time (length) is required to rationalize the conventional parameterization.

Considering temporally filtered wind speed in the surface flux parameterization is a plausible modification on numerical weather prediction models. Thus modified parameterization is implemented in LES. Its result is shown to realize more reasonable vertical profiles of wind speed variations in accord with the wind tunnel experiments: the log-law for the wind speed variations (Marusic et al., 2013) is realized.