On a surface flux model for atmospheric large eddy simulations

Atmospheric large eddy simulations (LES) are becoming increasingly popular in studies of various boundary layer processes, development of their parameterizations, and so on, since they give more reliable results by reducing uncertainties in the unresolved small-scale processes. However, there still remains a problem with respect to surface flux models used in the LES, since they are mostly designed based on field observation data averaged for sufficiently large duration of 10 minutes.

In order to examine the effects of surface flux models, we have performed a simple LES test in which two different surface momentum flux models are compared: The first model uses domain-averaged winds as in the original design, while the second model uses local winds at each grid point. The two models give nearly similar average winds. However, the difference in the surface flux models has a certain impact on the variance of winds, which is significantly decreased when the second model is used. We also made an additional LESs in which an isolated vertical vortex having a large tangential wind is introduced at a certain instant. The resulting temporal evolutions of the vortex also significantly differ between the LESs with the two different surface flux models. Thus the way the surface fluxes are modeled has a significant impact on gusty winds in the atmospheric LES.

To validate the surface flux model when turbulence is partially resolved as in the LES, we have made a direct numerical simulation (DNS) of a turbulent Ekman layer in which the Reynolds number based on the Ekman layer thickness and geostrophic flow is about 500. Using spatial filters of various horizontal scales on the DNS results, we have examined the relationships between winds and surface fluxes. The results suggest that, if the second model is used when turbulence is resolved, the bulk coefficient needs to be significantly reduced. An additional DNS experiment with an isolated vertical vortex suggests that the surface flux is not proportional to the local wind, and implies that the second model overestimates of surface fluxes for gusty winds.

We also have made an experimental study on a turbulent boundary layer in a wind tunnel at high Reynolds numbers. The results imply that there is little correlation between winds and surface fluxes which is assumed in the surface flux model.