抄録
Turbulence simulations in the neutral atmospheric surface layer with a numerical weather prediction model WRF have been performed under an ideal condition; the terrain was flat and the surface roughness was homogeneous. When the vertical turbulence diffusion coefficient was estimated by using a large-eddy simulation (LES) with a high-resolution grid (the horizontal-grid spacing was 100m), the predicted turbulence intensity of grid-scale velocity fluctuations agreed well with that of observations. On the other hand, the magnitude of grid-scale velocity fluctuations with a PEL scheme, which is usually used in weather forecasting, remained a small value even with the high-resolution grid. This is caused by a difference of prediction capability for turbulence structure in the surface layer, i.e., only a LES can capture coherence structures near the surface.
