Several field experiments to study subgrid- or sub-filter-scale modeling for large-eddy-simulation of the atmospheric boundary layer (ABL) have been conducted during the past several years (see separate papers by Porte-Agel et al. and Horst et al.). In the NCAR/JHU experiment held in Kettleman City (CA) in September 2000 the instrumentation setup consists of 14 3-dimensional sonic anemometers arranged in two parallel horizontal arrays in the atmospheric surface layer downwind of a long homogeneous fetch. From four weeks of continuous sampling with different array heights and widths, data encompassing a wide range of atmospheric stability conditions, and turbulence regimes are obtained. For details on the data sets and filtering operations, see companion talk by Horst et al.

The signals are filtered and sub-filter fluxes and resolved velocity gradients are evaluated. Energy dissipation arguments are used to compute model coefficients in 30 minute intervals. The dependence of model coefficients on dimensionless parameters, such as filter size normalized by measurement height (D/z) and filter size normalized by Monin-Obukhov-Length (D/L) is quantified from the data. Moreover, we study the level of intermittency of sub-filter fluxes as functions of these parameters. As expected, we find that intermittency increases with decreasing filter size. Perhaps surprisingly, the data suggests that intermittency decreases under stable conditions.