Large-eddy simulation (LES) provides three-dimensional, time-dependent fiel
ds of turbulent refractivity in the atmospheric boundary layer on spatial s
cales down to a few tens of meters. These fields are directly applicable to
the computation of electromagnetic (EM) wave propagation in the megahertz
range but not in the gigahertz range. We present an approximate technique f
or extending LES refractivity fields to the smaller scales needed for calcu
lating EM propagation at gigahertz frequencies. We demonstrate the techniqu
e by computing refractivity fields through 128(3) LES, extending them to sm
aller scales in two dimensions, and using them in a parabolic equation EM p
ropagation model. At 0.263 GHz the very small scale structure in the extend
ed fields has a negligible effect on the predicted EM levels. At 2 GHz, how
ever, it increases the predicted levels by 15-25 dB. We relate these result
s to the refractivity structure sampled by EM waves at 0.263 and 2 GHz. We
also show that at long range an EM field calculated through an LES-based re
fractivity field is generally less coherent and significantly weaker than o
ne computed from a "plywood" (i.e., stratified, range-independent) model of
the small-scale refractivity field. We give a physical explanation for the
differences in the EM fields computed in these two ways. Finally, although
the plywood model gives results that fit the EM levels observed in the rec
ent Variability of Coastal Atmospheric Refractivity (VOCAR) experiment, it
is not physically realistic. The instantaneous top of the atmospheric bound
ary layer is known to be sharp and horizontally varying, and we show that u
sing such a top also yields a fit to the VOCAR data.