Although knowledge of surface ground conductivity and dielectric
constant is much better than nothing, unfortunately it's still not
enough to get truly accurate results from modeling or calculation. Most
grounds are stratified, as anyone can discover by digging a few feet
down, with layers which can be vastly different. The skin depth at 3.5
MHz in average soil is 16 feet, and at 7 MHz and above about 13 feet(*),
which means that significant current flows to depths several times these
values. To make an accurate representation of the ground would require
knowledge of the ground characteristics to several tens of feet.

Fortunately, horizontally polarized antennas are quite insensitive to
ground characteristics except at high elevation angles. So for amateur
purposes, it's of interest mostly for vertical antennas and NVIS
operation with horizontal antennas.

To accurately predict the field strength for vertically polarized
antennas, you would need to know ground characteristics typically for a
few hundred feet from the antenna (depending on the height and current
distribution on the antenna and the elevation angle of interest) -- the
distance at which the field from the antenna strikes the ground and
reflects to interfere with the direct field. This is admittedly a
simplification of what really happens, but the distance where the
effects occur is approximately right. You also need to know the ground
characteristics in the close vicinity of the antenna to evaluate the
efficiency of an antenna having a radial system. Besides the problems of
stratified ground, many of us live in an urban environment where buried
utilities, pipes, reinforcing bar, and houses are likely to be within
the radius of interest.

Finally, even if you had exact information about each layer of a
stratified ground, I don't know of any readily available program which
can make use of that information. NEC based programs like EZNEC can
accommodate only a single ground characteristic, and assume that the
ground is homogeneous to an infinite depth. I'm not convinced that
there's any single value which can be substituted for a stratified
ground which will behave like the stratified ground. This is the weakest
point of currently available modeling programs. But even if they could
model stratified ground, you'd have to know values for each ground layer.

The method used in broadcasting, where the skin depth is on the order of
25 feet(*), of measuring the attenuation of a ground wave signal over a
lengthy path is a much more accurate way of determing overall path loss
than simple measurement of surface ground characteristics, since it
automatically takes into account all the factors I've mentioned.

Again, knowledge of surface ground characteristics is better than
nothing, but don't be fooled into thinking that it gives you real
information about the ground in which the RF current flows. Had Brown,
Lewis, and Epstein measured the surface ground conductivity, it wouldn't
have helped us much.

(*) You might notice that the skin depths don't follow the familiar rule
of decreasing as the square root of frequency. That's because the
familiar rule is valid only for good conductors. Average Earth acts like
a good conductor below about 7 MHz, but above that it's more like a
dielectric, where the skin depth becomes independent of frequency.
Consequently it's not less than 12 - 13 feet at any frequency for
average soil.

Roy Lewallen, W7EL