wrote
I think that comparison would be flawed though. That is not an optimum
"elevated" radial system, and I can see it losing to a 120 radial
ground system. For one thing, the radial system is not resonant.
This is quite important for an elevated system. Fairly critical
really.
Also, the manner of the wires being laid out like chicken wire mesh
is not an optimum use of wire. You would have better results using
the usual "spoke radial" system, due to more wire being concentrated
at the feedpoint, vs farther out.

However the study was not intended to model an ideal elevated radial system,
but one using an existing metal roof to serve as the "other conductor" of an
electically short, vertical dipole.

The mesh wires used in the study to simulate the metal roof have a density
of well below 1/10 wavelength, and for these conditions the roof would
appear essentially solid to r-f.

I often see and hear about MW comparisons using ground mount
vs elevated, and almost to a tee, most ruin the comparison by
applying substandard radial systems on the elevated antenna.
Either that, or they ignore the number of elevated radials needed
to equal the 120 on the ground. That of course varies with height
in wavelength.
At 1/2 wave, 3-4 radials will equal 120 on the ground.
At 1/4 wave, about 8-12 will be needed.
At 1/8 wave, about 60 or so at least..

And it just gets worse from there as you get lower and
lower. Less than 1/8 wave, and you will need a load of radials
to equal the 120 on the ground. Probably 80-100... ??

A NEC-2 model of a 1/4-wave vertical monopole (base at earth level) in the
AM broadcast band using just four, 1/4-wave radials elevated 12-15 feet
above a perfect earth shows a peak h-plane gain within tenths of a decibel
of the theoretical peak value for a 1/4-wave monopole over a perfect ground
plane, ie, at least 5 dBi.

Such systems have been installed by commercial AM broadcast stations where
the earth at the antenna site makes it impractical to use the standard 120
buried radials. These elevated radial systems permit those stations to
produce at least the minimum groundwave field at 1 km for 1 kW of applied
power that is required by the FCC for AM broadcast stations -- and with far
fewer than "80-100...??" radials elevated less than 1/8-wave, fortunately.

Here is a link to a paper about this
http://www.nottltd.com/ElevatedRadialSystem.pdf

Even if you equal the ground losses between the two, the
elevated still wins due to more clearance of ground clutter,
and a much better local ground/space wave signal, which
I assume is due to the radiator being high and in the clear.
More line of sight so to speak.. This also greatly improves
the lower angle DX coverage, being both use the same fairly
low angles of radiation.

My models were done over a perfect ground plane in order to show the field
of the monopole radiator as it is launched. This is the method used by the
FCC in AM broadcast practice.

Terrain, obstructions, r-f ground loss, nearby parasitic radiators etc --
and in the case of ham antennas, their height above the earth -- will have
an affect on system performance. But all of that needs to be evaluated
separately for the installation conditions, based first on a knowledge of
the field pattern launched by the radiator itself.

RF http://rfry.org