I second what George has said about Norton's papers (and those of his
associates). 1950s IRE is prob. where one should look.
At lower UHF, scattering can be an important mode. I did estimates
of these modes to predict interference to a radio-quiet site long ago.
As a general rule, lots of power needs to be involved.
At all of the frequencies, diffraction over the surface is
significant. A standard scheme was to assume that the earth had a
larger radius than is the case and them to plot LOS paths on that basis.
Used to do that with topo maps by hand correcting the height of each
contour (cutting the path) for how far away it was. The last time I did
it - to help a college radio station - I used electronic topo maps -
what progress!
73 Mac N8TT

--
J. Mc Laughlin - Michigan USA