There are two other downsides to a very low antenna. (1.) ground losses go up; less of the power put into the antenna is radiated. The lower the antenna, the more of the radiated power is absorbed in the earth. Ordinary dirt is NOT a good conductor of RF. (2.) The feedpoint impedance of a very low dipole can get quite low. This creates matching problems (to the transmission line) and the ratio of Radiation Resistance to Ohmic resistance goes down....meaning resistance losses in the wire go up.
Actually these two effects work against each other: the higher the ground losses, the higher the
impedance and the lower the SWR. Over some soils the SWR impedance actually stays pretty close
to 50 ohms even when the antenna is close to the ground.
But otherwise a good summary. Radiation patterns can be predicted easily for low heights, though
the magnitude of the losses may be a bit uncertain for antennas close to ground. (That was one of
the faults with some of the early studies of NVIS propagation - they relied on the ground model in
MININEC, which was quite inaccurate for such things and greatly underestimated the losses at
low heights. The ground models in NEC2 or NEC4 are much more accurate for such purposes.) For
a dipole the maximum radiation is straight up, tapering off at the lower angles to nearly nothing at
low angles. Except for losses, it really doesn't change much for heights below around 1/8 wavelength.
Such low antennas aren't nearly as good for DX due to the reduced radiation at low angles. They
are useful, however, on the lower bands (160m, 80m, and sometimes 40m) for local work when the
ionosphere reflects such high angle signals back down to Earth.
At no height does a horizontal antenna start acting like a vertical.
One way to measure your ground characteristics is to install a dipole at a low height and
measure the impedance of it. Otherwise there are descriptions of different soil types and
the measured characteristics thereof - from those you can get a pretty good guess of
the parameters for your soil.