Conventional wisdom based on the far-field (only) patterns calculated by M-o-M software such as NEC show that low-angle radiation from a horizontal dipole is very low, and zero in the horizontal plane. Is this belief valid?
The very small low-angle radiation noted above is the result of excluding the surface wave from the NEC calculations.
Also the commonly-used "far-field only" patterns calculated by NEC are based on an infinite distance from the radiator, and if a ground plane is included, it is considered to be flat. Both of these elements can lead to a misunderstanding of the real-world performance of antenna systems modeled in NEC.
Below is a comparison of NEC patterns including the surface wave. It shows the radiated field intensities at one wavelength in the horizontal plane for elevation angles from zero to about 17.2 degrees, with 100W of Z-matched r-f power at the antenna feedpoints.
Conclusions are left to readers to consider.
The reflection of an E field wave is well known to be 180 degrees out of phase with the incident wave, given a pure E field and a perfect reflector. Real world ground conditions, and the shift of the wave from pure E to a mix of E and H fields somewhat dilutes this effect, as shown by the chart.
If low angle signal strength is the objective, an interesting alternative is an H, or magnetic, wave generating antenna, as classical theory shows this energy is reflected in phase by the ground. Personal on-air 40 meter band amateur radio experience with such an antenna, a so-called magnetic loop, has borne this out. A 7 foot diameter loop 7 feet off the ground has consistently performed as well reaching DX as others' dipoles several times higher.