There are several different modes of operation in the HF Band (2-30 MHz), including:
a) Vertically Polarized Surface Wave (75-300+ miles over sea water, only a handful over land),
b) Bounce off of Meteor Burst/Trails (2.4+ kbps, but only 75-100 bps LONG term average),
c) Earth-Moon-Earth (EME) or simply "Moon Bounce". With Ham Radio power level limits, they need very high Gain TRANSMIT and RECEIVE Antennas...usually a Quad Array (or more) of very long 144 or 432 MHz Yagi's....and only good when both can "see" the moon:
http://www.nitehawk.com/rasmit/ws1_1.html
And there is an avid group of Amateur Radio enthusiasts who delight in posting screen grabs of very distant ANALOG TV signals that have fortuitous geometries to have a sufficient amount of energy escape into space and bounce back from the Moon to very high Gain Antenna Arrays....but may actually include E-Skip captures (see below). Not very many captures for the much more difficult DIGITAL waveforms:
http://home.iprimus.com.au/toddemslie/sydney-uhf-eme-tvdx.html
http://home.iprimus.com.au/toddemslie/dx-articles.html
and
d) Ionospheric Propagation, the old favorite for long range communications, which generally falls into either of three sub-types, ALL with imperfect reliability:
d1) NVIS (Near Vertical Ionosperic Skywave) where the signal propagates nearly straight up, with a small fraction of the signal bouncing off the layer (most goes THROUGH), bouncing back to a receiver. At mid-long range, the bounce is weaker, so handy to get over some mountains, but limited in range. Since TV Transmitters have almost ZERO signal directed upwards, it's currently not an option....and at VHF/UHF frequencies nearly all of the signal would pass right through the layer anyway. And you might as well use TV's usual LOS or Tropospheric Propagation modes.
d2) E-Skip Refraction off a fairly low and highly intermittent layer....comes when it wants to and mostly goes away. But when conditions are favorable, long range captures are possible...although any particular channel position may be a mush from multiple stations within the beam of the receive antenna:
http://home.iprimus.com.au/toddemslie/Es_distances.html
d3) Higher F1 and highest F2 Ionization Layer Refraction (glancing angle). Primary mode for long range communications, but low to non-existent layers during the day restrict range and utility...but is usually quite good at night....but only on a narrow range of operating frequencies between any given two locations
As a Comm Engineer, I've have experience using various Propagation Prediction Programs, including (Free) VOACAP , which I recommend if you want to get your feet wet....VOACAP is also used inside some extra cost Graphics "Shell" Programs. Layer height greatly affects the FOT (Frequency of Optimum Transmission) which is somewhat BELOW the predicted "foF2" (maximum frequency which will bounce directly back from overhead), as shown here for current Solar Flux (proportional to Sunspot number) and Time-Of-Day:
http://www.ips.gov.au/HF_Systems/6/5
http://www.ips.gov.au/HF_Systems/6/6/2
Since lowest TV frequency is 54 MHz (Ch2), this mode of propagation is NOT supported...not even during a highly active Solar Cycle.