At higher frequencies, even 915 MHz, moving the RF hardware 'up the tower' (so to speak) resulting in extremely short RF cables means that the cable loss due to lengthy coaxial cables is essentially eliminated. As you're contemplating, the backhaul down to 'ItSmoked HQ' can be by baseband cable or (as you've mentioned) WiFi.
Having suffered through at least two nearby lightning strikes, a wireless link (and local solar power) would be better (safer, lightning-wise) than cable(s).
Antennas: You can achieve fairly high gain and omni-directionality with a high gain vertical monopole (typically a collinear array). They squeeze the gain along the horizon, assuming that's where everyone is located. You could make one if you wish, but ideally some test equipment would be used to confirm the tuning / match.
Beware cheap commercial antennas. Many are junk. Even the ones where the antenna makes sense may have extremely lossy cable. I have a Wifi Yagi that is in need of better coax.
I wouldn't bother using the ground plane to try to twist the gain pattern. There are better approaches.
A nominal ground plane is about +5% bigger than quarter wavelength. Many applications would be best with a ground plane that is not horizontal, but sloped down. This improves the match to 50-ohms and pulls the gain down a bit to hit the horizon. General rule of thumb.
If I were you, just get it working and optimise the antenna as Phase 2. Don't let the antenna distract you from the other 99% of the project.