Wireless Backhaul - How to plan and test Point-to-Point feasability in potential antenna locations?

Hire a company to do a wireless survey for you.

The main reason I say this is because the specialist hardware and specialist software are likely to be unaffordable for you, unless you're making this your main business. Whilst the calculations are relatively straight-forward, a realistic answer can only really be achieved by setting up test hardware, and seeing how well it works (with quantitative measurement throughout).

All these calculations are relatively simple, you just have to know the formulas. Consequently, they're pretty easy to pop into a spreadsheet and reuse. Also, these days you'll want to over-provision for everything, so you don't have to be as accurate.

Fresnel zone

Behind and directly around antennas: λ / 2. At 2.4GHz that's about 2.5 inches. Higher frequencies are even less.

In the center of the propagation path: 8.65 * (km / GHz) ^ 0.5. So a 20km link at 2.4GHz would be 8.65 * (20 / 2.4) ^ 0.5 = 24m. That's ideal, but half of that would be acceptable.

Propagation Loss (free space)

Loss (in dB) = 20( log(km) + log(GHz)) + 92.5. So our 20km, 2.4GHz link is 20( log(20) + log(2.4)) + 92.5 = 126dB

Potential Speed

This one is really impossible to answer. The modulation technique, frequency, and bandwidth determine how many symbols can be carried by the radio wave. But, various radios work better or worse depending on what combination of the above you feed into them. Good manufacturers publish numbers for common combinations, so you'll want to look those up and work from there.

Link budget

The link budget is pretty easy to calculate, the radios will output as a certain level, you add for antennas, subtract for propagation loss, and you'll want 10-20 dB leftover for a safety margin.

For the longer links you'll probably need to look at topographical maps, or survey them yourself (a GPS that shows elevation helps here, "hikers" GPS generally show this info). You can farm it out to services that will evaluate existing topo data for you, but they're far from cheap. MicroDEM is a program that can aid in making radio maps of sites, but it probably much more complicated that you'll want (it is free if you have plenty of free time to play with it). Combining that with Google Earth and USGS Seamless Data and you'll get just about everything you need (binoculars are still a good way to verify, no matter how much tech you throw at the problem).

I tried to say away from product recommendations, but I've seen more than a few successful Ubiquity links put up, and their software is squarely aimed at people operating a dozen or five radios, which it sounds like the direction you're headed.