Depending on what digital modulation rates you are proposing, the in-ground fibre optic link would be virtually noise free (and relatively lossless) over that distance. As far as FSOC is concerned, Israel seems to be leading the work in that area and is particularly applicable due to the high optical availability of your clear air.
I agree with Rob..however, BER is also an important parameter. Again, if you use DWDM, you have to a bit careful about cross talk which will contribute to some unwanted distortions.
You both "linked" my question to communication considerations, though specifically this inquiry wasn't related to communication but rather to beam properties. In particular the wavefront phase, and as I alluded, do the deployed fibers involve changes in the phase at the output of the fiber due to earth motion - I would guess they do, but on what order are these changes. How much uncertainty would I have in the phase relative to a free-space link?
Working with fiber based interferometers, I can say that one must expect significant low frequency phase modulations from temperature influences, motion of the fiber and acoustic coupling to the fiber. Even when using PM fiber. Cannot say for large distances, but in lab dimensions this accounts to a few nanometers path length drift per meter fiber length. Even reaching micrometers for larger temperature changes.