[westbywest]

Fallback to other unlicensed bands like 24GHz or 60GHz would probably work better. The higher frequencies at least retain a compact dish size, smaller fresnel zone, and sharper directionality. That is, imagine hanging several of these optical devices + 5GHz fallback at ~1 meter vertical spacing on a pole. A weather event triggering failover on all optical links could lead to enough 5.8GHz self-interference to kill the fallback links as well. 5.8GHz devices that aren't synchronized in phase with each other often call for 3meter+ separation, eating up tower real estate.

[betterunix2]

Weather will also degrade 24GHz and 60GHz links. A friend of mine used to work at an ISP that relied on wireless links, and he said that even at 2.4GHz moisture was a significant problem (especially concentrated moisture in living things like trees). In general as the frequency increases attenuation due to water increases; at 60GHz particulates will also become a problem (consider that a selling point of 60GHz is that it does not penetrate walls, so you will have less interference in apartments or large office buildings).

(Edit: Looked it up; 2.4GHz is more readily absorbed by plants than 5.8GHz. Still, attenuation due to moisture seems to be a problem for microwave frequencies generally.)

[westbywest]

Indeed the 24GHz and 60GHz gear has its unique rain fade issues not shared by 5.8GHz. Increased atmospheric attenuation of higher bands is the general trade-off for higher throughput and more compact physical isolation requirements. UBNT 24GHz gear is being using for ~10km links, but looks like the ~5km is the preferred limit. My original comment would have been better written to stress need to prove out the signal path over all intended bands, e.g. primary link over (near) optical and fallback over whatever RF band proves viable.