Hi:

On second thought, rather than replicate treatments at paired sites, I
think I'd put all three treatments (control, top, bottom) in equal numbers
at each site.  That way, the within site variables will be the same - and
the excluders will be the single variable that has changed.  I'd intermix
the hives, etc.  Don't bunch them in groups by treatment.

Then, replicate over a couple of more sites (again with each site having
all three treatments).

Just a thought.

Jerry