Bill, it is the Reynolds number thingy.
Looks like you will need 3000 mph [not good] or 100 times the density.
Coarse basic mesh with a fine local mesh in the region of interest doesn't work? How large is too large? Just curious.
Flow sim turbulence model start to diverge at an R'number of about 6-8 million Well at least for flow over a sphere/cylinder. The modelling of planetary B'layers is a field on its own and the whole model would be within the planetary B'layer. What r u trying to figure out how persistent the effect of the fence is? Why do you need a model something so large?