[aerohead]

Perhaps a CFD expert will come forward.
I'm not sure that FLUENT can accurately model the flow,in light of the unmolested CAD model results.
A full-scale wind tunnel investigation with the logarithmic law layer ,with crosswind examination would be insightful.
So many of the masters thesis presented with CFD as the technique used for the investigation have shown significant deviations when backed up by full-scale tunnel testing.
Only Direct Numerical Simulation on a supercomputer is agreed to have the mesh size and time intervals necessary to develop a solution of high certainty for a 3-D bluff body in ground proximity.
Today,in 2015,BMW's CFD is off by 18%.
Today,in 2015,Mercedes-Benz is within 1%.Supercomputer run time is over 48-hours for a single iteration.
All we can say,is that on that CAD model,at that scale,using that CFD,we'd expect results like that.