nuSgs and non-uniform mesh, computational or visualization error
 

Hi,
I use pisoFoam to simulate the transient flow in an elementary street canyon.
The mesh is structured and non-uniform, being denser at solid surfaces.
The cell to cell expansion ratio is 1.048 and the max to min cell ratio is 100.
Viewing the results, there was nothing strange with U and UMean.
But when viewing the nuSgs results, I saw something weird.
The value of nuSgs is lower in smaller cells and larger in large cells.

Do you have any idea, if it's caused by errors in computations, or errors in visualization by paraView?
Additional data:
OpenFOAM edition 2.3.1, paraView (paraFoam) 4.1.0 (the edition with of240)
The mesh is 3D with 50 cells width.
The boundaries are
- cyclic for the inlet and outlet patches
- cyclic for the front and back patches (as you see them in the page)
- wall at the top
- wall at the street canyon buildings' walls