Scalar problem with refined mesh- HELP PLEASE
 

I am modeling a scalar which is a function of velocity. It works fine with a relatively coarse mesh (47 by 400), but when I have refined mesh (125 by 1800) the results become wierd near the wall. I am using low Reynolds number model, and the wall boundary conditin for the scalar is ADIABATIC (as I require it to be zero gradient at the wall). Does anyone have any idea why? The source terms are,

S1P = DEN * SQRT(U**2 + V**2) S2P = 0.0

The diffusivity term is,

DIF = 10.*(VISM+VIST)

I am modelling a flat plate (0.3m by 2m) with inlet, outlet and symmetry boundary conditions. Thanks