Inlet Turbulent Intensity Changing During Initialization

Why does initialization change inlet turbulent intensity?

The turbulence intensity (%) is defined as: TU = 100/U*SQRT(2/3*k)
where k is the turbulent kinetic energy. Now, for the purposes of ** boundary conditions ** the velocity U is defined as the mean velocity for the inlet boundary. For post-processing, however, U is defined as the Reference Velocity, which can be accessed from the Post-Processing tab --> Reference Values.

Why two different velocity definitions? The problem when you post-processes is that you may in fact have multiple inlet velocities, so you can't use the inlet velocity. You could use the mean velocity in the domain, but that really wouldn't be correct since the velocity varies in space. You could also use the local cell velocity, which would probably be best for assessing local TU levels. This is easily done using a custom field function (CFF), defining a local TU as 100/velocity-magnitude*SQRT(2/3*turbulent-kineticenergy).
So, if you redefine your Reference Velocity to be 2.5 m/s rather than the default 1.0 m/s, you will get 4.4% TU.