How can we include a change in the roughness parameter for the wall when using a k-epsilon two layer zonal model in Fluent 5?

The roughness parameter is introduced in the wall function approach in order to describe the shift of the velocity distribution of the log law. If You use the two layer zonal model, You calculate the velocity distribution near the wall but outside the viscous sublayer using a special turbulence model. Thus there is no empirical constant like the roughness parameter in Your model.