Abstract
A computationally turbulence closure scheme is formulated and evaluated. The scheme includes a correction to the redistribution terms in order to account for the influence of the underlying surface. Care is taken in order to ensure realizability in simulated fields. The formulation is implemented in a mesoscale model containing prognostic equations for mean momentum, heat and turbulent kinetic energy. All other second moments are obtained through diagnostic expressions. A one-dimensional version of this highly simplified scheme is evaluated against turbulence measurements from stable, neutral and convective atmospheric boundary layer. Inclusion of wall effects is shown to improve the performance of the closure when applied to the atmospheric boundary layer. It gives a better prediction of ratios between velocity variances, of importance particularly for air pollution applications. Discrepancies found in the upper 40% of the convective boundary layer suggest that the additional computational effort connected with the inclusion of a prognostic formulation for potential temperature variance should be considered if a detailed prediction in this layer is needed.