Calculation of winds disturbed by an array of fences

It is known that when an isolated porous windbreak is represented by a sink in the mean momentum equations, even the simplest turbulence closures lead to reasonably good simulations of the mean wind close to the barrier, irrespective of whether or not sources are introduced in the turbulent kinetic energy (TKE) and dissipation rate (ϵ) equations. However, unless the barrier is considered to dissipate TKE in addition to opposing the mean flow, the pattern of turbulence is poorly simulated. Here we examine the performance of simple Reynolds-averaged Navier–Stokes (RANS) wind models for the case of a windbreak network, relative to the field experiment of McAneney and Judd (15 porous fences, height h=2 m, spaced at Dx/h=6 along the mean wind). All closures we considered gave mediocre predictions even for the mean flow, contrasting with the case of an isolated barrier; while, just as for the isolated barrier, prediction of the TKE in a windbreak network hinges on ambiguous choices in the treatment of TKE (and ϵ) sources at the barriers. The additional turbulence generated by a sequence of barriers implies that a proper representation of the Reynolds stress is more critical than in the case of an isolated barrier, near which the pressure-gradient force dominates. This surely accentuates the importance of the turbulence closure, and our results may imply that no existing RANS turbulence closure is adequate for this type of flow. However, the problem could alternatively stem from the treatment of sinks parameterising the flow–fence interaction, not only in the TKE and ϵ-equations (where perforce such terms are heuristic), but even in the momentum equations.