The scope of this paper is to determine an optimal mesh adaptation strategy to compute turbulent flows in presence of solid bodies using RANS models. To this end we propose to use additionally model specific wall functions when the low-Reynolds turbulence model is not sufficiently resolved. Such wall functions degenerate to the low-Reynolds turbulence model they mimic when the mesh size tends to 0. This significantly improves solutions on coarse initial grids and fasten computations toward the final solution.