This dataset contains 16 simulation cases which are driven by the same large-scale forcing and exposed to a rough surface. Members of the parametric set of simulations differ by stable density stratification, which is imposed via a Dirichlet boundary condition (viz. constant temperature difference between upper and lower domain boundary). The cases are labelled by their ID (N, S001,..., S128, ..., S256P) according to the stratification measured by the Richardson number Ri_Λ; N stands for neutral stratification, S for stable. The simulations were carried out sequentially in time starting from the neutral case N, a suffix P indicates three concurrent runs for very stable stratification (S128P, S192P, S256P). Here the bulk Richardson number Ri_B is evaluated based on the boundary layer thickness of the neutrally stratified case N such that it does not evolve over the course of simulation. The choice of Re_Λ = 5x10^5 corresponds to Re_D = 1000 (Re_D = DG/ν, with the laminar Ekman-layer depth D = (2ν/f)^0.5 ) and a friction Reynolds number Re_τ ≈ 2700 for the neutrally stratified case. Two computational grids are used: (A) 3840x704x3840, (B) 3840x576x3840 with similar spatial resolution of approximately 2.6x1.0x2.6 wall units and a domain size in terms of the Rossby radius of 0.27x0.26x0.27 for grid (A), and 0.27x0.11x0.27 for grid (B). This dataset contains a PDF file outlining the simulation setup and description of the variables contained in the netCDF files. A detailed description of the methodology and implementation is found in the git repository https://github.com/turbulencia/tlab, which contains the relevant code and a manual.