RANS modelling of flows over a step change in roughness

Surface roughness is a primary contributor to frictional resistance on surfaces like ship hulls, increasing drag and operational costs. The spatial heterogeneity of this roughness presents a significant challenge, as the validity of Reynolds-Averaged Navier-Stokes (RANS) simulations is insufficiently established for such conditions due to their reliance on equilibrium turbulence assumptions. This study investigates the predictive capability of RANS for turbulent boundary layers over abrupt streamwise roughness transitions. Our simulations are first validated against experimental data for rough-to-smooth transitions, where the RANS results successfully predict mean velocity profiles and the development of the internal boundary layer. Crucially, the simulations predict extrema in the skin friction coefficient (C_f) profile near the transition, a phenomenon supported by literature but lacking quantitative validation from available experimental data. To address this deficiency, we perform a Direct Numerical Simulation (DNS) of a channel flow with an analogous roughness step. The DNS provides a benchmark dataset to quantitatively verify the RANS predictions of the extrema in skin friction. This work provides an evaluation of RANS performance and limitations, evaluating the contribution of the non-equilibrium physics of spatially developing flows to the overall drag estimation obtained from RANS.