Towards predictive models for transitionally rough surfaces

We analyze and model the previously presented decomposition for flow variables in DNS of turbulence over transitionally rough surfaces. The flow is decomposed into two contributions: one produced by the overlying turbulence, which has no footprint of the surface texture, and one induced by the roughness, which is essentially the time-averaged flow around the surface obstacles, but modulated in amplitude by the first component. The roughness-induced component closely resembles the laminar steady flow around the roughness elements at the same non-dimensional roughness size. For small --yet transitionally rough-- textures, the roughness-free component is essentially the same as over a smooth wall. Based on these findings, we propose predictive models for the onset of the transitionally rough regime.