Analysis of Roughness Correlations on Ice Geometries via Embedded Large-Eddy Simulation

This study uses embedded large-eddy simulations of a developing turbulent boundary layer over a heated flat plate with realistic ice roughness to assess local effects on velocity ([Formula: see text]) and thermal ([Formula: see text]) log-layer shifts. Validated against experimental data, the numerical setup is used to analyze the mean turbulent flow properties and ultimately assess a selection of equivalent sand-grain roughness [Formula: see text] correlations against computed roughness shifts. Several established models are taken into account that estimate equivalent sand-grain roughness distributions based on geometric properties of the nonuniform rough surfaces. The trend of [Formula: see text] and [Formula: see text] is observed by employing the different correlations. The analysis is complemented with a brief assessment of Stanton-number–friction factor analogies resulting from the simulation outputs. Finally, a multivariate approach is employed to investigate possible cross-dependencies between roughness geometrical properties and shifts. The results indicate that some existing models capture the trends of velocity shifts accurately only in the fully rough regime for [Formula: see text], while for [Formula: see text], they generally struggle to correctly capture the expected trends. Thermal shifts show inconsistent trends across the tested models. The same gap appears in [Formula: see text] estimates via friction analogies, showing no clear dependence on geometry features.