Evaluation of Reynolds-averaged Navier-Stokes turbulence models in open channel flow over salmon redds

This study evaluates computational fluid dynamics(CFD)turbulence closures for Reynolds-averaged Navier-Stokes(RANS)equations against experimental data to model complex open channel flows,like those occurring over dune-shaped salmon spawning nests called"redds".Open channel flow complexity,characterized by near-bed turbulence,adverse pressure,and free surfaces,requires suitable turbulence closure capable of capturing the flow structure between streambed and water surface.We evaluated three RANS models:Standard k-,shear-stress transport(SST)k-@and realizable k-,along with four wall treatments for the realizable k-:Standard,and scalable wall functions,enhanced wall treatment,and an unconventional closure combining standard wall function with near-wall mesh resolving the viscous sublayer.Despite all models generally capturing the bulk flow characteristics,considerable discrepancies were evident in their ability to predict specific flow features,such as flow detachments.The realizable k-model,with standard wall function and mesh resolving viscous sublayer,outperformed other closures in predicting near-wall flow separations,velocity fields,and free surface elevation.This realizable k-model with a log-layer resolved mesh predicted the free surface elevation equally well but lacked precision for near-wall flows.The SST k-@model outperformed in predicting turbulent kinetic energy and provided better predictions of the near-boundary velocity distributions than realizable k-closure with any of the conventional wall treatments but overestimated the separation vortex magnitude.The standard k-o model also overestimated near-wall separation.This study highlights the variability in accuracy among turbulence models,underlining the need for careful model selection based on specific prediction regions.