@article{ValeroBungErpicumetal.2022, author = {Valero, Daniel and Bung, Daniel B. and Erpicum, Sebastien and Peltier, Yann and Dewals, Benjamin}, title = {Unsteady shallow meandering flows in rectangular reservoirs: a modal analysis of URANS modelling}, series = {Journal of Hydro-environment Research}, journal = {Journal of Hydro-environment Research}, number = {In Press}, publisher = {Elsevier}, address = {Amsterdam}, issn = {1570-6443}, doi = {10.1016/j.jher.2022.03.002}, year = {2022}, abstract = {Shallow flows are common in natural and human-made environments. Even for simple rectangular shallow reservoirs, recent laboratory experiments show that the developing flow fields are particularly complex, involving large-scale turbulent structures. For specific combinations of reservoir size and hydraulic conditions, a meandering jet can be observed. While some aspects of this pseudo-2D flow pattern can be reproduced using a 2D numerical model, new 3D simulations, based on the unsteady Reynolds-Averaged Navier-Stokes equations, show consistent advantages as presented herein. A Proper Orthogonal Decomposition was used to characterize the four most energetic modes of the meandering jet at the free surface level, allowing comparison against experimental data and 2D (depth-averaged) numerical results. Three different isotropic eddy viscosity models (RNG k-ε, k-ε, k-ω) were tested. The 3D models accurately predicted the frequency of the modes, whereas the amplitudes of the modes and associated energy were damped for the friction-dominant cases and augmented for non-frictional ones. The performance of the three turbulence models remained essentially similar, with slightly better predictions by RNG k-ε model in the case with the highest Reynolds number. Finally, the Q-criterion was used to identify vortices and study their dynamics, assisting on the identification of the differences between: i) the three-dimensional phenomenon (here reproduced), ii) its two-dimensional footprint in the free surface (experimental observations) and iii) the depth-averaged case (represented by 2D models).}, language = {en} }