The secondary sedimentation tank (SST) is a sensitive and complicated process in an activated sludge process. Due to the importance of its performance, computational fluid dynamics (CFD) methods have been employed to study the underflow hydrodynamics and solids distribution. Unlike most of the previous numerical studies, in the present investigation, the performance of three different types of turbulence models, standard k-ε, RNG k-ε and Realizable k-ε, are evaluated. Firstly, two-dimensional axisymmetric CFD models of two circular SSTs are validated with the field observations. Next, comprehensive comparisons are presented of the model predictions of the key physical quantities, such as the concentration of effluent suspended solids (ESS), and returned activated sludge (RAS), sludge blanket height (SBH), turbulent properties and flow and concentration patterns. A surprising result shows that the prediction of the ESS concentration is not sensitive to the change of turbulence models; while remarkable prediction difference can be observed in the inlet zone and near-field of sludge hopper and SBH. The results suggest that more observations inside the inlet zone are needed to achieve better model calibration and correct application of the turbulence model, which can be crucial to optimizing the geometry of inlet structure and sludge hopper as well as changing return solids concentration for the operation. Copyright © 2018. Published by Elsevier Ltd.