Abstract: In this paper, a layer-averaged fully coupled model is applied for simulating both laboratory lock-release and constant-flux turbidity currents with the aim of analyzing the applicability and uncertainties of empirical relations for the water and sediment entrainment process. For this purpose, the performance of four empirical formulae (i. e., e_w59, e_w86, e_w87 and e_w01) for the water entrainment and five (i. e., E_s77, E_s86, E_s87, E_s93 and E_s04) for the sediment erosion are compared. The following understandings are drawn from numerical case studies. The water entrainment has mild effect on small-scale turbidity currents, though simulation results with the e_w01 relation are slightly better than those by others. It might be due to the fact that the e_w01 relation has taken into account of the effects of both bottom and upper interface resistances. Lock-exchange turbidity currents are mainly depositional and thus less sensitive to the choice of erosion relations. For constant-flux turbidity currents, the E_s87 and E_s93 relations appear to perform best. This must be attributed to the fact only these two erosion relations have been calibrated using experimental turbidity current data. This study should help further studies for developing and calibrating empirical relations of mass exchange for turbidity currents.