Abstract: The 2D depth-averaged model, quasi-3D model and 3D model are three important types of tools for simulation of cooling water. However, quite limited yet is our understanding of the key factors affecting the outputs and performances of these models of different dimensionality. In this study, a tank experiment was conducted in an artificial meteorological room to test open channel thermal discharging, and the experimental flows were simulated separately using the three models. The simulations were compared with the measured data. This study shows that the key factors in cooling water simulation are effective viscosity, effective diffusivity, and turbulence modeling. All the three models give quite similar temperature distributions in the far-field, but for the near filed of a buoyant jet flow, the 3D model performs better than the other two, more accurately capturing the flow characteristics including circulation zones and dilution mixing. The quasi-3D model produces too small vertical temperature gradients, especially in the far field, thus less accurate than the 3D model; the 3D simulations give a better shape of temperature distributions than the quasi-3D simulations do and much better than the 2D simulations do.