The coordinate and resolution dependence of three second moment turbulent closure models are studied using one-dimensional NCOM experiments and large eddy simulations at Ocean Station Papa. Our results suggest that finer resolution near the base of the mixed layer is critical for better model performance. A mixed layer enhanced vertical grid is proposed that outperforms both the uniform and the stretched grids with significantly fewer vertical layers used. For the new grid, the model accuracy is only dependent on the resolution near the base of the mixed layer regardless of the total number of vertical layers used. However, given the success of the new grid, the lack of representation for the near inertial gravity waves below the mixed layerhas hampered the ability of second moment turbulent closure models on accurate representation of turbulent mixing in the water column.