Parameterization of Frequency‐Dependent Internal Wave Drag in Global Ocean Models

A global unconstrained, tidally‐forced ocean model is implemented with the Modular Ocean Model version 6. In the model, a computationally efficient and memory friendly algorithm is employed to determine the instantaneous tidal velocities while the model is running. This filtering technique allows us to separately control the internal wave drag applied to different frequency bands and to improve the accuracy of multiple tidal constituents simultaneously in the same simulation. In the optimally tuned barotropic simulations, the global area‐averaged root mean square errors (RMSE) of the M 2 elevation, calculated against the TPXO solution, is 2.3 cm, whereas in the baroclinic simulations, the smallest RMSE is further reduced to 2.1 cm. This level of accuracy is comparable to that of the most accurate models in the literature, but was achieved without introducing any data assimilation processes or non‐physical forcing to the model. Moreover, the accuracy of individual tidal constituents is maintained as additional tidal constituents are incorporated into the model. Our simulations also revealed some limitations of wave drag parameterization schemes developed based on the linear wave generation theory. One of them is that the linear theory is not applicable to the prediction of tidal energy conversion occurring poleward of the critical latitudes. The comparison of barotropic and baroclinic simulation results suggests that, for the K 1 tide, the sub‐inertial, topographically trapped waves play a non‐negligible role in the global tidal energy budget.