Modeling internal tides over Fieberling Guyot: resolution, parameterization, performance

Terrain-following ocean models are being used to simulate baroclinic tides and provide estimates of the tidal fields for circulation and mixing studies. These models have successfully reproduced elevations with most of the remaining inaccuracies attributed to topographic errors; however, the replication of barotropic and baroclinic velocity fields has not been as robust. Part of the problem is the lack of an adequate observational dataset in the simulated regions to compare the models. This problem was addressed using a dataset collected during the Flow over Abrupt Topography initiative at Fieberling Guyot. To evaluate the capability of the Regional Ocean Model System (ROMS) to simulate baroclinic tidal velocities, the combined tides for four constituents, M-2, S-2, K-1, and O-1, were modeled over Fieberling Guyot. Model inputs, numerical schemes, and parameterizations were varied to improve agreement with observations. These included hydrography, horizontal resolution, and the vertical mixing parameterization. Other factors were evaluated but are not included in this paper. With the best case, semidiurnal baroclinic tides were well replicated with RMS differences between the model estimates and the observations of 1.85 and 0.60 cm s(-1) for the major axes of the tidal ellipses for M-2 and S-2, respectively. However, diurnal K-1 baroclinic tides were poorly simulated with RMS differences of 4.49 cm s(-1). In the simulations, the K-1 baroclinic tides remained bottom-trapped unlike the observed fields, which had free waves due to the contribution of the mean velocity to the potential vorticity. The model did not adequately simulate the mean velocity, and the K-1 tides remained trapped. A resolution of 1 km most accurately reproduced the major axes and mean velocities; however, a 4-km resolution was sufficient for a qualitative estimate of where baroclinic tidal generation occurred. Nine vertical mixing parameterizations were compared. The vertical mixing parameterization was found to have minor effects on the velocity fields, with most effects occurring over the crown of guyot and in the lower water column; however, it had dramatic effects on the estimation of vertical diffusivity of temperature. Although there was no definitive best performer for the vertical mixing parameterization, several parameterizations could be eliminated based on comparison of the vertical diffusivity estimates with observations. The best performers were Mellor-Yamada and three generic length scale schemes.