The Model

The MIT Ocean General Circulation Model employs the incompressible Navier-Stokes equations under the Boussinesq approximation and hydrostatic balance (Jamous, Hill, Adcroft, and Marshall,1997). The model domain is the northwest corner of the North Atlantic Ocean including the Labrador Sea proper as well as some of the outlying Atlantic Ocean. The domain goes as far east as the Mid-Atlantic Ridge and as far south as Newfoundland. The highly energetic North Atlantic Current flows through the southeast corner of the domain. The bottom topography is based on ETOPO5 data. The north and west boundaries are closed; hence, there is no Davis Strait inflow and a very small portion of the Denmark Strait is closed. The Davis Strait actually contributes 2 Sverdrups of very cold, fresh input to the Labrador Sea, and can not be considered a negligible effect (R. Pickart, personal communication). The south and east boundaries are open. The model resolution is [1/8] of a degree latitude by [1/5] of a degree longitude with 34 vertical levels. Therefore, the size of one snapshot of the field is 176 by 185 by 34. The gridpoint separation is between 8.4 km and 15.7 km zonally and 13.9 km meridionally. The Rossby radius of deformation is very small at high latitudes, which makes this nominally an eddy-resolving model.

The model is run for one year with a time step of 15 minutes with the intent to realistically simulate October 1996-September 1997. The model's temperature and salinity (T-S) structure is initialized with a Labrador Sea high resolution climatology produced by M. Visbeck, but with zero velocity. There is no spinup time for the model to reach an equilibrium state, therefore model drift must be investigated. The open boundaries have a 5-gridpoint sponge layer in which temperature and salinity are relaxed to the Visbeck climatology. The relaxation coefficient decreases linearly in the sponge layer to smoothly transition to the interior. The strongest relaxation to the Visbeck data set is in the outermost gridpoint with a relaxation timescale of 1 day. The Visbeck data set only includes a summer and a winter state so the time-varying boundary conditions are derived from a sinusoidally-varying interpolation of the full seasonal cycle. However, the sponge layer is also relaxed to velocity profiles from a global model run (courtesy D. Stammer). One source of inconsistency in the model may be the fact that the boundary conditions are acquired from two different sources. The boundaries will consequently not be in thermal wind balance. One suggestion to resolve this problem is to compute relative velocity from the M. Visbeck density profile and to acquire the vertical mean velocity from the D. Stammer global model. Output from 2 model runs (as adapted and implemented for the Labrador Sea by J. Sheinbaum) will be presented here: run 1 (10 day average output) forced by a 2 degree resolution global state estimation, and run 2 (1 day averaged output) forced by a 1 degree resolution global model run. The model runs of Stammer give output once a month. Therefore, the velocity boundary conditions are interpolated linearly in space and time to the higher resolution of this Labrador Sea model. Both the velocity and T,S boundary conditions extend to the full depth of the water column. Surface boundary conditions must also be supplied to the model. Twice-daily surface windstresses are used from the NCEP Reanalysis. Daily NCEP sensible and latent heat fluxes and precipitation values are used. Surface pressure is computed according to a free surface formulation and is equivalent to the weight of the sea surface height anomaly. Surface temperature and salinity are restored to the M. Visbeck data with a 30 day relaxation timescale.

Biharmonic parameterizations of momentum , temperature, and salinity are used in the model. Lateral diffusion is assumed to be proportional to $\nabla^{4}$ with a coefficient of $10^{11} m^{4}/s$. Vertical diffusion depends on the second vertical derivative of momentum, heat and salt with a coefficient of $10^{4} m^{2}/s$ . These diffusivities are independent of depth. Bottom friction is dependent on total absolute velocity with a coefficient of $10^{-3} m^{-1}$. Convective plumes are not explicitly resolved in the model and must be adequately parameterized in order to accurately predict water mass properties. The convective scheme checks for unstable profiles every 30 minutes (run 1) or every 12 hours (run 2) and mixes pairs of unstable levels once. A delicate balance exists between the frequency of convection and lateral diffusion (J. Sheinbaum, personal communication).

A comparison of the two model runs shows that run 1 handles the late winter convection much better. Run 2 , which used convective adjustment only every 12 hours, produced very noisy vertical velocity fields. Point-by-point comparisons of absolute sea surface height tend to diverge after the convective season. Therefore, run 1 will be used predominantly in comparisons with data. Run 2 will be used only in the spectral description of the model since the output is in 1-day averages and gives a wider frequency range. Special care will be used in judging the high frequency content of run 2.