In this section, we report on an extended run of the OPYC OGCM, which was forced by the wind stress and fluxes derived from the surface marine observations (see Miller et al., 1994a,b). The OPYC model, developed by Oberhuber (1993), consists of eight isopycnal interior ocean layers fully coupled to a surface bulk mixed-layer model, the latter also including a sea-ice model. The interior layers (fixed potential density) and the mixed layer (variable density) have prognostic thicknesses that vary as a function of space and time; the interior layers may have zero thickness, and the mixed layer has a minimum thickness of 5 m. The interior potential densities are specified in such a way as to yield increased vertical resolution (thinner layers) in the thermocline and less in the deep ocean (thick layers). The low horizontal resolution perforce disallows mesoscale eddy variability, which is probably unimportant in the evolution of large-scale SST anomalies. The ocean model solves the full primitive equations for mass, velocity, temperature, and salt for each layer in spherical geometry with a realistic equation of state. The domain is the Pacific Ocean from 70°S to 65°N and 120°E to 60°W. The grid resolution is 77 by 67 points. Resolution is enhanced near the equator and near the eastern and western boundaries. Open-ocean resolution in the middle latitudes is 4°, which is suitable for examining the large-scale variability. The conditions on horizontal solid boundaries are no-slip for velocity and thermally insulating for temperature. The model's Antarctic Circumpolar Current, however, has periodic boundary conditions and unrealistically connects to itself from 60°W to 120°E (half the global circumference). The surface boundary conditions for interior flow are determined by the bulk mixed-layer model, which is forced by the atmosphere. Frictional drag acts between each layer but most strongly along the bottom boundary, which has realistic topography. Horizontal Laplacian friction and diffusion with variable coefficients are also included.
A full discussion of the dynamics is given by Oberhuber (1993), who used the model in Atlantic Ocean modeling studies, and by Miller et al. (1992, designated MOGB below), who used an earlier version of this model for tropical Pacific Ocean circulation studies. Besides the differences in geometry and the forcing functions to be described, the present model differs from MOGB in the following ways: A horizontal finite-differencing scheme based on Bleck and Boudra (1981) that conserves both enstrophy and potential vorticity is implemented in this version of the model. In addition, for a more realistic mixed-layer depth (MLD) in the middle latitudes, the mean turbulent kinetic energy (TKE) input into the mixed-layer equation for entrainment velocity has been altered to provide for month-to-month variability, and is lower than that of the MOGB model. The net result is to reduce the MLD in the middle and high latitudes of the North Pacific relative to that of MOGB, with the present values being more typical of those observed.
Concerning the OPYC ocean general-circulation model, it is important to point out some potential defects that may obscure our interpretations (Miller et al., 1994a). First, horizontal advection may be underestimated because the model SST climatology is too weak, especially in the Northwest Pacific. Second, the Kuroshio is not well resolved, so the model cannot generate strong western boundary currents. This will impair its ability to advect SST anomalies to nearby open ocean regions. Third, a warm bias in the model winter SST field (approximately 4°C too warm) in the northern North Pacific results in an overly stable stratification that diminishes the effects of entrainment. Last, salinity variations were not included, and these may be important, particularly in high latitudes. Although the OPYC OGCM's rather low resolution has excluded the effects of oceanic mesoscale variability on SST anomaly generation, we expect that such anomalies would have spatial scales too small to provoke important reactions in the large-scale atmospheric fields (Klein and Hua, 1988; Halliwell and Cornillon, 1989; Miller, 1992).
Since no ocean model is perfectly realistic, any model forced by observed total heat fluxes (without any feedback) will establish an oceanic temperature climatology that will depart from the real observations. To circumvent this problem, we forced the model with observed anomalies of heat fluxes superimposed upon the model climatology rather than with the total observed flux fields. This scheme preserves the model ocean climatology, as follows. We first establish the model climatology by forcing with observed monthly long-term mean wind stresses, TKE input, and total heat fluxes derived from bulk formulae using monthly long-term mean atmospheric observations combined with model ocean temperature. After the oceanic system has reached an acceptably equilibrated state (gauged by the absence of drift in mean SST and MLD), monthly means of pertinent fields are saved for use in the later experiments.
Using the same strategy as that just discussed, the oceanic salinity field is stabilized to climatological average observations. However, the equilibrium salinity is very sensitive to the evaporation-minus-precipitation (E - P) field, which is not available at present from observations. Hence, we use Newtonian relaxation to the observed annual mean surface salinity field compiled by Levitus (1982) rather than attempting to specify E - P.
During a 35-year-long spin-up period, the model was forced by monthly long-term mean fields of atmospheric