Abstract

Three ocean-atmosphere models are constructed and evaluated for further studies of climatic variability and climatic sensitivity. All the models are based on the two-layer, highly truncated spectral atmospheric model of Held and Suarez modified to have a seasonal cycle of solar forcing and to interact with simple geographical distributions of land and sea. For simplicity, a 120° sector configuration is adopted.

The first coupled model has as its oceanic component two motionless layers, connected by vertical diffusion and convective adjustment. This is chosen to represent the seasonal cycle in the upper ocean more realistically than a constant-thickness mixed layer. The effect of wind stirring on mixed-layer thickness is added in the second coupled model. The third coupled model is one in which three-dimensional ocean circulation is included, using a relatively fine oceanic grid (1.5°). Long integrations of the three coupled models are carried out in order to understand their intrinsic dynamics.

The first coupled model exhibits many aspects of the seasonally varying atmospheric circulation, such as strong unstable wintertime westerlies, a summertime intertropical convergence zone at latitude 12°, and a low-latitude monsoonal circulation. Eddy transports of heat and momentum are realistic, except for a reduced contribution by stationary waves which is apparently due to the lack of mountains in the model. Eddy statistics are relatively insensitive to changes in the model parameters governing dissipation and precipitation.

Inclusion of a prognostic mixed layer in the second coupled model produces a realistic dependence of layer depth on season and on synoptic variations in wind forcing. The model's annual cycle of upper ocean heat storage and surface temperature agrees well with weather ship data. Statistically significant modifications in the summertime structure of zonal winds and atmospheric temperatures result from a reduced upper ocean heat capacity.

When active ocean circulation is included in the third coupled model, major current systems analogous to the Gulf stream, the oceanic interior flow, and the zonal equatorial currents, are produced on a seasonally averaged basis. Large barotropic oscillations are excited by the atmospheric forcing on synoptic time scales. Reduced ocean temperatures in the tropics result from equatorial upwelling. The annual-average oceanic heat transport is dominated by meridional overturning at low latitudes and is shared with gyre and diffusive components at midlatitudes. The time variation of net heat transport shows a semiannual oscillation in the tropics and an annual cycle at higher latitudes.

Further improvements in the coupled models are suggested. A program of future experiments on climatic variability and sensitivity is outlined.

Abstract

A sea-ice model based bulk-viscous plastic dynamics and 3-layer thermodynamics is coupled to a multilevel primitive equation model of the Arctic Ocean and Greenland Sea. The combined model is forced by inflow through the Faeroe-Shetland Channel and Bering Strait and by observed monthly atmospheric forcing and river runoff. A long-term integration produces a realistic cycle of ice cover, whose extent is strongly influenced by ocean heat transport. The wintertime maximum is controlled by northward heat transport of 0.4 petawatts in the Greenland Sea and by southward transport of ice and water through the Fram Strait. The summertime minimum extent of sea ice is influenced by subsurface flow through the Fram Strait of warm Atlantic water, which rises in winter and thins the ice lying over the Eurasian continental shelf and along the Alaskan and Siberian coasts. The oceanic circulation in the Canadian Basin is anticyclonic at all depths, but changes to cyclonic in the Eurasian Basin below 200 meters. Offshore ice transport in the Barents Sea promotes oceanic convection on the continental shelf through enhanced brine rejection, whereas surface heat loss in the ice-free Greenland Sea produces intermediate water sources similar in T-S characteristics to observed water masses of the region. Two short additional integrations of the coupled model show that the Arctic ice is vulnerable to the environmental effects of atmospheric C0² increase, but relatively insensitive to the maximum proposed amount of Soviet river diversions.

Abstract

A model is presented whereby the thickness and extent of sea ice may be predicted in climate simulations. A basic one-dimensional diffusion process is taken to act in the ice, with modifications due to penetration of solar radiation, melting of internal brine pockets, and accumulation of an insulating snow cover. This formulation is similar to that of a previous study by Maykut and Untersteiner, but the introduction of a streamlined numerical method makes the model more suitable for use at each grid point of a coupled atmosphere-ocean model. In spite of its simplicity, the ice model accurately reproduces the results of Maykut and Untersteiner for a wide variety of environmental conditions. In 25 paired experiments, annual average equilibrium thicknesses of ice agree within 24 cm for 75% of the cases; and the average absolute error for all cases is 22 cm. The new model has fewer computational requirements than one layer of ocean in the polar regions, and it can be further simplified if additional savings of computer time are desired.

Abstract

The circulation of the Arctic Ocean and Greenland Sea is simulated using the 1969 numerical model of Bryan and Cox. The coastline and bottom topography of the region are resolved by a 110 km horizontal grid spacing and by 14 vertical levels. The transfers of mass, heat and momentum at the ocean surface and at open lateral boundaries are specified from observations. In particular, the pattern of wind stress is obtained using a map of mean annual atmospheric pressure; and a scalar multiplier is applied to account for the nonlinear dependence of stress on wind speed. Three experiments with different values of this scalar multiplier are run to simulate the effect of high, medium and low wind stress. The first experiment is carried out for the combined Arctic Ocean and Greenland Sea, while the other two experiments are run for the Arctic Ocean only.

Many of the observed features of the Arctic circulation are reproduced by the simulations. The Greenland Sea exhibits cyclonic flow at all levels and deep convection in its central region. The Beaufort Sea shows anticyclonic flow at the surface and a stable stratification maintained by a halocline. The Arctic Ocean receives bottom water and an intermediate layer of warm Atlantic water through the Greenland-Spitsbergen Passage, and it exports surface water of low salinity into an intense East Greenland Current. The sense of circulation of the Atlantic layer in the central Arctic Ocean, although opposite to that usually inferred from water mass properties, seems to be in reasonable agreement with existing direct current measurements.

For computational reasons, an excessively large eddy viscosity is required in the experiments. As a result, predicted currents are too weak unless a large wind stress is used, but. then an excessive Ekman pumping makes the halocline too deep and erodes the temperature maximum in the Atlantic layer. These results indicate that simulations with finer resolution and reduced viscosity should be more realistic.

Abstract

The circulation of the western North Atlantic is simulated with a primitive equation model that has 5 levels and a horizontal grid size of 37 km. The idealized model domain is a rectangular basin, 3000 km long, 2000 km wide and 4 km deep, which is oriented so that the long axis of the basin is parallel to the east coast of the United States. The nearshore side of the basin has a simple continental shelf and slope, whereas the other sides are bounded by vertical wills. The model ocean is driven by a 2½ gyre pattern of steady zonal wind stress and by a Newtonian-type surface heating. After initialization from a 15-year spin-up with a coarser grid, two experiments are carried out, each of several years duration: the first uses a Laplacian formulation for the subgrid-scale lateral diffusions of heat and momentum, the second uses a highly scale-selective biharmonic formulation for these diffusions. Bottom friction is present in each case.

In both experiments, a western boundary current forms which separates from the coast and continues eastward as an intense free jet, with surface velocities >1 m s⁻¹ for almost 1000 km downstream. In the experiment with biharmonic closure, this simulated Gulf Stream develops large-amplitude transient meanders, some of which become cold-core cyclonic rings and warm-core anticyclonic rings that drift westward. In both experiments, transient mesoscale eddies also form in the broad westward-moving North Equatorial Current, where the simulated thermocline in the model ocean slopes downward toward the north. The remaining regions of the model ocean also contain transient mesoscale eddies, but they are of weaker intensity.

The dominant process of eddy kinetic energy production, in both experiments, is a baroclinic-barotropic instability which is concentrated in the part of the Gulf Stream that is over the continental slope. But where the Gulf Stream lies over the abyssal plains, there is a large reconversion of eddy kinetic energy into the kinetic energy of the time-averaged flow. Eddy kinetic energy is also produced by baroclinic instability in the North Equatorial Current, but at a much smaller rate. In the biharmonic experiment, the eddies transfer considerable kinetic energy downward, and bottom friction is the dominant process of eddy kinetic energy dissipation.

An analysis of the heat transports in the biharmonic experiment, shows that the horizontal transport of heat by eddies is much larger than the subgrid-scale horizontal heat diffusion. In the Gulf Stream region, the eddy heat transport is comparable to the effect of a lateral diffusion coefficient of 10⁷ cm² s⁻¹.

Abstract

A ten-year integration of the Held-Suarez climate model with simplified continents and prescribed, but seasonally varying, ocean temperatures produces mean climatic states that are qualitatively similar to observed seasonal climatology. However, the model's temperature gradients, zonal winds and interannual variability are of lower magnitude than observed.

Differences between these seasonal control climates and those obtained from other decadal integrations with fixed temperature anomalies superimposed on the ocean at different latitudes indicate that the seasonal response of the model atmosphere is a function of anomaly position. Thus, although there are increases in upward motion and precipitation in the vicinity of all ocean temperature anomalies, these changes are of considerably greater magnitude for the equatorial and subtropical anomalies than for the midlatitude anomaly. The vertical motion forced by the equatorial ocean temperature anomaly is that of a Walker-type circulation, with overturning to the east and west of the anomalous heating in all seasons.

Local changes in 750-250 mb thickness are much less sensitive to anomaly position: annual thickness increases in the vicinity of all the anomalies are roughly the same. However, there are large seasonal variations in the responses of the 750–250 mb thickness to each ocean temperature anomaly. In the case of the midlatitude anomaly, the maximum local and downstream increases in thickness occur in summer, while a much weaker response characterizes spring and winter. Local and remote changes in thickness forced by the subtropical anomaly exhibit the greatest seasonal variation, with large and extensive thickness departures in summer and autumn, but comparatively small differences in spring and winter. Significant changes in thickness associated with the equatorial anomaly are confined mainly to the tropics in spring and summer, but in autumn and winter there is substantial interaction with extratropical latitudes in the Southern and Northern Hemisphere, respectively. In winter an intensification of the Northern Hemisphere subtropical jet stream also results from the anomalous equatorial heating.

If the geographically extensive changes in thickness associated with the equatorial and subtropical anomalies in certain seasons are the result of the propagation of waves forced by the heat anomalies, these features appear to contradict the predictions of the linear theory of critical levels. Alternative explanations for these phenomena are proposed, whose merits are examined in Part II of this paper.

Abstract

We test the sensitivity of energy trapping by equatorial basin modes to the presence of mean currents in a series of numerical experiments. At low frequency, trapping persists in the form of modified basin modes only when flow is unidirectional to the east and centered on the equator. If return flow at higher latitudes is westward, severe leakage occurs along the eastern boundary, though mean flows in wide boundary layers can recycle some of this energy back into the interior before it propagates up the coast. If the boundary layers are thin, however, or if flow along the equator is westward, variability is dominated by leaky unstable waves. By contrast, high-frequency basin modes are robust regardless of the mean flow configuration. Thus basin modes may be a more important source of oceanic variability at high frequencies than at low.

Abstract

An equatorial ocean experiment has been carried out, using the primitive equation model of Semtner and Mintz (1977) with a highly conservative differencing scheme, with high horizontal resolution (Δx = 0.50°, Δy=0.25°) and with 14 levels in the vertical. A turbulent equilibrium state has been reached for a 3300 km × 2200 km equatorial ocean, driven by constant 0.5 dyn cm⁻² wind stress, heated at the surface and cooled at the northern and southern walls.

The predicted surface temperature field shows an upwelling-induced cold region along the equator. The temperatures at the equator near the eastern wall are as much as 6°C colder than in the subequatorial regions. Westward moving waves occur in the temperature field a few degrees north and south of the equator. These waves have periods of 33 days, wavelengths of 800 km, and are symmetric about the equator. Their structure is similar to that of equatorially trapped Rossby waves with n=1 in the vertical and m=1 in the horizontal. Shorter wavelength disturbances are found throughout the thermocline near the equator, and these have periods typical of equatorially trapped inertia-gravity waves. The horizontal temperature field at depth suggests that a number of high baroclinic modes are superposed.

The surface flow in the model is characterized by Ekman drift plus transient geostrophic flow off the equator and by weak and variable flow at the equator. A pressure gradient due to the tilt of the sea surface along the equator largely balances the wind stress on the surface layer. Below the surface, this pressure gradient drives an equatorial undercurrent, which slopes upward to the east and intensifies to a maximum of about 100 cm s⁻¹. The undercurrent meanders, with periods of 100 days or more, by as much as 100 km on either side of the equator. Below the current, westward moving cross-equatorial flows with periods of about 44 days sometimes link up the quasi-geostrophic circulations on opposite sides of the equator. These flows appear to be associated with an antisymmetric (in u and p) Rossby wave of the same period having m = 2.

An analysis of energetics shows that the disturbances on either side of the equator are maintained by baroclinic instability, whereas the equatorial undercurrent exhibits mainly barotropic instability. These instabilities lead to transient circulations whose characteristics are similar to those of equatorially trapped neutral waves. Frictional dissipation is concentrated at the equator, and most of the loss of energy is from the eddy circulations rather than from the mean flow.

Abstract

The trapped equatorial standing modes described theoretically by Gent (1979) are reproduced in a single vertical-mode numerical ocean model. integrations are carried out in domains whose longitudinal extents are characteristic of the widths of the Atlantic and Pacific Oceans, as well as in a narrow ocean in which the simplest possible standing mode can exist. The modes are shown to be very insensitive to small changes in basin width and to the inclusion of friction, and somewhat sensitive to the inclusion of the nonlinear terms and to rotation of the rectangular basin relative to the equator. Moreover, they arise spontaneously from simple atmospheric forcing or from random initial conditions. Typically, 20–40% of the energy input to the equatorial ocean remains trapped in a number of distinct standing modes after about nine years of integration time. The wider the ocean domain, the more energy remains trapped near the equator. These unexpected results have important implications for equatorial ocean dynamics and tropical air-sea interaction.

Abstract

The purpose of this paper is to compare two numerical models of vastly different complexity and computational requirements, which have been used recently in a number of midlatitude ocean simulations. Specifically, the two-layer quasi-geostrophic (QG) model of Holland (1978) is compared with the five-level primitive equation (PE) model of Semtner and Mintz (1977) for a wind-driven multi-gyre ocean, with effects of bottom topography and thermal forcing included. The dominant feature of the circulation predicted in the previous PE calculations is a strong free jet, with intense mesoscale transients which are maintained by baroclinic instability.

The configuration of the QG experiment is designed to approximate closely that of the PE experiment, while retaining as much of the simplicity of the Holland (1978) model as possible. The QG model spins up to a state of statistical equilibrium, which is characterized by a meandering jet and by mid-ocean mesoscale eddies with periods and wavelengths much like those in the PE experiment. The time-mean circulations and the distributions of eddy energy in both models are very similar. An energy analysis shows that the free jet in the QG model is more barotropically unstable than in the PE model; however, by reducing the QG upper layer depth to be closer to the thickness of the free jet in the PE model (200 m), this discrepancy disappears. Excellent agreement is also obtained between the volume-integrated energetics of the two models, provided one uses the same lateral diffusion coefficients for momentum and heat in both models.

To gain more insight into physical processes, the computational speed of the QG model is exploited to make additional experiments on the influences of bottom topography, thermal forcing and increased vertical resolution. Bottom topography is found to intensify the upper layer jet and to change substantially the pattern of the deep mean flow. While the presence or absence of topography does not alter the degree of baroclinic versus barotropic instability when the upper layer is 500 m thick, topography does cause a greater proportion of baroclinic instability when the upper layer is thinner. Thermal forcing strengthens the flow in both layers. The use of a three-layer QG model removes the arbitrariness associated with the choice of upper layer thickness: the dominant baroclinic instability of the free jet remains and is concentrated at the interface of the upper two layers.

The results of the present intercomparison suggest that QG simulations will produce the same basic dynamics as PE models in the type of problem considered, using a fraction of the computer time. The saying in computer resources can be profitably applied to understanding the important effects of parameter variations on the oceanic general circulation.