Abstract

In this study, a global isopycnal ocean model (GIM) is described and used for a simulation of variabilities of the global upper ocean during 1992–93. The GIM simulations are compared and validated with both the available observations and simulations with the Geophysical Fluid Dynamics Laboratory Modular Ocean Model (MOM). The observations include sea surface height from TOPEX/Poseidon (T/P), sea surface temperature (SST) from weekly National Centers for Environmental Prediction analysis, and vertical temperature profiles from gridded expandable bathythermographs (XBTs) data. The major differences between the GIM and MOM used in this study are the vertical coordinates, a Kraus–Turner mixed layer, and a tracer-transport velocity associated with an isopycnal-depth diffusion. Otherwise, the two models are formulated in the same parameter space, model configuration, and boundary conditions. The effects of these differences in model formulation on the model simulations are investigated.

Due to the difference in the orientation of interior flow and mixing, SST and the thermocline stratification in the eastern equatorial Pacific in GIM are more sensitive to the wind-driven upwelling than they are in MOM. In GIM there is no effective means to transfer heat between the upwelling cold water and the surrounding warm water since subsurface flow and mixing predominantly occur along isopycnic layers. As a result, the SST tends to be cold and the front tends to be sharp compared with the observations in the wind-driven upwelling region. The sharp front could potentially cause numerical instability in GIM. Thus, a large isopycnal-depth diffusivity has to be used to maintain the model stability since the isopycnal-depth diffusion is the most effective way to reduce the steep slope of isopycnals and the strength of the front associated with the cold upwelling in GIM. But the large isopycnal-depth diffusion results in excessive smoothing in the meridional isotherm doming in the equatorial and tropical thermocline. The trade-off between the numerical instability and the excessive isopycnal smoothing points to the necessity of improvement in the isopycnal-depth diffusion.

Sea level variabilities during 1992–93 simulated with both GIM and MOM are in good agreement with T/P observations. However, MOM poorly simulates the vertical distribution of the seasonal temperature anomalies in the upper ocean (the baroclinic component of the sea level variability) during 1992–93. Due to the lack of a realistic surface mixed layer, the MOM-simulated temperature profiles have a sharp subsurface gradient, which is not evident in both the GIM simulation and the XBT observation. As a result, the region below the subsurface gradient is almost insulated from the influence of the seasonal temperature variation. The Kraus–Turner mixed layer used in GIM helps to improve the model-simulated seasonal variations of the upper-ocean temperature and the background sea level variability. Implications of deficiencies in both GIM and MOM on the altimetric sea level data assimilation and transient tracer simulations are discussed.

Abstract

Output of an eddy-resolving model of the North Atlantic Ocean is used to diagnose the eddy thickness diffusivity coefficient, κ, defined by Gent and McWilliams in their quasi-adiabatic parameterization for transports by mesoscale eddies. The results suggest that κ has large spatial and temporal variations, with negative values about half of the time. The order of magnitude of κ shows a wide range in the western North Atlantic, varying from 10 m² s⁻¹ to 10⁷ m² s⁻¹. Also, the value of κ is considerably affected by the timescale used to define the high-frequency and low-frequency components. The results suggest that κ should be a diagnosed variable that reflects the strength of eddy mixing during a model integration.

Abstract

The structure of ocean-atmosphere annual cycle variability is extracted from the revised Comprehensive Ocean-Atmosphere Data Set SSTs, surface winds, and the latent heat (LH) and net shortwave (SW) surface fluxes using the covariance-based rotated principal component analysis method.

The coupled annual cycle variability is concisely described using two modes that are in temporal quadrature. The first, peaking in June/July (and Dec/Jan), represents monsoonal flow onto Indochina, Central America, and western Africa. The second mode peaks in September/October and March/April when it represents the extreme phases of the SST annual cycle in the eastern oceans.

Analysis of the surface momentum balance in the Pacific cold tongue core shows the equatorial flow, and in particular the zonal wind, to be dynamically consistent with the SST gradient during both the cold tongue's nascent (Jun/Jul) and mature (Sep/Oct) phases; the dynamical consistency improves when the impact of near-surface static stability variation on horizontal momentum dissipation is also considered. Evolution structure of the extracted annual cycle, moreover, shows the easterly wind tendency to lead SST cooling in the off-coastal zone. Taken together, these findings suggest that the Pacific cold tongue westward expansion results from local interaction of the zonal wind and zonal SST gradient, as encapsulated in the proposed “westward expansion hypothesis” -a simple analytic model of which is also presented.

Although positive LH flux tendency leads SST cooling in the off-coastal zone, its modest magnitude (∼5 W m⁻²/mo) indicates that its direct impact on SSTs, while additive, is secondary to the impact of equatorial upwelling. Comparison of the open ocean and coastal annual evolutions reveals that the northward expansion of the Pacific cold tongue likely results from the positive feedback between coastal meridional winds and the upwelled meridional SST gradient, but suggests that the reason for the nonobservance of equatorially antisymmetric SSTs is the counter LH-flux impact northward of the equator.

The comparatively modest SST annual cycle in the northern equatorial Indian Ocean is forced by the Asian-monsoon-driven (i.e., nonlocally forced) surface winds through coastal upwelling along the Somali coast and from the monsoon-cloudiness-impacted net SW surface flux and wind-speed-influenced LH flux in the off-coastal sector.

Abstract

Output of an eddy-resolving model of the North Atlantic is diagnosed in the vicinity of the Gulf Stream (GS), using quasigeostrophic potential vorticity (QGPV), Ertel’s potential vorticity (PV), and particle trajectories. Time series of QGPV show strong input of QGPV by the GS in the top 1000 m of the model ocean. Vigorous wave motions are observed in the vicinity of the model GS, mixing QGPV in the region. The time-mean horizontal QGPV structures show qualitative similarity to those of large-scale climatological PV calculated from hydrographic data by Keffer and that of Lozier. The top 1000 m of the model ocean is characterized by a tongue or an elongated island of high mean QGPV along the GS. It is demonstrated that the tongue is a product of strong QGPV input by the GS, vigorous mixing by eddies, and dissipation of QGPV along the path of the GS. At the intermediate depths, 1000–2500 m, a large region of nearly homogenized mean PV or weakly varying mean QGPV is found to the west of the Mid-Atlantic Ridge. It is located undernearth a region of strong near-surface eddy activity and is in qualitative agreement with a deep and large pool of nearly homogenized PV recently found by Lozier. Below the pool of nearly homogenized PV or weakly varying QGPV, the mean PV and QGPV show substantial horizontal gradient and some vertical gradient at deep levels. This structure is in qualitative agreement with results of idealized model experiments and a theory of baroclinic neutrality of the midlatitude atmosphere proposed by Lindzen that may well apply to this oceanic region of strong baroclinic wave activity.

Abstract

Output of an eddy-resolving model of the North Atlantic is diagnosed in the vicinity of the Gulf Stream (GS), using quasigeostrophic potential vorticity (QGPV), quasigeostrophic potential enstrophy (QGPE), and modified divergent eddy potential vorticity flux, ( V ^′ _h q′ *) _d . A tongue or an elongated island of large mean QGPV along the model GS in the top 1000 m is associated with predominantly downgradient ( V ^′ _h q′ *) _d , suggesting that the horizontal eddy fluxes are balancing a sink of eddy QGPE in most of the tongue or island by converting the mean QGPE into eddy QGPE. Some large upgradient ( V ^′ _h q′ *) _d is observed to the north of the center of the tongue or island, however, suggesting that some of the eddy fluxes in the northern half of the tongue or island of high QGPV are balancing a source of eddy QGPE there by converting eddy QGPE into the mean QGPE. At the intermediate levels of the model, under the GS, eddy QGPE is small, and the role of eddies appears to be mixed; they are forcing the mean to some extent and dissipating the mean to some extent. At the deep levels, the eddies show predominantly a dissipative role, tending to convert the mean QGPE into eddy QGPE. In the region of large time-mean meanders in the model GS, eddies are found to be reinforcing the meanders in a way very similar to that found in diagnoses of atmospheric blocking, which is essentially a large quasi-stationary meander in the subtropical jet. It suggests that the problem of the excessive meander amplitude in the model may be due to an imbalance between eddy forcing in the vicinity of the separation point and zonal acceleration of the GS simulated by the model or due to an unrealistically strong topographic stationary wave forcing in the model.

Abstract

An embedded bottom boundary layer (EBBL) scheme is developed to improve the bottom topographic representation in z-coordinate ocean general circulation models. The EBBL scheme is based on the combined techniques of an embedded topography-following slab, an explicit turbulent bottom boundary layer (BBL), and a generalized pressure gradient formulation. The coupling between the interior z-level model and the EBBL model is achieved by exchanging entrainment/detrainment and pressure gradients at the bottom layer surface, which allows temporal and spatial variations.

The EBBL is implemented into one of the most widely used z-coordinate models, the Modular Ocean Model (MOM). A test problem with a source of dense water on a slope is used. The new EBBL produces significantly more realistic plume spreading than the existing BBL scheme of Killworth and Edwards and is comparable to the results from a topography-following coordinate model (SCRUM), which is believed to be more suitable for such a problem. Calculation of the momentum budget demonstrates that the improved representation of the downslope pressure gradient formulation plays an important role in the simulations of dense slope flows.

Sensitivity experiments with different grid sizes, model parameters, and density contrast between the cold source water and the warm interior water are carried out to test the robustness of the EBBL scheme. In contrast to the BBL model of Killworth and Edwards, which tends to diffuse too much dense water along isobaths, the EBBL model allows dense water to sink across isobaths through a very thin bottom layer into the deep ocean. Even in the coarser-resolution case (¹/₄° and 15 levels) the EBBL produces more realistic deep water than the existing BBL with higher resolution (¹/₈° and 30 levels), and at only one-eighth the computational cost. It is therefore concluded that the EBBL scheme presented here is cost effective and robust to model resolution and mixing parameters, and should be easily implemented in any nontopography-following coordinate ocean model.

Abstract

A realistic oceanic general circulation model is forced with winds observed over the tropical Pacific between1967 and 1979. The structure of the simulated Southern Oscillation is strikingly different in the western andeastern sides of the basin, because the principal interannual zonal-wind fluctuations are confined to the westand are in the form of an equatorial jet. This causes thermocline displacements to have maxima offthe equatorin the west (where the curl of the wind is large) but on the equator in the east. Zonal phase propagation, bothon and offthe equator, is at different speeds in the west and east. The phase pattern is complex, and there is,on interannual time scale, no explicit evidence ofindividuai equatorial waves. These results lead to a modificationof the "delayed oscillator" mechanism originally proposed by Schopfand Suarez to explain a continual SouthernOscillation. The results also permit an evaluation of the various coupled ocean-atmosphere models that simulatethe Southern Oscillation and indicate which measurements are necessary to determine which models are most- relevant to reality.

Abstract

Although the winds on the equator at 28°W in the Atlantic and 140°W in the Pacific have similar seasonal variations, the current fluctuations have pronounced differences. In the Pacific the maximum speed of the Equatorial Undercurrent, attained in the northern spring, can exceed 140 cm s⁻¹, while the minimum speed, in the autumn, is less than 80 cm s⁻¹. In the Atlantic the maximum speed of 80 cm s⁻¹ hardly varies seasonally, although it tends to be largest in the autumn. Analyses of results from a realistic simulation of the equatorial currents indicate that the larger zonal extent of the Pacific, and the seasonal variations of the winds over the western Pacific, which can be out of phase with those in the east, are the principal reasons for the differences between the Atlantic and Pacific.

Abstract

One way to measure the skill of an ocean general circulation model is to evaluate its ability to simulate observed property distributions. Pressure, temperature, and salinity distributions generated by a ⅙° Atlantic Ocean general circulation model are compared with climatological fields on three potential density surfaces, representative of the upper, middepth, and deep ocean waters. The upper ocean property fields are relatively well simulated, a testimony to the model's ability to generally reproduce the wind-driven circulation in the North Atlantic. However, in the middepth and deep ocean, where wind forcing is negligible and buoyant flows associated with deep-water formation play a major role in establishing property distributions, the fields are poorly represented in the ⅙° Atlantic Ocean model. The comparison between the observed and modeled fields indicates several model deficiencies in the representation of intermediate and deep waters and their pathways. Possible model improvements to reduce the mismatch between model and data are proposed.

Abstract

An algorithm is proposed for the computation of streamfunction and velocity potential from given horizontal velocity vectors based on solving a minimization problem. To guarantee the uniqueness of the solution and computational reliability of the algorithm, a Tikhonov regularization is applied. The solution implies that the obtained streamfunction and velocity potential have minimal magnitude, while the given velocity vectors can be accurately reconstructed from the computed streamfunction and velocity potential. Because the formulation of the minimization problem allows for circumventing the explicit specification of separate boundary conditions on the streamfunction and velocity potential, the algorithm is easily applicable to irregular domains. By using an advanced minimization algorithm with the use of adjoint techniques, the method is computationally efficient and suitable for problems with large dimensions. An example is presented for coastal oceans to illustrate the practical application of the algorithm.