Search

Numerical mixing in fixed-coordinate ocean models such as NEMO is known to arise from several sources, including truncation errors in the advection scheme, leakage of the isoneutral explicit mixing into the dianeutral direction, and excessive entrainment of overflow waters, and this spurious mixing has been associated with significant drifts and biases in model simulations. We present results from an ensemble of ¼° forced NEMO simulations in which we investigate the sensitivity of the numerical mixing to various selections of numerical schemes and parameter settings in the model, including the z~ filtered ALE vertical coordinate, increased viscosity, changes in the order of tracer advection and the choice of isoneutral mixing schemes. We also discuss the sensitivity of numerical mixing to the choice of surface forcing product.

The Arctic can be seen as a two-layer ocean: thin (

A recognized deficiency of ocean models with a constant-depth vertical coordinate is for truncation errors in the vertical advection scheme to result in spurious numerical mixing of tracers, which can be substantially stronger than that prescribed by the model's mixing scheme. The z∼ vertical coordinate allows vertical levels to displace in a Lagrangian fashion on time scales shorter than a few days, but reverts to fixed levels on longer timescales, and is intended to reduce numerical mixing from transient vertical motions such as internal waves and tides. An assessment of z∼ in a 1/4° global implementation of the Nucleus for European Modeling of the Ocean model is presented. It is shown that, in the presence of near-inertial gravity waves in the North Atlantic, z∼ significantly reduces Eulerian vertical velocities with respect to those in a simulation with the default z* vertical coordinate; that the vertical coordinate approaches an isopycnal, or adiabatic, surface on short timescales; and that both tendencies are enhanced when the z∼ timescale parameters are lengthened with respect to the default settings. Analysis based on density transformation rates shows that numerical mixing is consistently reduced by of order 10% as the z∼ timescales are lengthened. The realism of the model simulation with different timescale parameters is assessed in the global domain, and it is shown that drifts in temperature and salinity, and the spindown in z* of the Antarctic Circumpolar Current, are reduced with z∼, without incurring significant drawbacks in other metrics such as the strength of the overturning circulation or sea ice cover.

Where earlier generations of ocean models with resolution of 1° or coarser tended to represent wintertime dense water formation in the North Atlantic mainly as a process of open water convection in the Labrador Sea and Nordic Seas, more recent models with higher resolution, in conjunction with observational programmes such as OSNAP, have presented us with a new, more complex, picture. Watermasses are progressively ventilated and lose buoyancy as they propagate cyclonically westward around the gyre, starting with the formation of Subpolar Mode Water close to the eastern boundary, and eventually leading to Labrador Sea Water, which forms part of the lower limb of the Atlantic meridional overturning circulation (AMOC).We present a set of hindcast integrations of a global 1/4° NEMO ocean configuration from 1958 until nearly the present day, forced with three standard surface forcing datasets. We use the surface-forced streamfunction, estimated from surface buoyancy fluxes, along with the overturning streamfunction, similarly defined in potential density space, to investigate the causal link between surface forcing and decadal variability in the strength of the AMOC. We confirm that surface heat loss from the Irminger Sea is the dominant mechanism for decadal AMOC variability, while that from the Labrador Sea has about half the amplitude. The AMOC variability is shown to be related to that of the North Atlantic Oscillation, primarily through the surface heat flux, itself dominated by the air-sea temperature difference, and we show that a metric based on the surface-forced streamfunction has predictive value for AMOC variability on interannual to decadal time scales.

The eddy-permitting 1/4° resolution in NEMO has been known to suffer from significant numerical diapycnal mixing. This arises from truncations in the advection scheme, which causes spurious mixing of tracers where there are transient vertical motions from internal tides and near-inertial waves, as well as from computational modes associated with partly-resolved mesoscale features. Suppressing the near-gridscale noise by increasing the viscosity has been shown to offer a useful reduction in that contribution to numerical mixing, but does not have a significant effect on tides and inertial waves.The z~ scheme replaces eulerian vertical tracer advection across the vertical coordinate surfaces, on time scales less than a few days, with displacements of the coordinate surfaces themselves, in a manner more consistent with the nearly adiabatic nature of near-inertial gravity waves and tides. This has been shown to give substantial reduction in numerical mixing in an idealised configuration, but has yet to be fully evaluated in a global ocean domain. It is shown, using a new prototype eORCA025 global NEMO configuration, that z~ with the default filter timescales reduces the effective diapycnal diffusivity and temperature drifts by only about 10%. Preliminary results will be presented for the sensitivity of the numerical mixing to the z~ timescale and other parameters. The application of z~ to a tidally-forced simulation will also be discussed.

Earth system models enable a broad range of climate interactions that physical climate models are unable to simulate. However, the extent to which adding Earth system components changes or improves the simulation of the physical climate is not well understood. Here we present a broad multivariate evaluation of the North Atlantic climate system in historical simulations of the UK Earth System Model (UKESM1) performed for CMIP6. In particular, we focus on the mean state and the decadal time scale evolution of important variables that span the North Atlantic climate system. In general, UKESM1 performs well and realistically simulates many aspects of the North Atlantic climate system. Like the physical version of the model, we find that changes in external forcing, and particularly aerosol forcing, are an important driver of multidecadal change in UKESM1, especially for Atlantic Multidecadal Variability and the Atlantic Meridional Overturning Circulation. However, many of the shortcomings identified are similar to common biases found in physical climate models, including the physical climate model that underpins UKESM1. For example, the summer jet is too weak and too far poleward; decadal variability in the winter jet is underestimated; intraseasonal stratospheric polar vortex variability is poorly represented; and Arctic sea ice is too thick. Forced shortwave changes may be also too strong in UKESM1, which, given the important role of historical aerosol forcing in shaping the evolution of the North Atlantic in UKESM1, motivates further investigation. Therefore, physical model development, alongside Earth system development, remains crucial in order to improve climate simulations.

Remarkably, global surface warming since 1850 has not proceeded monotonically, but has consisted of a series of decadal timescale slowdowns (hiatus periods) followed by surges. Knowledge of a mechanism to explain these fluctuations would greatly aid development and testing of near term climate forecasts. Here we evaluate the influence of ocean intrinsic variability on global ocean heat uptake and hence the rate of global surface warming, using a 50-member ensemble of eddy-permitting ocean general circulation model simulations (OCCIPUT ensemble) forced with identical surface atmospheric condition for the period 1960-2015. Air-sea heat flux, integrated zonally and accumulated with latitude provides a useful measure of ocean heat uptake. We plot the ensemble mean difference of this quantity between 2000-2009 (hiatus) and 1990-1999 (surge). OCCIPUT suggests that the 2000s saw increased ocean heat uptake of ~0.32 W m-2compared to the 1990s and that the increased uptake was shared between the tropics and the high latitudes. OCCIPUT shows that intrinsic ocean variability modifies the mean ocean heat uptake change by up to 0.05 W m-2or ±15%. Moreover composite analysis of the ensemble members with the most extreme individual decadal heat uptake changes pinpoints the southern and northern high latitudes as the regions where intrinsic variability plays a large role: tropical heat uptake change is largely fixed by the surface forcing. The western boundary currents and the Antarctic Circumpolar Current (i.e. eddy rich regions) are responsible for the range of simulated ocean heat uptake, with the North Pacific exhibiting a particularly strong signal. The origin of this North Pacific signal is traced to decadal timescale latitudinal excursions of the Kuroshio western boundary current.

Leading up to and during the summer of 2015 sea surface temperatures (SSTs) in the eastern North Atlantic Subpolar Gyre reached anomalously low values while in the subtropical gyre just to the SSTs were anomalously warm. Recent observation and modelling studies have found evidence showing that these SST anomalies can be linked to the heat wave experienced over Europe that summer. The latest observation based data still shows anomalously cold temperatures, as well as the anomalously fresh conditions that went along the 2015 cold blob in the upper layers of the eastern North Atlantic Subpolar gyre. A second heat wave over Europe occurred in the summer of 2018 where the SSTs reached another minimum value. Therefore, being able to predict the development, enhancement and persistence of such an anomaly is essential for good seasonal and longer predictions. At present several modelling systems have had difficulties in simulating/maintaining the 2015 cold blob. In this work we apply a novel initialization technique using anomalous initialization from a forced ocean simulation to simulate the 2015 cold blob. Initial results show that the model is able to maintain the cold blob as well as have a strengthening of the cold blob, however, it has difficulties capturing the timing of this strengthening.

The recent decline in the Atlantic meridional overturning circulation (AMOC) has attracted more than a little interest. The strongest AMOC recorded by the RAPID campaign at 26°N was at the start (2004/5), after which it declined about 3 Sv with a pronounced minimum in 2010. Proxies based on temperature and surface elevation have been used to extrapolate the AMOC strength before the RAPID era, and point reasonably reliably to a maximum strength in the mid 1990s, followed by a rise to a maximum at the start of the RAPID campaign in around 2005. Further back, less robust proxy data suggest that the AMOC gradually rose from the 1970s to the peak in 1990. This raises two questions: firstly, what drove these decadal variations in the overturning circulation (and hence of the ocean heat transport); and secondly whether there are observations that lead to useful predictive skill for changes in the AMOC. The surface-forced streamfunction, estimated from modelled/observed buoyancy fluxes, has been shown to be a reasonably good predictor of decadal changes in the overturning strength, preceding the latter with a lead time of about 5 years. although the reliability of the correlations before 2000 is limited by data sparsity, and especially so in the pre-satellite era.To verify a causal link between surface forcing and decadal variations in the AMOC over longer timescales, numerical simulations present a powerful tool. A set of hindcast integrations of a global 0.25° NEMO ocean configuration has been carried out from 1958 until nearly the present day, with a selection of standard surface forcing datasets (CORE2, DFS5.2 and JRA55). These show an evolution of the AMOC strength from 1970 onwards which is consistent, both between forcing datasets and with that inferred from observations. The surface-forced streamfunction is evaluated for these experiments and is used to relate the time evolution of the AMOC to changes in the individual components of the buoyancy flux, and the surface heat loss from the Labrador and Irminger Seas is found to be the dominant predictor of AMOC changes.

The decadal to multi‐decadal temperature variability of the intermediate (700 – 2000 m) North Atlantic Subpolar Gyre (SPG) significantly imprints the global pattern of ocean heat uptake. Here, the origins and dominant pathways of this variability are investigated with an ocean analysis product (EN4), an ocean state estimate (ECCOv4), and idealized modeling approaches. Sustained increases and decreases of intermediate temperature in the SPG correlate with long‐lasting warm and cold states of the upper ocean with the largest anomalous vertical heat exchanges confined to the vicinity of continental boundaries and strong ocean currents. In particular, vertical diffusive processes along the boundaries of the Labrador, Irminger, and Newfoundland basins are important drivers of the recent intermediate depth warming trend observed during 1996‐2014. The overall effect of those processes is captured by a 1‐dimensional diffusive model with appropriate boundary‐like parametrization and demonstrated through the boundary‐focused downward propagation of a passive tracer in a 3D numerical simulation. Our results imply that the slow and quasi‐periodic ventilation of intermediate thermohaline properties and associated heat uptake in the SPG are not strictly driven by convection‐restratification events in the open seas but also receives a key contribution from boundary sinking and mixing. Increased skill for modelling and predicting intermediate‐depth ocean properties in the North Atlantic will hence require the appropriate representation of surface‐deep dynamical connections within the boundary currents encircling Greenland and Newfoundland. Plain language summary The subarctic basins of the North Atlantic Ocean play a fundamental role in regulating the climate system. This occurs notably throughout direct connections between the ocean surface (and hence the atmosphere) and deep oceanic layers, which enable the long‐term sequestration and subsequent propagation of physical and biogeochemical anomalies (e.g. heat, carbon). Here, we employ a multi‐tool approach to investigate the mechanisms by which anomalous heat can penetrate downward in this region. Historical observations gathered during 1950‐2016 and combined with idealized modelling strategy suggest that decadal temperature trends in the intermediate layer strongly correlate with long‐lasting warm and cold states of the upper ocean, suggesting potential for first‐order predictability. Focusing on the well‐observed era (1992‐2016) using a realistic ocean reanalysis and tracer propagation experiments in a numerical model, we show that the associated downward penetration of temperature anomalies receives an important contribution from mixing and advection within the energetic boundary currents of the Labrador Sea, Irminger Sea, and Newfoundland basin.

The Panama Basin serves as a laboratory to investigate abyssal water upwelling. The basin has only a single abyssal water inflow pathway through the narrow Ecuador Trench. The estimated critical inflow through the Trench reaches 0.34 ± 0.07 m s−1, resulting in an abyssal water volume inflow of 0.29 ± 0.07 Sv. The same trench carries the return flow of basin waters that starts just 200 m above the bottom and is approximately 400 m deeper than the depth of the next possible deep water exchange pathway at the Carnegie Ridge Saddle. The curvature of temperature‐salinity diagrams is used to differentiate the effect of geothermal heating on the deep Panama Basin waters that was found to reach as high as 2200 m depth, which is about 500 m above the upper boundary of the abyssal water layer.

The Global Coupled 3 (GC3) configuration of the Met Office Unified Model is presented. Amongst other applications, GC3 is the basis of the United Kingdom's submission to the Coupled Model Intercomparison Project 6 (CMIP6). This paper documents the model components that make up the configuration (although the scientific description of these components are in companion papers), and details the coupling between them. The performance of GC3 is assessed in terms of mean biases and variability in long climate simulations using present-day forcing. The suitability of the configuration for predictabiity on shorter timescales (weather and seasonal forecasting) is also briefly discussed. The performance of GC3 is compared against GC2, the previous Met Office coupled model configuration, and against an older configuration (HadGEM2-AO) which was the submission to CMIP5. In many respects, the performance of GC3 is comparable with GC2, however there is a notable improvement in the Southern Ocean warm sea surface temperature bias which has been reduced by 75%, and there are improvements in cloud amount and some aspects of tropical variability. Relative to HadGEM2-AO, many aspects of the present-day climate are improved in GC3 including tropospheric and stratospheric temperature structure, most aspects of tropical and extra-tropical variability and top-of-atmosphere & surface fluxes. A number of outstanding errors are identified including a residual asymmetric sea surface temperature bias (cool northern hemisphere, warm Southern Ocean), an overly strong global hydrological cycle and insufficient European blocking.

Diabatic upwelling of abyssal waters is investigated in the Panama Basin employing the water mass transformation framework of Walin [1982]. We find that, in large areas of the basin, the bottom boundary layer is very weakly stratified and extends hundreds of meters above the sea floor. Within the weakly stratified bottom boundary layer (wsBBL) neutral density layers intercept the bottom of the basin. The area of these density layer incrops increases gradually as the abyssal waters become lighter. Large incrop areas are associated with strong diabatic upwelling of abyssal water, geothermal heating being the largest buoyancy source. While a significant amount of water mass transformation is due to extreme turbulence downstream of the Ecuador Trench, the only abyssal water inflow passage, water mass transformation across the upper boundary of abyssal water layer is accomplished almost entirely by geothermal heating.

A wide range of statistical tools is used to investigate the decadal variability of the Atlantic Meridional Overturning Circulation (AMOC) and associated key variables in a climate model (CHIME, Coupled Hadley-Isopycnic Model Experiment), which features a novel ocean component. CHIME is as similar as possible to the 3rd Hadley Centre Coupled Model (HadCM3) with the important exception that its ocean component is based on a hybrid vertical coordinate. Power spectral analysis reveals enhanced AMOC variability for periods in the range 15–30 years. Strong AMOC conditions are associated with: (1) a Sea Surface Temperature (SST) anomaly pattern reminiscent of the Atlantic Multi-decadal Oscillation (AMO) response, but associated with variations in a northern tropical-subtropical gradient; (2) a Surface Air Temperature anomaly pattern closely linked to SST; (3) a positive North Atlantic Oscillation (NAO)-like pattern; (4) a northward shift of the Intertropical Convergence Zone. The primary mode of AMOC variability is associated with decadal changes in the Labrador Sea and the Greenland Iceland Norwegian (GIN) Seas, in both cases linked to the tropical activity about 15 years earlier. These decadal changes are controlled by the low-frequency NAO that may be associated with a rapid atmospheric teleconnection from the tropics to the extratropics. Poleward advection of salinity anomalies in the mixed layer also leads to AMOC changes that are linked to processes in the Labrador Sea. A secondary mode of AMOC variability is associated with interannual changes in the Labrador and GIN Seas, through the impact of the NAO on local surface density.

We compare the variability of the Atlantic meridional overturning circulation (AMOC)\ud as simulated by the coupled climate models of the RAPID project, which cover a wide\ud range of resolution and complexity, and observed by the RAPID/MOCHA array at\ud about 26N. We analyse variability on a range of timescales, from five-daily to interannual.\ud In models of all resolutions there is substantial variability on timescales of a few days;\ud in most AOGCMs the amplitude of the variability is of somewhat larger magnitude than\ud that observed by the RAPID array, while the time-mean is within about 10% of the\ud observational estimate. The amplitude of the simulated annual cycle is similar to\ud observations, but the shape of the annual cycle shows a spread among the models.\ud A dynamical decomposition shows that in the models, as in observations, the AMOC is\ud predominantly geostrophic (driven by pressure and sea-level gradients), with both\ud geostrophic and Ekman contributions to variability, the latter being exaggerated and\ud the former underrepresented in models. Other ageostrophic terms, neglected in the\ud observational estimate, are small but not negligible. The time-mean of the western\ud boundary current near the latitude of the RAPID/MOCHA array has a much wider\ud model spread than the AMOC does, indicating large differences among models in\ud the simulation of the wind-driven gyre circulation, and its variability is unrealistically\ud small in the models. In many RAPID models and in models of the\ud Coupled Model Intercomparison Project Phase 3 (CMIP3), interannual\ud variability of the maximum of the AMOC wherever it lies, which is a commonly used model\ud index, is similar to interannual variability in the AMOC at 26N. Annual volume and heat\ud transport timeseries at the same latitude are well-correlated within 15--45N, indicating\ud the climatic importance of the AMOC. In the RAPID and CMIP3 models, we show that the\ud AMOC is correlated over considerable distances in latitude, but not the whole extent of the\ud north Atlantic; consequently interannual variability of the AMOC at 50N, where it is particularly\ud relevant to European climate, is not well-correlated with that of the AMOC at 26N, where it\ud is monitored by the RAPID/MOCHA array.

The control climates of two coupled climate models are intercompared. The first is the third climate configuration of the Met Office Unified Model (HadCM3), while the second, the Coupled Hadley–Isopycnic Model Experiment (CHIME), is identical to the first except for the replacement of its ocean component by the Hybrid-Coordinate Ocean Model (HYCOM). Both models possess realistic and similar ocean heat transports and overturning circulation. However, substantial differences in the vertical structure of the two ocean components are observed, some of which are directly attributed to their different vertical coordinate systems. In particular, the sea surface temperature (SST) in CHIME is biased warm almost everywhere, particularly in the North Atlantic subpolar gyre, in contrast to HadCM3, which is biased cold except in the Southern Ocean. Whereas the HadCM3 ocean warms from just below the surface down to 1000-m depth, a similar warming in CHIME is more pronounced but shallower and confined to the upper 400 m, with cooling below this. This is particularly apparent in the subtropical thermoclines, which become more diffuse in HadCM3, but sharper in CHIME. This is interpreted as resulting from a more rigorously controlled diapycnal mixing in the interior isopycnic ocean in CHIME. Lower interior mixing is also apparent in the better representation and maintenance of key water masses in CHIME, such as Subantarctic Mode Water, Antarctic Intermediate Water, and North Atlantic Deep Water. Finally, the North Pacific SST cold error in HadCM3 is absent in CHIME, and may be related to a difference in the separation position of the Kuroshio. Disadvantages of CHIME include a nonconservation of heat equivalent to 0.5 W m−2 globally, and a warming and salinification of the northwestern Atlantic.

We examine the sensitivity of the Weddell and Ross seas to vertical mixing and surface freshwater forcing using an ocean–sea ice model. The high latitude Southern Ocean is very weakly stratified, with a winter salinity difference across the pycnocline of only ∼0.2 PSU. We find that insufficient vertical mixing, freshwater supply from the Antarctic Ice Sheet, or initial sea ice causes a high salinity bias in the mixed layer which erodes the stratification and causes excessive deep convection. This leads to vertical homogenisation of the Weddell and Ross seas, opening of polynyas in the sea ice and unrealistic spin-up of the subpolar gyres and Antarctic Circumpolar Current. The model freshwater budget shows that a ∼30% error in any component can destratify the ocean in about a decade. We find that freshwater forcing in the model should be sufficient along the Antarctic coastline to balance a salinity bias caused by dense coastal water that is unable to sink to the deep ocean. We also show that a low initial sea ice area introduces a salinity bias in the marginal ice zone. We demonstrate that vertical mixing, freshwater forcing and initial sea ice conditions need to be constrained simultaneously to reproduce the Southern Ocean hydrography, circulation and sea ice in a model. As an example, insufficient vertical mixing will cause excessive convection in the Weddell and Ross seas even in the presence of large surface freshwater forcing and initial sea ice cover.

Offline particle trajectories are obtained for a quasi-global isopycnic-coordinate OGCM using an analytical method, adapted for use with online time-integrated isopycnal and diapycnal mass fluxes. The method is highly efficient, allowing the calculation of large ensembles of such trajectories. These ensembles can be used to establish pathways and transformations associated with the global circulation of water masses on timescales which are well in excess of any feasible model integration length. The method is here used to investigate the important, yet poorly observed, transformation of North Atlantic Deep Water (NADW) through slow spreading, upwelling and diapycnal mixing (defined when and where density decreases below a threshold value). A fundamental problem arises through unsteadiness in the thickness of NADW layers (due to various model flaws and/or intrinsic variability). Particles converge on gridboxes where layers inflate during the online time-integration period. Depending on the degree of layer inflation, only a fraction of NADW particles can be diagnosed as transformed at some point along their respective trajectories. However, the unsteadiness of layer thickness decreases during a 50-year spin-up, implying fewer converged trajectories and an increased fraction of transformed NADW. Using trajectories to trace NADW southward across the equatorial Atlantic, with mass fluxes from years 10, 30 and 50 of model spin-up, the transformed percentage (of NADW exported from the North Atlantic) increases from 17–18% (with fluxes from years 10 and 30) to 41% (using year 50 fluxes). In the latter case, about 30% of the NADW upwells south of 30°S after 500–1000 years. Most of the remaining 70% upwells in the South and North Pacific after 1000–2500 years.

We describe a new Global Ocean standard configuration (GO5.0) at eddy-permitting resolution, developed jointly between the National Oceanography Centre and the Met Office as part of the Joint Ocean Modelling Programme (JOMP), a working group of the UK's National Centre for Ocean Forecasting (NCOF) and part of the Joint Weather and Climate Research Programme (JWCRP). The configuration has been developed with the seamless approach to modelling in mind for ocean modelling across timescales and for a range of applications, from short-range ocean forecasting through seasonal forecasting to climate predictions as well as research use. The configuration has been coupled with sea ice (GSI5.0), atmosphere (GA5.0), and land-surface (GL5.0) configurations to form a standard coupled global model (GC1). The GO5.0 model will become the basis for the ocean model component of the Forecasting Ocean Assimilation Model, which provides forced short-range forecasting services. The GC1 or future releases of it will be used in coupled short-range ocean forecasting, seasonal forecasting, decadal prediction and for climate prediction as part of the UK Earth System Model. A 30-year integration of GO5.0, run with CORE2 (Common Ocean-ice Reference Experiments) surface forcing from 1976 to 2005, is described, and the performance of the model in the final 10 years of the integration is evaluated against observations and against a comparable integration of an existing standard configuration, GO1. An additional set of 10-year sensitivity studies, carried out to attribute changes in the model performance to individual changes in the model physics, is also analysed. GO5.0 is found to have substantially reduced subsurface drift above the depth of the thermocline relative to GO1, and also shows a significant improvement in the representation of the annual cycle of surface temperature and mixed layer depth.

The circulation of the equatorial Pacific is simulated with an implementation of the Miami Isopycnic-Coordinate Ocean Model in a Pacific basin extending from 40°N to 40°S. At a horizontal resolution of 1.25° the annual mean temperature structure on the equator in the model is close to the climatology, with a realistically sharp thermocline, but the speed and mass transport of the Equatorial Undercurrent (EUC) are only 50%–60% of those observed. Decreasing the Smagorinsky shear-dependent viscosity coefficient at this resolution does not increase the EUC speed or transport, but instead increases the grid-scale noise in the eastern part of the undercurrent. Sensitivity studies are carried out in which the resolution is increased in stages up to 0.3125° and the model's two viscosity parameters are reduced. Once the resolution is higher than 1°, the EUC becomes narrower, faster, and stronger as the viscosity is reduced, with the transport increasing more strongly with the decreasing viscosity than wit...

A global isopycnal coordinate GCM is used to investigate the processes that drive the meridional circulation, transformation, and interocean exchange of water masses in the Southern Ocean. The noneddy-resolving model (mesh size 1.25°) includes an active mixed layer, parameterized bolus transport, and seasonally varying surface fluxes. The model gives a plausible picture of the formation and circulation of subantarctic mode water (SAMW) and Antarctic Intermediate Water (AAIW). Progressively denser versions of SAMW and AAIW form in the Indian and Pacific Oceans as the Antarctic Circumpolar Current drifts south and loses buoyancy. SAMW forms predominantly in the Indian Ocean, at a rate of 20 Sv (Sv ≡ 106 m3 s−1), while AAIW forms mainly in the Pacific sector, at a rate of 8.5 Sv. Throughout the circumpolar zone 25°–42.5°S, there is a net formation of 11 Sv of SAMW, largely by surface cooling. This SAMW is exported northward across 25°S into the subtropical gyres. The properties, distribution, and re...