Thermohaline circulation is the large-scale movement and transformation of seawater associated with density differences arising from temperature and salinity. It connects surface waters with the deep ocean and helps redistribute heat, freshwater, oxygen, nutrients, and carbon between regions and depths. Its name combines the Greek roots for heat and salt. Dense waters form in a limited number of high-latitude regions, spread through the ocean interior, and eventually return toward the surface through pathways involving winds, mixing, and exchanges between ocean basins.
The term is often associated with the “global ocean conveyor belt,” a simplified picture of interconnected warm surface currents and cold deep currents. Actual circulation is more complex: currents branch, recirculate, mix, and change density, while different water parcels follow different routes and travel times. Thermohaline circulation is also not identical to the Atlantic meridional overturning circulation, or AMOC, which describes a measurable pattern of northward and southward transport in the Atlantic regardless of its driving mechanism.
Understanding this circulation is essential to explaining the distribution of ocean heat and carbon, the ventilation of the abyss, abrupt changes in past climate, and possible responses to ongoing climate change. Modern research therefore treats it as part of a coupled system involving the atmosphere, sea ice, ocean turbulence, and the global water cycle.
Meaning and dynamical framework
Thermohaline circulation is best understood as a description of processes associated with buoyancy and water-mass transformation, rather than as a completely separate current system. Oceanographers distinguish between circulation driven by surface buoyancy exchanges—heating, cooling, evaporation, precipitation, and ice formation—and circulation forced mechanically by wind and tides. These influences interact. A current cannot generally be assigned entirely to one category simply by observing its direction or depth.
A central distinction is between thermohaline forcing and meridional overturning. An overturning circulation is commonly calculated by adding the north–south velocity across an ocean basin and integrating vertically. It describes the exchange between upper and lower layers, but does not establish what powers that exchange. Both wind-driven transport and density-related pressure gradients contribute to the observed overturning. Furthermore, circulation expressed in depth coordinates can differ substantially from circulation expressed in density coordinates, particularly in the Southern Ocean, where density surfaces slope steeply.
Volume transport is usually measured in sverdrups, abbreviated Sv, with
A transport of 15 Sv therefore represents 15 million cubic metres of seawater passing through a section each second. Such values describe the combined movement across large areas, not the speed of an individual current or a literal pipe carrying an unchanging body of water.
Density differences establish pressure gradients, but those gradients alone do not explain the complete energy budget. In a rotating ocean, much horizontal flow is approximately in geostrophic balance: the horizontal pressure-gradient force balances the Coriolis effect associated with Earth's rotation. Friction, turbulence, boundary currents, and vertical motions allow water to move between different parts of the circulation. The ocean's stable stratification also resists the vertical displacement of dense water through lighter water.
Surface cooling can help create dense water that moves downward, but maintaining the global overturning also requires processes that return deep water toward the surface. Winds supply energy and drive upwelling, especially around Antarctica, while tides and winds generate internal motions that ultimately contribute to turbulent mixing. Consequently, the familiar statement that cold water sinks and thereby “pulls” a global conveyor is incomplete. The circulation depends on the joint operation of buoyancy transformation, mechanical forcing, rotation, and the geometry of ocean basins.
Density, stratification, and water-mass formation
Seawater density depends on temperature, salinity, and pressure. Cooling usually increases density, while adding salt also increases it. Pressure compresses seawater, so water at great depth is denser partly because it is compressed, not necessarily because it would sink beneath shallower water if both were brought to the same pressure. Oceanographers therefore use quantities such as potential density and related density variables to compare waters and trace their movement.
For small changes around a reference state, the density response can be represented approximately by
where is a reference density, is the thermal expansion coefficient, and is the saline contraction coefficient. These coefficients are not universal constants: they depend on the water's properties and pressure. In very cold seawater, thermal expansion is relatively weak, making salinity especially important in determining whether surface water becomes dense enough to mix downward.
Most of the ocean is stably stratified, with lighter water overlying denser water. Winter heat loss can erode this layering and deepen the surface mixed layer. If the surface becomes sufficiently dense, ocean convection redistributes heat, salt, and other properties vertically. Strong events can mix water through depths exceeding 1,000 metres in some subpolar regions. Nevertheless, a deep mixed layer is not equivalent to an equally large net downward volume transport: convection involves vigorous upward and downward motions, and newly transformed water must also be exported horizontally.
Freshwater exchanges alter buoyancy without directly adding or removing much heat. Evaporation raises salinity, whereas precipitation, river discharge, and melting ice generally lower it. When sea ice forms, much of the salt is excluded from the ice and returned to the surrounding water through brine rejection. This process is crucial around Antarctica, where it helps produce dense shelf waters. Conversely, melting ice can create a relatively fresh surface layer that suppresses vertical mixing.
A water mass is identified by a characteristic combination of properties inherited from its formation region and subsequently modified by mixing. Temperature–salinity relationships are especially useful, but oxygen, nutrients, and dissolved tracers provide additional information. Formation, sinking, and export need not occur at the same location. Water may acquire its properties over a broad surface region, flow along a shelf, descend a continental slope, and entrain surrounding water before joining a deep current. The resulting deep water is therefore a mixture rather than an unchanged sample of the original surface water.
The Atlantic overturning system
The Atlantic is unusual because substantial volumes of relatively warm upper-ocean water flow northward across the equator, while colder deep water returns southward. This arrangement produces a net northward ocean heat transport across much of the basin. The upper limb includes contributions from the Gulf Stream and its extensions, although the Gulf Stream also belongs to the wind-driven subtropical gyre. The current and the overturning circulation are related but are not interchangeable.
As Atlantic waters move northward, exchanges with the atmosphere cool them and modify their salinity. Their relatively high salinity helps them become dense enough to contribute to deep-water formation. Important transformation regions include the Labrador and Irminger seas and the Nordic seas between Greenland, Iceland, and Norway. Their contributions vary with atmospheric conditions, freshwater distribution, circulation, and the amount of pre-existing stratification. Deep mixing in a single region is therefore not a sufficient measure of the strength of the whole Atlantic overturning.
Dense waters formed in the Nordic seas cross the Greenland–Scotland Ridge through passages including Denmark Strait and the Faroe Bank Channel. As these overflow waters descend, they entrain surrounding seawater, increasing their volume and changing their properties. They combine with waters formed farther south to contribute to North Atlantic Deep Water, which occupies a broad depth range, approximately 1,500–4,000 metres in much of the Atlantic. Its constituent layers retain distinguishable temperature, salinity, oxygen, and tracer signatures.
The lower limb travels southward through boundary currents and interior pathways. A deep western boundary current provides an important route, but floats and tracers show that water can leave the boundary, recirculate, and cross the basin interior. Ridges, fracture zones, and other topographic features guide and divide these flows. Beneath much of the southward-flowing deep water lies colder water of Antarctic origin moving northward, giving the Atlantic more than a simple two-layer structure.
Continuous observations near , begun in 2004, have shown an overturning transport typically of the order of 15–20 Sv when averaged over substantial periods, with considerable variability. The measured strength depends on latitude, averaging period, and the definition of the overturning index. Changes in winds can produce rapid transport fluctuations without requiring an immediate change in high-latitude deep-water formation. Conversely, a buoyancy anomaly formed in the subpolar Atlantic may influence distant circulation only after adjustment through waves, currents, and changes in basin-wide pressure gradients.
Southern Ocean overturning and Antarctic bottom water
The Southern Ocean connects the Atlantic, Indian, and Pacific oceans without the continental barriers that interrupt comparable latitudes farther north. Its strong westerly winds support the Antarctic Circumpolar Current and help bring deep waters toward the surface. This makes the region a central junction of global overturning rather than simply a southern counterpart to North Atlantic sinking.
In the Southern Hemisphere, eastward winds drive a net near-surface Ekman transport to the left of the wind direction, broadly northward across much of the circumpolar belt. The resulting divergence helps draw deeper water upward. Because density surfaces rise toward Antarctica, some deep water can approach the surface by moving largely along those surfaces rather than crossing the full density stratification through vertical mixing. Mesoscale eddies partly oppose the wind-driven overturning, so changes in wind stress do not translate directly into proportional changes in net upwelling.
Once exposed to surface conditions, upwelled water follows different transformation pathways. Some gains buoyancy through heating, precipitation, or ice melt and contributes to northward-flowing upper and intermediate waters. Other water becomes denser through cooling and sea-ice formation. This creates distinct overturning branches: an upper branch that helps close the circulation involving North Atlantic Deep Water, and a lower branch associated with the production and export of Antarctic bottom water.
Antarctic Bottom Water forms from dense waters generated around parts of the Antarctic continental margin, including the Weddell Sea, Ross Sea, and sectors of East Antarctica. Brine rejection can create very saline shelf water even where the wider surface ocean is fresh enough to remain strongly stratified. Dense shelf waters descend the continental slope, mixing with surrounding water to form bottom water that spreads into the abyssal Atlantic, Indian, and Pacific oceans. Not all Antarctic coastal water becomes bottom water; local shelf geometry, sea-ice conditions, and access to the continental slope matter.
Interactions with an ice shelf further complicate these processes. Seawater circulating beneath floating ice can melt it, producing colder but fresher mixtures. Whether the resulting water contributes to dense outflow depends on its initial salinity, mixing history, and surrounding conditions. Changes in ice-shelf meltwater or sea-ice production can therefore affect bottom-water formation without a simple one-to-one relationship with local air temperature. Observed warming and freshening of abyssal waters demonstrate that Antarctic processes can transmit changes far beyond the coastal regions where those waters originate.
Interbasin pathways, mixing, and circulation times
The global circulation links different water masses through branching pathways rather than through a single closed loop. North Atlantic Deep Water spreads southward into the Southern Ocean, where it mixes with waters from other basins and contributes to circumpolar deep-water layers. Portions eventually enter the Indian and Pacific oceans, while others return toward the surface in the Southern Ocean. Antarctic-origin bottom waters spread northward beneath much of this deep circulation.
The North Pacific has no modern deep-water formation comparable in scale to that of the North Atlantic. Its surface waters are generally too fresh, and therefore too buoyant, to produce widespread abyssal ventilation through winter cooling. The Pacific nevertheless forms intermediate waters in subpolar regions and receives deep and bottom waters from the south. Long residence away from the atmosphere, together with the decomposition of sinking organic material, helps explain the high nutrient concentrations and relatively low oxygen found in much of the deep North Pacific.
Upper-ocean return routes include exchanges around southern Africa and through the Indonesian seas. Agulhas leakage transfers warm, saline Indian Ocean water into the South Atlantic around the southern tip of Africa. The Indonesian Throughflow moves water from the Pacific into the Indian Ocean through passages in Indonesia. These exchanges affect the heat and freshwater budgets of entire basins, but neither is a fixed conveyor segment: the waters involved mix, recirculate, and respond to regional winds and pressure differences.
Mixing is essential because it changes water density and permits movement between density classes. In the ocean interior, turbulent mixing is often weak, but it becomes much stronger near rough seabed topography, ridges, continental slopes, and some passages. Tidal currents and wind-generated motions produce internal waves that can break and generate turbulence. The uneven distribution of mixing means that the return of deep water cannot be represented accurately as uniform, extremely slow rising motion spread over every ocean basin.
Travel times are equally heterogeneous. Fast boundary-current segments, prolonged recirculation, and gradual mixing can coexist within the same broad pathway. Some connections operate over years to decades, whereas the ventilation ages of deep waters can reach centuries or more than a millennium. A single “conveyor-belt time” is therefore misleading. Moreover, an apparent age derived from a tracer is not necessarily the elapsed time since every molecule in a sample last contacted the atmosphere. Mixing combines waters with different histories, and some tracers are also affected by radioactive decay or changing atmospheric concentrations.
Heat, carbon, oxygen, and ecological effects
Overturning transports heat because upper and lower branches carry water with different temperatures. At a given latitude, the net heat transport depends on the distribution of velocity and temperature across the full ocean section, not simply on the volume transport of one current. Near subtropical North Atlantic latitudes, total northward ocean heat transport is of the order of a petawatt, or watts. The overturning is an important contributor, alongside horizontal gyre circulation.
This heat transfer influences the North Atlantic atmosphere and surrounding land, but western Europe is not warmed by ocean currents alone. Prevailing winds, atmospheric heat transport, continental geometry, and the ocean's seasonal heat storage also shape regional climate. Changes in overturning can alter sea-surface temperature patterns and air–sea heat exchange, which in turn affect atmospheric circulation. The climatic result is therefore a spatial pattern of changes rather than a uniform temperature adjustment around the entire Atlantic.
The circulation also connects the atmosphere with the ocean's large dissolved-carbon reservoir. Cold surface water can dissolve more carbon dioxide than warm water under otherwise comparable conditions. When newly ventilated waters move into the interior, they carry dissolved inorganic carbon, including carbon derived from human emissions. At the same time, organic matter produced near the surface sinks and decomposes at depth through the biological carbon pump. Overturning and mixing redistribute this regenerated carbon and determine when it may return to contact with the atmosphere.
Oxygen and nutrients provide another record of circulation. Newly ventilated waters usually acquire oxygen at the surface, whereas respiration consumes oxygen in the interior and releases nutrients such as phosphate and nitrate. Deep waters that have spent longer away from the surface commonly contain less oxygen and more regenerated nutrients. This broad relationship has important exceptions because oxygen solubility, initial saturation, biological productivity, and mixing also vary. Circulation changes can alter the supply of oxygen to deep habitats and the delivery of nutrients to productive surface regions.
Sea level is affected through both density changes and circulation dynamics. Warming expands seawater, while changes in currents modify the slopes of the sea surface required to balance pressure gradients. A weaker Atlantic overturning can contribute to regional sea-level rise along parts of the North American Atlantic coast, although the magnitude depends on winds, other circulation changes, and the details of the forcing. These dynamic effects are distinct from the addition of water by melting land ice. The combined influence on heat, carbon, and sea level makes overturning a major component of the Earth system rather than merely a mechanism for moving seawater.
Development of the concept and modern observations
The idea that polar cooling could sustain deep-ocean circulation developed from early measurements showing that deep tropical waters were unexpectedly cold. In the eighteenth century, the physicist Benjamin Thompson, Count Rumford, argued that cold water from high latitudes could spread beneath warmer surface waters. Nineteenth-century expeditions expanded temperature measurements across the ocean, while later surveys established the salinity and oxygen patterns needed to distinguish major deep-water masses.
Twentieth-century theory clarified both the importance and the limitations of density-based explanations. In 1961, Henry Stommel published a simple model showing how competing temperature and salinity effects could permit more than one circulation state under suitable conditions. In 1966, Walter Munk examined the balance between deep-ocean upwelling and vertical diffusion, helping frame the problem of how dense water returns upward. Wallace Broecker later popularized the global conveyor-belt image, especially during the 1980s and 1990s, linking ocean circulation with abrupt climate change.
Modern measurements combine several methods because no single instrument can observe the entire system. Ship-based hydrographic sections measure temperature, salinity, oxygen, nutrients, and tracers from the surface to the seabed. Repeated sections reveal changes in water-mass properties and ocean heat content. Current meters and moored instruments provide continuous records at selected locations, while ocean-bottom pressure measurements help constrain variations in the mass and pressure fields associated with transport.
The RAPID array near began monitoring the AMOC in 2004. It combines measurements of density near the basin boundaries and mid-ocean ridge, transport through the Florida Straits, and wind-derived Ekman transport. The OSNAP observing system, deployed across the subpolar North Atlantic in 2014, measures overturning and property transports closer to major water-transformation regions. Its observations have shown that variability in subpolar overturning cannot be inferred from Labrador Sea convection alone and have emphasized the importance of transformation east of Greenland.
The Argo program provides broad coverage of upper-ocean temperature and salinity, conventionally to about 2,000 metres, while Deep Argo extends measurements into the deeper ocean. Biogeochemical floats add oxygen and other variables. Satellite altimetry measures sea-surface height, offering information about circulation and changes in ocean heat storage, but it does not directly measure the full overturning.
Transient tracers, including chlorofluorocarbons and sulfur hexafluoride, help identify recently ventilated waters. Radiocarbon provides information over longer timescales. Together with numerical models and statistical reconstructions, these observations constrain pathways and transport. Nevertheless, the short duration of continuous basin-wide records, incomplete abyssal coverage, and large natural variability make long-term trend estimation substantially harder than detecting individual years of unusually strong or weak circulation.
Evidence from past climates
Past climates show that ocean overturning can reorganize substantially. Evidence comes from marine sediments, ice cores, corals, and other archives rather than direct current measurements. Each climate proxy records a particular environmental process, so reconstructions are strongest when several independent indicators agree. For example, sediment grain size can reflect near-bottom current strength, whereas carbon isotopes can help distinguish water masses and their accumulated biological signals.
During the Last Glacial Maximum, approximately 26,500–19,000 years ago, ocean conditions differed markedly from those of the present. Large continental ice sheets, lower sea level, altered winds, and extensive sea ice changed the boundary conditions for deep-water formation. Evidence indicates a generally shallower northern-sourced overturning cell in the Atlantic, with southern-sourced waters occupying more of the deep basin. However, the exact strength of the glacial circulation and its variation through time remain harder to establish than this broad change in water-mass geometry.
Abrupt glacial events provide evidence for connections between freshwater input, Atlantic circulation, and regional climate. Heinrich events involved major episodes of iceberg discharge into the North Atlantic, leaving distinctive layers of ice-rafted debris in marine sediments. Melting icebergs supplied freshwater that could strengthen surface stratification. Several of these episodes coincide with evidence of reduced northern deep-water influence, although the sequence and relative importance of ice-sheet dynamics, atmospheric changes, and ocean feedbacks varied among events.
The Younger Dryas, approximately 12,900–11,700 years ago, was a return to colder conditions around much of the North Atlantic during the last deglaciation. Reduced Atlantic overturning is a leading part of explanations for this event, but the origin, routing, timing, and magnitude of the freshwater disturbances remain active research questions. A different event about 8,200 years ago followed a large freshwater release associated with the drainage of glacial lakes in North America and produced a shorter-lived cooling in the North Atlantic region.
Changes in interhemispheric heat distribution are often described through the bipolar seesaw. When northward Atlantic heat transport weakens, the North Atlantic can cool while heat accumulates farther south, contributing to an Antarctic response on a different timescale. This is not a simple instantaneous exchange between the poles: ocean heat storage, atmospheric adjustments, and Southern Ocean processes affect the timing. Paleoclimate evidence establishes that circulation and climate can undergo large coupled changes, but glacial events cannot be transferred directly to a modern world with different ice sheets, greenhouse-gas concentrations, and freshwater sources.
Warming, freshwater forcing, and stability
Human-caused warming affects the overturning through several interacting processes. Surface heating makes seawater lighter and can strengthen upper-ocean stratification. Changes in rainfall, evaporation, river runoff, and land-ice melt alter salinity. Increased freshwater input to high-latitude formation regions can make surface water less likely to reach the density needed for deep mixing. However, the response depends on where freshwater enters, how currents redistribute it, and whether changes in heat loss, winds, or salt transport compensate for its buoyancy effect.
A major theoretical mechanism is the salt-advection feedback. In a circulation state that imports saline water into a northern sinking region, strong overturning can help maintain the conditions supporting its own continuation. If the circulation weakens, reduced salt import may allow further freshening and additional weakening. Under some model configurations, this feedback permits multiple equilibria and hysteresis: the forcing required to restore a strong circulation can differ from the forcing at which it weakened. Whether and how closely the present ocean approaches such thresholds depends on the full basin freshwater budget and other stabilizing or destabilizing processes.
Experiments with Earth system models generally project a weakening AMOC during the twenty-first century as greenhouse-gas concentrations rise. The Intergovernmental Panel on Climate Change's Sixth Assessment Report, published in 2021, assessed that the AMOC was very likely to weaken during that century under all emissions scenarios considered. It also expressed medium confidence that an abrupt collapse would not occur before 2100. These statements distinguish a projected decline in average strength from a transition to a substantially different circulation state; they do not provide a precise calendar date for a possible threshold.
Historical attribution is more difficult than future forced-response experiments. Direct continuous measurements cover only a small fraction of the timescales over which the circulation varies naturally. Reconstructions based on sea-surface temperature patterns, sediments, or other indirect indicators extend farther back but depend on assumptions about how the indicator relates to transport. A relatively cool region of the subpolar North Atlantic, for example, can be consistent with changes in ocean heat transport, yet temperature there is also influenced by atmospheric forcing, mixing, and heat exchange with the air.
Model uncertainty includes biases in surface salinity, deep-water formation, overflow representation, eddies, and freshwater exchanges between basins. Many important processes occur on scales smaller than the grids used in long climate simulations and must be represented approximately. Assessments using the Coupled Model Intercomparison Project therefore examine a range of model responses rather than treating one simulation as a prediction. Research on statistical early-warning indicators likewise depends on the suitability of the chosen observations and assumptions about noise, forcing, and equilibrium. Such indicators are not equivalent to direct measurements of distance from a circulation threshold.