6.2 Thermohaline Circulation
The thermohaline circulation (THC)—also called the meridional overturning circulation (MOC)—is the global "conveyor belt" driven by density differences arising from temperature and salinity variations. It ventilates the deep ocean, transports heat poleward, and can exhibit multiple equilibria with profound climate implications.
Atlantic Meridional Overturning Circulation (AMOC)
The AMOC is the Atlantic component of the global THC. Warm, saline surface water flows northward in the upper ocean, loses heat to the atmosphere at high latitudes, becomes dense, and sinks to form North Atlantic Deep Water (NADW). This dense water flows southward at depth (2000–4000 m), eventually upwelling in the Southern Ocean and Indian/Pacific basins.
NADW Formation
Forms primarily in the Labrador Sea (by deep convection in winter) and the Nordic Seas (Greenland, Iceland, Norwegian Seas), with dense water overflowing the Greenland–Scotland Ridge. Temperature ~2–4 °C, salinity ~34.9–35.0 PSU.
Antarctic Bottom Water (AABW)
The densest water mass in the ocean. Forms around Antarctica by brine rejection during sea ice formation (especially in the Weddell and Ross Seas). Temperature –0.5 to 1.7 °C, salinity ~34.7 PSU. Fills the abyssal ocean below NADW.
~18 Sv
AMOC strength at 26°N (RAPID array)
~1000 yr
Full overturning timescale
~1.3 PW
Northward heat transport at 26°N
Derivation: Density-Driven Flow from Pressure Gradient
Step 1: Hydrostatic Pressure in Two Columns
Consider two water columns (equatorial and polar) connected at depth. Each column is in hydrostatic balance. The pressure at depth $z$ in column $i$ is:
$$p_i(z) = p_{atm} + g\int_z^0 \rho_i(z')\,dz'$$
Step 2: Linear Equation of State
The density of seawater depends on temperature and salinity through the linearized equation of state:
$$\rho = \rho_0\bigl[1 - \alpha(T - T_0) + \beta(S - S_0)\bigr]$$
where $\alpha \approx 2\times10^{-4}$ K$^{-1}$ is the thermal expansion coefficient and $\beta \approx 7.6\times10^{-4}$ PSU$^{-1}$ is the haline contraction coefficient. Cold, salty water is denser.
Step 3: Horizontal Pressure Gradient at Depth
If the polar column is denser than the equatorial column ($\rho_{polar} > \rho_{eq}$), a horizontal pressure difference develops at depth. At the surface, sea level adjusts so that pressure is nearly equal. At depth $-H$:
$$\Delta p = g H\,\Delta\rho = g H\,\rho_0\bigl[\alpha\,\Delta T - \beta\,\Delta S\bigr]$$
Step 4: Geostrophic Overturning Flow
This pressure gradient drives a deep flow from high-latitude (high pressure at depth) toward low-latitude. In geostrophic balance on a rotating Earth, the thermal wind relation connects vertical shear to horizontal density gradients:
$$f\frac{\partial v}{\partial z} = -\frac{g}{\rho_0}\frac{\partial\rho}{\partial x}$$
Integrating vertically gives the baroclinic transport, which is the fundamental driver of the thermohaline overturning.
Step 5: Overturning Transport Scaling
A scaling argument for the MOC strength relates the overturning to the meridional density contrast and the depth of the thermocline $D$:
$$\Psi_{MOC} \sim \frac{g\,\Delta\rho\,D^2}{\rho_0\,f}$$
For $\Delta\rho/\rho_0 \sim 2\times10^{-3}$, $D \sim 1000$ m, and $f \sim 10^{-4}$ s$^{-1}$, this gives $\Psi_{MOC} \sim 20$ Sv, consistent with RAPID observations of ~18 Sv.
Stommel Two-Box Model & Multiple Equilibria
Stommel (1961) introduced an elegant two-box model that demonstrates the THC can have multiple equilibria and exhibit hysteresis. The model consists of a warm, saline equatorial box and a cold, fresh polar box connected by an overturning flow:
Stommel Model Equations
$$\frac{dT_1}{dt} = \gamma(T_1^* - T_1) - |q|\,(T_1 - T_2)$$
$$\frac{dS_1}{dt} = \gamma(S_1^* - S_1) - |q|\,(S_1 - S_2) + F_w$$
$$q = k(\alpha\,\Delta T - \beta\,\Delta S)$$
where $\gamma$ is the restoring rate, $\alpha$ and $\beta$ are the thermal and haline expansion coefficients, $q$ is the overturning strength, and $F_w$ is the freshwater forcing perturbation.
Bifurcation and Hysteresis
The model admits two stable states: (1) a thermally dominated mode with strong overturning (present-day-like), and (2) a salinity-dominated mode with reversed or collapsed overturning. As freshwater forcing increases (e.g., from Greenland ice sheet melting), the system can undergo a saddle-node bifurcation, abruptly transitioning to the off state. Crucially, reducing freshwater forcing does not immediately restore the AMOC—this is hysteresis.
Derivation: Stommel Box Model Stability and Multiple Equilibria
Step 1: Non-Dimensionalize the Box Model
Define anomaly variables $\Delta T = T_1 - T_2$ and $\Delta S = S_1 - S_2$. Subtracting the box equations and non-dimensionalizing time by the restoring timescale $1/\gamma$:
$$\frac{d(\Delta T)}{dt} = 1 - \Delta T - |\Delta T - \delta\,\Delta S|\,\Delta T$$
$$\frac{d(\Delta S)}{dt} = \mu - \Delta S - |\Delta T - \delta\,\Delta S|\,\Delta S$$
where $\delta = \beta/\alpha$ is the salinity-to-temperature density ratio and $\mu$ is the dimensionless freshwater forcing.
Step 2: Steady-State Condition
At steady state, setting the time derivatives to zero and defining the flow $q = \alpha\,\Delta T - \beta\,\Delta S$, the system reduces to a single equation for $q$. Using $\Delta T = \Delta T^*/(1 + |q|)$ and $\Delta S = \Delta S^*/(1 + |q|) + F_w/(1 + |q|)$:
$$q = \frac{\alpha\,\Delta T^* - \beta\,\Delta S^*}{1 + |q|} - \frac{\beta\,F_w}{1 + |q|}$$
Step 3: Cubic Equation for Flow Strength
Rearranging for the thermally-dominated regime ($q > 0$) and the salinity-dominated regime ($q < 0$) separately, one obtains a cubic equation in $|q|$:
$$|q|(1 + |q|) = \alpha\,\Delta T^* - \beta(\Delta S^* + F_w) \quad (q > 0)$$
$$|q|(1 + |q|) = \beta(\Delta S^* + F_w) - \alpha\,\Delta T^* \quad (q < 0)$$
For certain ranges of $F_w$, both equations have positive solutions, meaning two stable equilibria coexist.
Step 4: Linear Stability Analysis
Perturbing around a steady state $(\overline{\Delta T}, \overline{\Delta S})$ and linearizing, the Jacobian matrix $\mathbf{J}$ determines stability. The eigenvalues are:
$$\lambda_{1,2} = -\frac{1}{2}\text{tr}(\mathbf{J}) \pm \frac{1}{2}\sqrt{\text{tr}(\mathbf{J})^2 - 4\det(\mathbf{J})}$$
A steady state is stable when both eigenvalues have negative real parts. The system loses stability when $\det(\mathbf{J}) = 0$ (saddle-node bifurcation), which defines the critical freshwater forcing $F_w^{crit}$.
Step 5: Saddle-Node Bifurcation and Hysteresis
At the critical freshwater forcing, the thermally-dominated equilibrium (strong AMOC) collides with an unstable equilibrium and both vanish. The system jumps discontinuously to the salinity-dominated state (collapsed AMOC):
$$F_w^{crit} = \frac{\alpha\,\Delta T^*}{4\beta} \quad \text{(for the simplified symmetric case)}$$
Crucially, reducing $F_w$ below $F_w^{crit}$ does not restore the AMOC -- the reverse bifurcation occurs at a lower $F_w$, creating a hysteresis loop. This irreversibility is why AMOC collapse is considered a climate tipping point.
Step 6: Temperature vs Salinity Feedback
The key asymmetry is that temperature is restored toward the equilibrium by air-sea heat flux (negative feedback), while salinity responds to a fixed freshwater flux (no restoring feedback). This means:
$$\text{Temperature: } \frac{d(\Delta T)}{dt} \propto -\gamma\,\Delta T \quad \text{(self-correcting)}$$
$$\text{Salinity: } \frac{d(\Delta S)}{dt} \propto F_w \quad \text{(externally forced, no restoring)}$$
This mixed boundary condition (restoring for $T$, flux for $S$) is the fundamental reason for the existence of multiple equilibria in the thermohaline circulation.
Diapycnal Mixing & Abyssal Recipes
Munk (1966) showed that the observed abyssal density stratification requires a balance between downward diffusion of heat and upward advection of deep water. His "abyssal recipe" gives:
$$w\frac{\partial \rho}{\partial z} = K_v\frac{\partial^2 \rho}{\partial z^2}$$
Fitting observed profiles yields $K_v \approx 10^{-4}\;\text{m}^2/\text{s}$ and$w \approx 10^{-7}\;\text{m/s}$. The required mixing is much larger than observed in the open-ocean interior ($K_v \sim 10^{-5}\;\text{m}^2/\text{s}$), pointing to enhanced mixing over rough topography (ridges, seamounts) driven by breaking internal tides.
AMOC Slowdown & Younger Dryas
Modern Slowdown Evidence
RAPID array observations (since 2004) and proxy reconstructions suggest the AMOC has weakened by ~15% since the mid-20th century. Climate models project further weakening (but not collapse) under greenhouse warming due to freshwater input from Greenland ice melt and increased Arctic precipitation. The "warming hole" south of Greenland (a region that has cooled while the rest of the planet warms) is a fingerprint of AMOC slowdown.
Younger Dryas (~12,800–11,700 BP)
An abrupt return to near-glacial conditions in the North Atlantic region during the last deglaciation. Triggered by a massive freshwater pulse (likely from glacial Lake Agassiz) that shut down NADW formation. Greenland ice cores record a ~10 °C cooling in decades. AMOC recovery at the end of the Younger Dryas was equally abrupt, demonstrating the "flip-flop" nature of the system.
Meridional Overturning Streamfunction
The overturning is visualised by the meridional overturning streamfunction $\Psi(y,z)$, defined such that:
$$\Psi(y,z) = \int_{x_W}^{x_E}\int_{-H}^{z} v(x,y,z')\,dz'\,dx$$
The maximum of $\Psi$ in the North Atlantic (~18 Sv at ~1000 m depth and ~26°N) measures the AMOC strength. Positive values indicate clockwise overturning (NADW cell); negative values at the bottom indicate the AABW cell.
Python: Stommel Box Model with Bifurcation Diagram
Python: Stommel Box Model with Bifurcation Diagram
Python!/usr/bin/env python3
Click Run to execute the Python code
Code will be executed with Python 3 on the server
Fortran: Two-Box THC Model with Freshwater Perturbation
This Fortran program integrates the Stommel two-box model and performs freshwater forcing experiments to identify the bifurcation point where the AMOC collapses.
Fortran: Two-Box THC Model with Freshwater Perturbation
Fortran============================================================
Click Run to execute the Fortran code
Code will be compiled with gfortran and executed on the server
Water Mass Transformation Framework
The water mass transformation framework (Walin, 1982) quantifies how surface buoyancy fluxes convert water from one density class to another. The transformation rate at density $\rho$ is:
$$F(\rho) = -\frac{1}{\Delta\rho}\oint_{\rho=\text{const}} \frac{B_{surf}}{|\nabla\rho|}\,dl$$
where $B_{surf}$ is the surface buoyancy flux (from heat and freshwater exchange). This framework links air–sea fluxes directly to the overturning circulation: the formation of 18 Sv of NADW requires a corresponding surface buoyancy loss in the subpolar North Atlantic.
Major Water Masses
NADW: $T \approx 2$–4 °C, $S \approx 34.9$–35.0 PSU, depth 1500–4000 m. AABW: $T \approx -0.5$–1.7 °C, $S \approx 34.7$ PSU, fills the abyss below NADW. Antarctic Intermediate Water (AAIW): $T \approx 3$–7 °C, low salinity (~34.2 PSU), sinks to ~800–1000 m. Mediterranean Outflow Water (MOW): warm and salty, spreads at ~1000 m in the North Atlantic.
RAPID Array (26°N)
Since 2004, the RAPID-MOCHA array has continuously monitored the AMOC at 26°N using moored instruments spanning the full Atlantic basin. It measures the Gulf Stream transport through the Florida Straits (~32 Sv), the mid-ocean geostrophic transport, and the Ekman transport. The net AMOC strength averages ~17–18 Sv but shows significant interannual variability, including a notable weakening event in 2009–2010.
Heinrich Events & Dansgaard–Oeschger Oscillations
Paleoclimatic records reveal that the AMOC has undergone dramatic oscillations during glacial periods. Dansgaard–Oeschger (D–O) events are rapid warming episodes (~10 °C in Greenland over decades) recurring at ~1500-year intervals, likely linked to AMOC strengthening. Heinrich events are ice-rafting episodes where massive armadas of icebergs from the Laurentide Ice Sheet discharged into the North Atlantic, depositing distinct layers of detrital carbonate and lithic fragments (Heinrich layers) across the ocean floor.
AMOC Response to Freshwater Forcing
The freshwater flux from Heinrich events is estimated at $\sim 0.1$–$0.3$ Sv over 500–1500 years. This exceeds the critical threshold for AMOC shutdown in both the Stommel box model and coupled GCMs. The bipolar seesaw hypothesis explains the anti-phase temperature relationship between Greenland and Antarctica during these events: when AMOC weakens, the reduced northward heat transport cools the North Atlantic and simultaneously warms the Southern Ocean.
Key Equations Summary
Density from T and S (linear approximation)
$$\rho = \rho_0\left[1 - \alpha(T - T_0) + \beta(S - S_0)\right]$$
where $\alpha \approx 2 \times 10^{-4}\;\text{K}^{-1}$ and $\beta \approx 7.6 \times 10^{-4}\;\text{PSU}^{-1}$.
Munk's Abyssal Recipe Balance
$$w\frac{\partial T}{\partial z} = K_v\frac{\partial^2 T}{\partial z^2}$$
The exponential solution gives a thermocline scale depth $\delta = K_v / w \approx 1000$ m.