Impact of transient freshwater releases in the Southern Ocean on the AMOC and climate

Abstract

The bipolar ocean seesaw is a process that explains the competition between deep waters formed in the North Atlantic (NA) and in the Southern Ocean (SO). In this picture, an increase in the rate of formation of one of these water masses is made at the expense of the other. However, recent studies have questioned the effectiveness of this process. Namely, they show that adding freshwater in the SO can reduce deep water formation in the SO as well as in the NA. In this study, we explore the mechanisms and time scales excited by such a SO freshwater release by performing sensitivity experiments where a freshwater input is added abruptly in the ocean, south of 60°S, with different rates and durations. For this purpose, we evaluate the separate effects of wind, temperature and salinity changes, and we put the emphasis on the time evolution of the system. We find three main processes that respond to these freshwater inputs and affect the NA Deep Water (NADW) production: (i) the deep water adjustment, which enhances the NADW cell, (ii) the salinity anomaly spread from the SO, which weakens the NADW cell, and (iii) the increase in the Southern Hemisphere wind stress, which enhances the NADW cell. We show that process (i) affects the Atlantic in a few years, due to an adjustment of the pycnocline depth through oceanic waves in response to the buoyancy perturbation in the SO. The salinity anomalies responsible for the NADW production decrease [process (ii)] invades the NA in around 30 years, while the wind stress from process (iii) increases in around 20 years after the beginning of the freshwater perturbation. Finally, by testing the response of the ocean to a large range of freshwater release fluxes, we show that for fluxes larger than 0.2 Sv, process (ii) dominates over the others and limits NADW production after a few centuries, while for fluxes lower than 0.2 Sv, process (ii) hardly affects the NADW production. On the opposite, the NADW export is increased by processes (i) and (iii) even for fluxes smaller than 0.1 Sv. The climatic impact of the freshwater release in the SO is mainly a cooling of the Southern Hemisphere, of up to 10°C regionally, which increases with freshwater release fluxes for a large range of values.

Appendix: Formation–diffusion equilibrium

In this appendix, we explain how to obtain the Eq. 1 and how to calculate the terms of this equation in the model.

Following Speer et al. (2000), we firstly calculate a density flux defined as:

$$ f=-{\frac{\alpha g Q }{\rho_0 C_p }}+{\frac{\rho(T,0)\beta S} {\rho_0 (1-S)}} E $$

(3)

where S is the surface salinity, T is the surface temperature, C _p is the specific heat of seawater, ρ is the density, E is the net freshwater fluxes, α is the thermal expansion coefficient and β the haline contraction coefficient.

From the buoyancy flux, we can define the transformation by air–sea fluxes from one density class to another:

$$ {\fancyscript{T}}=\int\limits_{T} dt \int\limits_{S_b} dxdy f \delta ( \rho \prime -\rho) $$

(4)

Its derivative with respect to buoyancy (b = −gρ/ρ₀) is called the formation (${\mathcal{F}}$) and represents the amount of new formation of a given water mass by air-sea buoyancy flux on a period T.

$$ {\mathcal{F}}=-\partial_b{\fancyscript{T}} $$

(5)

We define ${\fancyscript{A}}$ as the net advective flux across isopycnals:

$${\fancyscript{A}}=\int\limits_{S_{b}}({\overrightarrow {V} }\cdot {\overrightarrow{n}})d\Sigma $$

(6)

where dΣ is the isopycnal surface area, ${\overrightarrow{V}}$ is the velocity and ${\overrightarrow{n}}$ is the unity normal vector to isopycnal.

We define ${\fancyscript{M}}$ as the mixing or diffusive flux across isopycnals:

$${\fancyscript{M}}=\int\limits_{S_{b}}-\kappa\partial_n{b}d\Sigma $$

(7)

where κ is the diapycnal diffusivity.

Ignoring volume fluxes across the sea floor and the sea surface, volume and mass conservation then write respectively:

$$ \partial_t{\delta v} = (-\partial_b{\fancyscript{A}}-\partial_b{\psi^c})\delta b $$

(8)

$$ \partial_t{\delta (vb)}= (-\partial_b{({\fancyscript{A}}b)}-\partial_b{\fancyscript{M}}-b\partial_b{\psi^c})\delta b $$

(9)

where ψ^c is the streamfunction at the control surface. These equations give:

$${\fancyscript{T}}-{\fancyscript{A}}-\partial_b{\fancyscript{M}}=0 $$

(10)

Lastly, we eliminate advection ${\fancyscript{A}}$ to obtain the volume balance of the isopycnal b (Speer et al. 2000) given by Eq. 11:

$$ \partial_t{V_{b}} =-\partial_{b}{\psi^c}-\partial_{b}{\fancyscript{T}}+\partial_{{b}^2}{\fancyscript{M}} $$

(11)

which allows to define the diapycnal mixing term as ${\mathcal{D}}=\partial_{{b}^2}{\fancyscript{M}}$ and the streamfunction export term as Ψ = ∂ _b ψ^c. With this notation, we obtain the Eq. 1.

We calculate the different terms of Eq. 1 (Table 2) for different numerical experiments using δρ = 0.05 kg/m³ with monthly mean output for the different oceanic variables, which is precise enough according to the analysis from Speer et al. (2000).