Sea-ice brine rejection

Why does ice forming at the surface of a polar sea set the deep ocean in motion?

Everything about sea ice happens at the top. The cold is above it, the ice floats on it, and the water below is, if anything, slightly warmer. Yet the heaviest water on the planet — the layer filling the abyss of the Atlantic, Indian and Pacific basins alike — is made in the top few metres of a polar sea in winter. How does an event at the surface reach four kilometres down?

The reflex answer is that cold water sinks. But push on it. Seawater near a polar surface in winter is already at its freezing point, around minus 1.8 degrees Celsius at ordinary ocean salinity, and you cannot chill it much further; the next thing cooling does is make ice, not colder water. Cooling has run out of room. Something else must be supplying the weight.

Ask what ice actually is. It is an ordered lattice of water molecules, and a sodium or chloride ion is the wrong shape and the wrong charge to sit in it. So freezing seawater is not a change of state so much as a separation: the lattice takes the water and refuses the salt.

Where does the refused salt go? It has nowhere to go but the liquid immediately beside the growing crystal, which therefore becomes saltier — and saltier water freezes at a lower temperature, so it stays liquid while the ice around it thickens. Real sea ice traps some of this concentrated liquid in pockets and channels between its crystal platelets, which is why young sea ice is not fresh but carries a fraction of the parent water's salt, commonly a third or less. As the floe ages those pockets drain downward and multi-year ice ends up nearly fresh enough to drink.

Now the density question. What makes a parcel of seawater heavy? Being cold, and being salty. But near the freezing point seawater expands and contracts very feebly with temperature, while it responds to salt just as strongly as ever. So at the poles, salinity is the effective lever, and the ocean has just pulled it: the brine draining out of the ice is at the freezing point and saltier than anything around it. It sinks — in fine plumes, streaming out of the underside of the floe.

And here is the objection that should have occurred to us already. Come summer the ice melts, gives its fresh water back, and the column reassembles. Freeze and melt in the same place, and the year cancels. So what breaks the cancellation?

Separation in space. The ice must be carried away before it melts. Along the Antarctic coast, katabatic winds pour off the ice sheet and shove new ice offshore as fast as it forms, holding open stretches of dark water — coastal polynyas — that freeze again immediately. The same patch of sea freezes over and over all winter, exporting the fresh part downwind and keeping the salt. The Ross Sea, the Weddell, the Adélie Depression and Cape Darnley work as ice factories in exactly this sense.

What they make is High Salinity Shelf Water, ponded in the shallow confinement of the continental shelf until it is dense enough to spill over the shelf break, run down the slope, mix in the water it drags with it, and settle out along the sea floor as Antarctic Bottom Water. The shelf matters twice over: it lets the dense water accumulate somewhere shallow instead of dispersing, and it gives it a slope to fall down.

How much each polynya contributes, and whether the whole system is weakening as the shelves freshen, is under active argument rather than settled — and the Arctic behaves differently again, where brine rejection mostly maintains a cold intermediate layer rather than reaching the floor.

Four refinements, each correcting a natural mis-picture.

The ocean gains no salt in any of this. Salt is not added, only concentrated into less water, and the total is unchanged; what has changed is where the salt sits relative to the fresh.

Nothing that you can see goes down. Ice floats, always. The thing that sinks is invisible — a cold, briny liquid draining out of the floe's underside.

The plume does not tunnel to the abyss. In the open deep ocean it stalls against the stratification below it. It reaches the bottom only because the shelf gives it a reservoir to fill and an edge to fall off.

And the whole process is a pump with a stroke, not a steady leak: freeze, export, melt somewhere else. Interrupt the export and the pump idles even though ice is still forming.

• Why coastal polynyas are called ice factories, and how satellites estimate how much ice they produce.
• The Arctic's cold halocline, and why it protects sea ice from the warmer Atlantic water beneath it.
• Whether the observed freshening and contraction of Antarctic Bottom Water is a trend or a swing.

Every term the collection defines is gathered in the glossary.