Turbidity currents

Why do seafloor cables snap hundreds of kilometres from an earthquake, hours after the shaking has stopped?

When the Grand Banks earthquake struck south of Newfoundland in 1929, the transatlantic telegraph cables nearest the epicentre parted at once. That is unremarkable — the ground moved. What is remarkable is what happened next: cables further out to sea kept failing, one after another, in strict order of increasing distance, over the following thirteen hours or so. The last were hundreds of kilometres from the shaking, and broke long after it had died away.

Seismic waves cross an ocean in minutes, not hours, so whatever cut those cables was not the earthquake. Something else set out from the epicentre and travelled — slow for a wave, very fast for anything else on the seafloor. What could move that way through water?

Ask first what could carry the necessary force. Water flowing over a cable does not usually break it. But mud is heavy. Suppose the shaking dislodged a mass of sediment perched on the continental slope, and suppose that sediment mixed into the water above it rather than sliding as a coherent block. What have you made? Not a landslide and not a current in the ordinary sense, but a cloud of water made denser than the water around it.

That density excess is the whole engine, and it is worth putting a number on it, because the intuition is misleading. Mineral grains have a density of roughly 2650 kilograms per cubic metre against seawater's 1027, so each one percent of sediment by volume adds about sixteen kilograms per cubic metre to the mixture — around one and a half percent excess density. That sounds trivial. It is not: the largest density contrasts driving the ocean's own circulation are far smaller than that. A modest suspension is, in oceanographic terms, enormously heavy, and it will run downhill.

Now follow the consequence. As the cloud accelerates it becomes turbulent, and turbulence is what holds the grains up. Ask what would happen without it: the sand settles, the density excess disappears, and the flow stops within minutes. So the flow's speed maintains the very suspension that makes it dense enough to be fast.

Push that one step further and the strange part appears. A fast, turbulent flow does not merely hold what it started with — it scours the bed and picks up more. More sediment means more excess density, which means more speed, which means more scouring. That is a positive feedback, and it can run away. Sedimentologists call the runaway ignition, and the point of it is that a flow triggered by a small failure high on the slope can arrive at the bottom carrying vastly more material than it began with.

But the same loop runs the other way. Reach a gentler gradient and the flow slows, turbulence weakens, the coarsest grains drop out first, and each grain lost makes the flow lighter and slower still. The current dies by the same mechanism that made it. So a turbidity current has two possible fates and a threshold between them, and which side of the threshold it lands on depends on the slope, the grain size, and how much loose sediment the bed has to give.

Does this account for the Grand Banks timing? It does, quantitatively enough to be a test rather than a story. If the breaks were caused by a flow, their times must increase monotonically with distance downslope, and the implied speeds must be highest on the steep upper slope and fall out onto the abyssal plain. Both hold. The speeds inferred are of the order of tens of kilometres an hour — I will not quote a single peak figure, because it depends on where along each cable the break is assumed to have occurred, and estimates in the literature differ substantially.

The instinct is to picture a wall of mud shoving cables aside. That is not quite the failure mode. Measurements in canyons where instruments have been moored in the path of real flows — Monterey Canyon off California and the Congo Canyon are the well-studied cases — show dense, fast basal layers that drag, undercut and bury objects on the bed. A cable fails from being scoured out of its trench, loaded and dragged, not from a single blow.

It is also worth separating what is settled from what is not. That a dense suspension flows downslope and deposits a graded bed is not in doubt: the deposits are found, each one fining upward from coarse base to fine top exactly as a decelerating suspension would drop its load, and they thin and fine with distance from the source. If a supposed turbidite showed no grading, or if the deposits coarsened away from the slope, the account would be in serious trouble. What is genuinely argued is how flows sustain themselves over such long distances, and how much of the material in a distal deposit was eroded en route rather than supplied at the head.

The attribution question deserves the same care. Because these events leave a countable layer, stacks of turbidites are used as a record of past earthquakes on margins such as Cascadia. But an earthquake is not the only trigger: river floods delivering water so muddy it plunges beneath the sea surface can start one, and so can storm waves or simple oversteepening under the weight of accumulating sediment. A single bed rarely carries the signature of its own cause. Confidence comes from correlation — the same bed, of the same age, appearing in cores from separate canyons that no single river flood could have fed at once — and the frequency estimates that follow are only as good as that correlation and the dating behind it.

• How turbidity current deposits build submarine fans, among the largest sedimentary bodies on Earth.

Every term the collection defines is gathered in the glossary.