The restless Earth

Walk south along the top of Almannagjá in Iceland and there's a wall of black rock on one side of you and a plain dropping away on the other. Guides will tell you that you're standing with a foot on two continents. That isn't quite true, and the truth is better.

The ground under Britain is going nowhere fast. It feels permanent, and we build our vocabulary out of that feeling — set in stone, on solid ground, terra firma. Every bit of it is wrong at the scale that matters.

The outside of this planet is broken into slabs. They're moving right now, roughly at the rate your fingernails grow, and over enough time that's fast enough to close oceans and shove seabeds five miles into the sky. Plate tectonics is the nearest thing geology has to a theory of everything, and until about 1966 hardly anybody believed it.

A cracked shell over something that flows

The rigid outer skin of the Earth — the crust, plus the cold top of the mantle welded to it — is called the lithosphere, and it's typically somewhere between 70 and 150 kilometres thick, thinner under oceans and thicker under old continents. It's cracked into about fifteen major plates and several dozen smaller ones, and the cracks don't follow coastlines. Most plates carry both continent and seafloor, which is why the Atlantic gets wider without Europe and America having any say in it.

Underneath sits the asthenosphere, and this is where the usual description goes wrong. It isn't liquid. It's hot solid rock, close enough to its melting point to creep — the way a glacier is solid ice that flows, or the way a bar of pitch will eventually run downhill if you leave it alone for years. Over a human lifetime it's rock. Over a million years it's a fluid.

The speeds are unimpressive individually. The Atlantic opens by roughly two to three centimetres a year. The fastest boundaries, out in the Pacific, manage ten or more. India is driving north into Asia at about five, which doesn't sound like much until you notice it's been doing it for fifty million years and the Himalayas are the receipt.

Nobody has to take this on trust any more. GPS stations bolted to bedrock measure the movement directly, year after year, and the numbers they give match what the rocks had already implied. The continents are on the move and the measurement is boring.

What's pushing them is not fully settled

Here's where an honest account has to slow down, because this is the part textbooks tend to gloss.

That the plates move is not in doubt. Why they move is still argued over in the literature, and the diagram you were shown at school — a tidy convection cell in the mantle with a plate riding passively on top like a shopping bag on a travelator — is at best a caricature.

Three forces are in the running. Slab pull is the weight of a cold, dense plate that has begun sinking at a subduction zone, dragging the rest of the plate along behind it like a tablecloth going over the edge of a table. Ridge push is gravity acting on the raised, hot topography at a mid-ocean ridge, sliding the plate away downhill. And there's basal drag, the mantle flowing beneath and either helping the plate along or, quite often, resisting it.

Most specialists now think slab pull dominates, and the evidence for that is fairly direct: plates with long subducting edges move several times faster than plates without them. The Pacific plate, which is sinking into trenches around most of its rim, is quick. The African plate, which mostly isn't, is sluggish. But the proportions are genuinely disputed, the mantle beneath is doing its own convecting on its own schedule rather than obediently pushing plates about, and the whole system is coupled in ways that are hard to separate.

So the mechanism isn't a mystery. It's just not a single clean number, and anyone who quotes you one is rounding off an argument that's still going.

Three kinds of edge

Almost nothing dramatic happens in the middle of a plate. The interior of Australia is geologically dull on purpose. The action is at the edges, and there are only three ways two plates can meet.

They can pull apart. Molten rock rises into the gap and freezes on, making brand new crust — which is what the mid-ocean ridge system does, a 65,000-kilometre volcanic mountain range running along the floors of every ocean, and easily the largest single feature on the planet's surface. Iceland is one of the few places it comes up for air, which is what Almannagjá is: a fault scarp on the western edge of a rift valley that's being stretched apart. The reason the two-continents line isn't quite right is that the pulling apart in Iceland is spread across a wide zone of fissures and faults, not concentrated in one photogenic crack.

They can collide. If one side is dense oceanic crust it dives beneath the other and gets recycled into the mantle — subduction — which builds a deep trench, a chain of volcanoes inland from it, and the biggest earthquakes there are. If both sides are buoyant continental crust, neither will go down, and the collision zone simply crumples upward. That's the Himalayas, and it's why Everest is still growing by a few millimetres a year.

Or they can grind past each other sideways, locking and slipping, as California's San Andreas fault does. No volcanoes there, no new crust, just accumulated strain that has to come out eventually.

Draw a dot for every earthquake of the last century and you'll have drawn the plate boundaries without meaning to.

How anyone worked this out

The idea was proposed, dismissed, and left for dead for half a century before it won.

Alfred Wegener, a German meteorologist with no geological credentials to speak of, laid out the case for drifting continents in 1912. He wasn't relying on the jigsaw fit that any schoolchild notices — he had matching rock sequences on facing coastlines, identical fossils of land animals split by thousands of kilometres of ocean, glacial scratches in places that are now tropical, and coal in Antarctica. It was a genuinely strong argument, and geologists rejected it with something close to contempt.

They had one good reason. Wegener couldn't say what force could possibly shove a continent through solid ocean floor, and the mechanisms he suggested were, correctly, shown to be far too weak. Without a mechanism, the rest looked like a heap of coincidences. He died on the Greenland ice sheet in 1930, aged fifty, still arguing.

The answer came from the sea, and mostly from the military. Wartime and post-war surveys equipped ships with echo sounders and magnetometers, and the ocean floor turned out to be nothing like the flat monotonous plain everyone had assumed. Marie Tharp, working at Columbia from soundings she wasn't permitted to go to sea and collect herself, drew the North Atlantic floor by hand and found a deep valley running down the middle of the ridge. Her colleague Bruce Heezen initially dismissed the observation as girl talk. It was a rift, and it changed the argument.

Harry Hess supplied the picture in 1962: new seafloor being made at the ridges and destroyed in the trenches, with the continents carried along as passengers rather than ploughing through anything. He knew how it sounded, and called his paper an essay in geopoetry.

The proof arrived a year later, from magnetism. Molten rock locks in the direction of the Earth's magnetic field as it cools, and that field flips polarity at irregular intervals of hundreds of thousands of years. Fred Vine and Drummond Matthews — and independently Lawrence Morley, whose paper was rejected by two journals — realised that if the seafloor was spreading, it should carry a barcode: symmetrical stripes of alternating magnetic polarity, mirrored either side of the ridge, like a tape recording running outward in both directions from the middle. The surveys found exactly that. By the late 1960s it was over.

The parts of the Earth nobody will ever see

We've never been inside. The deepest hole ever drilled is the Kola Superdeep Borehole in Russia, which reached 12,262 metres in 1989 and stopped because the rock down there was hotter and more plastic than expected and kept closing the hole. That's twelve kilometres into a planet with a radius of 6,371. It didn't even finish the crust.

Everything we know about the interior comes second-hand, from earthquake waves that travelled through it. Two kinds matter. P-waves are compressions and go through anything. S-waves are shear waves and can't pass through liquid at all, which turns out to be the single most useful fact in the subject.

Andrija Mohorovičić spotted the first boundary in 1909, working from a Croatian earthquake — a sharp jump in wave speed that marks the base of the crust, now called the Moho. Beno Gutenberg found the core-mantle boundary in 1913. And in 1936 the Danish seismologist Inge Lehmann, going through records of a New Zealand earthquake, noticed faint P-waves arriving in a shadow zone where nothing should have arrived at all. The only arrangement that explained them was a solid inner core sitting inside the liquid outer one. She published the finding under the title P', which may be the most understated paper title in the history of the subject, and she was right.

The liquid outer core is also the reason there's a magnetic field to record in the first place. Everything on this list connects to everything else, eventually.

Why some volcanoes explode and others just pour

Two volcanoes can sit on the same planet and behave nothing alike, and the variable is mostly silica.

Silica-rich magma is stiff. Silica-poor magma is runny. That single property decides almost everything about how a volcano behaves, because it decides whether gas dissolved in the magma can get out gradually or has to wait.

Runny magma lets gas escape as it goes, which gives you the fire fountains and slow lava flows of Hawaii — ruinous for property, survivable for anyone able to walk. Stiff magma traps gas until the pressure beats the strength of the rock holding it down, and then the whole thing lets go at once. Mount St Helens in 1980 and Vesuvius in AD 79 are the same mechanism at different scales, and the reason Pompeii is such an extraordinary archaeological site is that stiff magma doesn't give much notice.

You can read a volcano's temperament off its shape from a distance. Broad gentle slopes were built by fluid flows and belong to a volcano that pours. Steep symmetrical cones were built out of alternating layers of ash and lava, which is the signature of a mountain that erupts violently and then sits quiet for a long time.

Where the silica comes from is subduction. Water dragged down with a sinking slab lowers the melting point of the mantle above it, and the melt that rises picks up continental crust on the way. That's why the explosive volcanoes cluster along subduction zones and the placid ones tend not to.

The volcanoes in the wrong place

Hawaii ruins the tidy version, because it's sitting in the middle of the Pacific plate with no boundary anywhere near it.

The explanation is a hotspot — a persistent upwelling of unusually hot mantle that stays roughly put while the plate slides over the top of it, punching out a volcano, carrying it away, and punching out another one behind it. The Hawaiian islands get older and more eroded as you go north-west, exactly as that predicts, and the chain continues below sea level as a line of drowned seamounts running most of the way to Kamchatka. There's a distinct bend in it, dated to somewhere around fifty million years ago.

That bend was read for decades as the moment the Pacific plate changed direction. There's now decent evidence that the plume itself drifted southward instead, and the honest answer is probably some of each. Even the deep structure of hotspots — how far down they go, whether they're plumes in the textbook sense at all — is argued about more than the diagrams suggest.

Measuring the shaking

An earthquake is stored strain coming out. Two plates are trying to slide past each other, friction is preventing it, stress builds for years or centuries, and then the fault fails and the accumulated movement happens in seconds.

Magnitude is where people get caught, because the scale is logarithmic and the numbers are therefore lying to your intuition. Each whole number up means about ten times the ground movement and about thirty-two times the energy released. A magnitude 7 isn't a bad 6. It's more than thirty 6s. And a 9 releases something like a thousand times the energy of a 7, which is why the handful of magnitude-9 events on record are in a category of their own.

Charles Richter's original 1935 scale was designed for southern California and specific instruments, and it saturates badly at the top end — very large quakes all come out looking about the same. What's actually quoted now is moment magnitude, introduced in 1979, which is computed from the area of fault that broke, how far it slipped, and the strength of the rock. The numbers are deliberately kept close to Richter's so the old ones still mean something, and news reports still say Richter scale out of habit.

A tsunami is not a tidal wave

The old name has nothing going for it. Tides aren't involved at any point.

When a subduction earthquake ruptures, a slab of seabed jumps vertically, and it lifts the entire column of water above it rather than just ruffling the surface. That's a different animal from a wind wave. In deep ocean it may be under a metre high and hundreds of kilometres long, moving at something like the speed of an airliner, and a ship it passes under won't notice a thing.

It only becomes a wall of water in the shallows, where the front slows down, the back catches up, and all that energy gets forced upward into a shorter, taller, far more dangerous shape. The waves arrive minutes apart, and the first one is very often not the biggest.

The geometry also produces the one warning sign anybody can use. Because the trough frequently arrives before the crest, the sea can withdraw a long way from the shore several minutes before the wave lands, exposing seabed and stranded fish. It looks fascinating. It has killed a great many people who walked out to look. In December 2004, on a beach in Phuket, a ten-year-old British schoolgirl called Tilly Smith recognised what the retreating water meant from a geography lesson she'd had a fortnight earlier, told her parents, and the beach was cleared. It was one of the few along that coast with no fatalities.

What can and cannot be predicted

Knowing all this, why can't anyone say when the next big one will hit?

Because nobody can, and there's no credible method for it. Seismologists draw a hard line between prediction — naming a time, a place and a size, in advance, usefully — and forecasting, which assigns probabilities over long windows. Forecasting works, and it's what building codes are written from: a given fault segment has such-and-such a chance of a large rupture in the next thirty years. Prediction, in the sense of a warning days ahead that you could evacuate a city on, has never been demonstrated to work.

The record is not encouraging. Chinese authorities evacuated Haicheng in February 1975 on the strength of foreshocks and other signs, and a magnitude 7.3 duly followed — a genuine success, and it was never repeated. Eighteen months later Tangshan was destroyed with no warning at all. The US Geological Survey ran a careful, well-funded experiment at Parkfield in California, where moderate earthquakes had arrived at fairly regular intervals for over a century, and formally predicted the next one for the window between 1988 and 1993. It turned up in 2004, on a wired-up, instrumented fault, having shown none of the precursors anyone was watching for.

What has worked is engineering and early warning. Japan, Mexico and several other countries run systems that detect the fast, harmless P-waves radiating from a rupture and broadcast an alert before the slower, destructive S-waves and surface waves arrive. The lead time is seconds to tens of seconds.

Which sounds worthless, and isn't. It's long enough to stop a bullet train, shut a gas valve, halt a surgeon's hand, and get a child under a desk.

The machine is also the reason we're here

All of this reads as a list of hazards, and it's easy to come away thinking of the planet as hostile.

The same restlessness is what makes the place habitable. Subduction drags carbon-bearing rock and sediment down into the mantle and volcanoes hand carbon dioxide back to the atmosphere, and that loop has acted as a slow thermostat over hundreds of millions of years: hotter climates weather rock faster, weathering pulls CO₂ out of the air, and the planet cools back down. It works over geological time, which is precisely why it offers no help on the timescale of anything happening now.

Tectonics also builds the mountains that steer the weather, exposes fresh mineral-rich rock that soils are made from, drives the hydrothermal chemistry at ridge vents where some people think life began, and keeps the deep interior stirred and hot — which keeps the core convecting, which keeps the magnetic field up, which keeps the atmosphere from being stripped away.

Venus is roughly Earth's size and doesn't do any of it. Mars stopped.

So the ground isn't solid, isn't still, and isn't finished. It's a machine that has been running for about four billion years, and once every so often it makes enough noise to remind the people living on top of it that it never switched off.