Appearance
1.6 — Earthquakes
At 5:12 in the morning on 18 April 1906, about 477 kilometres of the San Andreas Fault let go at once. In places, fences that had run straight across the fault were found afterwards offset by more than six metres — one half of a farm had moved that far past the other half in under a minute. The shaking lasted about a minute. Then San Francisco burned for three days, because the same shaking that cracked the buildings also broke every water main in the city, and the fire brigade arrived at hydrants that produced nothing.
The investigation into that earthquake, led by Harry Fielding Reid, produced the explanation that still stands. This page works through it, and then through the four things that actually decide whether an earthquake kills nobody or kills a quarter of a million people.
What is actually happening
Rock is elastic. Push on it and it bends, store the push and it stores energy, and let go and it springs back. That is not obvious about granite, but it is true, and it is the whole mechanism.
Two plates are trying to slide past each other at a few centimetres a year. Friction locks them together along the fault. The plates keep moving anyway, so the rock on either side of the locked patch bends, more each year, storing elastic energy exactly as a bent steel ruler does.
Eventually the stored stress exceeds the friction holding the fault shut. The fault slips, the bent rock snaps back to straight, and all the energy stored over decades or centuries is released in seconds. Reid called this elastic rebound and it explains the 1906 offsets precisely: the fences had been bending along with the ground for fifty years, and the earthquake simply released the accumulated bend.
Three words to keep straight. The focus (or hypocentre) is the point underground where the rupture starts. The epicentre is the point on the surface directly above it. And the rupture does not happen at a point — it starts at the focus and tears along the fault at roughly 2 to 3 kilometres per second, which is why a great earthquake shakes for minutes rather than instants. The 2004 Sumatra rupture ran for about 1,300 kilometres and took roughly eight to ten minutes to finish tearing.
Where earthquakes happen, and why depth matters
Plot every recorded earthquake and the map draws the plate boundaries of Chapter 1.5 unaided. That was one of the confirmations of the theory.
Shallow earthquakes, down to about 70 km, happen at every kind of boundary. These are the dangerous ones for the simple reason that they are close to people.
Intermediate and deep earthquakes, from 70 down to about 700 km, occur only in subduction zones, inside the descending slab. They trace the slab's path down into the mantle. Below about 700 kilometres they stop, because the slab has warmed enough to deform without breaking.
Deep earthquakes are also a puzzle worth naming. At those pressures rock should not be able to fracture in the ordinary brittle way at all. The current explanation is that minerals in the slab suddenly collapse into denser crystal structures, and the sudden volume change lets the rock fail. It is an active research question, and this book states it as such.
And some earthquakes are nowhere near a boundary. Old faults inside continents can reactivate. The Latur earthquake in Maharashtra in 1993 killed around 9,000 people in a region considered stable, and the Bhuj earthquake in Gujarat in 2001 killed around 20,000 — neither on a plate boundary. Intraplate earthquakes are rarer but often deadlier, precisely because nobody built for them.
The magnitude scales, and what they actually mean
This is where popular understanding goes wrong most often, so it is worth doing carefully.
Charles Richter's 1935 scale measured the largest wiggle on a particular type of seismograph, corrected for distance. It is logarithmic, meaning each whole number is a factor of ten in ground motion amplitude. A magnitude 6 shakes the ground ten times as far as a magnitude 5.
But energy release grows faster than amplitude. The relationship is
\log_{10} E \approx 1.5 M + 4.8
with E in joules and M the magnitude. Read aloud: the logarithm of the energy is about one and a half times the magnitude plus 4.8. What that means in practice is that one whole magnitude step is about 32 times the energy, since 10^{1.5} \approx 31.6, and two steps is about a thousand times.
| Magnitude | Energy compared with M5 | What it does |
|---|---|---|
| 5 | 1× | Damage to poorly built structures |
| 6 | ~32× | Serious damage over ~100 km |
| 7 | ~1,000× | Major destruction, wide area |
| 8 | ~32,000× | Great earthquake, regional catastrophe |
| 9 | ~1,000,000× | Among the largest ever recorded |
A magnitude 9 releases about a million times the energy of a magnitude 5. People hearing "9 versus 5" imagine roughly double. It is not roughly double.
Modern reporting does not use the Richter scale. It uses moment magnitude (M_w), because Richter's scale saturates — above about 7 it stops distinguishing, since the instrument response maxes out. Moment magnitude is computed from the physics of the rupture itself:
M_0 = \mu \, A \, d
The seismic moment M_0 equals the rigidity of the rock \mu (how hard it resists shearing), times the area A of fault surface that slipped, times the average distance d that it slipped. In words: how strong the rock was, times how much of it broke, times how far it moved. That is a direct measure of the work done, it never saturates, and it is why the largest earthquakes can be compared at all.
The largest ever instrumentally recorded is Chile in 1960, at magnitude 9.5. Its rupture was about 1,000 kilometres long. There is an upper limit set by geography — you cannot exceed magnitude 9-and-a-bit because there is not a long enough continuous fault anywhere on Earth to rupture at once.
Magnitude is not damage. A separate scale, intensity (the Modified Mercalli scale), describes what the shaking actually did at a particular place, from I (unfelt) to XII (total destruction). One earthquake has one magnitude and many intensities. A magnitude 6 directly beneath a city of unreinforced masonry is a catastrophe; a magnitude 8 under empty ocean floor 300 kilometres down may be barely felt.
The four things that decide whether people die
This is the part that matters, and it is not the part that gets reported.
1. What the ground is made of. Loose, wet, sandy sediment shakes far harder than bedrock — several times harder — because seismic waves slow down when they enter soft material, and slowing down means the same energy has to be carried by larger motion. Worse, saturated sand can undergo liquefaction: the shaking packs the grains closer, the water between them takes the load, and the ground temporarily behaves as a liquid. Buildings sink or topple intact. Mexico City in 1985 is the standard case — the earthquake was 350 kilometres away, but the city sits on the bed of a drained lake, and the soft clay both amplified the shaking about fivefold and rang at a period that matched the natural sway of the city's 6-to-15-storey buildings, which are the ones that fell.
2. What the buildings are made of. Unreinforced masonry — brick or stone with no steel through it — is the great killer. It has no ability to bend; the wall cracks, the roof it was holding comes down, and stone floors and roofs land on the occupants. Reinforced concrete with proper detailing, or wood frame, or steel, all bend and stay up. The comparison that settles the argument: the 2010 Haiti earthquake was magnitude 7.0 and killed somewhere between 100,000 and 300,000 people; a magnitude 8.8 in Chile six weeks later, releasing roughly 500 times the energy, killed about 500. The difference was building codes and their enforcement.
3. What time it is and what else breaks. Night quakes kill more because people are indoors and asleep. Fire after shaking destroyed San Francisco in 1906 and killed around 100,000 people in Tokyo in 1923, where much of the city was wood and paper and it happened at lunchtime with thousands of cooking fires lit. Landslides, dam failures and disease in the aftermath all add.
4. Whether there is a warning, and whether anyone knows what it means. Which brings us to tsunamis.
Tsunamis
A tsunami is not a wind wave. Wind waves are a disturbance of the top few metres of the ocean. A tsunami is the entire depth of the ocean moving, and that is why the energy is on a different scale.
How one is generated. A subduction earthquake with vertical motion of the sea floor lifts or drops the entire water column above it — sometimes by several metres over an area hundreds of kilometres across. Gravity pulls that displaced water back to level, it overshoots, and waves radiate outward. Undersea landslides and volcanic collapses do the same thing.
In deep water it is almost invisible. Wave height perhaps half a metre, wavelength up to 200 kilometres, period of tens of minutes. A ship in the open ocean will not notice one pass beneath it. It travels at the speed v = \sqrt{gD} — the square root of gravity times water depth — which in 4,000 metres of water is about 200 metres per second, or 700 km/h, comparable to a jet airliner.
In shallow water it transforms. As depth falls the speed falls, the front of the wave slows while the back is still moving fast, and the energy that was spread through 4 kilometres of water depth is compressed into a few metres. The wave piles up. What arrives is not a curling breaker but a fast-rising wall of water that does not stop, followed by more of them, arriving minutes apart for hours — and the outflow between them is frequently as lethal as the inflow.
The 2004 Indian Ocean tsunami killed about 227,000 people in fourteen countries. The earthquake was magnitude 9.1 off northern Sumatra. Waves reached Sri Lanka and southern India in about two hours and Somalia in about seven. There was no warning system in the Indian Ocean at the time, and there is one now — the Indian National Centre for Ocean Information Services in Hyderabad runs India's, operational since 2007.
The one piece of knowledge most worth carrying: if you are on a coast and the sea suddenly withdraws far beyond the normal low tide, exposing sea floor, that is the trough of a tsunami arriving and you have minutes. Move inland and uphill immediately. In 2004 large numbers of people walked out onto the exposed sand to look. A ten-year-old British girl named Tilly Smith, who had learned this in a geography lesson two weeks earlier, recognised it at a beach in Thailand and got her family and the hotel's beach cleared. Everyone on that stretch survived.
Can earthquakes be predicted?
Short-term prediction — this fault, this week, this magnitude — has never been achieved and there is no method that works. This should be stated plainly because a great deal of nonsense is published to the contrary. Claimed precursors — radon gas in wells, unusual animal behaviour, strange lights, changes in electrical resistivity — have all been reported, and none has ever produced a track record better than chance in prospective testing. The one apparent success, the evacuation of Haicheng in China in 1975, was followed by the Tangshan earthquake in 1976 with no warning at all, killing at least 242,000 people.
What does work is different, and it works well.
Long-term hazard assessment. From the fault map, the measured plate motion, and the geological record of past ruptures, you can state the probability of a given level of shaking at a given place over the next fifty years. That is exactly the input a building code needs, and it saves enormous numbers of lives.
Early warning, which is not prediction. When a rupture starts, the fast P-waves outrun the slower, far more destructive S-waves and surface waves. Detect the P-wave near the epicentre, and you can send an electronic alert ahead of the damaging shaking. The lead time is short — seconds to a couple of minutes depending on distance — but it is enough to stop trains, close gas valves, halt surgery and get people under a table. Japan, Mexico and the US west coast all run such systems, and Android phones now participate in one using their own accelerometers.
Where this shows up in your life
If you live in a seismic zone — the Himalayan front, Gujarat, the northeast of India, Japan, Chile, California, Turkey, Iran, Indonesia — the useful facts are these. During shaking, drop, cover and hold on: get under sturdy furniture and stay there. Do not run for a doorway, which is advice left over from adobe buildings and is wrong for modern ones. Do not use a lift. Once shaking stops, expect aftershocks, which follow a well-established statistical pattern (roughly, the number decays as one over time since the mainshock) and can be nearly as large as the original. On a coast, treat strong or long shaking as the tsunami warning itself and go uphill without waiting for an official one.
And there is a chapter of history in it. The Lisbon earthquake of 1 November 1755 struck on All Saints' Day while the churches were full, killed tens of thousands, and was followed by a tsunami and a fire. The moral shock of it — a devout city destroyed at prayer — ran straight into European philosophy. Voltaire wrote a poem about it and then Candide; Rousseau replied by noting that the deaths came from people crowding into tall stone buildings; and Kant wrote three essays trying to explain it physically. Modern seismology and a large part of the Enlightenment argument about whether the world is governed benevolently both trace to that morning. Chapter 8.9 picks that thread up.
What the next page covers
The other thing that happens where plates meet is that rock melts and comes out. Chapter 1.7 covers volcanoes — why some pour out rivers of lava you can outwalk while others erase a mountain and cool the whole planet for a year, what a pyroclastic flow actually is, why Indonesia and the Pacific rim get the worst of it, and the eruptions that changed human history.