ÐÓ°ÉÔ­´´

AT AROUND noon one day in late June, 1838, Charles Brown was struck
on the back of the head by a swinging vat of lard as he entered his cabin
in the virgin redwood forests of the southern San Francisco peninsula. Rushing
to the door, he saw the land outside ‘rising and falling in solid waves’,
‘redwoods rocked like lake-side reeds: thousands of them broken off and
hurled through the air’. After the shaking had died down, he found the ‘ground
cracked in all directions. One immense opening extended from near Lone Mountain
(close to modern San Francisco) to the mission at San Jose’ – some 60 kilometres.
New York born Brown, lumberjack, fortune-hunter and naturalised Mexican,
was the first recorded witness to movement on the San Andreas Fault.

Less than 70 years later, on the morning of 18 April 1906, the San Andreas
Fault made itself notorious by destroying San Francisco. Since then, studies
of this 1100 kilometre long fault have provided much of the basis for the
modern understanding of earthquakes, and the fault itself has become part
of Western culture. It is not just the most famous fault in the world, but
for most non-geologists, the only fault. The fault became so familiar that
most San Franciscans had stopped taking it seriously, until a few minutes
after 5 o’clock on 17 October last year when everything changed.

The ‘Loma Prieta’ earthquake, of magnitude 7.1, which crumpled Oakland’s
freeways and turned the soil of San Francisco’s Marina district into a quicksand,
has revived Californians’ fears of the ‘Big One’, an earthquake 10 times
as big. Where will such a huge earthquake occur, and when? Geologists and
seismologists must look to the past, the history of California’s earthquakes,
to give them a key to the future.

The pattern of past movements along the fault provides the best picture
of the future, but the view is foreshortened by California’s lack of history.
For Californians, the year 1838 AD might as well be BC; their state was
part of Mexico until 1848. Nine years later, on 9 January 1857, there was
a great earthquake, whose centre lay in the mountains northeast of the village
of Los Angeles. Eye-witness reports said that ‘the line of disturbing force
was marked by a fracture of the Earth’s surface, continuing in one uniform
direction for a distance of 200 miles. In some places the sliding movement
seems to have been horizontal.’ One account claimed that a circular sheep
corral changed into ‘a rude S-shaped figure’.

It was not until 18 April 1906 that these anecdotes became credible.
The earthquake, magnitude 8.25, that destroyed San Francisco involved horizontal
movement along more than 400 kilometres of the San Andreas Fault. In the
open country along the northern Californian coast, the fault revealed itself
as cracks, ridges, shallow ponds or lines of loose earth known as ‘mole
tracks’; as if a giant plough had dragged through the earth. Roads, fences,
tunnels, railways, and pipelines had all been broken along the line of the
fault, and moved sideways. The amount of slip varied; up to 6 metres to
the north of San Francisco, and between 3 and 4 metres in the northern San
Francisco peninsula, but to the southeast of Palo Alto, it was no more than
1.5 metres.

This movement was horizontal, and dextral or ‘right-handed’: stand facing
the fault, and the other side moves to your right. Such a slip would also
account for the distortion, in 1857, of a circular sheep-pen into an S-shape.

Following the earthquake, the state organised a geodetic survey, using
triangulation points that had been installed soon after 1850. The horizontal
distortion died away farther than 6 kilometres from the fault, but distant
survey points on opposite sides of the fault had also moved relative to
one another, suggesting that the crust had become distorted over a large
region.

Most puzzling was that this regional distortion of the Earth’s crust
had also happened before the 1906 quake. The surveyors had measured the
triangulation network twice before,between 1851 and 1865, and between 1874
and 1892. The Farallon Islands, 40 kilometres west of the Golden Gate Bridge,
had shifted 1.4 metres to the northwest (relative to the mountains of the
mainland) between the first and second surveys. They had then moved an additional
1.8 metres by the time of the survey of 1906-7, after the San Francisco
earthquake. Such rapid distortion of the crust was completely inexplicable.
The state’s geodesists attributed the earlier distortion to the effects
of an earthquake east of San Francisco Bay in 1868. This explanation did
not satisfy everyone.

In 1910, America’s foremost geological physicist, Harry Fielding Reid,
offered an alternative explanation: movement so far from the fault could
not have happened at the time of the earthquake, but must be accumulating
continuously. Reid realised that the crust around the fault behaves like
an elastic spring. Between earthquakes, the crust deforms elastically, storing
energy which is released when the crust breaks along the fault, and the
two sides snap back. From the change in the position of the Farallon Islands,
Reid calculated that the elastic distortion released by the 1906 earthquake
on the San Andreas Fault, 6 metres, was approximately twice that built up
across the region in the previous 50 years, a total of 3.2 metres. Geophysicists
now know that the survey data on which Reid based his ideas was inaccurate,
but his model remains the foundation of the theory of earthquake generation.

Reid later became a fierce critic of the theory of continental drift
pioneered by Alfred Wegener, the German geophysicist, even though this offered
a convenient explanation for the steady horizontal movement. In 1926, a
geologist mapping southern California was widely ridiculed when he proposed
that the San Andreas Fault had built up a cumulative offset of at least
25 kilometres. By the late 1950s, geologists were discussing offsets of
several hundred kilometres.

In the mid-1960s, the theory of plate tectonics finally made sense of
the San Andreas Fault: it was no longer an aberration, but a beautiful example
of a type of plate boundary known as a transform fault. It connects an active
volcanic ridge, an extensional plate boundary, which runs in a series of
broken sections down the centre of the Gulf of California, to a triple junction
where the American and Pacific plates meet the small Juan da Fuca plate
off Cape Mendocino, northern California.

The fault itself is no more than the main line of the plate boundary.
It and a series of branch-line faults together carry the relative movement
of the plates. There is a continuous zone of movement deep in the crust,
but, closer to the surface, most of the slip takes place on the San Andreas
Fault. The rest is distributed over several roughly parallel faults. The
overall movement is dextral: the Pacific plate, carrying parts of southern
California, is moving north. In 15 million years or so, the smog of Los
Angeles will blot out San Francisco’s spectacular sunsets.

The San Andreas Fault is a series of short, often overlapping, faults
in a zone that can be anything between metres and kilometres across. Where
these faults step to the right, a gap opens between the two strands of the
fault, making a depression. This can be as small as a pond or as large as
the Salton Sea, at the southern end of the San Andreas Fault. Where the
faults step to the left, the intervening crust is squeezed together and
upwards. The repeated uplift from thousands of earthquakes builds mountains
out of mole-tracks. In southern California, where the whole San Andreas
Fault swings to run more nearly east-west, the horizontal motion has partly
converted to compression, raising the Transverse Ranges to 3600 metres.
Another small range, the Santa Cruz mountains, lies at a more subtle bend
in the Fault, and rises to 1200 metres. This was the site of the Loma Prieta
earthquake of 1989.

The geology of the Santa Cruz mountains reveals a series of tight folds,
more like sticklebacks than whalebacks, aligned almost parallel to the San
Andreas Fault. The earthquake of 17 October revealed that the fault itself
is not vertical, as people had assumed, but inclined at about 70 degrees,
dipping towards the southwest. Movement in that earthquake was neither simply
vertical nor horizontal, but oblique. The southwest side of the fault moved
up and to the north, a total of 1.7 metres horizontally and 1.3 metres vertically.
The Santa Cruz mountains to the southwest of the fault moved up about 0.6
metres, and the town of Santa Cruz rose some 0.4 metres. The most prominent
faults at the surface after this earthquake had offsets of up to a metre,
but they were sinistral – moving to the left – in the opposite sense to
the overall movement. The earthquake also tightened folds in the ground
above the fault; researchers think that the sinistral faulting results from
the folding, rather than simply from the San Andreas Fault below.

The overall movement taking place between the American and Pacific plates
around California is 56 millimetres a year, a figure that researchers have
established independently from the velocities of plates worldwide. The long-term
rate of movement of the San Andreas Fault is 37 millimetres a year along
the central section of the fault, and 25 millimetres a year in the Transverse
Ranges. To the south of San Francisco it is around 15 millimetres a year.

For the past half century, seismologists have been learning how to read
the nature of movements on faults from the unique pattern of the earthquake
vibrations. Now geologists are helping to interpret the pattern of fault
ruptures from the shape of the landscape. The advantage of landscape is
that it retains the earthquake ‘signature’ long after the short-term vibrations
have rippled away. Topography can reveal the bends and offsets in the line
of the fault, where individualfault ruptures start and finish, and which
fault strand carriesthe greatest or least slip where there are several strandsclose
together.

In the mid-1970s, Kerry Sieh, now at the California Institute of Technology,
mapped the fault movements of the great earthquake of 1857. A 400 kilometre
length of the fault ruptured at that time, and the horizontal offset along
the centre of the fault break was 9.5 metres, half as long again as the
greatest seen in 1906. Just as in 1906, the displacement remained almost
constant for long sections of the fault, yet jumped abruptly from one section
to the next. Sieh also explored the traces of earlier movements along this
section of the San Andreas Fault, preserved in the near-desert climate.
He found evidence that, for any given section of the fault, earlier offsets
had been similar to those of 1857. Therefore, on a particular section of
the fault, the amount of slip repeats itself from one earthquake to the
next.

Sieh’s discovery, that individual sections of the fault repeated the
same ‘characteristic’ earthquake, appeared to confirm Reid’s model, developed
70 years earlier. Reid had attempted to show that the 6 metres of displacement
in the San Francisco earthquake of 1906 had taken around 100 years to accumulate.
His work provides the basis of modern methods of forecasting earthquakes.

Ideally, a fault accumulates its overall displacement in infrequent
sudden movements, which generate earthquakes. Divide the rate at which the
fault is moving, averaged over many centuries, by the displacement in an
individual earthquake, and you have the recurrence interval. Add this recurrence
interval to the date of the last earthquake and you have forecast the next.

Because the characteristic displacement varies from one section of the
fault to the next, the recurrence interval must also vary. Some short sections
moved independently between the expected repetitions of the 1857 earthquake.
Yet this minor complication does not upset the model. Researchers can use
this model to estimate the date when the fault’s displacement budget has
passed into deficit.

Sieh also checked the recurrence intervals using radiocarbon dating.
He found traces of past fault movements in a trench cut through a peat bog
at Pallett Creek, on the line of the 1857 earthquake. Instead of regular
recurrence, he found a range of different intervals lasting between 55 and
275 years. Sieh thought the data represented two recurrence intervals, with
a long and a short period. But more recent statistical work has suggested
that the spread of dates is random around a mean of some 180 years.

His results seemed contradictory. Even if the displacements appear characteristic,
the intervals between events decidedly are not, at least at Pallett Creek.
There is only one section of the San Andreas Fault where researchers can
test this model of earthquake recurrence from the historical record: the
120 kilometres between San Francisco and San Juan Bautista.

No one knows the full extent of rupture along the San Andreas Fault
in 1838. But only 68 years later in 1906, the section near San Francisco
managed to move between 3 and 4 metres, when the accumulated deficit in
slip should have been only between 1 and 1.5 metres. At the southern end
of the 1906 rupture, an earthquake involving 0.5 metres of slip had happened
only 16 years before the 1906 movement of 1.2 metres. This behaviour looks
decidedly uncharacteristic. Yet in 1983, Allen Lindh of the US Geological
Survey at Menlo Park, California, showed that, assuming a slip rate of 15
millimetres per year south of Palo Alto, the San Andreas Fault would go
into deficit on its budget in 1990. By that reckoning, the Loma Prieta earthquake
arrived 75 days early! It also ruptured along only half the distance predicted
by Lindh, so another earthquake now appears likely. It will probably be
to the northwest of the 1989 rupture, passing close to Silicon Valley, and
could have a magnitude of 6.5. Yet the fact that this section did not break
on 17 October suggests that it is not yet fully ripe.

To the southeast of the Santa Cruz mountains there is another short
unbroken section, ready for a magnitude 6 earthquake. Beyond San Juan Bautista,
the fault is moving in a different way. Instead of the infrequent massive
jolts that make earthquakes to the north and south, the fault here is moving
almost continuously, in many tiny judders. Geologists call this slow, almost
steady slip, ‘creep’. The fault creeps where it is greased with the natural
mineral lubricant, serpentine (hydrated magnesium silicate), plentiful in
the great slab of ocean crust that underlies the western edge of the San
Joachin Valley. Seismologists locate the creeping faults of California from
concentrations of small earthquakes. The largest earthquakes and biggest
fault movements have been along those sections of the fault that show hardly
any tiny earthquakes. There is a moral in this that people in most parts
of the world might prefer to ignore: big earthquakes happen precisely where
small ones do not.

At the southeastern end of the creeping section of the San Andreas Fault,
close to the town of Parkfield, the fault becomes locked. Loaded at a regular
rate by the creeping fault to the northwest, it responds as near to clockwork
as any fault known. Moderate earthquakes, of magnitude between 5 and 6,
recur on average every 22 years. So far, they have happened in 1881, 1902,
1922, 1934 and 1966. This area provides an ideal laboratory for studying
how individual fault ruptures start, and how they might be predicted.

The Parkfield earthquakes all start at a patch on the fault 12 kilometres
below the surface, and 8 kilometres long, known as the ‘preparation zone’.
This is where the rock is strong, and stress builds up more than on the
fault nearby. The most highly stressed rock breaks first when an earthquake
begins. These patches might reflect a difference in the mineralogy of the
rock, or more often, a bend in the fault itself. But once broken, the rupture
can grow quickly through the weaker rock nearby. Rocks become stronger when
the pressure is higher, so the deeper down they are in the crust, the stronger
they are. At depths of between 10 and 25 kilometres, this changes, and the
rocks start to become weaker again. This happens because the rocks become
so hot that they can flow more easily than fracture. As a result of this
increase of strength with depth, preparation zones tend to skulk at the
bottom of faults. The Parkfield preparation zone takes 22 years to break;
the rupture that starts out at the zone expands through some 20 kilometres
of fault to the south-east, creating the ‘characteristic’ Parkfield earthquake.

Researchers had not identified the Loma Prieta preparation zone before
last October’s earthquake. It is hidden 18 kilometres down at the very bottom
of the deepest section of the fault, beneath the highest peak of the Santa
Cruz mountains, Loma Prieta, that gave the quake its name. The mountains
here have grown very quickly. As they built up, the crust became thicker.
A root of relatively cold crust pushes down into the mantle below. The rocks
there are strong to a greater depth than usual, and the faults break farther
down than is typical of the San Andreas Fault (except in the Transverse
Ranges, east of Los Angeles).

Eventually, researchers may identify a handful of preparation zones
where major earthquakes begin, and where they can direct all their efforts
to short-term prediction. The problem remains one of timescales; now that
seismologists have found the Loma Prieta preparation zone, it is unlikely
to fire off again for about 80 years. It is their great-grandchildren who
will monitor the next Loma Prieta quake, in the middle of the 21st century.

Another problem with earthquake forecasting is the way that a moderate
earthquake could transform itself into the ‘Big One’, a repeat of the events
of 1906 or 1857. Great earthquakes are, in effect, no more than a series
of moderate quakes that fire off in quick succession as the rupture sweeps
along the fault, unchecked for hundreds of kilometres. One theory proposed
by Sieh for the 1857 earthquake is that it was started by the standard 22-year
Parkfield earthquake, which loaded the adjacent section of the fault to
breaking point. The central section of the San Andreas Fault is creeping
at only 40 per cent of the long-term rate of slip and, theoretically, if
all the various sections of the fault were reaching maturity about the same
time, a rupture could break the whole 1100 kilometre fault in the ‘Biggest
One’, a quake with a magnitude of 8.5.

In the face of a concentrated scientific assault, the San Andreas Fault
has yet to be tamed. The 1989 earthquake lent some support to the model
of characteristic earthquakes which was beginning to seem vulnerable. In
other respects it was something of a surprise. The Parkfield earthquake
is being stalked by a well-equipped army of seismologists, but it is already
overdue. If it refuses to appear, they will have to change their models
yet again. Other preparation zones will be targeted, as soon as geologists
can recognise them. Perhaps they will be able to predict the repeat of the
1906 earthquake, due in the 22nd century, that researchers believe originates
offshore from the Golden Gate Bridge.

But a repeat of the quake of 1838, when freeways could dance like Charles
Brown’s redwood trees, may be only a few decades away. Even Superman will
not be able to prevent a major earthquake before the year 2000 on the southeastern
end of the San Andreas Fault. This, at least, will ensure that the plate
boundary snaking through their state will continue to give Californians
‘the edge’ – a frisson to life without which they might as well be living
in Massachusetts.

* * *

Finding faults with earthquakes

FAULTS are planar breaks in the rocks of the Earth’s crust. The movement
of one side relative to the other often happens abruptly, generating the
vibrations felt at the surface as earthquakes. Most quakes come from small
movements on faults as far as 20 kilometres underground. The greater the
area of the fault that breaks, the bigger the earthquake.

The size of an earthquake is expressed as its magnitude, a logarithmic
measure of the size of the vibration. The shaking is amplified and recorded
by seismometers, and the magnitude scale takes into account the distance
of each instrument from the source of the quake. Researchers estimate the
area of the fault that has actually broken by mapping the locations of the
smaller tremors that follow an earthquake: these tend to cluster around
the broken section of the fault. An earthquake of magnitude 7 involves a
fault that may have an area of around 500 square kilometres, one of magnitude
8, a fault break of around 5000 square kilometres.

The largest fault breaks emerge at the Earth’s surface, where geologists
can map the relative positions of broken roads, pipelines, fences and so
on, to find the size and direction of the displacements that occurred at
the time of the earthquake. They have to work quickly: in cleaning up the
damage, people can obliterate the very traces that the researchers need.
Many faults move almost vertically, but the San Andreas Fault is a ‘strike-slip’
fault, with horizontal movement.

Robert Muir Wood is a consultant for Yard Ltd, a division of the FEMA
Group, and Editor of the journal Terra Nova