POSTED 3 NOVEMBER 2005
The ground truth
Let's start talking about earthquake prediction by disposing of one bothersome detail. Nobody can predict earthquakes -- nobody can say a quake of any given size will happen in a particular place on a particular date.
Even the more general quest -- to anticipate the statistical earthquake hazard at a given location -- remains a tough problem, as Earth continues dishing up surprises. It's becoming clear that places once thought stable actually have this nasty habit of shaking in a big way, every once in a long while.
Socially, economically, even geologically, the American Midwest seems a paragon of stability. But survivors of the New Madrid [Mo.] earthquakes of 1811 and 1812 "reported that the earthquakes caused cracks to open in the earth's surface, the ground to roll in visible waves, and large areas of land to sink or rise."
The Pacific Northwest has its share of scary volcanoes, but it was once considered rather immune to seismic shaking. We now know that in 1700, a giant earthquake jolted Oregon, Washington and British Columbia. "Twenty years ago, nobody thought there was a major fault, but geologic records indicate that there have been magnitude 9 earthquakes offshore," says seismologist Thorne Lay of the University of California. The quake may not have been recorded in North America, but the tsunami it apparently sparked swept clear across to Japan. The tsunami can be dated from coastal geological evidence, and written records in Japan, Lay says.
A third caution against the accuracy of earthquake hazard assessments comes from the 2004 Sumatra quake. This, the largest earthquake since Alaska, 1964, struck a section of fault that had never, to our knowledge, caused a great earthquake. The absence of quakes is not just based on the seismological record. Geologists have reported no evidence of sudden uplift, submersion or tsunamis along the Indian Ocean coast during the last 1,000 years.
How rare is rare?
Earth science is all about records. Climate studies routinely deal with the relative brevity of climate records, and a similar shortage has hobbled our understanding of earthquake hazards. Although the first seismograph was invented in China in AD 132, detailed records of ground shaking date only to the late 19th century. So even though the Earth moves according to geologic time -- scales of thousands to millions of years -- seismologists are forced to understand by relying on short-term records.
The huge, deadly rupture on the northern section of the Sunda Fault may be quite a rare occurrence, but "rare" translates into "hard to predict." The 2004 quake, says Lay, "taught us that in some regions, the time between events is so long that we are basing the anticipation of future earthquakes on a historical record that is very limited. So we may miss some very large earthquake regions."
If it can be difficult to see earthquakes in hindsight, it's much tougher to predict them in the future. Even in heavily instrumented Parkfield, a location on the San Andreas Fault that may be home to a world record stash of seismological instruments, predictions have been a bit dodgy. Magnitude 6 quakes were thought to have occurred 22 years apart, on average. But when the most recent magnitude 6 quake finally occurred in 2004, it was 16 years behind schedule.
In Parkfield and other shaky locations, scientists have assessed many factors in their quest to predict earthquakes:
Foreshocks -- small shocks that may precedes a larger one
Changes in groundwater chemistry
Pressure of fluids in underground rocks
Creep -- slow movement along a fault
Earthquakes in adjacent parts of a fault
Statistical analysis of earthquake frequency
Changes in the local magnetic field
Strain in underground rocks
And even though most of these factors have been examined in detail on the San Andreas, there is no indication that anything seemed out of the ordinary before the 2004 quake shook Parkfield. "The lack of anybody jumping up and down about it now tells me what I expected, that efforts to find a predictive sign were a failure," says Thurber. "We are a long way away from making predictions."
After so many years of study, earthquakes remain the natural disaster that is the most difficult to predict. It's fairly easy to pinpoint locations where quakes are likely, and often possible to forecast the general level of future danger. But in terms of saying, "we expect a magnitude X quake to occur on the Y of July at location Z," don't look for that any time soon.
Ironically, even as seismologists admit that they can't predict earthquakes, data from seismology plays a crucial role in helping volcanologists predict eruptions -- a science that is a bit further along than earthquake prediction. But earthquakes still amount to a bolt from the blue, and, as we have learned in the past year, they can be the deadliest of surprises.
Further fundamental research is needed into the processes that hold faults together, then allow them to suddenly tear apart, indicates seismologist Thorne Lay. "We have to study, are they even predictable at all? Until we can understand the physics well enough, we can't put into place any reliable predictive scheme."
See what's shaking in our bibliography.
Megan Anderson, project assistant; Terry Devitt, editor; S.V. Medaris, designer/illustrator; David Tenenbaum, feature writer; Amy Toburen, content development executive