Annals of the Former World

by John McPhee

  Along the San Andreas Fault, the average annual rate of slip is enough to transport something one nautical mile in sixty thousand years. Locally, the jump in 1906 represented roughly two hundred years. In some parts of the fault system, the motion assumes an almost steady creep, but, over all, most of the slip is staccato in time and occurs in elastic rebounds. The strain is essentially constant as the Pacific Plate tugs northwest. In response, earthquakes occur annually in the tens of thousands, most of them below the threshold of human sensitivity. Where the two sides of the fault are most tightly locked, the strain builds highest before it goes. The event of 1906 was what is now known as a large plate-rupturing earthquake. Vertically, the earth broke all the way down to the lower crust. Laterally, it opened the surface like a zipper—from the epicenter northward, where the fault trace for the most part lies just offshore and parallel to the coast, and southward, throwing up what appeared to be a plowed furrow through the rifted hills of Marin, tearing the seafloor where the fault passes west of the Golden Gate, opening the cliff at Mussel Rock, splitting the San Francisco Peninsula, and stopping near San Juan Bautista, east of Monterey Bay. Only once in the historical record has a jump on the San Andreas exceeded the jump of 1906. In 1857, near Tejon Pass outside Los Angeles, the two sides shifted thirty feet.

  Kerry Sieh, a San Andreas specialist at Caltech, has dug trenches in numerous places across the fault zone near Los Angeles in order to examine the evidence in the exposed sediments. He has established that twelve great events have occurred on the southcentral San Andreas Fault in the past two millennia, with intervals averaging a hundred and forty-five years. The Tejon event of January 9, 1857, is the most recent. One does not have to go to Caltech to add a hundred and forty-five to that.

  In 1992, the United States Geological Survey completed a series of studies of the fault segment near San Francisco, and concluded that earthquakes on the order of magnitude of the 1906 event—it has been estimated at 8.3—probably recur every two hundred and fifty years. To the human eye, such a number appears in dim light. In a country where people get up in the dead of night to see what has happened on the Tokyo market, who is worried about two hundred and fifty years? When your complete range of concern begins with your grandparents and stops with your grandchildren, one of your safest bets is elastic rebound. In the prodigious roster of earthquakes on the San Andreas Fault, nearly all of them affect no one. The plates drift, the people with them. Fourteen times a year, an earthquake on the order of the 1989 event near Loma Prieta in the Santa Cruz Mountains occurs somewhere in the world. That might seem to thicken the risk. But not much. In California, only thirteen events have occurred at that level since 1769. So why not move in, spread out, build up, lay back, occupy this incomparable terrain?

  It is said that if a cow lies down in California a seismologist will know it. In Iceland, which is seismically one of the most active countries in the world, there are fewer than thirty monitored seismographs. In California, there are seven hundred. Among countries of the world, only Japan has more seismographs than California. To track plate motions on both a large and a local scale, geologists also use very-long-baseline interferometry, in which the arrival times of noise from quasars at widely separated stations on earth are used to measure distances (even very long distances) with an error margin of less than a centimetre. (They can measure the actual distance that Africa and South America move apart in one calendar year.) The seismographs, the V.L.B.I.s, and other devices enabled Lynn Sykes and Stuart Nishenko, of Columbia University’s Lamont-Doherty Earth Observatory, to predict in 1983 that “the segment of the San Andreas fault from opposite San Jose to San Juan Bautista, which ruptured less than 1.5 m in 1906 and which probably also broke in 1838, is calculated to have a moderate to high probability of an earthquake of magnitude 6¾ to 7¼ during the next 20 years.” In 1989, after that particular stretch of the San Andreas Fault produced a magnitude 7.1 temblor, television interviewers rolled their eyes toward soundproof ceilings when geologists told them that predictions could not be more exact than a time frame of twenty years. This was too much for the anchor flukes, who think in airtime. But a ratio of one moment to twenty years actually represented an amazing juxtaposition of human time and geologic time. To predict a major earthquake within twenty years was like shooting out a candle flame at five thousand yards.

  Sykes and Nishenko also noted that the segment of the San Andreas Fault between Tejon Pass and the Salton Sea was another place quite likely to experience a major event. In the lower part of that region, east of Los Angeles and San Diego, there has not been a great earthquake—8.0 and larger—in historic time. A loud announcement that such an event may be forthcoming was made by the Joshua Tree, Landers, and Big Bear earthquakes of 1992, which occurred fairly near that part of the San Andreas and effectively increased the stress upon it. The combined power of Landers and Big Bear, which came on the same day, caused oil to flow heavily from mountain seeps a hundred and fifty miles west, further endangering the endangered threespine stickleback (a fish). Astonishingly, the Landers earthquake and one that occurred only a few weeks earlier at Cape Mendocino rank among the twelve largest earthquakes that have happened in California in the twentieth century. With three earlier and lesser earthquakes, Joshua Tree and Landers form a straight line pointing north. They have broken open a new fault. Like a river seeking a straight path, the San Andreas seems to [587] want to shift direction and go north through the Mojave Desert and up the east side of the Sierra Nevada (a probability mentioned to me in 1978 by Ken Deffeyes). The Landers earthquake, as if to emphasize the significance of the new fault vector, reached 7.3 and would have been extremely destructive had it happened in a setting other than a desert. Even before 1992, the accumulated strain on the nearby San Andreas was thought to be enough to open a plate-shattering rupture two hundred miles long. In 1981, the Federal Emergency Management Agency published a warning that a temblor of 8.3 or better was likely to occur on that part of the San Andreas before the end of the century. The agency said that property losses would amount to roughly twenty billion dollars, large numbers of people would be hospitalized, and as many as fourteen thousand would be dead. In 1982, the California Division of Mines and Geology chimed in with a special publication describing the same putative earthquake as “an event expected to take place during the lifetime of many of the current residents of southern California,” and going on to say that “two of the three major aqueduct systems which cross the San Andreas fault will be ruptured and supplies will not be restored for a three- to six-month period.”

  In the same year, the California Division of Mines and Geology issued a companion publication called Earthquake Planning Scenario for a Magnitude 8.3 Earthquake on the San Andreas Fault in the San Francisco Bay Area. The scenario predicted that the Bay Bridge would withstand the shaking. So would the Golden Gate. Of a viaduct running through the Marina the scenario said it would “collapse.” Of the Nimitz Freeway in Oakland the scenario said, “The hydraulic fills used to construct miles of freeway along the east shore of the Bay in Alameda County may liquefy during heavy shaking, with long sections becoming totally impassable … . The elevated section through downtown Oakland is expected to be extensively damaged.” The earthquake of the prediction was thirty-five times as intense as the earthquake that actually came.

  By the time of these scenarios, the rock offsets along the San Andreas had been explained, and the role of earthquakes at the plate boundary was understood. In 1906, the great earthquake was an unforeseeable Act of God. Now the question was no longer whether a great earthquake would happen but when. No longer could anyone imagine that when the strain is released it is gone forever. Yet people began referring to a chimeric temblor they called “the big one,” as if some disaster of unique magnitude were waiting to happen. California has not assembled on creep. Great earthquakes are all over the geology. A big one will always be in the offing. The big one is plate tectonics.

  At one time and another, for
the most part with Moores, I have travelled the San Andreas Fault from the base of the Transverse Ranges outside Los Angeles to the rocky coast well north of San Francisco. In clear weather, a pilot with no radio and no instrumentation could easily fly those four hundred miles navigating only by the fault. The trace disappears here and again under wooded highlands, yet the San Andreas by and large is not only evident but also something to see—like the beaten track of a great migration, like a surgical scar on a belly. In the south, where State Route 14 climbs out of Palmdale on its way to Los Angeles, it cuts across the fault zone through two high roadcuts in which Pliocene sediments look like rolled-up magazines, representing not one tectonic event but a whole working series of them, exposed at the height of the action. On the geologic time scale, the zone’s continual agitation has been frequent enough to be regarded as continuous, but in the here and now of human time the rift extends quietly northward through serene, appealing country: grasses rich in the fault trough, ridges intimate on the two sides—a world of tight corrals and trim post offices in towns that are named for sag ponds.

  Farther north, it loses, for a while, its domestic charm. Almost all water disappears in a desert scene that, for California, is unusually placed. The Carrizo Plain, only forty miles into the Coast Ranges from the ocean at Santa Barbara, closely resembles a south Nevada basin. Between the Caliente Range and the Temblor Range, the San Andreas Fault runs up this flat, unvegetated, linear valley in full exposure of its benches and scarps, its elongate grabens and beheaded channels, its desiccated sag ponds and dry deflected streams. From the air, the fault trace is keloid, virtually organic in its insistence and its creep—north forty degrees west. On the ground, standing on desert pavement in a hot dry wind, you are literally entrenched in the plate boundary. You can see nearly four thousand years of motion in the bed of a single intermittent stream. The bouldery brook, bone dry, is fairly straight as it comes down the slopes of the Temblor Range, but the San Andreas has thrown up a shutter ridge—a sort of sliding wall—that blocks its path. The stream turns ninety degrees right and explores the plate boundary for four hundred and fifty feet before it discovers its offset bed, into which it turns west among cobbles and boulders of Salinian granite.

  You pass dead soda ponds, other offset streams. The (gravel) road up the valley is for many miles directly on the fault. Now and again, there’s a cattle grid, a herd of antelope, a house trailer, a hardscrabble ranch, a fence stuffed with tumbleweed, a pump in the yard. A daisy wheel turns on a tower. Down in the broken porous fault zone there will always be water, even here.

  With more miles north come small adobes, far apart, each with a dish antenna. And with more miles a handsome spread, a green fringe, a prospering ranch with a solid house. The fault runs through the solid house. And why should it not? It runs through Greater San Francisco.

  Of the two most direct routes from southern to northern California, always choose the San Andreas Fault. If you have adequate time, it beats the hell out of Interstate 5. Nearly always, some sort of road stays right in the fault zone. Like a water-level route through rough country, the fault is a place to find gentle grades and smooth ground. When the fault makes minor turns, they are nothing compared to the bends of a river. With more distance north, the desert plain yields to hay meadows and then to ever lusher country, until vines are standing in the fault-trace grabens and walnuts climb the creaselike hills. Ground squirrels appear, and then ever larger flocks of magpies, and then cottonwoods, and then oaks in thickening numbers, and velure pastures around horses with nothing to do. In age and rock type, the two sides of the fault are as different as two primary colors. Strewn up the west side are long-transport gabbroic hills and deracinated ranges of exotic granite. Just across the trough is Franciscan mélange—stranger, messier, more interesting to Moores.

  Near Parkfield, you cross a bridge over the San Andreas where Cholame Creek runs on the fault. The bridge has been skewed—the east end toward Chihuahua, the west end toward Mt. McKinley. Between Cholame and Parkfield, plate-shattering ruptures have occurred six times since 1857, an average of one every twenty-two years, and the probability that another would occur before 2003 had been reckoned at ninety-eight per cent. Thirty-seven people live in Parkfield. If the population is ever to increase, seismologists will be the first to know it, for the valley here is wired like nowhere else. Parkfield has attracted earthquake-prediction experts because the brief interval time on this segment of the fault suggests that if they monitor this place they may learn something before they die. Also, the Parkfield segment has—in Moores’ words—“relatively simple fault geometry.” And the last three earthquakes have had a common epicenter and have been of equal magnitude.

  An average of one plate-shattering earthquake every twenty-two years works out to forty-five thousand per million years. The last big Parkfield event was in 1966. It broke the surface for eighteen miles. Words on the town’s water tower say “Parkfield, Earthquake Capital of the World, Be Here When It Happens.” The actual year doesn’t matter much. The instrumentation of Parkfield assumes that a shock is imminent. Its purpose is not to confirm the calculated averages but to develop a technology of sensing—within months, days, hours, or minutes—when a shock is coming. Even a minute’s warning, or five minutes’, or an hour’s, let alone a day’s, could (in highly populated places) save many lives and much money. Accordingly, the Cholame Valley around Parkfield—between Middle Mountain, to the north, and Gold Mountain, to the south—has been equipped with several million dollars’ worth of strain gauges, creepmeters, earth thumpers, laser Geodimeters, tiltmeters, and a couple of dozen seismographs. It is said that the federal spending has converted the community from Parkfield to Porkfield. Some of the seismographs are in holes half a mile deep. Experience suggests that rocks creep a little before they leap. The creepmeters are sensitive to tens of millionths of an inch of creep.

  If ever there was a conjectural science, it is earthquake prediction, and as research ramifies, the Tantalean goal recedes. The maximum stress on the San Andreas Fault—the direction of maximum push—turns out to be nearly perpendicular to the directions in which the fault sides move, like a banana peel’s horizontal slip when pressure comes upon it from above. A fault that moves in such a manner must be weak enough to slide—must be, in a sense, lubricated. Among other things, the pressure of water in pores of rock in the walls of the fault has been mentioned as a lubricant, and so has the sudden release of gases that may result from shaking. Such mechanisms would tend to randomize earthquakes, diminishing the significance of mounting strains and temporal gaps. Those who practice earthquake prediction will watch almost anything that might contribute to the purpose. A geyser in the Napa Valley inventively named Old Faithful seems to erupt erratically both before and after large earthquakes that occur within a hundred and fifty miles—an observation that is based, however, on records kept for not much more than twenty years. In 1980, the United States Geological Survey began monitoring hydrogen in soils. Two years later, near Coalinga, about twenty miles northeast of Parkfield, the hydrogen in the soil was suddenly fifty times normal. It appeared in bursts, and such bursts became increasingly numerous in April, 1983. In May, 1983, a 6.5 earthquake occurred on a thrust fault under Coalinga. Releases of radon are watched. So are patterns and numbers of microquakes, especially those that are known as the Mogi doughnut. In the mid-nineteen-sixties, a Japanese seismologist noticed on his seismograms that microquakes occurring in the weeks before a major shock sometimes formed a ring around the place that became the epicenter. Mogi’s doughnut is a wonderful clue, but—like hydrogen bursts and radon releases—before most major shocks it fails to appear.

  People who live in earthquake country will speak of earthquake weather, which they characterize as very balmy, no winds. With prescient animals and fluctuating water wells, the study of earthquake weather is in a category of precursor that has not attracted funds from the National Science Foundation. Some people say that well water goes d
own in anticipation of a temblor. Some say it goes up. An ability to sense imminent temblors has been ascribed to snakes, turtles, rats, eels, catfish, weasels, birds, bears, and centipedes. Possible clues in animal behavior are taken more seriously in China and Japan than they are in the United States, although a scientific paper was published in California Geology in 1988 evaluating a theory that “when an extraordinarily large number of dogs and cats are reported in the ‘Lost and Found’ section of the San Jose Mercury News, the probability of an earthquake striking the area increases significantly.”

  Earthquake prediction has taken long steps forward on the insights of plate tectonics but has also, on occasion, overstepped. Until instrumentation is reliably able to chart a developing temblor, predictors obviously have a moral responsibility to present their calculations shy of the specific. The mathematical equivalent of a forked stick will produce such absurdities as the large earthquake that did not occur as predicted in New Madrid, Missouri, on December 2, 1990. A U.S.G.S. geologist and a physicist in the United States Bureau of Mines whose research included (among other things) the study of rocks cracking in a lab predicted three great earthquakes for specific dates in the summer of 1981, to take place in the ocean floor near Lima. The largest—9.9—was to be twenty times as powerful as any earthquake ever recorded in the world. A few hundred thousand Peruvians were informed that they would die. Nothing happened.