So, Can We Actually Predict Earthquakes?
On Waiting for the Big One, and the Imprecise Science of Seismology
March 17, 2014, Emily Brodsky and Thorne Lay were having coffee in UC Santa Cruz’s Seismology Lab. The two scientists do this pretty much every morning. They call it “seismo coffee” (even though Brodsky usually drinks tea), and they invite bedraggled graduate students to join them in a well‐worn conference room cluttered with academic journals, Tupperware containers, and a few beer bottles. Mostly, the two talk about their research or unpack new scholarly articles. On this particular morning, the handful of doctoral students sitting around the table instead listened as Brodsky and Lay speculated about a series of earthquakes off the coast of Chile that had begun the day before. Could they be foreshocks heralding a megaquake?
These quakes were occurring not far from the Tarapacá region of Chile in a part of the Pacific known for having what geologists call a “recognized seismic gap,” which is to say a place known to be capable of quakes but one that has not had an event for quite some time. It usually means that pressure and energy have been building up there for some time and are just waiting to explode.
Brodsky says she and Lay had a feeling that release was about to occur—and in epic fashion. “We were asking each other daily, ‘Has it happened? Has it happened?’”
And then it did happen.
On April 1—sixteen days after the first tremor—a major quake rocked Chile. Residents in Iquique, Tarapacá’s capital city, reported a cacophony of sounds created by the magnitude 8.1 quake—collapsing buildings, causing landslides, tearing apart roads—before they were drowned out by the sound of tsunami warning sirens. By the end of the day, six people were confirmed dead—a number that no doubt would have been higher, had the epicenter of the quake been near a more densely populated region of the country.
As the number of aftershocks continued to grow, Brodsky and Lay began work on a scientific paper about those first tremors and whether they might have been what seismologists call nucleation, or precursory activity before a big quake. If so, then maybe similar events could be used to forecast future seismic activity.
Prediction is the holy grail of the geological world. It’s also a four‐letter word in that community and the easiest way to start a barroom brawl among seismologists. British geologist John Milne, one of the fathers of the field, wrote in 1880: “Ever since seismology has been studied one of the chief aims of its students has been to discover some means which would enable them to foretell the coming of an earthquake.” This was important, he wrote, not just for scientific integrity but also for appeasing the layperson, who, as it turned out, often held the purse strings to scientific inquiry. “What the public imagine they would like to know about an earthquake is the time at which it might occur. If this could be stated, and at the same time something about the character of the expected disturbance in earthquake districts, seismology would be liberally supported.”
By Milne’s time, science had given it a serious college try: They’d tracked cloud formations and changing weather patterns, lunar tides and the behavior of goats. In the decade or two that followed, they’d also investigate the tilt of the earth, the presence of magnetic fields, the temperature of hot springs, and the possibility that the earth may well have an arrhythmic pulse.
With the Cold War still very much raging, the United States and the Soviet Union entered into a joint earthquake‐prediction program in 1972, which was announced at a bilateral summit hosted by Nixon and Brezhnev. Widely touted as the “most successful scientific partnership” between these two chilly nations, the project brought together American and Soviet geologists in the name of wresting some control back from the planet. Why? Because, the scientists wrote in their first of official report, every year our ability to increase geological hazards grows larger. They cited mining, the capture of fossil fuels, quarry blasting, and growing reservoirs as examples. Natural earthquakes were bad enough. Man‐made earthquakes were making things much worse. “Of all geologic hazards,” they warned, “earthquakes can cause the greatest loss of life and the largest property damage.” In an attempt to thwart this destructive power, they hung out in one another’s labs; they blew up stuff together; they visited one another’s dams and oil fields and power plants; they nosed around their seismic arrays. The end result was a 529‐page document entitled The Soviet-American Exchange in Earthquake Prediction, which was, unfortunately, very light on the subject of earthquake prediction.
Just to be safe, we also launched our own earthquake‐prediction program. To the casual observer, it seemed to be going swimmingly. In September 1975, Time ran a cover story boasting that leading nations, including the United States, China, and the Soviet Union, were consistently predicting quakes with impressive accuracy. The US National Research Council urged Congress to fund a long‐term program aimed at promoting seismic prediction. It was a decade of profound optimism and more than a little scientific hubris—the same decade Nixon announced his war on cancer and predicted a cure would soon be found, the same decade scientists predicted that all the world’s oil would be depleted by 2000.
But sustained seismic prediction soon proved more difficult than anyone had thought.
The Soviet Union invested heavily in earthquake prediction beginning in the early 1950s, and their Complex Seismological Expedition was viewed by many as the best science on the subject. In 1969, a group of US geologists traveled there and were so impressed with the Soviets’ findings that they snapped photos of all their diagrams and research. Japan had been conducting similar research. The real leader in the field, though, appeared to be China. There, government‐funded projects included charting everything from water table levels to electrical currents below the surface of the earth to, famously, the behavior of wild and domestic animals.
Then, in February 1975, the Chinese government issued an earthquake warning and evacuation for the area surrounding the city of Haicheng. Within hours of its issuance, a magnitude 7.3 earthquake struck the region, destroying over 90 percent of the buildings there. Just over 2,000 people died. A lot, to be sure, but scholars estimated the death toll would have been 150,000 had the warning not been issued.
Prior to the earthquake, residents reported seeing snakes rise out of the ground and livestock behaving “strangely,” though there was never much said about what that behavior looked like. Nevertheless, that was enough for China to launch a study on animals and earthquake prediction. The USGS started to go down this road as well. They even sponsored a symposium on the subject in 1975 (though you’d be hardpressed to find any evidence of it on their website or in their promotional materials). Without any major quakes, it was hard to say anything definitive. And yet examples of animals anticipating earthquakes continue to make the news. Just five or ten minutes prior to that M 5.6 earthquake that struck Oklahoma in 2016, weather radar captured tens of thousands of birds taking flight. Zookeepers at the National Zoo reported similarly atypical behavior of their charges, including lemurs, flamingos, and Iris, the charismatic star of the orangutan exhibit, who began screaming and then climbed to the top of her enclosure in a panic just minutes before the 2011 quake struck there.
In general, seismologists remain skeptical about the degree to which any animal can predict an earthquake. And certainly the Chinese prediction program has witnessed some major failures. In 1976, a magnitude 7.6 earthquake rocked the city of Tangshan. Scientists never saw it coming. Neither did the approximately 1 million people living in the city. A quarter of them died that day. Another 164,000 were injured. That same year, the Chinese government issued another warning for a major quake. Residents streamed out of their homes and camped in fields and streets. The earthquake never came.
However, neither of these incidents did much to cool the prediction fervor. The Soviet Union was producing its own excitement in Garm, an ice‐swept section of Siberia that had witnessed a major quake in 1949 that killed 12,000 people. Since that disaster, Soviet geologists had focused their study there. In 1971, they announced that they were able to anticipate quakes by examining P wave activity. US geologists began learning Russian, figuring their Cold War nemesis had the best lead on predictive techniques.
In November 1974, seismologist Malcolm Johnston wowed members of California’s Pick and Hammer Club when he predicted that a quake would soon hit the small town of Hollister, located east of Monterey Bay. The next day—Thanksgiving—an M 5.2 quake struck the area just as families were carving their turkeys and mashing their potatoes.
Congress approved the US National Earthquake Hazards Reduction Program in 1977. Almost half of its $30 million budget was reserved for prediction research. “People were measuring everything from cockroach activity to radon emissions, seismic patterns to strange electrical currents,” John Filson, former chief of the USGS Office of Earthquake Studies, told Earth magazine.
Most of the systems, though, relied upon mathematical theories and series of dichotomous keys—statements like “the area has a history of earthquake activity” to which researchers can answer “true” or “false.” UCLA’s Vladimir Keilis‐Borok quickly became the recognized leader in the field. The Russian‐born geophysicist founded the International Institute of Earthquake Prediction Theory and Mathematical Geophysics in 1981. There, he and his team used everything from chaos theory to geodynamical statistics to create an algorithm capable of detecting earthquake patterns. To test the program, he also began using it to predict US presidential elections. He died in 2013, but his colleague Allan Lichtman, who helped to develop the election keys, continues the presidential‐election‐prediction work. They still maintain 100 percent accuracy—including the successful prediction of Donald Trump over a month before his upset victory over Hillary Clinton.
The earthquake side of this work can’t boast the same batting average. To date, they’ve successfully predicted just one earthquake—the M 8.1 quake that struck Hokkaido, Japan.
Allan Lindh has been at the center of US earthquake prediction since its inception. Lindh joined the USGS in 1970 and was a reluctant scientist. He’d spent most of his teens and twenties traveling the West Coast with a tattered copy of Jack Kerouac’s On the Road as his guide. He eventually landed in British Columbia, where he stacked lumber in what he calls “crummy little sawmills” before meeting a geologist who told him there were easier ways to earn a living. A few months later, he arrived at Stanford. Ten years after that, he had a PhD in geophysics and a position with the USGS.
Lindh, who is now 70, says he was a “young punk—a nasty, obnoxious person” when he entered the federal agency. This is hard to believe of a man who wears a long white beard and even longer whiter locks in the spirit of Walt Whitman or Charlton Heston’s Moses—a man who spends most of his time today tending to his laying hens and writing polemics against nuclear war, whose LinkedIn profile lists his current occupation as an “apprentice at life” and whose telephone number I found, incidentally, on a website dedicated to meditation groups in what may be California’s crunchiest town. What is easier to believe is that the USGS was, in Lindh’s words, a largely autocratic organization when he was hired. In 1977, he sat in one of the organization’s field offices and listened as his supervisors detailed the list of projects they’d be undertaking in the coming year.
“I stood up and told them they should be ashamed of themselves. It was the dumbest list of experiments I’d ever seen. The USGS is supposed to be about public interest, not esoteric science. My supervisors told me if I was so smart I should go write my own project proposal. So I did.”
What resulted was the Parkfield prediction experiment.
The town of Parkfield, California, isn’t much of a town at first glance: an unincorporated community located in the Central California mountains about halfway between LA and San Francisco, it has a population of either 34 or 18, depending on whom you ask. There isn’t much in the way of infrastructure there either. The nearest gas station is 25 miles away. The closest grocery store is even farther.
But what Parkfield lacks in amenities, it more than makes up for in seismic activity. Parkfield sits atop San Andreas. That in itself isn’t all that unusual in the Golden State, but the particular patch of fault on which this little ranching community sits is. Parkfield exists at the nexus of two very different types of fault activity: To its south, San Andreas is loaded, much like the Red Canyon and Hebgen Faults were before they erupted in 1959. The section of San Andreas north of Parkfield, on the other hand, is slowly creeping. This nexus of activity offered Lindh and other seismologists something they couldn’t find anywhere else: reliable earthquakes. A magnitude 6.0 earthquake had occurred there in 1966. And in 1934. Also in 1922 and 1901. Logic dictated another one would happen around 1988.
That offered a once‐in‐a‐lifetime opportunity: to record and study an earthquake in real time—and, more important, to figure out what happens in the days and hours before the quake.
“We figured it was shooting sh in a barrel,” says Lindh.
Working with the state of California and several branches of the university system there, Lindh and his colleagues at the USGS installed arrays of instruments, including devices that sound like they belong in Adam West’s Batcave: creepmeters, proton magnetometers, motion accelerometers, trilateration arrays. They measured strain and ground deformation. They studied the earth’s magnetic field. Basically, says Lindh, they installed any instrument they could think of that might contribute data to the study of how quakes begin and behave. The tiny town began to look like a NASA flea market.
“We told the people at Parkfield we didn’t know what we were doing,” recalls Lindh. “They kept their cool. They knew we were just doing our job.”
More than that, says Lindh, the residents of Park eld were exactly the kind of people to get behind this sort of thing. “They weren’t the kind of people who were going to freak out the minute you started to predict earthquakes around them.” Instead, most of them thought it was pretty great. The town had a one‐room schoolhouse that looked like a decrepit mobile home. Its lone employee, who also served as a de facto mayor, was charged with manning some of the instruments. Other ranchers offered up their land for further instrumentation. There was a kind of ebullience around the whole thing, says Lindh. It felt like a community effort.
Jack Varian is one of those residents. He and his wife, Zera, moved to the area in 1961, after learning the hard way that mountain ranching is a lot more struggle than it is success. They made a decent go of it until drought struck in the 1980s, making ranching hard even in places like Parkfield. With few prospects and four kids, the couple decided they needed a new plan to breathe life into the region.
For over a decade, the area’s only tourism was predicated almost entirely on the untimely death of James Dean: He’d been driving his Porsche convertible, nicknamed “Little Bastard,” to a race in Salinas when he T‐boned another driver just outside of town. The Hollywood star died en route to the hospital. A memorial quickly sprang up just down the road from the scene of the crash and just across from the Jack Ranch Cafe. People came to pay their respects and eat tuna melts. Varian figured he could start a café of his own and built the Parkfield one in 1985. It wasn’t much of a café, really—just a barbecue pit outside a log cabin decorated with junk he and Zera found around the area.
His timing was far from random. That same year, Lindh and the Parkfield project issued its of official prognostication: Parkfield would see a magnitude 6.0 quake by 1993. Given the historical regularity of quakes there, this was far from an audacious claim. But it was revolutionary if for no other reason than that US scientists had never made a prediction like this before.
The prediction, says Lindh, was based on a single computational model. He and the rest of the Parkfield team took it as a given that the next Parkfield earthquake would look like the others—that it would rupture in the same direction, and that it would be preceded by the same sequence of foreshocks. They assumed that sections of fault like the one at Parkfield have an identifiable threshold for stress—that you can depend upon it snapping whenever that threshold is met. There was no reason to believe that wouldn’t occur in the same time frame it had for over a hundred years there.
Varian liked the sound of it. He came up with the town’s motto, “Be here when it happens,” which he painted in large letters on the iconic water tank outside the café. He built a seven‐room guest lodge with the slogan “Sleep here when it happens.” And he orchestrated the café’s menu accordingly (these days, its most popular item is the “Shakin’ Burger,” made with beef raised on Varian’s ranch).
Years went by. People continued to wait for the Parkfield quake. Meanwhile, the state of California’s Office of Emergency Services began dabbling in short‐term earthquake warnings. They issued their first one in 1985 when the USGS saw an increased risk of a moderate quake occurring in San Diego. The state issued another advisory in June 1988, after a magnitude 5.1 earthquake erupted near San Francisco. The USGS announced there was a five percent chance another would follow in the next five days. It didn’t. They issued a similar five‐day prediction after a similar quake rumbled through the region in August 1989. The quake didn’t come in the window promised, but when Loma Prieta slammed the region two months later, people began wondering if there was something to the process.
Meanwhile, a New Mexico–based climatologist named Iben Browning claimed to have predicted the Loma Prieta quake using computations based on the earth’s gravity and tidal patterns. Half‐way across the country, the USGS had become simultaneously interested in the New Madrid fault. The organization published estimates stating there was anywhere between a 13 and 65 percent chance the fault would rupture again before 2000. Browning took it one step further: He predicted there’d be a major quake there on December 3, 1990. Ads began appearing in papers like the St. Louis Post-Dispatch proclaiming Browning a “remarkable man” and including a quote from Southeast Missouri State University geologist David Stewart calling Browning “the most intelligent person” Stewart had ever met. For just $39, readers could purchase a VHS tape with 30 minutes of excerpted interviews with Browning in which he explained his predictive process. The full 90 minutes could be purchased for a mere $99. I called the 1‐800 number listed on the ad and got a vague recording asking me to call back—no matter when I called.
Browning’s theory was based on ideas of tidal loading, which is to say the way in which tides redistribute ocean mass and the subsequent shift in stress that occurs. Browning had noted that, on December 2 and 3 of 1990, high tide would occur when the earth and moon were at their closest point and when the earth, sun, and moon were in alignment. These conditions had occurred a few other times that century without any noticeable increase in seismic activity, but that didn’t dissuade Browning. Neither did the resounding rejection of his theory sent up by the seismological community. Geologists and seismologists at regional universities all made public pronouncements against Browning’s claims. An ad hoc community comprising scientists from the USGS, the California Division of Mines and Geology, and the Argonne National Lab did the same. They all cited the lack of precedent for this theory—and Browning’s clear lack of scientific method.
Nevertheless, media coverage of his prediction exploded. Days before the predicted earthquake, dozens of news trucks, tilting under the weight of their giant satellites, parked in every available space in the tiny town of New Madrid. Reporters from tiny local papers rubbed elbows with correspondents from The New York Times, the Miami Herald, the LA Times, Newsweek, and Time. Over 200 media outlets in total had camped out in the town, including Good Morning America and the Today Show, both of which scored interviews with Browning. Even the BBC showed up.
Local emergency managers took note. I remember. I was a sophomore in high school then, and the usual tornado and fire drills were suddenly augmented by earthquake protocols as well. We did a lot of those drills that year. Communities around the region, including the cities of St. Louis and Louisville, participated in their own disaster drills. State national guards called up their troops. Homeowners rushed to buy earthquake insurance. Then, on September 26 of that year, the town of Cape Girardeau, Missouri, experienced an M 4.7 quake: Clearly, speculated many in the region, this was a foreshock heralding the validity of Browning’s prediction. The frenzy intensified.
By December 2 of that year, the town of New Madrid—unsure whether to plan for a party or a mass funeral—had decided to embrace both. The chamber of commerce sold T‐shirts and coffee mugs; local restaurants trotted out earthquake specials; a bluegrass band with the name Faultline Express played on a stage erected expressly for them. Meanwhile, other residents hung “SOS” banners from their homes and sat for doomsday sermons given by local preachers. Rumors that the state had ordered truckloads of body bags prompted thousands of parents to keep their kids home from school.
December 3 came and went. The local McDonald’s handed out free coffee. The Red Cross distributed earthquake preparedness kits. The quake never manifested. David Stewart resigned his post as director of the SMSU earthquake center. Iben Browning began to fade into oblivion. He died of a heart attack six months after the quake that never happened.
The same year Browning made his prediction, Alan Lindh was more confident than ever about Parkfield, California’s chances for a big event.
In 1990, Lindh authored an article in MIT Technology Review in which he put the chance of an earthquake hitting Parkfield by 1993 at 95 percent. He was interviewed by The New York Times in January of that year. The reporter characterized him as a “modern‐day Jeremiah,” one of the major prophets of the Old Testament. Lindh corrected him: “It’s not like being Jeremiah,” he told the reporter. “It’s like being a high school teacher explaining AIDS to teenagers.” Times readers took the comment in stride.
Months went by. Still no earthquake.
Andrew Michael was a field geologist for the USGS working in Parkfield then. He says the overall footprint of the project was never all that big: It began with two technicians working in a small ranch house. As the years went on, they were down to one employee and a single trailer home. Most of the data from the instruments was, instead, patched back to the USGS of offices in Menlo Park. Geologists like Michael wore pagers that would alert them to any activity.
When a Japanese film crew arrived in Parkfield to document the experiment, they were incredulous—and more than just a little suspicious.
“They kept demanding, ‘Why are you hiding your control room from us?’” says Michael. “It was inconceivable to them that we didn’t have one.”
Meanwhile, project supervisors were under intense pressure to step down the experiment even further.
“A lot of people wanted the money for Parkfield redirected for other things,” says Michael. “We had one of the best strong motion arrays there, and there were definite pressures to move it to Southern California where there were a lot more people.”
The National Earthquake Prediction Evaluation Council responded by conducting an extensive review of the Parkfield project. They concluded that, even though Parkfield hadn’t yet seen another quake, it was still “the most likely place yet identified to trap a moderate earthquake in a densely instrumented region and the best locale identified to answer a number of important scientific questions about the seismic source.” The project, they recommended, should be continued. A 1995 fact sheet issued by the USGS promised that earthquake forecasting had advanced from a research frontier to an emerging science—and one “reliable enough to make official earthquake warnings possible.”
For his part, Lindh began reciting his favorite quotation by Charles Richter: “Only fools, liars, and charlatans predict earthquakes.” It’s still a line he uses a lot today. Michael says he gradually began focusing on other projects. But as far as he was concerned, Parkfield was still the most likely place for a next earthquake.
They finally got it on September 28, 2004. Michael remembers feeling surprised.
“I was like, ‘Really?’ The earthquake wasn’t something I was thinking about very often at that point.”
Lindh was. He says that, even though the quake was over ten years late, it happened. And it brought with it the opportunity to do some real on‐the‐ground science. Seismologists learned a lot from the delayed quake, in terms of what it looked like both as its own event and as part of a historical sequence. The 2004 quake ruptured differently than its predecessors, in terms of both the geographic direction of the quake and the rumblings preceding it. Those variables, says Lindh, have offered important contributions to seismic modeling and assessing hazard. If we are ever going to predict quakes, he says, we’re going to need a lot more in the way of sophisticated modeling to do so.