The Signal and the Noise

by Nate Silver

  No earthquake hit Sulmona that day. After the prediction failed, Giuliani was reported to local authorities for procurato allarme (bringing about alarm)—in essence, having yelled fire in a crowded theater. He was forced to remove his predictions from the Internet for fear of triggering further panic.

  Authorities in L’Aquila told the residents the earthquake swarm* was nothing to worry about; the fault was helpfully discharging energy, explained Bernardo De Bernardinis, the deputy chief of Italy’s Civil Protection Department,2 reducing the threat of a major earthquake. He agreed with a reporter that they should sit back and enjoy a glass of wine;3 De Bernardinis recommended a local specialty, a Montepulciano.

  A major earthquake did hit L’Aquila, however. Measuring at magnitude 6.3, it came at 3:32 A.M. local time on Monday morning. Shaking houses from their foundations, caving in roofs, and turning furniture into projectiles, the quake killed more than 300 residents, left another 65,000 homeless, and caused more than $16 billion in damage.4

  What We Do When Our Foundations Are Shaken

  L’Aquila ought to have been better prepared. The city sits near a particularly violent type of fault known as a subduction zone, where the African Plate, one of the eight major tectonic plates that cover the earth’s surface, slips slowly and inexorably beneath the Eurasian one. Its first significant earthquake was recorded in 1315, and earthquakes struck again in 1349, 1452, 1461, 1501, 1646, 1703, and 1706;5 the most serious one, in 1786, had killed more than 5,000 people. Each time, often on direct order of the pope,6 the town was rebuilt and repopulated.

  Since then, L’Aquila had tempted fate for more than two centuries. An earthquake hit in 1958, but it was fairly minor—magnitude 5.07—and only the town’s oldest residents would have remembered it. The 2009 earthquake was much more powerful. The magnitude scale is logarithmic; a one-point increase in the scale indicates that the energy release has multiplied by thirty-two. Thus, the 2009 earthquake, magnitude 6.3, was about seventy-five times more powerful than the one that had hit L’Aquila in 1958. And it was about 3,000 times more powerful than the tremors—foreshocks to the major earthquake—that L’Aquila had experienced earlier that evening.

  Still, while the 2009 earthquake was large by Italian standards, it was barely a hiccup on the global scale. The earthquake that devastated Japan in 2011 measured at magnitude 9.0 or 9.1—almost 11,000 times more powerful. And the largest earthquake recorded since reliable estimates were possible, which hit Chile in 1960 and measured magnitude 9.5, was about 60,000 times stronger than the L’Aquila quake.

  Why, then, did L’Aquila—a fairly well-to-do town in a wealthy, industrialized nation—sustain such significant damage? One reason was the city’s geology—L’Aquila sits on an ancient lake bed, which tends to amplify the earth’s shaking. Mexico City was also built on an ancient lake bed,8 and 10,000 were killed there in 1985 from an earthquake whose epicenter was more than two hundred miles away.

  But the major reason was simply that the town had become complacent about the seismic danger that lay just fifteen kilometers underground. There was nothing resembling the proper level of earthquake readiness:9 building codes, emergency supplies, community drills. Not only were centuries-old buildings leveled by the tremor, but so too were many modern ones, including a wing of a hospital that had been erected as recently as 2000. A little bit of warning would have saved untold lives there.

  Had Giampaolo Giuliani provided that warning? In the Italian tabloids, he had become something of a savant and a martyr. Soft-spoken and disheveled, and often wearing the colors of the local soccer team, he played the role of the humble civil servant or absentminded professor whose insights had been ignored by the scientific establishment. He claimed that he had warned friends and family about the L’Aquila quake and was prevented from telling others only because of the police order against him. He demanded an apology from the authorities—not to him, he said, but to the people of L’Aquila.

  Never mind that Giuliani had not actually predicted the earthquake. His prediction had been very specific: Sulmona, not L’Aquila, was at greater risk, and the earthquake would come in March rather than April. In fact, he had suggested to a local newspaper that the danger had passed. “To simplify the concepts,” he said before launching into a rambling explanation about the lunar cycle, “the Earth-Moon system has come to visit at perihelion . . . the minimum distance from Earth, and aligned with the planet Venus. . . . I feel I can reassure my fellow citizens because the swarm will be diminishing with the end of March.”10

  Perihelion with the planet Venus? Radon gas? What did any of this have to do with earthquakes? And what about Giuliani’s failed prediction in Sulmona? It didn’t matter. When catastrophe strikes, we look for a signal in the noise—anything that might explain the chaos that we see all around us and bring order to the world again. Giuliani’s rambling explanations were the closest thing available.

  No type of catastrophe is more jarring to our sense of order than an earthquake. They quite literally shake our foundations. Whereas hurricanes descend upon us from the heavens and have sometimes been associated with metaphors for God’s providence,* earthquakes come from deep underneath the surface and are more often taken to be signs of His wrath,11 indifference,12 or nonexistence. (The Lisbon earthquake of 1755 was a major spark for the development of secular philosophy.13) And whereas hurricanes—along with floods, tornadoes, and volcanoes—can often be forecasted in advance, earthquakes have defied centuries of efforts to predict them.

  Magic Toads and the Search for the Holy Grail

  Pasadena, California, has long been the world’s epicenter for earthquake research. It is home to the California Institute of Technology, where Charles Richter developed his famous logarithmic scale in 1935. The United States Geological Survey (USGS) also has a field office there, where most of its earthquake specialists reside. I traveled there in September 2009 to meet with Dr. Susan Hough, who is one of the USGS’s top seismologists and who has written several books about earthquake prediction. She had watched Giuliani’s television interviews with suspicion and had written a blistering editorial in the New York Times14 that criticized both Giuliani and the attention paid to him.

  Hough’s editorial argued that Giuliani’s success was merely coincidental. “The public heard about Mr. Giuliani’s prediction because it appears to have been borne out,” she wrote. “But there are scores of other [incorrect] predictions that the public never hears about.”

  If you have hundreds of people trying to make forecasts, and there are hundreds of earthquakes per year, inevitably someone is going to get one right. Giuliani’s theories about radon gas and lunar cycles had been investigated many times over15 by credentialed seismologists and had shown little or no ability to predict earthquakes. Giuliani had been lucky: the monkey who typed Shakespeare; the octopus who predicted the World Cup.

  Hough’s office at the USGS sits near a quiet corner of the Caltech campus where there are more eucalyptus trees than students. She seemed a little road weary when I met her, having just returned from a trip to Turkey where she’d been to study a system of earthquake faults. She has soft features and frizzy hair and her eyes are dark, tired—skeptical. “What’s your day job?” she quizzed me a few moments after I greeted her.

  At one point, she pulled a pocket-size globe off her desk, the sort that looks like it was bought at an airport gift shop. She took her index finger and drew a line across the surface of the globe, starting in the Sea of Japan and moving east–southeast.

  “They are really concentrated in this belt—stretching from southern China through Greece,” Hough explained, referring to the world’s most destructive earthquakes. “It’s a complicated earthquake zone, a lot of buildings with vulnerable construction. If you put a big earthquake under Tehran, you could kill a million people.”

  Indeed, almost all the deadliest earthquakes in modern history (figure 5-1) have occurred along the path that Hough outlined, one which passes through the
Cradle of Civilization in the Middle East and through some of the most densely populated regions of the planet, including China and India. Often poor and crowded, these areas lack the luxury to prepare for a once-per-three-hundred-year catastrophe. But the death tolls can be catastrophic when earthquakes hit, stretching into the hundreds of thousands.*

  FIGURE 5-1: DEADLIEST EARTHQUAKES SINCE 1900

  Earthquakes kill more people than hurricanes, in fact,16 despite seeming like the rarer phenomenon.17 Perhaps that is because they are so seldom predicted successfully. Whereas the landfall position of hurricanes can be forecasted at least three times more accurately now than they were even twenty-five years ago, the science of earthquake forecasting seems barely to have evolved since the ninth century A.D., when the Japanese first claimed to be able to anticipate earthquakes by looking at the behavior of catfish.18 (Cows, pigs, eels, rats, parakeets, seagulls, turtles, goldfish, and snakes have also been reported at various times to behave unusually in advance of an earthquake.)

  Kooks like Giuliani are still taken seriously, and not just in the Italian tabloids.19 The California Earthquake Prediction Council receives hundreds of unsolicited earthquake forecasts per year, most of which, the agency says, “discuss the strange behavior of household pets, intuition, Aunt Agatha’s aching bunions, or other mysterious signs and portents that scientists simply don’t understand.”20 Meanwhile, some of the stuff in academic journals is hard to distinguish from ancient Japanese folklore. A 2010 paper21 in a relatively prestigious journal, The Journal of Zoology, observed that toads in a pond fifty miles from L’Aquila had stopped spawning five days before the major earthquake there.22 Remarkably, it asserted that this was evidence that they had predicted the earthquake.

  It’s research like this that exhausts Hough. “If you look back in time, certainly going back to the 1970s, people would come up with some idea—they’d be optimistic—and then you wait ten years and that method would be debunked,” she told me. “Ten years later, you have a new method and ten years later it’s debunked. You just sort of sense a theme. Most top scientists at this point know better than to chase after a Holy Grail that probably doesn’t exist.”

  But while Giuliani’s close encounters with Venus or the toads are easy to dismiss, is there really no way at all to predict an earthquake? What about the swarm of smaller quakes around L’Aquila just before the Big One hit? Was that just a coincidence? The seismological community has a reputation for being very conservative. It was very slow to accept the theory of plate tectonics, for instance23—the now broadly accepted notion that the shifting of the earth’s continental plates is the primary cause for earthquakes—not adopting it into their canon until the 1960s even though it was proposed in 1912. Had Hough’s skepticism crossed the line into cynicism?

  The official position of the USGS is even more emphatic: earthquakes cannot be predicted. “Neither the USGS nor Caltech nor any other scientists have ever predicted a major earthquake,” the organization’s Web site asserts.24 “They do not know how, and they do not expect to know how any time in the foreseeable future.”

  Earthquakes cannot be predicted? This is a book about prediction, not a book that makes predictions, but I’m willing to stick my neck out: I predict that there will be more earthquakes in Japan next year than in New Jersey. And I predict that at some point in the next one hundred years, a major earthquake will hit somewhere in California.25

  Both the USGS and I are playing some semantic games. The terms “prediction” and “forecast” are employed differently in different fields; in some cases, they are interchangeable, but other disciplines differentiate them. No field is more sensitive to the distinction than seismology. If you’re speaking with a seismologist:

  A prediction is a definitive and specific statement about when and where an earthquake will strike: a major earthquake will hit Kyoto, Japan, on June 28.

  Whereas a forecast is a probabilistic statement, usually over a longer time scale: there is a 60 percent chance of an earthquake in Southern California over the next thirty years.

  The USGS’s official position is that earthquakes cannot be predicted. They can, however, be forecasted.

  What We Know About How Earthquakes Behave

  If you explore the USGS Web site, in fact, you’ll find that it makes lots of tools available to help you forecast earthquakes. One particularly neat one is an application that lets you type in the longitude and latitude at any point in the United States; it will estimate the long-term probability of an earthquake there.26 In figure 5-2, I’ve listed the probabilities for earthquakes in a variety of major U.S. cities as provided by the USGS Web site.

  We all know that California is very seismically active; the USGS estimates that an earthquake of magnitude 6.8 or higher will hit San Francisco about once every thirty-five years. Many of you will also know that Alaska has many earthquakes—the second largest one in recorded history, magnitude 9.4, hit Anchorage in 1964.

  FIGURE 5-2. FREQUENCY OF A MAJOR (>= MAGNITUDE 6.75) EARTHQUAKE WITHIN A 50-MILE RADIUS OF SELECT U.S. CITIES

  Anchorage

  1 per 30 years

  San Francisco

  1 per 35 years

  Los Angeles

  1 per 40 years

  Seattle

  1 per 150 years

  Sacramento

  1 per 180 years

  San Diego

  1 per 190 years

  Salt Lake City

  1 per 200 years

  Portland, OR

  1 per 500 years

  Charleston, SC

  1 per 600 years

  Las Vegas

  1 per 1,200 years

  Memphis

  1 per 2,500 years

  Phoenix

  1 per 7,500 years

  New York

  1 per 12,000 years

  Boston

  1 per 15,000 years

  Philadelphia

  1 per 17,000 years

  St. Louis

  1 per 23,000 years

  Atlanta

  1 per 30,000 years

  Denver

  1 per 40,000 years

  Washington, DC

  1 per 55,000 years

  Chicago

  1 per 75,000 years

  Houston

  1 per 100,000 years

  Dallas

  1 per 130,000 years

  Miami

  1 per 140,000 years

  But did you know about Charleston, South Carolina? It is seismically active too; indeed, it experienced a magnitude 7.3 earthquake in 1886. The USGS estimates that there will be another big earthquake there about once per six hundred years. If you live in Seattle, you should probably have an earthquake plan ready; it is more earthquake-prone than many parts of California, the USGS says. But you don’t need one if you live in Denver, which is a safe distance away from any continental boundaries.

  This seems like an awful lot of very specific and user-friendly information for an organization whose party line is that it is impossible to predict earthquakes. But the USGS’s forecasts employ a widely accepted seismological tool called the Gutenberg–Richter law. The theory, developed by Charles Richter and his Caltech colleague Beno Gutenberg in 1944, is derived from empirical statistics about earthquakes. It posits that there is a relatively simple relationship between the magnitude of an earthquake and how often one occurs.

  If you compare the frequencies of earthquakes with their magnitudes, you’ll find that the number drops off exponentially as the magnitude increases. While there are very few catastrophic earthquakes, there are literally millions of smaller ones—about 1.3 million earthquakes measuring between magnitude 2.0 and magnitude 2.9 around the world every year.27 Most of these earthquakes go undetected—certainly by human beings and often by seismometers.28 However, almost all earthquakes of magnitude 4.5 or greater are recorded today, however remote their location. Figure 5-3a shows the exponential decline in their frequencies, based on actual records of earthquakes from January
196429 through March 2012.30

  It turns out that these earthquakes display a stunning regularity when you graph them in a slightly different way. In figure 5-3b, I’ve changed the vertical axis—which shows the frequency of earthquakes of different magnitudes—into a logarithmic scale.* Now the earthquakes form what is almost exactly a straight line on the graph. This pattern is characteristic of what is known as a power-law distribution, and it is the relationship that Richter and Gutenberg uncovered.

  Something that obeys this distribution has a highly useful property: you can forecast the number of large-scale events from the number of small-scale ones, or vice versa. In the case of earthquakes, it turns out that for every increase of one point in magnitude, an earthquake becomes about ten times less frequent. So, for example, magnitude 6 earthquakes occur ten times more frequently than magnitude 7’s, and one hundred times more often than magnitude 8’s.

  What’s more, the Gutenberg–Richter law generally holds across regions of the globe as well as over the whole planet. Suppose, for instance, that we wanted to make an earthquake forecast for Tehran, Iran. Fortunately, there hasn’t been a catastrophic earthquake there since its seismicity began to be measured. But there have been a number of medium-size ones; between 1960 and 2009, there were about fifteen earthquakes that measured between 5.0 and 5.9 on the magnitude scale in the area surrounding the city.31 That works out to about one for every three years. According to the power law that Gutenberg and Richter uncovered, that means that an earthquake measuring between 6.0 and 6.9 should occur about once every thirty years in Tehran.