The press and the public will go toward the suggestion of prediction like hogs to a trough.
—From the never-delivered retirement speech of Charles Richter
ONCE UPON A TIME, ALL DISASTERS—STORMS, FLOODS, HURRICANES, earthquakes, eruptions—arrived without warning.
Afterwards, the survivors would reflect on whether there had been some unusual precursors. On the volcano Hekla in Iceland, the sheep farmers noticed that the streams dried up before the start of each new eruption.1 Thirty-six centuries ago, the inhabitants of Akrotiri on the Greek island of Santorini correctly interpreted the signs that a cataclysmic volcanic explosion was brewing and evacuated their town.2
Of all classes of catastrophe, volcanoes prove to show the earliest premonitory signs. The question is then what action to take. A successful evacuation requires a leader willing to risk his or her reputation and a people prepared to abandon their fields and homes. The largest eruptions have the strongest signature, but also require retreating to the greatest distance.
The city of St. Pierre was founded in the early seventeenth century on the northwest lee of the island of Martinique, along a narrow coastal shelf of land beneath the steep slopes that rise up to two mountains, one of them the highest peak on the island, 4,600-foot (1,400-meter) Mont Pelée. Through the nineteenth century, the town flourished, gaining streets of two- or three-storied, balconied houses lined with tropical trees and gardens—St. Pierre became “the little Paris of the Caribbean.”3 By 1900, the population was 26,000.
At the start of 1902, plumes of steam were spotted rising out of the ground on the flanks of Mont Pelée. The emanations increased in intensity through April, and there was a strong smell of sulfur in the city.4 There were many small earthquakes, and the underwater cable connecting St. Pierre to Dominica and Guadeloupe mysteriously broke. On April 26, walkers visiting the summit found a lake of hot water filled from a boiling water spring on a new 50-foot (15-meter) ash cone. The local newspaper called off the picnic on the mountain planned for May 4. Then hundreds of deadly fer-de-lance snakes were spotted coming into the city.
On May 5, the crater wall containing the lake collapsed, and a hot mudflow descended the Blanche River Valley, burying the twenty-five workers at the Guerin sugar factory. People in outlying villages around the northern tip of the island started to arrive in St. Pierre, where they were housed in churches by the priests. During the night, the city’s electricity gave out.
Everyone was aware that an eruption was brewing. The governor at Fort-de-France, on the south side of the island, convened a panel of “experts,” who concluded that the deep valleys immediately beneath the volcano would channel the mudflow or ash cloud from any eruption directly to the sea, protecting the city. To impress the nervous citizens, on May 6 the governor and his wife arrived to stay in St. Pierre. That day an eruption of ash fell over the city, muffling the steps of people walking through the streets. Rumblings from the volcano were growing louder, intermixed with roars from avalanches of rock descending the valleys immediately beneath the volcano. People were beginning to panic. On the afternoon of May 7, the mayor requested soldiers to help keep order. Without such restraint, many people would have already left the town. A reassuring statement was published in the St. Pierre newspaper telling people not to worry. The more fearful went to the cathedral to spend the night praying.
On the bright morning of Ascension Day, May 8, at 7:52 a.m., a plug of lava blocking the mouth of the volcano suddenly burst out, releasing a foam of molten magma, ash particles, and superheated gases. As viewed from a cable-repair ship moored off the coastline, the mountainside beneath the crater suddenly ripped open, and a dense dark cloud shot out horizontally, accompanied by a second dense black cloud that soared upwards into a gigantic mushroom formation that caused darkness to descend for 25 miles around. The horizontal black cloud plunged down the western flank of the volcano, reaching the city in less than a minute. The cloud was glowing red-hot at an estimated 1,000 degrees Celsius. As it reached the city, everything in its path caught fire. Thousands of barrels of rum burst, sending streams of flames into the streets. All vegetation and buildings were incinerated over eight square miles, and all 30,000 inhabitants and visitors to the city were killed, suffocated, and burned—all except a prisoner, Ludger Sylbaris, who had been held deep underground for the night in the town’s jail. The black cloud was afterwards known as a nuée ardente—a burning, glowing avalanche. The theory advanced by the governor’s experts that the valleys descending from the volcano would protect the city failed because such an avalanche can ride over ridges as it plunges downhill.5 The governor and his wife died trying to convince the citizens that the city was safe.
The lessons from St. Pierre are still relevant today. What is considered a “safe refuge” may simply be someone’s optimistic theory, whether hurricane shelters built two feet above sea level on New Providence Island in the Bahamas or the evacuation destinations overwhelmed by the tsunami along the Tohoku coast of Japan in 2011.
After Mont Pelée erupted, no city administrator in the Lesser Antilles Islands would again dare to ignore the symptoms of an emerging eruption. However, volcanoes can be fickle: sometimes they come back to life with only explosions of superheated steam (so-called phreatic eruptions). The critical question then concerns whether the gases or ash particles reveal that fresh magma is rising, in which case the steam explosions would be preparatory to a much more catastrophic eruption.
In the second half of 1975, a series of earthquakes were recorded beneath the Grande Soufrière volcano on Basse-Terre, the southwest part of the island of Guadeloupe, sister département to Martinique.6 The tremors increased into the early summer of 1976; then, at 8:55 a.m. on July 8, there was a large phreatic (steam) explosion at the volcano. A thick cloud laden with ash descended the mountain and plunged the town of Saint Claude, at the upper end of the city of Basse-Terre, into twenty minutes of darkness. With the folk memory of St. Pierre still strong, the people panicked: over the next two hours, 25,000 fled to the Grande Terre peninsula to the northeast of the island. On July 13, the head of the French volcanological service, Haroun Tazieff, arrived. A Polish-born and Belgian-educated former mining engineer, Tazieff was a charismatic daredevil who had made television programs about his exploits climbing down to lava lakes and standing next to fumaroles. (He had been declared one of the six most famous men in France.) He announced that, without any sign of fresh magma in the eruption on July 8, there was unlikely to be a Mont Pelée–style catastrophe involving the dreaded nuée ardente.7 Tazieff returned to Paris on July 17, leaving a small monitoring team. On the 24th and 28th, further steam explosions erupted from the volcano. Meanwhile, Tazieff set off on a long-planned expedition to the volcanoes of Ecuador. In his absence, he transferred responsibility to the director of the Seismic Research Unit in Trinidad, John Tomblin, who had twenty years’ experience observing Caribbean volcanoes.
Upon his arrival on the island on August 3, Tomblin declared that the volcano was capable of another Mont Pelée.8 Back in Paris, on August 8, Claude Allègre, a very capable thirty-nine-year-old geochemist, was named head of the Institut de Physique du Globe de Paris (IPG), a position that made him Tazieff’s new boss. On August 9, there was a large, explosive steam-and-ash eruption at the volcano that projected blocks weighing up to 200 pounds (100 kilograms). Tomblin wrote that there was a one-in-four chance that this would end in a magma eruption, and that it was a coin toss whether such an eruption would be cataclysmic. On August 12, when there were more explosions, a French petrologist sent to Guadeloupe, Professor Robert Brousse, made a critical announcement: contradicting Tazieff, he said that there were traces of fresh magma in the ash from the July eruption. The prefect of the island commanded that the population of Basse-Terre start evacuating; by the 15th, 76,000 people had left their homes all around the volcano.
Professor Brousse was quoted in headlines saying that the course was now irreversible: they were heading toward a great eruption that could be 30 megatons in size. The prefect upped the stakes and declared that there was only a 1-in-2,000 chance they would avoid a significant eruption.
On August 26, Claude Allègre arrived in Guadeloupe, and on the 29th Tazieff returned from Ecuador to discover he had a new boss. He immediately expressed his amazement that the “so-called” experts had termed the situation “irreversible” and declared that, from the evidence to hand, he could not see any immediate danger.
On August 30, Tomblin, Tazieff, Allègre, and other scientists made a journey to the summit of the volcano. The petrologist Brousse stayed behind, as he had forecast a significant explosive eruption on that day. At 10:30 a.m., an eruption occurred while they were standing close to the crater. Projectile blocks of up to a cubic meter were hurled into the air. Three of the party, including Tazieff, were injured by the raining stones and taken by helicopter to the hospital, but they were soon discharged. Tazieff’s action-man image was burnished as he told the press it was a miracle that none of them had been killed.9
On September 1, Tazieff gave a press conference at which he declared that Professor Brousse, who “had never studied a volcanic eruption previously,” had “panicked.” If Tazieff had been around, he would never have authorized the evacuation. Now, in direct conflict with his new boss, whose job was to ensure that there was a coordinated scientific commentary, Tazieff left Guadeloupe.
Tazieff was such a celebrity that even for this level of insubordination he was hard to dismiss. On September 6, Allègre attempted to dilute Tazieff’s influence by announcing that a “team of experts”—of which Tazieff could be a member—would be established to monitor the volcano. On September 9, Tazieff gave an interview to the local Guadeloupe radio station in which he described the pressure being put on him to stay silent and accused both Professor Brousse and Claude Allègre of incompetence. He also encouraged the local inhabitants to return, declaring it was safe to do so.10
On September 19, Tazieff was vindicated when analyses from laboratories at Los Alamos in New Mexico showed that there had been no fresh magma in the July eruption. On the 21st, the latest analyses of gases also confirmed that no new magma was rising.11 On October 11, the prefect announced the reopening of schools and reoccupation of villages to the east of the volcano. Most people returned on October 27, the same day Tazieff was fired from his job as head of volcanology at the IPG.
Meanwhile, Tazieff had become a hero in Guadeloupe. On November 20, the consul general arranged a series of meetings—billed as “For Truth, for Science, for Guadeloupe”—at which Tazieff spoke in front of thousands of Guadeloupians. Three days later, 6,000 Guadeloupians had signed a petition demanding that Professors Brousse and Allègre be banned from any further role in monitoring volcanoes on the island and that this function be handed to Tazieff.12
The last village up the mountain from Basse-Terre, St. Claude, reopened on December 1. There were a couple of steam explosions early the next year, and then it was over. The evacuation had lasted three months and cost several hundred million dollars in economic disruption.
The lesson taken from Guadeloupe concerned the thanklessness of organizing an evacuation when there might not be an eruption. Now the pendulum would swing the other way. Almost inevitably, people would die because of the fear of an unnecessary evacuation. It took less than a decade for this scenario to unfold.
Through 1985, the glacier-covered 17,500-foot (5,300-meter) Nevado del Ruiz volcano in Colombia, 80 miles (130 kilometers) west of Bogotá, showed signs of reawakening.13 After weeks of agitation, the volcano erupted shortly after 3:00 p.m. on November 13, 1985. The eruption was relatively small but generated enough heat to melt the summit glaciers and send lahar mudflows—a mix of volcanic ash and meltwater—racing down the slopes at up to one kilometer per minute and into the steep river valley ravines that led out into the surrounding plains.
A month before the eruption, hazard maps had been circulated showing that the town of Armero, in the plain just beyond the mountain, was at high risk from lahar floods.14 Local politicians accused the scientific and civil defense agencies of damaging the local economy with their scaremongering. After the ashfall from the afternoon eruption, the mayor and priest of the town set out to reassure the people. When civil defense officials from other towns in the path of the lahars witnessed the arrival of the massive muddy floods and tried to alert Armero, they could not get through because the electricity was down. Later that evening, the mudflow hit the town, destroying 5,000 houses and drowning three-quarters of the 29,000 inhabitants.
The next opportunity to test evacuation procedures for a big volcanic eruption came in 1991 on the island of Luzon in the Philippines. This time the procedure worked more or less to plan.15 Earthquakes and steam eruptions gave several months’ warning that the volcano Mount Pinatubo was coming back to life. Gas samples showed a big uptick in sulfur dioxide emissions, the signature of fresh magma rising. The swelling of the mountain was picked up by tilt meters. When, on June 7, an eruption burst 4 miles (7 kilometers) into the sky, everyone in the area, convinced that the risk was real, began formal evacuations. The presence of a nearby US airbase and a very well connected head of the Philippine volcano science agency ensured good coordination. Evacuation zones were designated at different radii around the volcano, out to almost 24 miles (40 kilometers), and were progressively depopulated as the eruption approached. Forty thousand people lived within 12 miles (20 kilometers) of Pinatubo, and another 370,000 lived within 24 miles. Although many people had little confidence that the government would provide them with adequate shelter and sanitation at their unknown destinations, by early June 250,000 people had been evacuated. The mountain exploded on June 15 in the second-largest eruption of the twentieth century. An estimated 847 were killed—mostly those who had chosen to stay in their village and then died from collapsing roofs laden with ash—but tens of thousands of lives were saved.16
After the success at Pinatubo, one might conclude that protecting people from devastating eruptions is now a problem solved. Yet in many ways Pinatubo was an ideal case. Evacuating ahead of the next big eruption in a populated region, wherever it is, may not be so straightforward. In April 2012, after scientists warned of an impending eruption of Popocatépetl in Mexico, only half the population agreed to evacuate.17 The remainder trusted in the power of the church bells to call the faithful to prayer to calm the volcano. Fortunately the eruption was not a big one.
How would today’s volcanologists handle the great 1815 eruption at Mount Tambora on the island of Sumbawa in Indonesia if that mountain had held off exploding for another two centuries? Would the evacuation begin in time? Would the evacuation destinations be set far enough away? Today there are 1.4 million people on the island of Sumbawa. On Bali and Lombok, how many people would die because of the collapse of their roofs from the weight of the ash blown by the easterly winds? When all the crops failed, would enough food and water be available to support more than 4 million inhabitants of Bali for a year or longer?
If the great 1883 eruption of Krakatoa had been delayed for 150 years, would the volcano forecasters identify tsunami as the most lethal hazard? Would people on the surrounding coasts really believe that they had to stay 100 feet (30 meters) or more above sea level?
Metropolitan Naples, home to almost 3 million people, is situated between two volcanoes: the mountain of Vesuvius broods to the east, while to the west lies the Campi Phlegraei—the 6-mile (10-kilometer)-wide caldera (“fiery fields”)—centered on Pozzuoli Bay.18 Evacuation plans exist (in principle at least) for removing the 550,000 inhabitants at greatest risk from pyroclastic avalanches in the “red zone” around Vesuvius within an assumed seventy-two-hour warning time.19 Each of eighteen municipalities in the red zone is twinned with an Italian province expected to transport and shelter the evacuees through the emergency. The evacuation plans are far too big and complex to be tested in a practice run, and the evacuation will be triggered by a largely untested eruption forecast procedure. There are few volunteers for the role of chief forecaster: the last time Italian geophysicists became linked with an over-optimistic disaster forecast, they were successfully prosecuted for manslaughter. Any future evacuation may well be overcautious and premature. While the eruption could happen faster than anticipated, it might also start and stop, with people returning, before it starts again.
The Campi Phlegraei Vent is believed to be the source of the enormous 40,000-year-old Campanian Ignimbrite Tuff, which has a volume of 120 cubic miles (500 cubic kilometers)20—more than 100 times the size of the AD 79 Vesuvius eruption—and covers 2,600 square miles (7,000 square kilometers)21 of the Campanian Plain to the east of the Bay of Naples, where in places it is 330 feet (100 meters) thick.22 There is no conceivable evacuation plan to handle a repeat.
Of all the places in the world with the potential for a supervolcano eruption, the most populated is the island of Kyushu in western Japan.23 There have been seven giant eruptions there in the past 120,000 years. The latest, at Mount Aso, erupted 144 cubic miles (600 cubic kilometers) of materials, leaving a 16-by-11-mile (25-by-18-kilometer) caldera, among the largest in the world. Seven million people live in areas that could be buried meters deep in pyroclastic flow avalanches. There is no evacuation plan.
Someone should create a video game in which eruption forecasters could explore all the dimensions of their challenging role of intervening between the willful volcano and the unforgiving politicians. Catastrophic eruptions are so rare that the national officials who get charged with issuing warnings are typically novices, with no previous experience. The lead eruption forecaster will have only a single opportunity to get it right, with dire political and personal ramifications for being too early or too late or for underestimating the required evacuation distance.
It might be better if there was a global agency tasked with providing this function. Such an agency could learn much from the long-running (1995–2010) dome collapse avalanches and explosions at the Soufrière Hills volcano on the island of Montserrat, where scientists coordinated a single perspective and made the talk of eruption probabilities as everyday as the weather forecasts. We must teach people how to respond to the uncertainties of a forecast and not be like the president of Haiti in 1963—who banned a hurricane forecast for fear it might induce panic.24
AFTER SPENDING TWO MONTHS IN CALIFORNIA EXPLORING THE San Francisco earthquake, in August 1906 Fusakichi Ōmori returned to Japan to cope with some insubordination at his Tokyo department. An assistant professor of seismology who was two years younger than Ōmori, Akitsune Imamura, believed himself every bit the equal of his boss.25 To prove the point, in 1905 Imamura published a textbook on seismology.
The same year Imamura wrote in a popular journal that sometime in the next fifty years Tokyo would be hit by a destructive earthquake in which 100,000 could die if the city once again caught fire. The Sagami Bay faults offshore to the south of the city had last generated a catastrophic earthquake in 1703. Based on recurring historical earthquakes further to the southwest, he believed that the time was approaching for a repeat.
Imamura had not checked with Ōmori before publishing his forecast. The following year Ōmori wrote an article for the same popular journal in which he scathingly compared Imamura’s prediction to the folkloric legend of the fire horse—that when the astrological symbols for “fire” and “horse” were aligned, “many cities would burn.” He wrote: “The theory that a large earthquake will take place in Tokyo in the near future is academically baseless and trivial.”
The feud simmered for years. In 1915 Imamura arranged another public clash over his earthquake forecast for Tokyo. This time Imamura was rebuffed by his boss so badly that he had to retire from the university for months and return to his home village.
Issuing forecasts of impending earthquakes was considered very much part of the scope of Japanese seismology in this period.26 However, as Ōmori wrote in a paper in 1920, “the repetition of destructive shocks necessarily at one and the same point is a great fallacy.”27
Ōmori’s certainty about Tokyo’s earthquake potential became progressively tested. On December 8, 1921, Tokyo experienced its largest earthquake in twenty-eight years, which came close to severing the city’s water supply. The shock on April 26, 1922, was even larger, damaging buildings, cutting the telephone service, and interrupting trains. Anxiety in the city was rising. To help reassure the public Ōmori publicized his official view that two successive strong earthquakes in the city were not unprecedented, but that the crisis was now over. When another earthquake occurred on January 14, 1923, he penned a scientific paper in which he asserted that “Tokyo may be assumed to be free in future from the visitation of a violent earthquake like that of 1855, as the later shock originated right under the city itself, and as destructive earthquakes do not repeat from one and the same origin, at least not in the course of 1,000 or 1,500 years.”28
In early August 1923, Ōmori set off by boat to travel to Australia to attend the Second Pan-Pacific Science Conference. On the afternoon of September 1, he was being shown a seismic recorder in the Sydney Observatory when he saw the needle stir and then wildly swing to record a trace of a large distant earthquake. The next morning he learned that he had witnessed the signature of the earthquake that had just destroyed Yokohama and Tokyo. The initial news suggested tens of thousands of casualties.
When he was interviewed by the local Australian papers, Ōmori declared that he “thought reports of the disaster had been exaggerated.” He traveled to Melbourne to embark on the Tenyo Maru, the first ship leaving for Tokyo. He became ill on the voyage, although he had sufficient energy to give a lecture on earthquakes to the passengers as they approached Japan. His deputy and rival Imamura had been the daily spokesperson with the Tokyo press during Ōmori’s absence, and he came to meet Ōmori on the quayside when the ship arrived on October 4, more than a month after the earthquake. In a poignant and humiliating moment, Ōmori apologized for having spent twenty years scorning his deputy’s earthquake forecast.
However, Ōmori was getting progressively more incapacitated and had to be hospitalized in a ward filled with earthquake victims. He was diagnosed with a brain tumor and died on November 8 at the age of fifty-five. His paper on the January 1923 earthquake discounting any prospect of Tokyo being hit by a catastrophic earthquake for 1,000 years, written before he left for Sydney, was cynically allowed to be published in the year after the earthquake. Imamura, hoping to heap further ignominy on his boss’s reputation, was the likely instigator.
In an article published in 1924, Imamura wrote that he had discussed the anticipated calamity at Tokyo in great detail, “but people refused to believe me. There was even an eminent scientist who ridiculed my opinion once (in 1905) and again in 1915 as nothing other than a rumor which might cause general panic.” Imamura took over Ōmori’s job as professor of seismology, with a reputation that would be hard to quench in Japan—as the seismologist who had finally mastered the mysteries of earthquake forecasting. These inflated expectations have been loaded on Japanese seismologists ever since.
IN THE TWENTIETH CENTURY, CALIFORNIA EXPERIENCED TWO OUTBREAKS of earthquake prediction fever.
Geodetic surveying after the 1906 earthquake showed that the sudden movement of up to 18 feet along the San Andreas Fault had relieved decades of accumulated strain. In 1924 the Coast and Geodetic Survey published its latest survey results for southern California: over the thirty years since the previous survey, an impressive 24 feet of movement had accumulated.29
Seeing the survey results, Bailey Willis, a former geology professor at Stanford University, wrote an article declaring that a major earthquake was imminent on the southern San Andreas Fault.30 He took his forecast to the insurance industry, which was responding to the dramatic demand for earthquake insurance that followed the 1925 Santa Barbara earthquake. In response to the news, insurers raised their earthquake insurance rates for the region; in fact, some were doubling them, and one even put up the price by a factor of twenty. The Los Angeles Chamber of Commerce was outraged at this attempt to panic people into believing that their city shared the earthquake perils that blighted San Francisco.31
In 1927 the Coast and Geodetic Survey discovered that its earlier survey work was flawed—the 24 feet of movement had melted away. The chamber commissioned the geologist Robert Hill to write a reassuring 1928 book, Southern California Geology and Los Angeles Earthquakes, in which he identified and mapped the faults of the Los Angeles metropolitan area and then one by one dismissed them as “inactive” and “largely things of past geologic time.”32 Of the coast-parallel Inglewood Fault, Hill noted, “it cannot be said there is any great menace.” Five years later, the Inglewood Fault ruptured in the 1933 Long Beach earthquake. Bailey Willis told a meeting of insurance industry executives in New York City that the 1933 earthquake had vindicated his prediction that “within 3, 7 or 10 years” a major earthquake could be expected, neglecting to mention that his forecast had been for a completely different fault.
This story would be echoed in the “Palmdale Bulge” saga of the 1970s, when a US Geological Survey geologist, Bob Castle, revisited old and new geodetic survey results and concluded that a region of land to the north of Los Angeles had become raised.33 In 1976 an article in Popular Science asked suggestively: “What is 10 inches high in some places, covers more than 4500 square miles, and worries the hell out of laymen and professionals alike?” The Palmdale Bulge gained $2 million of earthquake prediction research funding before two scientists at UCLA identified potential errors in the survey findings, and at a 1979 science meeting there was nearly a brawl as supporters and detractors of the bulge fought over the evidence.34 By the mid-1980s, all evidence for the Palmdale Bulge had evaporated.
The mid-1970s enthusiasm for California earthquake prediction was driven by news out of China and also by the idea that the United States was being left behind in earthquake prediction research. On the evening of February 4, 1975, a Magnitude 7.3 earthquake occurred 16 miles (25 kilometers) south of the industrial city of Haicheng in Manchuria, 380 miles (600 kilometers) northeast of Beijing. The Chinese authorities issued a press release claiming to have predicted the earthquake, saving tens of thousands of lives.35 This story played to some potent themes: the Chinese had superior older wisdom; the Communist authorities could look after their citizens more effectively than leaders in the West; and the Chinese would lead the way in this key new area of practical science.
There had been prominent foreshocks, changes in well levels, and ground movements in the lead-up to this earthquake. There had also been a particular intuition that the larger earthquake was about to arrive, leading to the evacuation of one city. However, the Chinese were not prepared to submit their evidence to full scientific scrutiny.
The Chinese pride in their earthquake prediction prowess would last only eighteen months; then a larger and unforecasted Magnitude 7.8 earthquake occurred directly beneath the city of Tangshan.36 In the following years, it became clear that whatever had facilitated the forecast at Haicheng was the exception and not the rule.
In 1977, when earthquake prediction enthusiasm was at its peak, Brian Brady, an expert on rock bursts in mines who worked with the geologist William Spence, based out of the US Geological Survey office in Colorado, made the bold prediction that a Magnitude 8.4 earthquake would hit Lima in late 1980. The shock would then be followed, he said, by an extraordinary Magnitude 9.2 earthquake in October or November 1981.37 Brady and Spence’s prediction was based on extrapolating the patterns seen among smaller earthquakes.
In late 1979, the news media in Peru got wind of the threatened catastrophe. In February 1980, the head of the Peruvian Red Cross appealed to the United States for 100,000 body bags. The letter was leaked to the press in Lima. As the nation seemed to march inexorably toward Armageddon, tourists cut back on their planned visits to Lima and the economy started to suffer. In January 1981, the Peruvian government asked the US National Earthquake Prediction Evaluation Council to vet the prediction. After reviewing the underlying science, the council reported that, although the possibility of an earthquake occurring on any given day could not be discounted, this particular prediction appeared to lack merit.
By April 1981, Brady had extended the window of the first great earthquake to June 28. The Peruvian authorities asked John Filson, head of the US Geological Survey’s earthquake studies program, to come to Peru. Filson arrived on June 25 and soon discovered that the prediction was daily front-page news. People were leaving town, property prices had fallen, and hotel bookings were down by two-thirds. June 28 was quiet, and Filson left on the 29th, but Brady had now extended the dates of his great earthquake forecast up to July 10. When that date passed peacefully, Brady finally retracted the prediction on July 20.
Filson saw the need for more sober and scientific prediction research and was a driving force behind the program to lay a net of recording instruments to catch an earthquake at Parkfield, a rural backwater in central California. Moderate-sized Magnitude 6 earthquakes on the San Andreas Fault at Parkfield appeared to have fired off like clockwork—on average every twenty-two years all the way back to the mid-nineteenth century. The previous earthquake had been in 1966, so the instrumental trap was set for 1988.38 The US Geological Survey seismologists gave themselves a little leeway by stating confidently that the Parkfield earthquake would strike within four years of 1988. The months passed and then the years; the prediction window came and went, the research funding drained away, and the equipment started to rust or was moved to other sites. The Parkfield prediction had become a scientific joke. The earthquake finally arrived in 2004, sixteen years late. The willful behavior of even the most regular and modest of earthquakes marked the end of significant scientific funding for California earthquake prediction.
The subject will never go away, but the likelihood that we will ever achieve reliable earthquake prediction seems lower now than it was forty years ago.39 Only Ōmori’s aftershocks (which can sometimes be bigger than the original main shock), situated around the main-shock source, with their consistent exponential decay in numbers, have proven amenable to forecasting. Many seismologists are actually quite pleased with this situation because any hint of earthquake prediction has led to a minefield of unintended consequences.
IT WAS AMATEUR EARTHQUAKE PREDICTION THAT WAS THE UNDOING of scientists at L’Aquila, Italy, in 2009.40 Giampaolo Giuliani had been a technician at the Gran Sasso physics laboratory for forty years and had built four homemade radon detectors that he placed in water springs around the region. Based on the radon levels, he predicted earthquakes.
From the start of 2009, a swarm of earthquakes occurred around the city of L’Aquila, one or two on average every day, and their numbers increased through March. Giuliani started issuing predictions of specific larger earthquakes, which captured the local headlines. Faced with a rattled and fearful population, a meeting was organized in the town on March 31 between civil defense officials and national seismologists, and then a televised press conference was held by the two officials, away from the scientists and their cautions. One of the officials made the mistake of saying that the seismic situation in L’Aquila was “certainly normal,” posing “no danger,” and that “the scientific community continues to assure me that, to the contrary, it’s a favorable situation because of the continuous discharge of energy.” No one at the press conference pointed out the fragility of the town’s buildings, both ancient and modern, or the fact that a swarm of tremors had preceded the last catastrophic earthquake in the city, in 1703.
Late in the evening on April 5, there was a strong shock, after which some people (mostly men) chose to spend the night outside in their cars, while others, remembering the calming official statements, stayed in their beds—to their cost. At 3:30 a.m., a Magnitude 6.3 earthquake occurred beneath the city, destroying many buildings and taking more than 300 lives. The survivors rued that they had not taken greater heed of the warnings and wished they had not believed the soothing blandishments of the officials. The scientists and civil defense officials who had attended the fateful meeting on March 31 were tried in a local court and all found guilty of manslaughter (a verdict overturned on appeal).41
PREDICTION INVOLVES FORECASTING THE SIZE, LOCATION, AND time of an earthquake. Take out the timing and we have the more general question of defining the shaking hazard. What constrains the maximum size of all potential earthquakes?
In the 1950s, Japan, with no indigenous sources of oil or gas, developed a craving for nuclear electricity. Rivers in Japan were short, their flows too intermittent and seasonal, and therefore reactor cooling would have to come from the sea. And then there was the problem of earthquakes. Most notorious was the southern coast of Honshu and Shikoku—already the site of three massive earthquakes and tsunamis in the twentieth century. In contrast, the coastline of eastern Honshu, north of Tokyo, appeared benign: there had been no large earthquakes there for at least three centuries. In the early 1960s, one of the first nuclear reactors in Japan was planned for this coastline in Fukushima Prefecture.
At the time, in July 1967, when the foundations were being dug for the site, “plate tectonics” had not even been named. By the time the reactor was operational in 1971, there was a whole new theory to explain earthquake origins.42 The oceanic trench off northeast Japan was a classic “subduction zone” plate boundary megafault dipping down beneath the coastline.43
Reactor units 2 and 3 were built in 1974 and 1976. Eventually there would be ten reactors in two sites at Fukushima and a total of fourteen along this coastline, making up almost 30 percent of all the reactors in Japan. Engineers believed that the largest earthquake along this coast would be below Magnitude 8.
Through the 1980s, seismologists developed a theory for where the largest (Magnitude 9) earthquakes could and could not occur worldwide. Giant earthquakes only happened, the theory proposed, where the two sides of the subduction zone megafault were tightly locked together, a condition that required the most rapid plate motions and geologically young ocean crust.44 The theory was taught to a whole generation of students.
For almost forty years there were no giant earthquakes. Then, on the morning of December 26, 2004, a fault started to break along a subduction zone to the west of northern Sumatra and continued to break north for ten minutes, by which time the rupture was 750 miles (1,200 kilometers) long and the earthquake was greater than Magnitude 9.45 The seafloor moved tens of meters and generated a monster tsunami that overwhelmed the Sumatran city of Banda Aceh and sped across the Indian Ocean to inundate the beaches of Thailand to the east and Sri Lanka to the west. Yet this great earthquake happened in a location that the theory said was not possible—the subduction zone was neither young enough nor fast enough to be “strongly coupled.”
That should have been the end of the theory. An alert should have immediately gone out to those living in all the world’s subduction zones—such as the one along the Pacific coast of Japan—who were relying on the theory to protect themselves from giant earthquakes and tsunamis. Meanwhile, evidence was being uncovered to suggest that there had already been such an earthquake.
In the late ninth century, the Imperial Court of Japan was consolidating control over northern Honshu.46 In the middle of the summer night of July 13 in the year AD 869, an official with the court reported a huge earthquake. As the people looked out over the ocean, a powerful glow lit up the night sky. Then, within half an hour, the ocean began to flood the land, filling the whole coastal plain of Sendai; it overwhelmed a castle at the town of Tagajo, ten miles to the east of Sendai, destroyed all the houses, and took an estimated 1,000 lives.
At the end of the 1990s, a team of geologists from Osaka City University, Tokyo University, and the local Tohoku University dug down and found traces of the destroyed town of Tagajo. By 2001, they had located the remains of many buildings of the eighth and ninth centuries, covered by sands typical of those left by a tsunami, which could be traced up to 3 miles (4.5 kilometers) inland.47 Beneath the AD 869 tsunami sand were two earlier tsunami sand layers deposited within the past 3,000 years.
In 2007 the researchers presented a paper at a meeting of the Japanese Society of Engineering Geology, and they also talked to the press about their findings.48 By digging trenches, they had found the same tsunami sand in marshes all along the coast, reaching several kilometers inland. The “Jogan Sanriku” tsunami, as they now called it, “may have been the strongest seismic disturbance ever to strike Japan,” possibly even reaching the exalted Magnitude 9.49 From the evidence of the earlier tsunamis, a repeat of such an earthquake was now “overdue.”50
In 2008, after a shock close to another Japanese nuclear power plant, the Japanese nuclear regulator initiated a review of earthquake and tsunami hazards at all of the seventeen reactor sites in Japan.51 For each site, the reviewers had to decide which was the most significant hazard. At Fukushima, they chose earthquake. Seven experts on earthquakes met twenty-two times to consider all the evidence. At the end of June 2009, a seismologist asked what would happen in a repeat of the AD 869 earthquake and mega-tsunami, but his question was rebuffed by an executive from the Tokyo Electric Power Company. At the next meeting, the earthquake and tsunami safety report for the Fukushima Daiichi nuclear power plants, declaring the safety features at the facility to be sufficient, was approved.52
And then, as if playing out some elemental tragedy—1,142 years after its previous incarnation but only four years after its scientific reconstruction—the colossal AD 869 earthquake and accompanying tsunami happened again. It was as if the final act of scientific reconstruction had made the whole event come back to life—the genie had come out of the bottle. This time it was not in the middle of a summer night but midafternoon on a cold late winter day, so the tsunami—50 feet (15 meters) high at Sendai, and higher to the north—was not just a black tide in the night but an event witnessed and filmed.
At the Fukushima Daiichi nuclear power complex, the vibrations caused water to slosh out of the fuel storage ponds on top of the buildings and to run in contaminated waterfalls down the stairs to the control room. Yet the tsunami was the more potent avenger. The wave came in at around 40 feet (12 meters), high above the 16-foot (5-meter) seawall protecting the site. Offsite the shaking had already broken the external power supply. The backup generators, located behind the walls in front of the plant, were knocked over by the tsunami wave, saturated in saltwater, and disconnected. (It was later admitted that, for the cost of $10,000, they could all have been located at a higher elevation.) Two reactor units were already shut down for maintenance, and a third was in cold shutdown, but reactors 1, 2, and 3 were all operating at full power. Without electricity to drive the pumps, the three reactors could not be cooled, and within a few hours the fuel rods overheated and a pool of molten fuel started to develop in the base of the reactors. Meanwhile, water decomposed to form hydrogen, which forced out seals in the reactor vessels, releasing radioactive clouds and setting off explosions.
While Fukushima was ground zero of this disaster, the tsunami walls were overwhelmed all along the coast. In some coastal towns even the evacuation destinations and disaster command posts were destroyed. At Rikuzentakata, the fire chief sent forty-five young firemen to close the tsunami gates. They all drowned.53
The 2011 Tohoku earthquake toppled a philosophy of overconfident mastery of the natural world. Science no longer has a theory to discriminate among the 35,000 miles (55,000 kilometers) of worldwide subduction zones. So where in the world will the next mega-tsunami happen? A safe bet is somewhere that has not been hit by such an earthquake for at least 200 years, and probably much longer. It could be in the eastern Caribbean, south of Turkey, or along the Cascadia coast in the northwest United States.
The safety of hundreds of thousands of future lives will depend on providing education and alerts about mega-tsunamis for coastlines with no recorded history of any such disaster. As in Japan, there may be strong resistance from developers and politicians to publicizing this message.
ONE FORM OF LIMITED EARTHQUAKE FORECASTING HAS FLOWERED in the new millennium. The initial vibrations are recorded as close as possible to the earthquake source, and then the signal is automatically read so that people farther away can be warned of the vibrations coming toward them at a speed of 2 to 4 kilometers per second.54 For a large offshore earthquake, seafloor recorders can give tens of seconds’ warning to shut down trains or dangerous equipment. The system installed in Japan after 2007 also rings all mobile phones with a sinister tone.55 Yet where the largest earthquakes happen on land, as in California, those people located close to the source will not have sufficient time to receive an effective warning.
ONCE UPON A TIME, ALL CATASTROPHES—STORMS, FLOODS, HURRICANES, earthquakes, volcanic eruptions—arrived without warning. That was before satellite observations, supercomputing, numerical weather prediction models, ensembles, Doppler radar, and Bayesian forecasting schemes. Today we have come to expect at least a week’s warning for an eruption, two or three days for a hurricane, twenty-four hours for its storm surge, at least twelve hours for (faster and harder-to-forecast) intense windstorms, six hours for flash floods, twenty minutes for a tsunami, and at least five minutes for a tornado.
But the deadly earthquake remains strangely, remarkably, almost admirably, resistant to all that forecasting science has thrown at it.