Seeking the Impossible--The Science of Earthquake Prediction

Then again, hindsight is always 20/20 and the officials were in a no-win situation. On the one hand, there is no scientific basis for claims that animal migrations augur the arrival of earthquakes. On the other, the Sichuan Earthquake Administration's repeated erroneous warnings about aftershocks following the quake put the public on edge and led to complaints that the agency should be renamed the Aftershock Administration.

Sichuan's experience shows just how hard it is to predict quakes. Nine years ago, Taiwan's inability to provide advance warning of the massive Chichi earthquake similarly led to huge casualties. More recently, scientists have begun seeking reliable indicators of imminent earthquake activity in an effort to reduce the huge financial and human costs of major quakes.

In early February 1975, frequent small earthquakes of magnitude 2-3 struck the Haicheng County-Yingkou County area of China's Liaoning Province. When seismologists suggested that a major quake was on the way, the provincial government began making preparations to minimize the damage. At 10 a.m. on 4 February, officials issued an emergency evacuation order.

The government's preparations succeeded in limiting the death toll from that evening's magnitude 7.3 quake to just 1,300 people.

According to the United Nations, the Haicheng quake was the first major temblor to be accurately predicted. Unfortunately, it remains the only one. China, then a technological laggard, was proud of its achievement. But the following year's Tangshan earthquake quickly laid to rest the myth that scientists had a firm grasp on quake forecasting. Unheralded by foreshocks, the magnitude 7.8 quake caught China's earthquake authorities napping and left them watching awestruck as it leveled Tangshan and killed more than 242,000 people.

In recent years, scientists have been studying changes in the electron density of the ionosphere as a possible earthquake precursor. In the May 2008 electron density graph at left, the red lines represent daily changes in the ionosphere. The image above shows the magnitude five or greater earthquakes from the same month. A comparison of the two reveals significant declines in the electron density above Taiwan one to five days before quakes.

Today's technology is not yet able to predict the location, time, and magnitude of earthquakes with anything like the accuracy with which we predict the weather. But the scientists have not given up. Historical data show that major earthquakes are usually heralded by a number of abnormal phenomena, including foreshocks like those seen at Haicheng or abnormally low numbers of small quakes, as seen in northeastern Taiwan over the three to four years preceding the Chichi earthquake. In the case of the Chichi quake, northeastern Taiwan experienced only one quake of magnitude 4 or less in June 1998 versus 18 in June 2008. This suggests that if the Earth's crust in a given area moves a great deal more or a great deal less than usual, a strong quake may be on the way.

Large-scale animal migrations and unusual animal activities are also potential indicators. Just before the 12 May Sichuan temblor, hundreds of thousands of toads poured out into the roads of Sichuan's Tanmu Village, less than 100 kilometers from the quake's epicenter. Unfortunately, the local forestry bureau blithely shrugged off the occurrence, stating that the local environment was particularly well suited to toads. Still other quakes have been preceded by explosive rises or dramatic declines in the water table, or by changes in the density of free electrons in the ionosphere.

All of these possible earthquake precursors are being studied, but scientists are quick to admit that their greatest challenge is determining whether a causal relationship exists between these potential precursors and major quakes. That is, while it is certainly possible for water tables to rise or fall, for the composition of the atmosphere at the Earth's surface to change, or for the electron density of the ionosphere to drop before an earthquake, these phenomena can also be triggered by a number of other manmade and natural events, including rainfall, water pressure, over-extraction of groundwater, and geomagnetic storms.

In late May 2005, Taiwan's Central Weather Bureau (CWB) found itself in a very awkward position after acting on one of these "precursors." A researcher testing wells in the Liuchia area of Tainan County observed a surprising 88-centimeter drop in the local water table. Reacting to the report, the CWB passed along a "friendly reminder" about the possibility of an earthquake to the disaster preparedness units of the central and local governments. When the news leaked to the public, some schools were closed and residents of Chiayi and Tainan Counties panicked. No quake materialized, and it turned out that pumping by Taiwan Sugar had caused the drop in the water table. The CWB was roundly criticized for the situation, and suffered a heavy blow to its reputation.

Today's technology is not yet able to predict the location, time, and magnitude of earthquakes with anything like the accuracy with which we predict the weather. But the scientists have not given up. Historical data show that major earthquakes are usually heralded by a number of abnormal phenomena, including foreshocks like those seen at Haicheng or abnormally low numbers of small quakes, as seen in northeastern Taiwan over the three to four years preceding the Chichi earthquake. In the case of the Chichi quake, northeastern Taiwan experienced only one quake of magnitude 4 or less in June 1998 versus 18 in June 2008. This suggests that if the Earth's crust in a given area moves a great deal more or a great deal less than usual, a strong quake may be on the way.

Large-scale animal migrations and unusual animal activities are also potential indicators. Just before the 12 May Sichuan temblor, hundreds of thousands of toads poured out into the roads of Sichuan's Tanmu Village, less than 100 kilometers from the quake's epicenter. Unfortunately, the local forestry bureau blithely shrugged off the occurrence, stating that the local environment was particularly well suited to toads. Still other quakes have been preceded by explosive rises or dramatic declines in the water table, or by changes in the density of free electrons in the ionosphere.

All of these possible earthquake precursors are being studied, but scientists are quick to admit that their greatest challenge is determining whether a causal relationship exists between these potential precursors and major quakes. That is, while it is certainly possible for water tables to rise or fall, for the composition of the atmosphere at the Earth's surface to change, or for the electron density of the ionosphere to drop before an earthquake, these phenomena can also be triggered by a number of other manmade and natural events, including rainfall, water pressure, over-extraction of groundwater, and geomagnetic storms.

In late May 2005, Taiwan's Central Weather Bureau (CWB) found itself in a very awkward position after acting on one of these "precursors." A researcher testing wells in the Liuchia area of Tainan County observed a surprising 88-centimeter drop in the local water table. Reacting to the report, the CWB passed along a "friendly reminder" about the possibility of an earthquake to the disaster preparedness units of the central and local governments. When the news leaked to the public, some schools were closed and residents of Chiayi and Tainan Counties panicked. No quake materialized, and it turned out that pumping by Taiwan Sugar had caused the drop in the water table. The CWB was roundly criticized for the situation, and suffered a heavy blow to its reputation.

In recent years, scientists have been studying changes in the electron density of the ionosphere as a possible earthquake precursor. In the May 2008 electron density graph at left, the red lines represent daily changes in the ionosphere. The image above shows the magnitude five or greater earthquakes from the same month. A comparison of the two reveals significant declines in the electron density above Taiwan one to five days before quakes.

Though research into earthquake precursors is still a long way from producing reliable predictions, there have been some tantalizing preliminary results.

Over the six days preceding the Sichuan quake, Taiwan's Formosa-3 satellite observed a noticeable decline in the electron density of the ionosphere above Sichuan. And America's National Aeronautics and Space Administration (NASA) reported observing unusual infrared radiation above the quake's epicenter before the quake occurred. NASA planned to follow up on its discovery by working with scientists in various countries to study the feasibility of constructing a comprehensive earthquake warning system based on atmospheric observations.

Liu Jann-yeng, a professor in the Graduate Institute of Astronomy at National Central University, made a similar discovery of his own-on three of the four days immediately preceding the Chichi earthquake, the electron density of the ionosphere above Taiwan dropped significantly below the median for the preceding 15 days. When Liu went on to look at the atmospheric data on 184 Taiwanese earthquakes of magnitude 5 or greater, he found that significant declines occurred one to five days prior to 90% of the quakes with hypocenters at depths of 35 km or less.

The ionosphere is the portion of the atmosphere 50-2,000 km above the Earth's surface. X-ray, ultraviolet, and far ultraviolet radiation from the Sun ionize molecules within it, creating a plasma of free electrons and positively charged ions.

Liu says that there are a number of hypotheses about why earthquakes cause a drop in electron density. One theory suggests that changes in stress resulting from the relative movement of the crustal plates prior to an earthquake generate electric current, like a battery. This creates a positive charge (electrical "holes") that flows to the surface, creating an electrical field that stretches up into the sky. The field attracts free electrons, reducing their density in the ionosphere.

Another hypothesis holds that rock fractures and pressures between rocks generate electric currents that give rise to powerful electric and magnetic fields and large electromagnetic waves. When these waves reach the ionosphere, they cause rapid heating and expansion of the ionospheric plasma, reducing the density of particles in the ionosphere.

The CWB and National Central University are now making daily observations of the ionosphere above Taiwan from latitudes 21 to 25 North in the hope that changes in the ionosphere will let them know when an earthquake is imminent.

Kuo Kai-wen, head of the CWB's Seismological Center, admits that this research suffers from numerous limitations. The principal one is that after you exclude geomagnetic storms, violent thunderstorms, and typhoons, we don't know why most of the remainder of the sudden declines in the electron density of the ionosphere occur.

"A scientist who fired projectiles into the bedrock at medium to low speeds discovered that doing so caused fluctuations in the electron density of the ionosphere above the spot," says Kuo. "This suggests that these electron densities are very sensitive to change."

The fact that major quakes can occur anywhere within very large ranges-including the entire length of faults or even whole tectonic plates-is another problem. For example, the 2008 Sichuan quake could have arisen anywhere within a 2900-km-long fault zone. Consequently, the areas above which the ionosphere might change are also very large. Even if we were to measure a sharp decline in the electron density over Taiwan, with present-day technology, we wouldn't be able to tell whether the coming quake was going to be centered in Taipei, Pingtung, or Hualien.

"If we were to observe electron densities dropping to significantly less than their 15-day median all we could say was there was a 50% likelihood of a magnitude 5 or greater quake somewhere in Taiwan over the next five days," says Kuo. "At this stage, the information is so imprecise and has such a margin of error that it wouldn't make any sense to issue a public warning. The only thing we could do would be to remind our Seismological Center colleagues to be extra vigilant."

Though research into earthquake precursors is still a long way from producing reliable predictions, there have been some tantalizing preliminary results.

Over the six days preceding the Sichuan quake, Taiwan's Formosa-3 satellite observed a noticeable decline in the electron density of the ionosphere above Sichuan. And America's National Aeronautics and Space Administration (NASA) reported observing unusual infrared radiation above the quake's epicenter before the quake occurred. NASA planned to follow up on its discovery by working with scientists in various countries to study the feasibility of constructing a comprehensive earthquake warning system based on atmospheric observations.

Liu Jann-yeng, a professor in the Graduate Institute of Astronomy at National Central University, made a similar discovery of his own-on three of the four days immediately preceding the Chichi earthquake, the electron density of the ionosphere above Taiwan dropped significantly below the median for the preceding 15 days. When Liu went on to look at the atmospheric data on 184 Taiwanese earthquakes of magnitude 5 or greater, he found that significant declines occurred one to five days prior to 90% of the quakes with hypocenters at depths of 35 km or less.

The ionosphere is the portion of the atmosphere 50-2,000 km above the Earth's surface. X-ray, ultraviolet, and far ultraviolet radiation from the Sun ionize molecules within it, creating a plasma of free electrons and positively charged ions.

Liu says that there are a number of hypotheses about why earthquakes cause a drop in electron density. One theory suggests that changes in stress resulting from the relative movement of the crustal plates prior to an earthquake generate electric current, like a battery. This creates a positive charge (electrical "holes") that flows to the surface, creating an electrical field that stretches up into the sky. The field attracts free electrons, reducing their density in the ionosphere.

Another hypothesis holds that rock fractures and pressures between rocks generate electric currents that give rise to powerful electric and magnetic fields and large electromagnetic waves. When these waves reach the ionosphere, they cause rapid heating and expansion of the ionospheric plasma, reducing the density of particles in the ionosphere.

The CWB and National Central University are now making daily observations of the ionosphere above Taiwan from latitudes 21 to 25 North in the hope that changes in the ionosphere will let them know when an earthquake is imminent.

Kuo Kai-wen, head of the CWB's Seismological Center, admits that this research suffers from numerous limitations. The principal one is that after you exclude geomagnetic storms, violent thunderstorms, and typhoons, we don't know why most of the remainder of the sudden declines in the electron density of the ionosphere occur.

"A scientist who fired projectiles into the bedrock at medium to low speeds discovered that doing so caused fluctuations in the electron density of the ionosphere above the spot," says Kuo. "This suggests that these electron densities are very sensitive to change."

The fact that major quakes can occur anywhere within very large ranges-including the entire length of faults or even whole tectonic plates-is another problem. For example, the 2008 Sichuan quake could have arisen anywhere within a 2900-km-long fault zone. Consequently, the areas above which the ionosphere might change are also very large. Even if we were to measure a sharp decline in the electron density over Taiwan, with present-day technology, we wouldn't be able to tell whether the coming quake was going to be centered in Taipei, Pingtung, or Hualien.

"If we were to observe electron densities dropping to significantly less than their 15-day median all we could say was there was a 50% likelihood of a magnitude 5 or greater quake somewhere in Taiwan over the next five days," says Kuo. "At this stage, the information is so imprecise and has such a margin of error that it wouldn't make any sense to issue a public warning. The only thing we could do would be to remind our Seismological Center colleagues to be extra vigilant."

Rapid changes in groundwater levels are also receiving attention as a potential earthquake precursor.

Chia Yee-ping, a professor with the Department of Geosciences at National Taiwan University, says that most earthquakes are caused by the movement of tectonic plates compressing or dilating faults. When a fault slips or ruptures, the stresses on a given area change, altering local water pressure and, consequently, groundwater levels. These so-called "coseismic water-level changes" can be widespread. For example, Taiwan saw fluctuations in some of its groundwater levels after the Sichuan quake even though the wells in question were located several thousand kilometers from the quake's epicenter.

Scientists believe that the rapid accumulation of stresses in the area around a fault prior to a quake can also cause fluctuations in groundwater levels.

"But where is this rapid accumulation of pre-quake stresses taking place?" ask Chia. "No one dares guess because the Earth's structure isn't uniform. The groundwater level fluctuations might be occurring near the future epicenter, but they also might be occurring quite far away in an area that has no obvious relation to it. This is the main challenge to water-level-based quake prediction."

Chia explains that heavy rains, tides, water extraction, and even engineering works all can affect groundwater levels. After accounting for human and natural causes, data gathered at more than 600 measuring stations set up around Taiwan since 1997 by the Water Resources Agency and the Central Geological Survey recorded only two abnormal pre-quake fluctuations in water levels.

In the first instance, the monitoring station at Chushan Township's Sheliao Elementary, located about 1.5 kilometers from the Chelungpu fault in Nantou County, recorded a rapid 4-cm drop in the water table just three hours before the Chichi quake. Then at the time of the quake the station recorded a coseismic water-level rise of 14 cm before water levels resumed their fall. The other instance was related to the magnitude 6.7 quake that occurred near Jen-ai Township, Nantou County, on 11 June 2000. One hour before that quake, the Pingting No. 1 monitoring station in Linnei Township, Yunlin County, recorded a 5-cm drop in water levels. This was followed by a coseismic water-level rise of 22 cm.

Chia says that though both stations showed unusual water-level declines followed by coseismic increases, they were located at very different distances from the epicenters of the respective quakes, demonstrating just how much hydrological and geological conditions, and the resulting water-level fluctuations, can vary with location.

Even if a monitoring station observes an abnormal change in water level, it is very difficult to determine if the coming quake will be centered in the station's vicinity. For example, the Pingting No. 1 station mentioned above is situated near several faults, yet these faults remained quiet, while the quake that did occur was actually centered tens of kilometers away at Jen-ai. "We need longer-term data, and we need a clearer understanding of the links between water levels and hydrological- and geological conditions," says Chia.

Rapid changes in groundwater levels are also receiving attention as a potential earthquake precursor.

Chia Yee-ping, a professor with the Department of Geosciences at National Taiwan University, says that most earthquakes are caused by the movement of tectonic plates compressing or dilating faults. When a fault slips or ruptures, the stresses on a given area change, altering local water pressure and, consequently, groundwater levels. These so-called "coseismic water-level changes" can be widespread. For example, Taiwan saw fluctuations in some of its groundwater levels after the Sichuan quake even though the wells in question were located several thousand kilometers from the quake's epicenter.

Scientists believe that the rapid accumulation of stresses in the area around a fault prior to a quake can also cause fluctuations in groundwater levels.

"But where is this rapid accumulation of pre-quake stresses taking place?" ask Chia. "No one dares guess because the Earth's structure isn't uniform. The groundwater level fluctuations might be occurring near the future epicenter, but they also might be occurring quite far away in an area that has no obvious relation to it. This is the main challenge to water-level-based quake prediction."

Chia explains that heavy rains, tides, water extraction, and even engineering works all can affect groundwater levels. After accounting for human and natural causes, data gathered at more than 600 measuring stations set up around Taiwan since 1997 by the Water Resources Agency and the Central Geological Survey recorded only two abnormal pre-quake fluctuations in water levels.

In the first instance, the monitoring station at Chushan Township's Sheliao Elementary, located about 1.5 kilometers from the Chelungpu fault in Nantou County, recorded a rapid 4-cm drop in the water table just three hours before the Chichi quake. Then at the time of the quake the station recorded a coseismic water-level rise of 14 cm before water levels resumed their fall. The other instance was related to the magnitude 6.7 quake that occurred near Jen-ai Township, Nantou County, on 11 June 2000. One hour before that quake, the Pingting No. 1 monitoring station in Linnei Township, Yunlin County, recorded a 5-cm drop in water levels. This was followed by a coseismic water-level rise of 22 cm.

Chia says that though both stations showed unusual water-level declines followed by coseismic increases, they were located at very different distances from the epicenters of the respective quakes, demonstrating just how much hydrological and geological conditions, and the resulting water-level fluctuations, can vary with location.

Even if a monitoring station observes an abnormal change in water level, it is very difficult to determine if the coming quake will be centered in the station's vicinity. For example, the Pingting No. 1 station mentioned above is situated near several faults, yet these faults remained quiet, while the quake that did occur was actually centered tens of kilometers away at Jen-ai. "We need longer-term data, and we need a clearer understanding of the links between water levels and hydrological- and geological conditions," says Chia.

Chemical monitoring of gases in the crust and at the surface near active faults is another means of keeping abreast of earthquake activity.

Frank Yang, the NTU geochemist leading Taiwan's research into the phenomenon, says that the gases residing deep in the Earth's crust-primarily carbon dioxide, methane, radon, and helium-are quite different from the nitrogen and oxygen that predominate at its surface. But faults or fractures in the crust sometimes allow gases from beneath the earth to bleed into the atmosphere.

Yang says that his group's approach is to seek fault activity by monitoring changes in the atmosphere at the Earth's surface. To this end, they have established monitoring stations at sensitive locations above faults. "The accumulated stresses in the crust begin changing before a quake strikes, providing radon and other gases from deep beneath the earth with an opportunity to vent upwards along the fault."

NTU's Department of Geosciences and the Central Geological Survey have set up four gas monitoring stations, concentrated in western Taiwan. Their Hsinchu Tapingti Station, located near the Hsincheng Fault, has so far shown the greatest promise. In the 2006-7 period, shallow quakes of magnitude 4 or greater usually struck northern Taiwan one to seven days after the station recorded abnormally high radon levels. But the method has limitations: no radon level changes have been recorded for quakes centered more than 100 km from the monitoring stations or at depths greater than 70-80 km.

And though studies into radon changes have already yielded good preliminary results, the method suffers from the same problems as the other approaches-there's still no way to precisely predict the quake's epicenter, magnitude, intensity, and potential for damage.

Right now, these various approaches to quake prediction have advanced to about the same point: measuring changes in electron densities, water levels, and atmospheric composition, and calculating the frequency of quake activity can all indicate the possibility of a quake, but none can predict one with certainty.

Being able to anticipate major earthquakes and ameliorate the terrible suffering they bring would be an enormous boon. Consequently, earthquake prediction remains an important area of research in spite of the myriad unknowns that continue to plague the field. If we were to establish a national earthquake warning bureau to collate all of these studies and undertake long-term analysis of their results, earthquake forecasting might turn out to be not so impossible after all.

Chemical monitoring of gases in the crust and at the surface near active faults is another means of keeping abreast of earthquake activity.

Frank Yang, the NTU geochemist leading Taiwan's research into the phenomenon, says that the gases residing deep in the Earth's crust-primarily carbon dioxide, methane, radon, and helium-are quite different from the nitrogen and oxygen that predominate at its surface. But faults or fractures in the crust sometimes allow gases from beneath the earth to bleed into the atmosphere.

Yang says that his group's approach is to seek fault activity by monitoring changes in the atmosphere at the Earth's surface. To this end, they have established monitoring stations at sensitive locations above faults. "The accumulated stresses in the crust begin changing before a quake strikes, providing radon and other gases from deep beneath the earth with an opportunity to vent upwards along the fault."

NTU's Department of Geosciences and the Central Geological Survey have set up four gas monitoring stations, concentrated in western Taiwan. Their Hsinchu Tapingti Station, located near the Hsincheng Fault, has so far shown the greatest promise. In the 2006-7 period, shallow quakes of magnitude 4 or greater usually struck northern Taiwan one to seven days after the station recorded abnormally high radon levels. But the method has limitations: no radon level changes have been recorded for quakes centered more than 100 km from the monitoring stations or at depths greater than 70-80 km.

And though studies into radon changes have already yielded good preliminary results, the method suffers from the same problems as the other approaches-there's still no way to precisely predict the quake's epicenter, magnitude, intensity, and potential for damage.

Right now, these various approaches to quake prediction have advanced to about the same point: measuring changes in electron densities, water levels, and atmospheric composition, and calculating the frequency of quake activity can all indicate the possibility of a quake, but none can predict one with certainty.

Being able to anticipate major earthquakes and ameliorate the terrible suffering they bring would be an enormous boon. Consequently, earthquake prediction remains an important area of research in spite of the myriad unknowns that continue to plague the field. If we were to establish a national earthquake warning bureau to collate all of these studies and undertake long-term analysis of their results, earthquake forecasting might turn out to be not so impossible after all.