Since the beginning of man, there has been a desire to predict the
future. The advent and evolution of science and technology has made
predictions in many areas possible. In fact, the successful scientific
predictions in areas such as meteorology or astronomy have heightened
the expectations of the public. Physicists and geologists are under
tremendous pressure to predict the time, place, and strength of one of
the earth’s most devastating phenomena: earthquakes.
In the past, people have tried to predict earthquakes in any way possible like watching the behavior of animals or watching for a certain kind of weather, but now there are more scientifically and mathematically sound methods. This paper will examine four of those methods in brief: the seismic gap theory, fault modeling, electrical measurements, and the dilatancy-fluid diffusion model. However, there are inherent problems in predicting earthquakes with the methods available, though the social benefits would be great.
Why should scientists try to predict earthquakes?
There are, of course, the rather obvious and egoistic reasons
scientists might try to predict earthquakes, such as fame and wealth,
as well as the satisfaction from accomplishing the seeming impossible.
Aside from these, there are more benevolent reasons.
If earthquakes could be predicted in the long term (a warning period of several months), the most prominent results would be a reduction in property damage and the number of deaths. Bolt suggests a downside to long term prediction in that the communities in the area might “suffer social disruption and economic decline” for the simple reason that investment would drop and people would wish to relocate for safety (1993 , p. 204).
If short term warnings (days to hours) were the only possibility, this would still allow for life (and money) saving preparations. Gas lines could be shut off before they were damaged. People wouldn’t be caught in dangerous places like unstable buildings and bridges. Emergency services could be put on alert. Still, even this has problems from “the rescheduling of public events, the cessation of work activity, the closing of schools” et cetera (Bolt, 1993, p. 204).
The benefits of any kind of warning of an imminent earthquake can be summed up by simply looking at the number of lives saved in both the events during the earthquake and the events afterward.
Methods and Theories for the Prediction of Earthquakes
Seismic Gap Theory
The seismic gap theory provides long term predictions based on the conditional probability of another earthquake based on past ones (Aki, 1995). The basic theory is this: each segment of a fault builds up stress over time. During this build up, there is a period of seismic inactivity. Suddenly, the earth gives way and releases the strain in an earthquake. Afterwards, the process begins again (Stix, 1992, p. 48).
Immediately following an earthquake, the probability of another is very small. However, by statistically examining real earthquake data, there is a tendency to cluster earthquake occurrences (Aki, 1995). These trends are not necessarily in conflict because “one is for the whole catalog of earthquakes and the other is for . . . characteristic earthquakes,” though assuming the existence of characteristic earthquakes may be wishful thinking (Aki, 1995).
A basic problem of applying this theory to any one fault segment is the lack of data “from which the recurrence interval distribution and hence, the forecast probability can be estimated” (Nishenko, 1987, p. 1382).
An attempt to apply this theory to seismic zones in the Pacific rim “failed to forecast the location of 40 large earthquakes” in a decade beginning in 1979 (Stix, 1992, p. 48).
It may be that fault rupturing is influenced by “chaotic phenomena” that throw off the straightforward model of repetitive strain build-up and release that this theory proposes. If the system is chaotic, then large stresses that might suggest an earthquake to scientists may not trigger an earthquake at all or small stresses and events that scientists would not worry about may set off the big one. In a chaotic system, intervals between earthquakes could be short or long and have no meaning for prediction (Stix, 1992, p. 52).
The fault system near Parkfield, California, had produced earthquakes (around magnitude 6) almost every 22 years. This system should have been ideal for modeling. Scientists issued a prediction in 1984 which was endorsed by the NEPEC (National Earthquake Prediction Evaluation Council). To date (April 1998), this long awaited earthquake has not yet occurred. Some scientists believe that “the earth’s crust is such a remarkably complex system that chaos overwhelms predictability.” One major assumption was made about the system, and that was that a fault segment is an isolated system. After the 6.5 magnitude Coalinga earthquake in 1983, scientists calculated that this earthquake may have relieved some of the building tension in the area which would delay the forecasted earthquake (Kerr, 1993, p. 1120). Because of external forces on an individual system, scientists have to compensate by acknowledging a more complex system.
A study was done by Pepke, Carlson, and Shaw that looked at the use of algorithms in predicting earthquakes (1994). These algorithms should provide a way to objectively assess “the probabilities of large earthquakes based on a collection of precursor functions. . .” (Pepke et al, 1994, p. 6770). The first was algorithm CN which used “nine characteristics of the earthquake sequence in a region” with certain characteristic combinations being typical and others atypical of the time of increased probability (Keilis-Borok et al, 1990, p. 1461). The second was algorithm M8 which used a “set of uniform, overlapping areas, which are either circles or squares” and is used to cover the region being studied (Keilis-Borok et al, 1990, p. 1462). These algorithms account for the more complex system by including several possible precursors.
Even using the algorithms for only one precursor, they were still more effective than other long-term prediction methods. However, it is important to keep in mind that these algorithms were used on an ideal model which is systematic compared to the dynamic earth (Pepke et al, 1994, p. 6771).
A technique named VAN (after the last names of the developers: Varotsos, Alexopoulos, and Nomicos) which involves taking readings of precursory electrical signals from electrodes. Varotsos says that he can predict quakes by weeks. While it is true that the earth can transmit small electrical signals over long distances, other researchers believe that this technique has no predictive ability (Schneider, 1998, p. 23).
In 1995, Varotsos predicted the Kozáni, Greece, earthquake. However, the earthquake occurred to the north and east of his prediction and was also the wrong size. Varotsos said “This is purely a misunderstanding” because the location of the epicenter had previously been aseismic (Schneider, 1998, p. 24). Sylvie Gruszow tried to duplicate Varotsos’s results. After finding a signal that resembled Varotsos’s, she and her colleagues came to the conclusion that the signals were man-made (Schneider, 1998, p. 24).
In a press release by the American Geophysical Union, four of the criticisms that are directed at the VAN technique were listed (1996). They are: (1) the success rate is no better than that of chance, (2) the hypothesis that the theory is based on has changed several times making a statistical evaluation of the method pointless, (3) some of the claims of success have been based on “misrepresentation of the facts”, and (4) the actual predictions are vague and “should not be considered ‘earthquake predictions’ in the first place” (AGU, 1996).
Dilatancy-Fluid Diffusion Model
This method is based on precursory changes in seismic waves. Because p-wave velocities are sensitive to changes in fluid saturation, the ratio of Vp to Vs changes as the stresses on the rocks change (Pakes, 1995). As the stress increases, dilatancy (“an inelastic volume increase caused by the formation of microcracks within the rock”) may increase the “pore volume” in the rock. When this begins, the rock will be undersaturated, and the wave velocity ratio will change. An estimate of seismic probability can be made from this change. This model has been used world-wide, but it is important to remember that “seismic probability is not to be misconstrued with earthquake prediction” (Pakes, 1995).
Problems Hindering Earthquake Prediction
There are several inherent problems with predicting random events like earthquakes:
- detailed and reliable records of seismic activity are too short (timewise) compared to earthquake frequency
- differences in fault geometry and dynamics makes defining earthquake precursors difficult
- there are still things about the mechanics of earthquakes, like how they start, that we do not completely understand
- “knowledge of the strain distribution and yield points along faults is insufficient to indicate the locations of future epicenters”
-Pepke et al, 1994, p. 6769
These factors make earthquake prediction difficult because they determine how predictable each fault system is. Each fault system may be completely unique in its precursors and wave propagation due to the surrounding geology (rock types, folding, and other faulting). There may not even be any precursors for a given area which might make research in that subject useless.
Criticisms of these and other methods of earthquake prediction exist in abundance. The main arguments of those in opposition are as follows. For earthquakes to be predicted, the events surrounding them would have to be unusual and be the results of very specific circumstances. The systems of the earth are inherently chaotic. Prediction is limited by unknown conditions and the inability to know quantitative details. There is no exhaustive catalog of precursory phenomena, and precursors are usually interpreted after the fact (Geller et al, 1997, p. 1616). I believe that Charles Richter said it best when he said “Only fools, liars, and charlatans predict earthquakes” (Stix, 1992, p. 52).
Aki, Keitti. “Earthquake prediction, societal implications.” Rev. Geophys. Vol. 33 Suppl., 1995. (http://earth.agu.org/revgeophys/aki00/aki00.html).
Geophysical Union. “Earthquake Prediction Method Topic of Special
Issue of Geophysical Research Letters.” Press Release. 23 May 1996.
Bolt, Bruce A. Earthquakes and Geological Discovery. New York: W. H. Freeman and Company, 1993.
Geller, Robert J., et al. “Earthquakes Cannot Be Predicted.” Science. 14 March 1997: 1616 - 1617.
Keilis-Borok, V. I., et al. “Intermediate-term Prediction in Advance on the Loma Prieta Earthquake.” Geophysical Research Letters. Aug. 1990: 1461 - 1464.
Kerr, Richard A. “Parkfield Quakes Skip a Beat.” Science. 19 Feb. 1993: 1120 - 1122.
Nishenko, S. P., and R. Buland. “A Generic Recurrence Interval Distribution for Earthquake Forecasting.” Bulletin of the Seismological Society of America. Aug. 1987: 1382 - 1399.
Pakes, Angela M. “The Dilatancy-Fluid Diffusion Model for Predicting Earthquakes.” 13 Aug. 1995 (http://www.engr.wisc.edu/epd/uer/angelap/briefand.html).
Pepke, S. L., J. M. Carlson, and B. E. Shaw. “Prediction of Large Events on a Dynamical Model of a Fault.” Journal of Geophysical Research. 10 Apr. 1994: 6769 - 6788.
Schneider, David. “On Shaky Ground.” Scientific American. Apr. 1998: 23 - 24.
Stix, Gary. “Finding Fault.” Scientific American. Dec. 1992: 48 - 52.