|
| |
Can Earthquakes Be Predicted?
From: Cambridge University Press
| By:
Peter M. Shearer |
EDITOR'S INTRODUCTION |
In the dark of night or in the heat of rush-hour traffic, earthquakes happen suddenly and without warning. Why is it that, for all our scientific sophistication, we do not seem to be able to second-guess the movements of the earth? Peter M. Shearer, in this extract from his Introduction to Seismology, reviews the theories and suggests that our best option for now may be to adapt our lifestyles to the uncertainty of nature. |
 espite their usefulness as a research tool for illuminating Earth structure, earthquakes are generally considered harmful because of their potential for causing death and destruction. It is therefore unfortunate that the most useful thing that seismologists could do--predict earthquakes--is what they are least able to do. Although many ideas for earthquake prediction have been explored, the sad truth is that reliable prediction of earthquakes is not currently possible on any time scale.  |
Searching for precursors
Short- to intermediate-term prediction (minutes to months) has proven especially problematic. Here the focus has been to search for anomalous behavior that can be observed prior to earthquakes that would provide warning that an event was imminent. Despite extensive searches for possible precursors, extremely few reliable examples have been found. This is not to say that claims for evidence of precursory phenomena have not been made. In the history of seismology such claims have occurred many times, often receiving great attention, only to be later discredited upon more careful and comprehensive study. |
One of the most famous examples occurred in the early 1970s, when several studies seemed to observe large (10 to 20%) changes in seismic velocity before earthquakes (e.g. A. M. Semenov, Izv. Acad. Sci. USSR Phys. Solid Earth 4 (1969), pp. 245-48; Y. P. Aggarwal, Nature 241 (1973), pp. 101-4; Y. P. Aggarwal et al., Journal of Geophysical Research 80 (1975), pp. 718-32; J. H. Whitcomb et al., Science 180 (1973), pp. 632-35; R. Robinson et al., Science 184 (1974), pp. 1281-283). Such observations appeared to have a physical basis in laboratory studies of rock samples, which showed that when rocks are compressed until they fracture, a phenomenon termed dilatancy often occurs for a short time interval immediately before failure. Dilatancy is caused by microcracks forming in the sample, resulting in a slight volume increase and a change in the bulk seismic velocities. These results formed the basis of the dilatancy theory of earthquake prediction which briefly was the focus of great excitement (see A. Nur, Bulletin of the Seismological Society of America 62 (1972), pp. 1217-222; C. Scholz et al., Science 181 (1973), pp. 803-10; D. Anderson and J. Whitcomb, Journal of Geophysical Research 80 (1975), pp. 1497-503). |
However, it soon became apparent that accurate measurements of changes in seismic velocity are difficult using naturally occurring events. Much greater precision can be achieved using artificial sources with repeatable locations and waveforms. Studies of records from quarry blasts and nuclear explosions found no evidence for velocity changes before earthquakes down to levels of 1 to 2%, providing limits that are an order of magnitude smaller than the changes that were claimed in the earlier studies (e.g. T. V. McEvilly and L. R. Johnson, Bulletin of the Seismological Society of America 64 (1974), pp. 343-53; D. M. Boore et al., ibid. 65 (1975), pp. 1407-18; H. Kanamori and G. Fuis, ibid. 66 (1976), pp. 2027-37; B. Bolt, ibid. 67 (1977), pp. 27-32; C. W. Chou and R. S. Crosson, Geophysical Research Letters 5 (1978), pp. 97-100). In addition to these observational constraints, it is also now clear that the average stress level on faults is surprisingly low, much lower than that used in laboratory experiments to fracture unbroken rocks. This implies that dilatancy, if it occurs before earthquakes, is likely to be confined to small areas of high stress concentration and not spread over significant volumes of rock where it might more readily be observed. |
Other possible precursors that have received varying degrees of attention over the past few decades are changes in seismicity patterns, variations in the rate of radon gas emissions, and electromagnetic anomalies. Tantalizing suggestions of precursory behavior have often been seen for individual events, but more comprehensive studies have not been able to establish clear evidence for a link to the earthquakes. The history of these studies shows a familiar, if depressing, pattern. An apparent precursor will receive publicity as a possible method to predict earthquakes. Only rarely is there a clearly defined physical mechanism that might be causing the precursory behavior, and so the debate centers on the character of the observations and if the anomalies can indeed be correlated with earthquake occurrence. A number of papers will appear, some supporting the method and others challenging it. Typically the arguments then become embroiled in statistical arguments regarding the significance of the result and the exact way in which the method should be applied. These exchanges eventually become so technical as to be of little interest to anyone outside of the groups involved. The end result is that the proposed method is not completely discredited but sufficiently clouded that most researchers move on to other things. |
Foreshocks
The only definitively established earthquake precursor is the occasional occurrence of foreshocks, events close in time and space to a subsequent mainshock. These occur too often to be attributed to random chance; they must be related in some way to the larger event (just as the aftershocks that follow the mainshock are not randomly occurring). The existence of foreshocks made possible the most important earthquake prediction of recent times--the Chinese order to evacuate the city of Haicheng prior to the Ms = 7.4 earthquake of February 4, 1975. A series of small events occurred immediately prior to the mainshock, and, when the earthquake struck, most of the population had left their homes and very few lives were lost. At the time this was touted as a great achievement for the Chinese earthquake prediction program, but the outcome owes its success mostly to the existence of the foreshock swarm and other precursory anomalies. |
Unfortunately, most large earthquakes are not signaled by easily recognized foreshock sequences. In some cases, there are no foreshocks, while in others the foreshocks are small in magnitude and not easily distinguished from the many naturally occurring clusters of events that do not lead to larger earthquakes. On July 27, 1976, a Ms = 7.8 earthquake struck the Chinese city of Tangshan, only 200 km away from Haicheng. No foreshocks preceded this event and the population received no warning. The death toll was the greatest of any earthquake in modern times; the official count is 255,000 people killed, with unofficial estimates going much higher. |
Are earthquakes unpredictable?
In California, no clearly recognizable precursor has been observed prior to any of the large earthquakes in the past few decades, despite the widespread deployment of seismometers and other instrumentation. Why should this be so? Why should such incredibly powerful events as major earthquakes apparently have no detectable precursors? |
One possible explanation was described by J. N. Brune (Journal of Geophysical Research 84 (1979), pp. 2195-98), who proposed that earthquakes may be inherently unpredictable since large earthquakes start as smaller earthquakes, which in turn start as smaller earthquakes, and so on. In his model, most of the fault is in a state of stress below that required to initiate slip, but it can be triggered and caused to slip by nearby earthquakes or propagating ruptures. Any precursory phenomena will only occur when stresses are close to the yield stress. However, since even small earthquakes are initiated by still smaller earthquakes, in the limit, the region of rupture initiation where precursory phenomena might be expected is vanishingly small. Even if every small earthquake could be predicted, one is then still faced with the impossible task of deciding which of the thousands of small events will lead to a runaway cascade of rupture composing a large event. |
Brune proposed his model only as a possible scenario for earthquake unpredictability, but subsequent results have tended to support his idea. The average level of shear stress on major faults is now thought to be quite low, far below the levels at which laboratory experiments suggest rock failure and precursory phenomena should occur. Modern ideas about self-organized criticality in nonlinear systems suggest that it is to be expected that faults should be in a stress state such that even small events can initiate rupture to long distances. (The classic physical model for self-organized criticality is a sand pile in which individual grains of sand are continually added. The slope of the pile reaches an angle close to the maximum angle of repose. Additional sand grains may then trigger landslides of varying sizes, but it is difficult to predict in advance which grains will cause the largest slides.) Block-slider models have been devised that exhibit self-organized criticality (e.g. P. Bak and C. Tang, Journal of Geophysical Research 94 (1989), pp. 15,635-37). |
Finally, studies of the beginnings of earthquakes of varying size have shown no difference between small and large events (e.g. J. Anderson and Q. Chen, Bulletin of the Seismological Society of America 85 (1995), pp. 1107-116; J. Mori and H. Kanamori, Geophysical Research Letters 23 (1996), pp. 2437-40). That is, there is no way to tell from the initial part of a seismogram how large the event will eventually become. This supports the notion that large earthquakes do not necessarily originate in anomalous source regions but are triggered by rupture from a smaller event (for an opposing view, see W. L. Ellsworth and G. C. Beroza, Science 268 (1995), pp. 851-55). If Brune's hypothesis holds up, and the evidence for it is particularly strong in California, then short-term earthquake prediction may inherently be so difficult as to be impossible in practice. Barring dramatic new developments, an earthquake prediction program that promises timely, accurate warnings of future events, with a minimal number of false alarms, is unlikely to be achieved in the foreseeable future. |
Conclusion
In any case, seismology's most direct benefits to society are more likely to be achieved through identifying those regions most at risk from major earthquakes and encouraging suitable engineering and construction practices. People are rarely killed directly by earthquakes; rather the casualties arise from the failure of buildings and other structures. Through well-designed and rigorously enforced building codes, the death toll from earthquakes can be minimized. An example is provided by a comparison between two recent earthquakes: the Ms = 6.8 Armenian earthquake of December 7, 1988, and the Ms = 6.6 Northridge (southern California) earthquake of January 17, 1994. The Armenian event killed over 25,000 people and left 500,000 homeless, whereas only 60 were killed at Northridge. The shaking was slightly stronger in Armenia (maximum intensity of X on the modified Mercalli scale, compared to IX at Northridge), but most of the difference in the outcome can be attributed to the weaker construction practices that prevailed in Armenia. Future earthquakes are inevitable and there are few, if any, areas in the world that are completely free of earthquake risk. However, catastrophic loss of life can be prevented through sensible land use planning and the construction and maintenance of earthquake resistant structures. |
|
| |
