An earthquake results from sudden slip on a geological fault. [HN10] Such fracture and failure [HN11-13] problems are notoriously intractable. The heterogeneous state of the Earth and the inaccessibility of the fault zone to direct measurement impose further difficulties. Except during a brief period in the 1970s (2), the leading seismological authorities of each era have generally concluded that earthquake prediction is not feasible (3). Richter [HN14-15], developer of the eponymous magnitude scale, commented as follows in 1977: "Journalists and the general public rush to any suggestion of earthquake prediction like hogs toward a full trough... [Prediction] provides a happy hunting ground for amateurs, cranks, and outright publicity-seeking fakers" (4). This comment still holds true.
For large earthquakes to be predictable, they would have to be unusual events resulting from specific physical states. However, the consensus of a recent meeting [HN5] (5) was that the Earth is in a state of self-organized criticality [HN16] where any small earthquake has some probability of cascading into a large event. This view is supported by the observation that the distribution of earthquake size (see figure) is invariant [HN17] with respect to scale for all but the largest earthquakes. Such scale invariance is ubiquitous in self-organized critical systems (6). Whether any particular small earthquake grows into a large earthquake depends on a myriad of fine details of physical conditions throughout a large volume, not just in the immediate vicinity of the fault (7). This highly sensitive nonlinear dependence of earthquake rupture on unknown initial conditions severely limits predictability (8,9). The prediction of individual large earthquakes would require the unlikely capability of knowing all of these details with great accuracy. Furthermore, no quantitative theory for analyzing these data to issue predictions exists at present. Thus, the consensus of the meeting was that individual earthquakes are probably inherently unpredictable.
Figure 1, PostScript , a slightly modified version
Figure 1, PDF , a slightly modified version
Critical quakes. Number of earthquakes from 1 January 1977 to 30 June 1996 in the Harvard catalog [HN27] (24) with magnitude greater than M for shallow (0 to 70 km), intermediate (71 to 300 km), and deep (301 to 700 km) earthquakes. Dotted lines are power-law curves modified by an exponential taper for the largest magnitudes [equation 3 of (8)]. Analyses of smaller earthquakes show that self-similarity extends to magnitudes as small as zero (25). Such power-law curves are characteristic of systems in a state of self-organized criticality.
Empirical earthquake prediction would require the existence of observable and identifiable precursors [HN18] that would allow alarms to be issued with high reliability and accuracy. There are strong reasons to doubt that such precursors exist (10). Thousands of observations of allegedly anomalous phenomena (seismological, geodetic, hydrological, geochemical, electromagnetic, animal behavior, and so forth) have been claimed as earthquake precursors, but in general, the phenomena were claimed as precursors only after the earthquakes occurred. The pattern of alleged precursors tends to vary greatly from one earthquake to the next, and the alleged anomalies are frequently observed at only one point, rather than throughout the epicentral region. There are no objective definitions of "anomalies," no quantitative physical mechanism links the alleged precursors to earthquakes, statistical evidence for a correlation is lacking, and natural or artificial causes unrelated to earthquakes have not been compellingly excluded (11). In other fields threshold signals have often been erroneously claimed as important physical effects (12); most if not all "precursors" are probably misinterpreted as well. Unfortunately, each new claim brings a new set of proposed conditions, so that hypothesis testing, which is what separates speculation from science, is nearly impossible.
Chinese seismologists claimed that the 4 February 1975 Haicheng (magnitude = 7.3) earthquake was successfully predicted and that "very few people were killed" (13). However, an official publication in 1988 (14) states there were 1328 deaths and 16,980 injured. This disparity casts doubt on claims [HN19] for the Haicheng prediction. China's Cultural Revolution was still taking place in 1975. An American delegation's report (15) captures the remarkable atmosphere: "Earthquake prediction was not a minor experiment.... Indeed, belief in earthquake prediction was made an element of ideological orthodoxy that distinguished the true party liners from right wing deviationists." The possibility that political pressures caused inaccuracies in claims for the Haicheng prediction cannot be excluded. An intense swarm of microearthquakes, many of which were large enough to be felt by local residents, began over 24 hours before the main shock (15). These microearthquakes might well have induced some spontaneous evacuation. At least 240,000 people died in the 1976 Tangshan, China, earthquake, which was not predicted.
Varotsos [HN20] and co-workers claim to be able to predict earthquakes in Greece on the basis of geoelectrical observations (16), but our analyses show their claims to be without merit (17,18). Some of the geoelectrical signals are artifacts of industrial origin (19), and there is no compelling evidence linking any of the geoelectrical signals to earthquakes. Controversy lingers primarily because Varotsos's claims have not been stated as unambiguous and objectively testable hypotheses (20).
Is prediction inherently impossible or just fiendishly difficult? In practice, it doesn't matter. Scientifically, the question can be addressed using a Bayesian approach [HN21] (21). Each failed attempt at prediction lowers the a priori probability for the next attempt. The current probability of successful prediction is extremely low, as the obvious ideas have been tried and rejected for over 100 years (17). Systematically observing subtle phenomena, formulating hypotheses, and testing them thoroughly against future earthquakes would require immense effort over many decades, with no guarantee of success. It thus seems unwise to invest heavily in monitoring possible precursors.
Seismology can, however, contribute to earthquake hazard mitigation. [HN22-25] Statistical estimates of the seismicity expected in a general region on a time scale of 30 to 100 years (22) [as opposed to "long-term predictions" of specific earthquakes on particular faults within a few years (23)] and statistical estimates of the expected strong ground motion are important data for designing earthquake-resistant structures. Rapid determination of source parameters (such as location and magnitude) can facilitate relief efforts after large earthquakes. Warnings of tsunamis [HN26] (seismic sea waves) produced by earthquakes also contribute significantly to public safety. These are areas where earthquake research can greatly benefit the public.
Discussion from note
Discussion from note
A group of prediction researchers established validation criteria (including a precise definition of the anomaly, an explicit statement of the signal-to-noise ratio, detection at more than one station, and full disclosure of both negative and positive results) and invited nominations of precursor candidates. Only 31 nominations were submitted; none of these fully satisfied the validation criteria. As these nominations were presumably the cream of the crop, the fact that not one fully met the validation criteria is strong empirical evidence against the existence of the type of precursors required for prediction. (Three of the 31 precursor nominations were placed on a "preliminary list of significant earthquake precursors," despite failure to fully meet the validation criteria: One lacked a clear definition of what constitutes an "anomaly" and a comprehensive statistical evaluation; a second was not supported by a quantitative analysis, and the number of false alarms and missed events was not evaluated; and a third was seen for one event at only one station, and there was no quantitative definition of what constituted an anomaly.) Further evaluations of precursor case studies by the above group of prediction researchers are presented by M. Wyss, Ed., Pure Appl. Geophys. 149, 3 (1997).
Long term predictions were issued in 1976 for the Tokai region in Japan [the initial publications were all in Japanese; see K. Ishibashi in, Earthquake Prediction: An International Review, D. W. Simpson and P. G. Richards, Eds. (Ewing Monograph Series, Am. Geophys. Union, Washington, DC, 1981), pp. 297-332, for a discussion in English and references] and in 1985 for the Parkfield region in California [W. H. Bakun and A. G. Lindh, Science 229, 619 (1985)]; both predictions have failed, as no large earthquakes have occurred. In contrast, severely damaging earthquakes in California [Loma Prieta in 1989 (see below), Landers in 1992, Northridge in 1994] and Japan (Okushiri Island in 1993, and Kobe in 1995) occurred on faults for which long-term predictions had not been issued. J.C. Savage [Bull. Seismol. Soc. Am. 83, 1 (1993)] discusses and criticizes the "Parkfield prediction fallacy." Y. Y. Kagan [Tectonophys. 270, 207 (1997)] questions the claim that quasi-periodic "characteristic earthquakes" regularly occur at Parkfield. After the 1989 Loma Prieta earthquake there was a claim that a relatively general long term seismicity forecast--as opposed to a long term prediction for the particular fault that ruptured--had been successful [U.S. Geological Survey Staff, Science 247, 286 (1990)]. But this claim proved controversial [R. A. Kerr, ibid. 249, 860 (1990)], and a statistical analysis strongly argues against this claim [J. C. Savage, Geophys. Res. Lett. 19, 709 (1992)].
Discussion from note
David Jackson's Home Page summarizes his research interests and lists his publications.
Yan Kagan's Home Page summarizes his research interests and lists his publications.
F. Mulargia can be reached via the Department of Physics' Home Page at the Universita di Bologna, Italy.
The Royal Astronomical Society and its affiliated Joint Association for Geophysics held a discussion meeting in November 1996 on Assessment of schemes for earthquake prediction.
The US National Academy of Sciences held a symposium (attendance and presentation by invitation only) on earthquake prediction in February 1995, for which abstracts are available. The introduction Earthquake prediction: The scientific challenge was given by L. Knopoff.
K. Aki gave a review of earthquake prediction in Reviews of Geophysics, vol. 33, 1995, as part of the U.S. National Report to International Union of Geodesy and Geophysics 1991-1994.
The USGS Cascades Volcano Observatory offers definitions and descriptions of Magnitude, Intensity, and the Modified Mercalli Scale.
Michigan Technological University, Department of Geological Engineering and Sciences, presents an Earthquake Magnitude Scale and Classes Chart as part of UPSeis, a new program created to teach young people about Earth.
The fundamentals of faulting are reviewed in Earthquake ABCs at the Southern California Earthquake Data Center.
The Landers Earthquake page has links to MPEG movies of the rupture and animations of the aftershocks. Related pages about the Northridge earthquake and the faults of Southern California are also available.
M. Willemse (Stanford University, Department of Geological and Environmental Sciences) provides links to images of fracture patterns, strike-slip faults, and normal faults.
K. M. Cruikshank (Geology Department, Portland State University) has a comprehensive bibliography on faulting.
A description of the Richter scale is provided by NORSAR, a geophysics research institution supported by the Research Council of Norway.
The online pages of the science radio series Earth and Sky has information on Charles Richter and the magnitude scale he developed.
The Santa Fe Institute presents a discussion of Self-Organized Criticality that includes applications to sandpiles and document delivery over the Web.
R. Devaney of the Boston University Math Department has a Web page about chaos and fractals, including a discussion of self-similarity and scale invariance.
At the November meeting of the Royal Society, I. Main discussed the difficulties of defining precursory phenomena.
The official position of the government of the People's Republic of China on recent prediction research is outlined in the China Science and Techology Newsletter (The State Science and Technology Commission). See also the review by Aki on prediction claims (earlier hypernote).
A special issue of Geophysical Research Letters (27 May 1996) edited by R. J. Geller contains reports by Varotsos and his collaborators, along with reports critical of his methods.
A brief definition and a simple example of using Bayes' Theorem is presented in the Statistics for Engineers course at the Faculty of Engineering, University of Wollogong, Australia.
A lecture on the hazards of earthquakes is offered by the Department of Earth Sciences & The Institute of Tectonics, University of California, Santa Cruz.
A guide to international building codes that are designed to mitigate earthquake damage is provided by the National Center for Earthquake Engineering Research of SUNY Buffalo.
The Western States Seismic Policy Council has images of the aftermath of the 1995 Kobe earthquake among others.
The US Geological Survey's Homepage for Earthquakes points to a variety of hazard topics, such as the National Seismic Hazard Mapping Project.
The Japanese word tsunami is written as two characters meaning "harbor wave." The Tsunami Web site is an online information resource about these great waves.
The Harvard Centroid-Moment Tensor (CMT) database is a catalog of large earthquakes maintained by the Harvard Seismology group. A query page for the CMT database is available at the Earthquake Research Institute, University of Tokyo.
Also see the archival list of Enhanced Perspectives