Next: The Gap Theory Up: Earthquake prediction, societal implications Previous: Earthquake prediction, societal implications
``If I were a brilliant scientist, I would be working on earthquake prediction.'' This is a statement from a Los Angeles radio talk show I heard just after the Northridge earthquake of January 17, 1994. Five weeks later, at a monthly meeting of the Southern California Earthquake Center (SCEC), where more than two hundred scientists and engineers gathered to exchange notes on the earthquake, a distinguished French geologist who works on earthquake faults in China envied me for working now in southern California. This place is like northeastern China 20 years ago, when high seismicity and research activities led to the successful prediction of the Haicheng earthquake of February 4, 1975 with magnitude 7.3. A difficult question still haunting us [ Aki, 1989] is whether the Haicheng prediction was founded on the physical reality of precursory phenomena or on the wishful thinking of observers subjected to the political pressure which encouraged precursor reporting. It is, however, true that a successful life-saving prediction like the Haicheng prediction can only be carried out by the coordinated efforts of decision makers and physical scientists.
Earthquake prediction research in the U.S. has been stagnant since the early optimism regarding precursory phenomena was dissipated by negative observations. In recent years, however, we see a revival of interest in prediction research at several fronts in a more subdued manner but rooted on more solid scientific ground.
First, sufficient data began to accumulate for testing the validity of certain prediction methods and hypotheses underlying them. For example, the seismic gap theory for long-term (decades) prediction first formulated by Fedotov  and developed by Sykes and his colleagues has been tested by Kagan and Jackson . Likewise, the method of intermediate-term (years) prediction developed by Keilis-Borok and his colleagues [e.g. Keilis-Borok et al., 1988] in the former Soviet Union using earthquake catalog data is being tested by researchers in the U.S. (e.g. Healy , and Minster and Williams ). Furthermore, the IASPEI (International Association for Seismology and Physics of Earth's Interior) sub-commission on Earthquake Prediction evaluated claims of earthquake precursors [ Wyss, 1991]. It is essential for a healthy development of earthquake prediction research to formulate a prediction method testable by others.
Another trend is the increasing use of probabilities in communicating earthquake information to the public. Instead of predicting the time, place and magnitude of a future earthquake, recent public-policy documents attempt to estimate an earthquake probability in a given window of time, space and magnitude. Examples are the reports of the Working Group on California Earthquake Probabilities[1988, 1990] based on the recurrence data of ``characteristic earthquakes'' [ Schwartz and Coppersmith, 1984] on major fault segments in California. Currently, a major report (to be published in April, 1995) is being prepared by SCEC (Southern California Earthquake Center) scientists to characterize earthquake sources in southern California by integrating the geologic data on faults, catalog data on historic earthquakes and GPS (Global Positioning System) data on crustal strain [ Ward, 1994; Jackson et al. 1993]. Through this work we found that the method of probabilistic seismic hazard analysis [ Cornell, 1968] was useful not only as a means for integrated transmission of multidisciplinary earth science data to the user community, but also for promoting interactions among the different disciplines and identifying critical issues that can be resolved only by a multidisciplinary cooperative work.
Probability is a relatively new concept in human history. The origin of the theory of probability goes back to the 17th century, when B. Pascal and P. de Fermat exchanged letters on dice throwing [ Encyclopedic Dictionary of Mathematics, 1980]. The concept appears to be useful in dealing with difficult problems in human society.
In recent years, several earthquake predictions made officially by the U.S. Geological Survey were in terms of probabilities. For example, the 1988 and 1990 working group mentioned above evaluated the probability of fault rupture in the next 30 years, and the Parkfield experiment issued a short-term alert on the basis of a 72-hour probability.
In addition to the above new trends, namely, the hypothesis testing and the use of probabilities, I recognize a more fundamentally important new trend to simulate the complex space-time-magnitude behavior of earthquake occurrence by physical modeling. A central issue in these attempts is whether the complex behavior of seismicity is caused by the non-linear dynamics inherent to earthquake rupture or by the interaction of rupture with heterogeneities of fault zone structure. This issue is critical to the validity of ``characteristic earthquake'' concept, and is being debated vigorously [ Rice, 1993]. I see in this trend a beginning of interaction between geology (heterogeneous earth) and physics (non-linear dynamics of fault rupture) needed for a sound development of physical theory of earthquake prediction.