Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.
METHODS OF ATTEMPTING PREDICTION Two methods of study are being used by scientists in their efforts to predict earthquakesâstatistical methods and geophysical methods. The first uses the catalogued history of earthquakes in a region as a key to estimating when and where such future events may occur. The second involves the observation and interpretation of certain changes in the physical environment in earthquake-prone regions as indicators of an impending event. STATISTICAL METHODS The occurrence of earthquakes, especially those that have caused casual- ties or damage, has been documented throughout historical time. In China, earthquake catalogues span several thousand years. In California, statistically useful catalogues span only a few decades. The quality of statistical analyses of catalogued earthquakes improves with the acqui- sition of more information. Unfortunately, the number of recorded earthquakes is insufficient to allow statistical evaluations for small areas and small times periods. A number of earthquakes sufficiently large for statistical inference may be tabulated for a large area over a short time interval or for a small area over a long interval, but seldom are data available in ade- quate numbers for a small area over a short time interval. For example, the worldwide locations of earthquakes for only a short interval of time (Figure l) yields considerable information about the non-uniform distribution throughout the world of the regions in which earthquakes are likely to occur. About l40 earthquakes of magnitude 6 or greater occur in many parts of the world each year. Statistically, the map (Figure l) can tell us that a large earthquake is likely to occur during a relatively short time interval somewhere within a number of known areas totaling between l and l0 percent of the earth's surface. Such a statistical statement is obviously unsatisfactory as an indicator or predictor of a specific hazard. Similarly, from a catalogue spanning several decades or longer, for a specific seismically active region measuring a few hundred kilometers on a side, it is possible to estimate the likelihood of occurrence of a large earthquake. However, such a likelihood statement, or "generalized
l0 *a â¢d 3 N fl M <S CQ o n fc ~ V ⢠-P rH 0) ⢠0 0 -rf z Oil w » 01 w m 8 . w 0 -H vO O O^ rH 3l 4J - (U £ -H Zi u M 0 5 S 41 $ IM 0 O -H ^J w -H -H U 0) 0) -u W 0) -U 'M 0) -H 3
ll prediction," is usually inadequate to permit taking specific preparatory measures in advance of any hazard. In Southern California, for example, about 5 earthquakes of magnitude 6 or greater occur in 20 years, but without more precise information no immediate preparatory action has been possible. It is possible to provide information about probabilities of occur- rence of large events in selected restricted intervals of both space and time (e.g., in a given area during a 5-year period). Unfortunately, these probabilities are so low that such estimates cannot permit pre- parations for specific events. Whether the rates of occurrence of earthquakes in a specific location and time period are high or low, such estimates can only give information about likelihoods of occurrence or risk. As the space-time windows are narrowed, the reliability of sta- tistical risk assessments becomes poorer. But statistical risk .statements have value. They can be used as input in developing criteria for the design of earthquake-resistant structures and to other measures for reducing the damaging effects of earthquakes, in planning for remedial action following earthquakes, and in the development of a strategy of deployment of often scarce instru- mental resources for use in more-direct prediction studies. As noted in the previous section, however, such generalized probability state- ments are a far cry from the short-term predictions that are visualized by most people as the primary objective of the present research effort. The simplest analysis of the occurrence of earthquakes in a specified interval of time and space usually leads to the recognition that earth- quakes are occasionally clustered; that is, they are not randomly dis- tributed throughout this space-time interval. For example, in a region as small as Southern California, both the space and time sequences for earthquakes can be shown to be significantly non-random. One reason for space non-randomness is that earthquakes tend to be concentrated on specific earthquake faults. Much of the time, non-randomness is due to aftershocks that are known to be clustered in a short time interval following the parent event. When aftershocks are removed from the Southern California catalogue so that only a residual catalogue of large events remains (i.e., aftershock-producing events), the residue is random. A statistical result that yields randomness is discouraging for the identification of specific space or time intervals in which to concentrate scientific effort. In another example, the map of world-wide earthquake epicenters shows that earthquakes are not spread out randomly over the surface of the earth but instead are concentrated in specific regions. This fact has contributed to the construction of the plate-tectonics model of the motions of the surface of the earth. The earth's surface is divided into a relatively small number of rigid plates, each of large areal extent. These plates are in motion relative to each other, probably driven by large-scale processes in the earth's interior. When the plates come into contact, the effect of "friction" at the edges results in earth- quakes. Thus, the locations of earthquakes shown in Figure l delineate the edges of the plates.
l2 An effort is now being made to identify clustering or anti-clustering, i.e., the presence of a concentration of events or a scarcity of events, in space and time. This effort uses generalizations drawn from funda- mental geophysical processes to modify the statistical analysis. Efforts to discern such clustering include the following: l. Studies of seismicity "gaps" along plate margins. These are regions where no earthquakes have taken place in recent times, although there is relative motion of adjacent blocks. These temporarily quiescent regions are likely zones for future large events. Parts of the San Andreas fault zone in California currently display the seismic-gap phenomenon. 2. Studies of migrations of epicenters along and near plate margins, both on the scale of thousands of kilometers associated with plate mar- gins and of hundreds of kilometers associated with smaller, extensively faulted regions. It is known that one earthquake can trigger another; aftershocks are clear evidence of this. After a large earthquake, there is a redistribution of the stress field both in the short-distance range, causing aftershocks, and to a smaller extent at long distances. The altered stress field might cause a distant region to reach criti- cality earlier than it would if it were isolated from neighboring earth- quake zones. 3. Studies of triggering of earthquakes by non-seismic effects. Many attempts have been made to correlate earthquakes with obvious astronomical effects such as time of day, and season of the year. These have been unsuccessful. Attempts are continuing to seek associations of earthquakes with more-subtle geophysical phenomena that can "pres- tress" (or add stress to) a seismic zone that is already near the critical state. One phenomenon being investigated as a possible trig- gering stress is the periodic tide in the solid earth. It is unlikely that the extension of earthquake catalogues in time would result in a large improvement in the precision of statistical analysis, since such extensions usually involve only a small number of years of data. One exception to this is the possibility that the dates of large earthquakes of the distant past can be obtained by historical methods, including archaeology. Here, analysis of records from those few places where long-term historical catalogues are available, e.g., China, Japan, and the Middle East, may be of value. But, in general, in order to make up for the lack of comprehensive long-term historical records in the United States and in other parts of the world, we must improve statistical analysis by taking geophysical models into account. GEOPHYSICAL METHODS Geophysical methods involve searching for, identifying, and monitoring changes in the physical state of the earth that are precursory to earthquakes. Unlike statistical methods, these observations and inter- pretations have the ultimate capability of leading to prediction of
l3 magnitude, time, and place of individual events. The statistical methods, at present, are being applied almost independently of the physical state of the earthquake environment; the geophysical methods, conversely, are concerned in the main with the detailed short-term physical state of matter and at present have little relationship to the long-range statis- tical evidence of earthquake occurrence. Current interest in the observation of precursory phenomena was sparked by contacts between American and Soviet scientists in l97l during the quadrennial Assembly of the International Union of Geodesy and Geophysics. Soviet seismologists, after 25 years of intensive ob- servations in the Garm region of Tadjikistan, reported a l0 percent decrease in the ratio of the velocity of compressional waves relative to that of shear waves (Vp/Vs) for some time period prior to earthquakes of moderate size, in comparison with the normal value of this ratio. The impending earthquake was signaled by a return of the Vp/Vs ratio to a normal value immediately before the earthquake. The duration of these precursory anomalous seismic-velocity ratios appeared to be longer for larger earthquakes. Subsequently, it has been found that the duration of the anomaly may be a few days for a magnitude-3 earthquake and, by extrapolation, 40 years or more for a magnitude-8 earthquake. The Soviet scientists also observed other precursory phenomena. Spanning an even longer period, Japanese scientists observed that significant changes in elevation of the ground surface preceded some large earthquakes. Most notable among these observations were precur- sory elevation changes associated with the Niigata earthquake of l964. Historical accounts tell of significant changes in elevation associated with other huge historical earthquakes. Stimulated by these and other reports of precursory anomalies in grophysical measurements, an effort was mounted in the United States to seek such anomalies in this country as well. Changes in the ratio of velocities of seismic waves were reported prior to earthquakes in the Blue Mountain Lake area of New York; these changes had many of the characteristics of the Soviet observations. By reviewing seismic records for the period preceding the San Fernando earthquake of February l97l, it was later found that this earthquake was preceded by similar anomalous seismic velocity ratios. Since then, anomalous changes in tilt directions, variations in radon concentration in ground water, variations in compressional velocity, anomalous magnetic fields, anomalous electrical resistivity, the relative abundances of large and small earth- quakes, and the overall level of seismic activity have all been pro- posed, and on occasion used, either to forecast or "hindcast" earthquakes. Only infrequently have anomalies observed by several methods shown precursory indications of the same earthquake. This is partly because of an inability to focus all the different types of experiments on the same area. Part of the problem of studying these phenomena is the relatively short time-baseline for the accumulation of the needed data. This has led to false alarmsâthat is, anomalies corresponding to no subsequent earth- quakes. It has also led to failure to predict, since some events have occurred without evident precursory anomalies.
l4 In the discussion above, the geophysical methods have been presented as phenomenological in character. Models for the change in physical state of the material in the vicinity of an earthquake focus have been proposed that are capable of accounting for the observed precursory anomalies. Such models take advantage of the anomalous behavior of materials as they approach the critical fracture condition. Investiga- tors in the United States have proposed that the observed changes in the velocity ratio can best be explained by the phenomenon of rock dilatancy. In this model, the velocity ratio decreases initially be- cause of the growth of cracks and the related increase in volume of the rock mass near the focal region as stresses build up prior to the earth- quake. This phase is followed by one in which the velocities return to normal prior to failure. Two hypotheses have been proposed to explain the velocity recovery. In one, water flows into the cracked region immediately preceding the earthquake, causing the velocity ratio to return to normal; the increase in water pressure also serves to weaken the rock. This hypothesis has been termed the dilatancy/diffusion model. In the other hypothesis, most of the cracks close up in the dilated region prior to fracture because of the growth of certain naturally selected cracks. These closures increase the velocities and also increase the pressure of the water in the pore spaces in the rock. This has been termed the dilatancy/crack-closure, or dilatancy/ instability, model. Further experimental and theoretical work must be done to resolve the differences between the two models or to develop a more definitive model that might, when much more is known, even be dif- ferent from those discussed above. In any case, it is likely that the focal region is highly complex, studded with anisotropy and non-linear rheology and with complex electrical and mechanical properties. It has been proposed that these dilatancy models can explain the wide variety of observed precursory phenomena, including changes in seismic velocity and electrical resistivity, land uplift and tilt, fluid flow, rate of radon emanation, frequency of occurrence of small earthquakes, and the relative abundance of large and small earthquakes. As yet, only the roughest of linkages exist between the models of precursory phenomena and observations of them in the field. Further- more, no satisfactory models yet exist for determining the extent to which a given fault will tear, once rupture has been initiated, and hence for estimating the probable magnitude of an earthquake. However, the empirical data suggest that the magnitude of an earthquake may be predetermined by the extent of the physically anomalous zone and the duration of the anomalous episode.
