How earthquakes are predicted

Many methods have been used to try to predict earthquakes. These have ranged from attempts to identify typical ‘earthquake weather’ to observations of strange animal behaviour prior to an earthquake. Since the 1960s much scientific effort has been directed to the problem of prediction, particularly in the USA. the Soviet Union. Japan and China. It must be confessed, however, that little success has been achieved.

The absence of any obvious precursors to the moderately large earthquake (magnitude 5.7) that occurred in August 1979 near Coyote Lake. 100 km south-east of San Francisco, suggests that some earthquakes may be extremely difficult or even impossible to predict. The most publicized lack of forewarning was the tragic earthquake which on 27 July 1976 almost razed Tangshan. an industrial city in China of one million people. Unofficial reports estimated a death toll of about 650000. including about a hundred people in Peking, some 150 km to the west. It is estimated that an additional 780 000 people were injured.

Historical studies of world seismicity patterns have made it possible to predict the probable place at which a damaging earthquake can be expected to occur. However, such records do not enable us to forecast a precise time of occurrence. Even in China, where between 500 and 1000 destructive earthquakes have occurred within the past 2700 years, statistical studies have not revealed any clear periodicities between great earthquakes, although they do indicate that long periods of quiescence can elapse between them.

Many different forces, including severe weather conditions, volcanic activity and the gravitational pull of the Moon, Sun and planets, have been suggested as earthquake triggers. Numerous catalogues of earthquakes, including complete lists for California, have been searched for such precipitators without any convincing results. Some of the more promising lines of research are the detection of strain in crustal rocks by geodetic surveys (of the shape of the Earth), the identification of suspicious gaps in the regular occurrence of earthquakes in both time and space, and the observation of foreshocks.

In recent years the major earthquake prediction effort has been directed to more precise measurements of fluctuations in the physical properties of crustal rocks in seismically active continental areas. Special sensing devices have been installed to observe long-term changes in the rocks, but the number of measurements is still limited, and results have thus far been conflicting. In some, unusual behaviour was seen before a local earthquake, while in others nothing significant occurred before the event or, alternatively variations were observed that were not associated with earthquakes.

If some properties of rocks change before an earthquake, then the speed of seismic waves may also vary. Some of the first information on precursory changes in the travel time of waves in moderate earth- quakes was obtained in 1962 in the Tadzhikistan region of the Soviet Union. Measurements there suggested that P velocities changed by about 10-15 per cent before the occurrence of local earthquakes.

Fieldwork since then, in the Soviet Union and elsewhere, has indicated that in some cases the velocity of P waves in the focal region decreases by about 10 per cent for a time and then, just before the main shock occurs, increases again to a more normal value. More detailed checks have now been made in a number of countries, with mixed results. In the USA, in 1971. researchers at the Lamont Doherty Geological Observatory, working on quite small earthquakes in the Adirondacks in New York, detected increases in the travel time of P waves before three small earthquakes.

In contrast similar tests carried out in central California have shown that fluctuations in travel times before a number of small to moderate earthquakes along the San Andreas fault were not significant. This negative result indicates that any precursory changes in the velocity of P waves before small to moderate earthquakes are probably highly localized around the region.

A second variable that has been used for prediction is precursory changes in ground level, such as ground tilts in earthquake-prone regions. A third possible variable is the release into the atmosphere of the inert gas radon, particularly from deep wells along active fault zones. It has been claimed that significantly increased concentrations of radon have been detected just before earthquakes in some parts of the Soviet Union. However, because so few measurements in different geological circumstances are available it is impossible to say whether the observed increases are exceptional rather than normal variations in radon concentration.

A fourth variable, to which a good deal of attention has been paid, is the electrical conductivity of the rocks in an earthquake zone. It is known from laboratory experiments on rock samples that the electrical resistance of water-saturated rocks such as granite changes drastically just before the rocks fracture at high pressure. A few field experiments to check this result have been made in fault zones in the Soviet Union, China, Japan and the USA, and decreases in electrical resistance before earthquakes have been reported.

Part of the difficulty in predicting earthquakes is the lack of a proper understanding of how the two faces of a fault slide by one another. Unlike a theoretical model in which a fault is thought of as two surfaces flush to one another, real faults tend to be rough and irregular. Such irregularities extend several kilometres below the surface at which earthquakes actually occur. These irregularities may be bends and short offsets in the direction of the fault, bumps or rough spots on the faces of the fault or differences in the material caught up between the fault faces.

Any or all of these irregularities could hinder movement along the fault. One result would be that instead of stress being evenly distributed along the fault it would tend to be concentrated at the irregularities. When the stresses reach a critical level the fault might rupture at one of these stress concentrations before most of the fault is ready to break. Most of the jolt' of the earthquake could come from the small rupture, but the stress released there could also cause slippage elsewhere on the fault.