The highly uncertain art of earthquake forecasting

Recent devastating earthquakes in Italy and Algeria have again raised the question: Can earthquakes be predicted, can warning be given to evacuate in time?

The answer is that earthquakes have been successfully predicted. But the skill involved is still too uncertain to be trusted, as the following examples show.

On Feb. 4, 1975, an earthquake of Richter magnitude 7.3 occurred near Haicheng in Liaoning Province, northeast China. Following a chain of precurosors that had begun five years earlier and continued until the main shock , the Chinese were able to narrow down the prediction area and time of occurrence. The final warning to evacuate the 90,000 inhabitants came only a few hours before the quake destroyed or severely damaged 90 percent of the buildings in Haicheng. This was the first successful prediction of a major earthquake.

Nearly a year and a half later, an earthquake of magnitude 7.8 occurred at Tangshan, a city also located in northern China. A succession of precursors very much like those of the Haicheng quake had been monitored, leading to a long-range warning, then to an intermediate range-warning. At this point, a various observations judged to be preliminary to a quake ceased. A few days later, on July 28, 1976, the earthquake struck without warning, reportedly killing more than 600,000 people.

The success and the failure point up the recalcitrant nature of quake prediction. They highlight our poor understanding of the earthquake mechanism upon which predictions are based.

Most earthquakes are generated in zones where the huge plates that constitute the outer shell of the earth interact. Their relative motions are resisted by friction. This builds up stress that eventually makes the plates slip or fractures the rock, producing an earthquake, Seismic gaps, or gaps in the occurrence of large earthquakes along major plate boundaries, may delineate regions where such unrelieved stress is building up. This may be why these gaps seem to be one of the better indicators of future earthquake sites.

Real faults are not like two smooth blocks flush to each other. They are rough and irregular. Moreover, it now appears that these irregularities extend several kilometers below the surface to the level where earthqukes actually occur.

Fault movement can be held up at these points. They may consist of bends or short offsets in the direction of the fault. They may be rough spots or masses of rock of different composition from that of surrounding material on the faces of the fault. In time, enough stress may accumulate at one of these areas to cause a small rupture before the fault as a whole is ready to break. Either the slippage is confined to a small region centered at that point of stress concentration, or the stress released there induces slippage elsewhere along the fault.

This mechanism has several implications for earthquake prediction. The existence of small areas of high stress makes precursors much harder to find -- and thus quakes much harder to predict -- that would be the case were stress evenly distributed along a fault. If a particular precursor is generated only at a critical stress, it may appear only at the point of initial rupture, and nowhere else along the fault that eventually breaks. Large quakes are interpreted not as simply huge versions of moderate ones, but as a near-coincident succession of smaller fault ruptures. These greatly destructive multiple-event earthquakes may then be no easier to predict than small earthquakes.

The Chinese have long believed that animals may sense phenomena related to a coming earthquake. These phenomena may include low frequency sound generated by foreschocks too weak to be detected by instruments, changes in the electrostatic field, or gas emissions from the earth.

A period of "seismic calm" followed by increased activity just prior to the event also is characteristic of many large quakes.

There is frequently an abrupt departure from the normal value of the ratio of the S-and P-wave travel times to nearby seismic stations. (These are the two general types of seismic waves: P-waves are longitudinal, or compression, waves that travel from their source through the earth's interior; S-waves are more slowly traveling transverse shear waves that are restricted to the solid portions of the earth.)

Other localized events signaling an impending earthquake include anomalous changes in the volume of the crustal rock that cause ground level changes observed by tiltmeters and repeated surveying. Changes in the water level of deep wells and in concentration of the radioactive gas radon in this well-water and fluctuations of the electrical resistivity of the rock are other indicators.

The occurrence of many of these precursors can be explained by the "dilatancy hypothesis." Its essential by the "dilatancy hypothesis." Its essential feature is that many --ceded by a progressive opening of tiny cracks and pores in the critically strained rock -- that is, that rock dilates. But as long as the cracks remain dry, the frictional forces along the fault prevent the slippage necessary to induce an earthquake. The earthquake only occurs after the cracks have been filled with water: this reduces the friction across the fault and allows it to move.

In the first stage of this process, the number of tiny cracks is increasing rapidly and hence producing undersaturation in the dilatant region. This drying-out acts to inhibit faulting and thereby create a quiescent period, to increase electrical resistivity (dry rock is a poorer electrical conductor than wet rock), and to lower the ratio of P-wave velocity to S-wave velocity (compression waves travel more slowly through gas than through a liquid or solid). As water diffuses into the undersaturated region, the electrical resistivity drops, the velocity ratio rises, and radon is carried upwards from the rock below. Small tremors occur in this second stage as frictional forces are weakened, and the main shock follows.

The dilatancy hypothesis has enjoyed considerable success. In particular, injection of wastewater into a deep well near Denver in 1965 triggered small quakes. By successive injection and withdrawal of fluids in a well of the Rangely oil field in Colorado, scientists were able to turn seismic activity on and off.

But dilatancy effects appear to be irrelevant in the case of deep-focus earthquakes. And there is some doubt that this hypothesis accounts for the magnitude of the velocity-ratio decrease, the precursor on which much hope for prediction is based. More discouraging though, is the fact that these precursors may not appear, or at least may not be recognized, prior to an earthquake.

With this in mind, increasing attention is being given to measuring physical properties that should be more closely related to the actual failue of a fault. The accumulation of stress at a point may cause observable deformation of adjacent rock, for example.

Measurement of the stress itself by pressurizing a section of a drill hole, however, is very difficult, as it must be done five to ten kilometers below the surface where earthquakes occur. Another way of inferring stress changes independence of dilatancy effects is to examine how the magnetization of strongly magnetic minerals, such as magnetite, changes with time in regions of stress. These minerals have the property that a decrease in volune causes a slight increase in magnetization. This effect, called piezomagnetization, though observable in theory, has not been useful as yet.

It is an open question as to when accurate earthquake prediction will become possible. The presence of seismic gaps in an area with a history of quakes already provides a crude method for narrowing down the choices of when and where to look for signs of an imminent large earthquake. The successful prediction of the location and size of the November 1978 earthquake near Oaxaca, Mexico, using this indicator, demonstrates its utility.

In some cases, the various precursors can be monitored up to the time of the event, as the Chinese were able to do at Haicheng. But in other cases anticipated warning signs have been ambiguous or absent. It is apparent that our understanding of earthquakes is still very limited. Indeed, the ultimate test of our understanding of the nature of earthquakes will be our ability to predict them.

We want to hear, did we miss an angle we should have covered? Should we come back to this topic? Or just give us a rating for this story. We want to hear from you.