Predicting earthquakes: act wisely on long-term forecasts

For seismologists trying to develop reliable means of earthquake forecasting, the recent Mexican shocks have reinforced an important point. The most useful kind of forecast they now can make is to delineate areas of earthquake hazard. The most effective use to which such forecasts can be put is to help local authorities set up building codes, land use zoning, and emergency response plans that will minimize the potential danger. Short-term prediction -- the kind that warns of an impending quake a few weeks or days ahead of time -- remains an elusive goal. The short-term processes leading up to a quake, which must be understood to make forecasting possible, are largely unknown. But the kind of long-term forecasting that led seismologists to anticipate the Mexican quake and to place instruments in the suspect region is founded on a substantial and growing understanding of basic earthquake mechanisms and of the historical ``repeat

cycle'' of major shocks in quake-prone areas.

This is a major achievement of geophysical research; the implications for planning have yet to be fully appreciated.

It indicates, for example, that southern California should be preparing aggressively today for a great earthquake, which is quite likely to occur along the southern San Andreas Fault within the next 50 years. It suggests that the complacent Eastern United States should also be establishing quake-resistant building and land-use standards. Strain is accumulating in the triple fault system that caused the massive New Madrid, Mo., shocks during the winter of 1811-1812. Such quakes today could break windows as far away as Washington, D.C., and cause damage from the Great Lakes to the Gulf of Mexico.

Some 95 percent of earthquakes occur along the edges of 10 or so great plates making up Earth's crust. Sometimes two plates slide past each other, as the Pacific Plate slides past the North American Plate in California. Elsewhere, one plate dives beneath another in a process called subduction, as the Cocos Plate dives beneath the North American Plate along Mexico's west coast. Where the moving plates stick, strain accumulates to be released eventually in an earthquake.

Careful study of earthquake records shows that plate boundaries are well outlined by earthquakes over a period of several centuries. However, over shorter periods there may be zones along these margins where quakes have not happened for some time. These so-called gaps indicate areas where a quake may be expected. The Mexican shock was foreshadowed by such a gap.

Also, the research field of paleoseismology, which works with the records of ancient quakes left in underground soil structures, can also give long-term warnings. Organic remains, such as peat, which can be dated, sometimes allow researchers to trace the recurrence of earthquakes in an area. Kerry E. Sieh, who has pioneered this research at the California Institute of Technology, has used such geological records to find that great earthquakes recur along the southern San Andreas Fault about once every 1 50 years. The last such quake was in 1857. Taken together with geodectic measurements indicating that strain is accumulating along that part of the fault, this has led to the estimate that southern California is likely to have a great earthquake within the next 50 years -- a quake that could occur at any time.

Even for those earthquakes that occur in the middle of a plate, such as the New Madrid series, faults are being located and recurrence cycles established. In the Eastern United States these cycles seem to run to several hundred years or more. The likelihood of a quake is much less than in California. Yet it now is known to be substantial enough for earthquake hazard to be considered in setting building standards.

The most precise use of this new understanding has been made at Parkfield, Calif., about halfway between Los Angeles and San Francisco. Here moderately severe earthquakes have recurred with a regular rhythm about every 22 years. The US National Earthquake Prediction Evaluation Council and the California Seismic Safety Commission have officially endorsed the US Geological Survey (USGS) prediction that a quake will shortly recur along this section of the San Andreas Fault. The most likely time

is January 1988, with a 95 percent probability of a quake within five years of that date.

Parkfield now is the focus of a massive USGS-sponsored research effort. The fault here is heavily instrumented both to monitor the expected quake and to trace any signs of strain buildup leading to it. Researchers are particularly interested to see if there are any distinctive precursor signs -- such as land deformation or preliminary shocks -- which could serve as short-term earthquake warnings.

So far, such warning signs have been equivocal. Sometimes important quakes are preceded by foreshocks and sometimes not (the Mexican quakes were not, according to the USGS). Abnormal animal behavior, release of certain gases such as the radioactive gas radon from strained rocks, and other precursory signs have not proved reliable guides. In short, geologists do not yet know if short-term earthquake forecasting may ever become a widely useful operational technique.

But they do know with some certainty many areas where earthquakes are more or less likely to happen. This is knowledge upon which people can act now. Buildings can be built to resist expected earthquake stresses. Building can be prohibited on unstable soils. Such land is better used for parks, highways, and other ``open space'' purposes. There is no need to wait for seismologists to perfect their forecasting skills.

A Tuesday column. Robert C. Cowen is the Monitor's natural science editor.