The magnitude 7.4 Izmit earthquake, which struck north-central Turkey on August 17, killed at least 15,000 people. Yet the catastrophe also helped to validate a relatively new technique in earthquake science, known as stress-transfer analysis, which may save lives in the future. The practitioners of this technique attempt to gauge the likelihood of earthquakes by studying how faults interact with one another over time and space.
When a segment of a fault ruptures, explains geophysicist Ross S. Stein of the U.S. Geological Survey in Menlo Park, Calif., the stress on that segment drops, but part of the released stress goes to nearby regions. This transfer--a consequence of the elasticity of the earth's crust--affects adjacent segments as well as other faults in the vicinity. Depending on each fault's location, orientation and direction of slip, its likelihood of rupture may increase or decrease.
Typically Stein and his colleagues find that the transferred stresses are quite small--only a few percent of the total stress that accumulates on a fault from one rupture to the next. Even so, when the group examined the seismic history of several regions of California, they found a marked tendency for earthquakes to occur selectively on those faults that had experienced a stress increase as a result of a prior earthquake nearby.
About three years ago Stein, USGS colleague James H. Dieterich and geologist Aykut A. Barka of Istanbul Technical University turned their attention to Turkey's North Anatolian fault. This 1,400-kilometer-long (870-mile-long) fault is the line along which the Anatolian microplate is rotating westward with respect to the Eurasian plate. Since 1939 a sequence of disastrous earthquakes has progressed westward along the fault, reaching the area east of Izmit in 1967. Earthquakes have also progressed eastward from the 1939 rupture, though in a less orderly fashion.
According to the group's analysis, most of the ruptures started at points on the fault that had experienced stress increases as a result of previous ruptures. They also found that the yet unbroken segments near Izmit had been subjected to higher stress as a result of the ruptures to the east of the city. They estimated a 12 percent probability that a magnitude 6.7 or larger earthquake would strike the Izmit area within 30 years. With the benefit of hindsight, this prediction might seem excessively cautious. In the notoriously controversial business of earthquake forecasting, however, it represents a modest success.
Unlike the North Anatolian fault, California's San Andreas fault is embedded in a dense network of other active faults. Geophysicist Steven N. Ward of the University of California at Santa Cruz uses the stress-transfer approach to model the behavior of this network. Within the safe confines of his computer, Ward allows the faults to rupture repeatedly over thousands of years, and he looks for spatiotemporal patterns in the resulting "earthquake movie." He finds that stress transfers between faults largely prevent the San Andreas fault from breaking in orderly, progressive sequences. The same phenomenon may explain why earthquakes along the eastern part of the North Anatolian fault form a less orderly sequence than they do to the west.
Still, significant patterns emerge from Ward's movie. A major rupture on the northern San Andreas fault, for instance, tends to decrease the likelihood of earthquakes on other San Francisco Bay Area faults for several decades. In fact, the Bay Area has enjoyed just such a period of seismic quiescence since the great San Francisco earthquake of 1906. But a recent increase in the number of small earthquakes, as well as the 1989 Loma Prieta earthquake, have signaled that the truce is coming to an end.