events will occur. And since even highly accurate predictions will not prevent widespread damage, an improved ability to forecast the consequences of catastrophic events, and hence to prepare for them, is at least as important as predicting them.
Earthquakes at their worst are extreme catastrophes. The 1556 Shaanxi, China, earthquake killed over 800,000 people in a matter of minutes. By some estimates the next large earthquake under Tokyo could cause trillions of dollars in direct economic losses. These consequences could be mitigated if earthquakes could be predicted over timescales short enough to allow an effective response. However, this goal remains elusive.
The nature of earthquakes makes them uniquely terrifying. No one sees an earthquake coming. It is a matter of seconds from the time shaking first becomes perceptible until it becomes violent. At any locale, damaging earthquakes occur infrequently on human timescales, which means that most people caught in a major earthquake have no previous experience. It is also profoundly disturbing when our usually stable reference frame, the planet beneath us, does not hold still. The unpredictability and sudden onset of earthquakes also mean that once an earthquake begins, it is generally too late to do much more than duck and cover. The combination of unpredictability, abrupt onset, rarity, and unfamiliarity means that the risk posed by earthquakes is difficult to manage, for both individuals and governments.
The goal of earthquake prediction is to specify where and when a significant earthquake will occur. Where future earthquakes will occur is largely understood, with some important exceptions (summarized in Beroza and Kanamori, 2007). Predicting when they will strike is much more difficult, though progress has been made and promising avenues of research have emerged. The term “significant” is a subtle but important part of the definition of earthquake prediction, and it brings up the important question of what controls earthquake size. Finally, even the word “earthquake” needs definition. Scientists use the term to describe the fault rupture that generates seismic waves. However, the public views an earthquake more broadly as both the faulting and the waves. We will use this more general definition and discuss prediction of the faulting event and the shaking that accompanies it.
Scientists have long recognized that some regions are seismically active, while others are not. By the middle of the 20th century, seismologists had produced a remarkably complete earthquake atlas (Gutenberg and Richter, 1954) that chronicled systematic features of global seismicity, but they lacked a framework to understand those features. The advent of plate tectonics soon changed that and also enabled the first steps toward earthquake prediction to be taken. For example, plate tectonics theory led to the recognition that Cascadia should be subject to large earthquakes, despite no history of earthquake activity. This expectation was confirmed by the discovery, using stratigraphic and other evidence, of a magnitude (M) ~ 9 earthquake in Cascadia in January 1700 (Figure 4.1; Atwater, 1987).
Plate tectonics holds that Earth’s lithosphere consists of large plates that move relative to one another at speeds of several centimeters per year (Question 5). Relative plate motion is accommodated on plate-boundary faults or, more typically, on complex fault systems. These faults are frictionally locked between earthquakes, causing ongoing plate motion to deform the crust around them, storing elastic strain energy in the process. Once friction is overcome and the fault starts slipping in an earthquake, this stored energy is converted to other forms, most notably energy radiated away from the fault as seismic waves.
Most earthquakes occur at plate boundaries, and the type of boundary plays a role in controlling the nature of earthquake activity. Extension at divergent boundaries is accommodated by normal faulting and formation of new crust through basaltic volcanism, with much of the deformation taking place aseismically. Horizontal motion across transcurrent plate boundaries takes place on strike-slip faults. Most transcurrent boundaries are oceanic, but when they traverse continents, they pose significant seismic hazard. All of the largest earth-