Predicting Aftershocks


Within a few days of the 1994 Northridge earthquake, Caltech seismologist Egill Hauksson was able to predict with surprising accuracy where years of aftershocks would occur.

There were hundreds of them in that first week, mostly centered in three areas: the northwestern San Fernando Valley near Chatsworth, the northeastern Valley near San Fernando and the Santa Susana Mountains between the Santa Clarita and Simi valleys.

Eight years after Northridge, the quake has produced nearly 16,000 aftershocks, including 57 in the magnitude-4 range and 11 of magnitude-5 or greater. Almost all of them were in the zones Hauksson noted in those first days.

Hauksson, working with a Swiss scientist, Stefan Wiemer, and Matt Gerstenberger of the Swiss Federal Institute of Technology in Zurich, has honed the method over the last eight years and has developed a mapping system that can predict where the most powerful aftershocks will occur within four days of a large quake.


The forecasts rely on precise monitoring of the first aftershocks. By tracking their number, size, distribution and location, Hauksson and Wiemer say they can plot the shakes and shimmies beneath the Earth’s surface on a map showing specific areas where there is a high probability of strong aftershocks.

The method, outlined in an article appearing in a special issue of the Bulletin of the Seismological Society of America today, represents a significant improvement over the current system of aftershock prediction, which now only forecasts the strength and frequency of aftershocks for a whole region.

“It’s a great idea to go from the statistics of aftershocks to mapping out probable sites and the strength of ground motions,” said Andy Michael, a geophysicist for the U.S. Geological Survey at its Western regional headquarters in Menlo Park. “They’re doing a great thing, but there is a question of reliability.”

Progress has definitely been made, said Tom Heaton, a professor of earthquake engineering at Caltech. He said the authors “have brought together existing technologies into a synthesis that gives our best state-of-the-art estimate of aftershock likelihood. The change is that it is specific as to location and offers a quantitative estimate as to the strength of the impending shaking.”


The prediction of aftershocks has long been a tricky problem because of its extreme complexity. Every earthquake is different, and a long series of aftershocks can traverse varied terrain of different soils, topography and fracture structures--all factors that can play an important role.

But such predictions are a crucial element in dealing with one of nature’s most violent and destructive forces. In the past, aftershocks have often posed as serious a threat as the main shock of an earthquake. For example, a magnitude-5.8 aftershock of the 1999 magnitude-7.4 quake in Izmit, Turkey, killed seven people and injured 420.

Another example of a fatal aftershock occurred a month and a day after the July 21, 1952, Tehachapi earthquake. This magnitude-7.7 temblor was followed on Aug. 22, 1952, with an aftershock near Bakersfield that killed two people, injured 35 and did $10-million in damage.

The strongest aftershocks of big temblors are typically about one order of magnitude lower than the initial quake, but they come when buildings are already weakened.


The authors said knowing where aftershocks are most likely to occur can help emergency planners deploy building inspection and recovery personnel to the most vulnerable areas.

A better understanding of how stress triggers aftershocks in certain directions from the “mainshock” has allowed the scientists, once they know the dimensions of the aftershock zone, to calculate where the strongest aftershocks will occur. The size of the original quake, plus the size of the initial aftershocks, allows them to calculate just how big they will be.

They are then able to fashion maps of “probabilistic aftershock hazard” that show, according to a color code, where the most significant aftershocks are most likely to strike.

The system has worked in the magnitude-7.3 Landers quake in 1992 and the 7.1 Hector Mine quake in 1999, but the authors readily acknowledge that it remains to be tested in other big quake sequences to be certain that it works consistently for all types of quakes.


The system can be employed only in places such as California that have extensive networks of quake monitoring stations already installed, since it requires a very close delineation of where the first aftershocks, even quite small ones, have occurred.

A key element of the predictive system is the “rate of decay” of the aftershocks, Hauksson said. The term refers to the falloff in the number of aftershocks within that first period, day by day.

This is an important clue to how many aftershocks may occur later and how long the aftershock sequence will last. (In the Northridge quake, the aftershocks are just now believed to be ending. There was a significant swarm of aftershocks in January.)

Some quakes have a more robust aftershock sequence than others. Based on the monitoring data, Hauksson and his colleagues believe they can make a reliable estimate of the 30-day hazard for further aftershocks, and they can predict beyond then as aftershocks provide more data.


The direction in which a fault ruptures is also an important indicator of where the most damaging aftershocks will occur. It is a factor known as the “directivity” of the quake.

Just because a ground rupture goes from south to north, as in the Landers quake, doesn’t mean that all the damaging aftershocks will be to the north. Many of the strongest Landers aftershocks were to the south of the epicenter. The most damaging, a magnitude-6.5 temblor in the Big Bear area, was to the southwest of the epicenter.

Similarly, in the magnitude-7.1 Hector Mine quake of 1999, the biggest aftershocks occurred in the opposite direction from the quake’s southward rupture.

The reason for this is not altogether clear to the scientists, but it is based on their observations of what happened after the Landers and Hector Mine quakes.


Both earthquakes were categorized as horizontal strike-slip earthquakes on vertical faults--a kind of temblor in which one side of the surface rupture slides in one direction and the other slides in another.

In the Northridge quake, the rupture was on what is known as a dipping-thrust fault--a slanting rupture that rises from deep inside the Earth like a kind of ramp.

Most of the aftershocks were to the north of the epicenter, in the same apparent direction as the rupture. But Hauksson said the proper way to view the earthquake would be to look at the fault as if it were turned on its side. From this perspective, the aftershocks occurred in the opposite direction of the rupture.

In the most striking illustration of this, the Weimer-Hauksson team has provided a 30-day aftershock hazard map they developed for the Hector Mine earthquake and superimposed on it the magnitude-4 and 5 aftershocks that occurred.


All but one of these large aftershocks occurred within the areas that had been identified as having the highest probability of such aftershocks.