Stress transfer theory and earthquake prediction on the North Anatolian Fault

The Aegean- Anatolian Plate is moving towards the west at an average rate of around 25 mm/y relative to the adjacent Eurasian Plate. This motion takes place along the 1600 km long North Anatolian Fault (NAF), and there have been many large earthquakes in this area, records of which go back at least as far as the 10th century. The NAF is one of the best understood fault zones in the world, for several reasons, including the following:

Coulomb Stress Transfer Theory, which was first applied to earthquakes in the late 1980's (eg. Roth, 1988; King et al., 1994), has been applied with increasing success to the understanding of earthquakes in recent years. The principle behind the application of stress transfer theory is an understanding of the stress regime within the rocks adjacent to a dynamic plate boundary, and the ability to create a mathematical model of how the stress changes when there is sudden movement (an earthquake) along some part of a fault at the boundary. While failure on one part of a fault plane will tend to reduce stress in the rocks surrounding the rupture zone, it is likely to increase the stress in some other part of the fault plane, or on other nearby fault planes. The immediate result of this stress transfer is the generation of aftershocks. Most of these occur within seconds of the original shock, but some are delayed for minutes, hours, days, months and even decades.

There are two sources to the stress which exists between plates along a fault zone. One of these is the steady build-up of stress caused by the constant motion of the plates on either side of a locked boundary - known as 'secular' stress. The other is the abrupt change in stress - either positive or negative - related to a rupture along some part of the boundary. Secular stress increases more or less evenly on all parts of the fault zone over time, and hence an understanding of this factor alone does not normally provide much insight into where an earthquake may happen next. Stress changes related to fault rupture affect specific areas of the fault zone in different ways, and if the physics of the fault is sufficiently well understood it should be possible to determine which areas have been subjected to stress reduction and which to stress increase following an earthquake. Areas of stress increase could be considered to be more susceptible to an earthquake in the future. If the stress increase itself is enough to overcome friction on the fault plane than a triggered earthquake could occur very soon (seconds, days etc.). If the stress increase is not sufficient to overcome the friction, then a triggered earthquake will only happen when enough additional secular stress has accumulated (years, decades etc.). Whether it occurs within seconds or decades, such a triggered earthquake can be considered to be an "aftershock" in the sense that it is the direct result of a previous earthquake.

The westward-progressing series of earthquakes which occurred along the NAF between 1939 and 1992 gave geoscientists a tantalizing hope that they might be able to predict future quakes in this area if they had a clear understanding of the fault geometry, reliable data on the displacement associated with the historical earthquakes, and good measurements of the rate of plate motion. Recent work by various Turkish geologists (notably Aykut Barka and others from Istanbul Technical University) resolved some of the outstanding issues, and the new data were incorporated into a 1997 paper by Stein and Dieterich (of the USGS) and Barka. Based on a modeling of stress transfer it was concluded that there was a significant probability of a large earthquake in the Izmit area within the next 30 years. This prediction was fulfilled sooner than expected, with the devastating M 7.4 earthquake of August 1999.

The August 1999 earthquake is described in a short September 1999 paper by Barka. The paper includes a map of the extent of the rupture zone, and shows two areas in which stress was increased by the earthquake. One of these areas is near to Dzce, which was hit by a M 7.2 earthquake on the 12th of November 1999.

The pattern of historical earthquakes and the present stress regime of the western part of the NAF have been examined in detail in a March 2000 paper by Hubert-Ferrari and others. A westward-progressing series of earthquakes started at the eastern end of the NAF in 1668 and reached the Izmit area by 1719. The 1719 earthquake, which had many similarities to the August 1999 Izmit quake (extent, magnitude, area affected), was followed by a 1754 quake in the Marmara Sea, immediately south of Istanbul, and then by two other quakes in 1766 farther to the west. There has not been a major earthquake on the segment of the NAF south of Istanbul since 1766, and it is estimated that there is currently 5.5 m of accumulated strain on this part of the fault. The 1999 earthquakes resulted in an increase in the stress on the segment south of Istanbul, and it now seems very likely that this city of 20 million people will be hit with a major and devastating earthquake within the next few decades.


Barka A., The 17 August 1999 Izmit Earthquake, Science, V. 285, p. 1858-1859, (September 1999).

Hubert-Ferrari A., Barka A., Jacques E., Nalbant S., Meyer B., Armijo R., Tapponnier P. and King G., Seismic hazard in the Marmara Sea Region following the 17 August Izmit earthquake, Nature, V. 404, (March 2000)

King G., Stein R. and Lin J., Static stress changes and triggering of earthquakes, Bulletin of the Seismological Society of America, V. 84, p. 935-953 (1994)

Roth F., Modelling of stress patterns along the western part of the North Anatolian fault zone, Tectonphysics, V. 152, p. 215-226 (1988)

Stein R., Barka A. and Dieterich J., Progressive failure on the North Anatolian fault since 1939 by earthquake stress triggering, Geophysical Journal International, V. 128, p. 594-604, (1997) [also available on line]

Steven Earle, 2000. Return to Earth Science News