Karen R. Felzer, Thorsten W. Becker, Rachel E. Abercrombie, Göran Ekström, James R.
Rice

American Geophysical Union, Fall Meeting, 2001.

Many studies over the last decade have used the static Coulomb stress change produced by a mainshock to predict the locations of triggered earthquakes.  This method has shown some success, but often fails to predict the locations of 20% to 40% of the aftershocks of a given mainshock.  We use statistical Monte Carlo modeling to show that this amount of failure is consistent with the perturbation to the stress field provided by the aftershocks themselves.  Although most aftershocks are more than a magnitude unit smaller than their mainshocks, the ability of earthquakes of all magnitudes to affect large static stress changes
at short range, and the pronounced clustering of aftershock hypocenters, means that many aftershock hypocenters in a sequence may be primarily stressed by a previous aftershock rather than by the mainshock itself.   The exact percentage stressed by previous aftershocks increases with the activity of the aftershock sequence, the magnitude of the mainshock, and the time since the mainshock.  Our model predicts that two days after the average California M7 earthquake, for example, over 50% of new aftershocks are primarily in response to stress changes from previous aftershocks.  This means that the majority of
the new aftershocks are most likely to occur near previous aftershocks, and not necessarily within regions of Coulomb stress increase from the mainshock.  The same happens three days after the average M6, and three weeks after the average M5 mainshock.  Our statistical modeling uses Omori's Law for aftershock decay, the Gutenberg-Richter magnitude frequency relationship, Bath's Law, and the assumptions that earthquakes of all sizes are capable of generating aftershocks and that the timing of each aftershock is essentially determined by a single mainshock.  We apply our model to the 1999 M7.1 Hector Mine earthquake, which may be classified as an aftershock of the 1992 M7.3 Landers earthquake.  Our modeling shows that at the time of the Hector Mine earthquake, at least 82% of Landers aftershocks were occurring in response to stress perturbations from previous aftershocks, not the mainshock.  Correspondingly, the Hector Mine earthquake's hypocenter did not clearly fall within an area of static Coulomb stress increase from the Landers mainshock, but did fall within a tight cluster of previous Landers aftershocks.