The Critical Point Model for Large Earthquakes Revisited (Invited)

The central tenant of the critical point model for earthquakes is that a large earthquake on a fault or network of faults is only possible when the crust is in a critical state defined by a long-range spatial correlations of high stress patches that allow a rupture, once nucleated, to grow into a major event. The earthquake then lowers the regional stress thereby moving the region away from criticality. Tectonic loading and stress transfer from adjacent regions then raise the stress until the region once again becomes critical and another large earthquake is possible. The hope has been that the approach to a critical state is reflected in the regional seismicity in terms of larger precursory events and some measure of a spatial correlation between events; it is the rational for what has become known as the accelerating moment release (AMR) approach to intermediate term forecasting. Although many examples of such precursory seismicity have been documented, their statistical significance is weak - mainly because the ARM is due to a relatively small number of intermediated events and their aftershocks. Recent seismicity studies using precisely located small events have found that most lie very close to major fault planes and define a core of uniform activity about 100 meters wide bordered by activity which decreases as a power law of distance extending a few hundred meters from the core. For the case of the San Andreas Fault near Parkfield CA, it is well established that a significant fraction of the seismicity occurs in tight clusters of periodically repeating events. Analysis of the temporal and spatial evolution of these clusters over 7395 days prior to the 2004 M6 event shows a systematic decrease in the repetition rate of new clusters with time, which can be interpreted as the gradual activation of more slowly loaded asperities until a critical fraction of fault area becomes activated. This new focus on individual fault zones has the advantage that it does not rely on the small number of intermediate events that limited prior regional analyses.