Depth variation of coseismic stress drop explains bimodal earthquake magnitude-frequency distribution

[1] An essential part of seismic hazard analysis is the earthquake size-frequency relationship, used to estimate earthquake recurrence time, and thus probability. A key feature of those distributions is their bimodal character: small to moderate magnitude earthquakes follow the Gutenberg-Richter (GR) inverse power-law relation while large magnitude (characteristic) earthquakes are more frequent than anticipated from GR, following approximately a gaussian distribution around the maximum magnitude limited by fault geometry. Using a numerical earthquake simulator, we show that the temperature dependence of friction behavior and therefore the depth-variation of coseismic stress drop, derived from laboratory friction experiments, presents a simple and comprehensive explanation for the observed bimodal seismicity distribution. Characteristic earthquakes are the result of an abrupt increase in rupture width at the transition from small to large earthquakes.