Abstract
The magnitude of great subduction megathrust earthquakes is controlled mainly by the number of adjacent asperities failing synchronously and the resulting rupture length. Here we investigate experimentally the long-term recurrence behavior of a pair of asperities coupled by static stress transfer over hundreds of seismic cycles. We statistically analyze long (c. 500 ka) time series of M8-9 analogue earthquakes simulated using a seismotectonic scale model approach with two aims: First, to constrain probabilistic measures (frequency-size distribution, variability) useful for hazard assessment and, second, to relate them with geometric observables (coseismic slip pattern, locking pattern). We find that the number of synchronized asperity failures relative to the number of individual asperity failures as well as the coefficients of variation of recurrence intervals and seismic moment scale with the logarithm of stress coupling between the asperities. Accordingly, tighter packed asperities tend to recur more periodically and with a more characteristic magnitude while more distant asperities show clustering of more variable sized events. The probability of synchronized failures seems to be controlled to first order by geometrical relations (i.e., spacing and offset of asperities). The effects of rheological properties are evident but it remains to be explored to which extent they vary in nature and how sensitive the system is to those.