Spatio-temporal evolution of earthquake potential constrained by a physical and statistical approach: Application to the Chilean subduction zone

To become very large earthquakes, seismic ruptures that saturate the seismogenic width (M>8.3 in subduction zones) need to propagate long distances along-strike. Multiple factors can hinder this propagation, among them the available energy on the fault. A recent extension of Linear Elastic Fracture Mechanics theory to elongated ruptures provides a framework to estimate when a portion of a fault has enough potential energy, and is hence sufficiently loaded, to generate a large earthquake. Based on this framework, we present a method that takes into account the along-strike distribution of available energy to evaluate, using a probabilistic approach, the timing and magnitude of potential future large earthquakes, and thus the seismogenic potential of the fault. This approach assumes that the ruptures have already saturated the seismogenic width of the fault. We apply and assess this method on the Chilean subduction zone. We first perform a sensitivity test and explore the impact of the uncertainties of model parameters on the timing Tc at which a section of a fault is ready to host large ruptures. This initial test shows that Tc is controlled by the uncertainty of the parameter B, a coefficient involved in the scaling between fracture energy and final slip, which controls the energy consumed by the rupture. We further constrain B by comparing the observed interevent time between ~M9.5 earthquakes on the Valdivia segment and the one predicted from our model, assuming that such earthquakes occur as soon as the fault is ready to host it. Fixing B to this constrained value, we then estimate the evolution of the probability of earthquakes exceeding M8.5 over the whole Chilean subduction. Along-strike heterogeneity of the available energy arises from the heterogeneity of the loading rate, based on an geodetically-inferred coupling map, and from the along-strike changes of the seismogenic width. Our results highlight that the earthquake potential on a specific segment can be significantly altered by the occurrence of earthquakes on neighboring segments. This is illustrated by the drops in the probability of >M8.5 events on the Copiapo segment after the 2010 Maule and 2015 Illapel earthquakes. By combining our estimates with the rate of events that saturate the seismogenic zone, we are able to estimate the probability of occurrence of >M8.5 events. Such physics-based modeling is a novel approach to time-dependent seismic hazard analysis.