Bayesian parameter estimation for space and time interacting earthquake rupture model using historical and physics-based simulated earthquake catalogs

This paper presents a robust parameter estimation technique for a probabilistic earthquake hazard model that captures time and space interactions between earthquake mainshocks. The approach addresses the existing limitations of parameter estimation techniques by developing a Bayesian formulation and leveraging physics-based simulated synthetic catalogs to expand the limited datasets of historical catalogs. The technique is based on a two-step Bayesian update that uses the synthetic catalog to perform a first parameter estimation and then uses the historical catalog to further calibrate the parameters. We applied this technique to analyze the occurrence of large-magnitude interface earthquakes along 650 km of the central subduction zone in Peru, located offshore of Lima. We built 2,000-years-long synthetic catalogs using quasi-dynamic earthquake cycle simulations based on the rate-and-state friction law. The validity of the synthetic catalogs was verified by comparing their annual magnitude exceedence rates to those of recorded seismicity and their predicted areas of high interseismic coupling to those inferred from geodetic data. We show that when the Bayesian update uses the combination of synthetic and historical data, instead of only the historical data, it reduces the uncertainty of model parameter estimates by 45% on average. Further, our results show that the time-dependent seismic hazard estimated with the both datasets is 40% smaller than the one estimated with only the historical data