Bayesian modelling of the temporal evolution of seismicity using the ETAS.inlabru R-package

TL;DR

This paper introduces a Bayesian approach using INLA for seismicity modeling with the ETAS model, implemented in the ETAS.inlabru R-package, enabling fast, scalable, and uncertainty-quantified parameter estimation for earthquake forecasting.

Contribution

The paper presents a novel Bayesian method with INLA for ETAS modeling, implemented in an R-package, improving speed, scalability, and uncertainty quantification over existing methods.

Findings

The method requires quiescent periods for reliable parameter estimates.

It is robust to stochastic uncertainties and initial conditions.

Including historic earthquakes improves inversion quality.

Abstract

The Epidemic Type Aftershock Sequence (ETAS) model is widely used to model seismic sequences and underpins Operational Earthquake Forecasting (OEF). However, it remains challenging to assess the reliability of inverted ETAS parameters for a range of reasons. The most common algorithms just return point estimates with little quantification of uncertainty, and Bayesian Markov Chain Monte Carlo implementations remain slow to run, do not scale well and few have been extended to include spatial structure. Here we present a new approach to ETAS modelling using an alternative Bayesian method, the Integrated Nested Laplace Approximation (INLA). We have implemented this model in a new R-Package called ETAS.inlabru, which builds on the R packages R-INLA and inlabru . Whilst we just present the temporal component here, the model scales to a spatio-temporal model and may include a variety of spatial covariates. Using a series of synthetic case studies, we explore the robustness of our ETAS inversion method. We demonstrate that reliable estimates of the model parameters require that the catalogue data contains periods of relative quiescence as well as triggered sequences. We explore the robustness under stochastic uncertainty in the training data and show that the method is robust to a wide range of starting conditions. We show how the inclusion of historic earthquakes prior to the modelled domain affects the quality of the inversion. Finally, we show that rate dependent incompleteness after large earthquakes has a significant and detrimental effect on the ETAS posteriors. We believe that the speed of the inlabru inversion, which include a rigorous estimation of uncertainty, will enable a deeper exploration of how to use ETAS robustly for seismicity modelling and operational earthquake forecasting.