Global search inversion of multiple data to image the rupture process of M6 to M9+ earthquakes

Abstract:

Earthquake source reveals spatial and temporal heterogeneities at all scales. Understanding the complexity of seismic sources, beside its importance from a theoretical standpoint, responds to the fundamental challenge of improving the calculation of the ground motion and tsunami inundation variability for future earthquakes, especially in the near-field where usually major damage occurs.

In the last 10 years, the surge of great earthquakes, the increasing number and quality of near-field observations and the exponential growth of computational resources boosted the seismologist capability of reconstructing the spatio-temporal evolution of rupture processes, even for strong earthquakes.

Here we present an inverse method we developed in recent years and applied to study several events occurred in different tectonic environment and spanning a wide range of magnitude, from strong to great earthquakes. Our inverse technique belongs to a well-known class of optimization methods based on simulated annealing algorithms. One of the key point of this inverse method relies on its capability of performing an efficient importance sampling of the model space, the latter being characterized by high dimensionality (>100), to build up an ensemble of good data fitting models. Then, the algorithm performs a statistical analysis of the model ensemble to extract stable features of the earthquake source model together with the associated uncertainty.

A further key point of the method is the possibility to jointly invert for different datasets (strong motion, static and high rate GPS, InSAR data, tide gauges and ocean bottom pressure gauges), each probing different characteristics of the source process and having different spatio-temporal resolution. Relevant case studies will be presented to discuss ongoing and future improvements.