Regional-scale forward modeling and inversion of 3D wave propagation and dynamic rupture processes with nonlinear mechanical models of rocks and soils

Niu, Zihua (2025): Regional-scale forward modeling and inversion of 3D wave propagation and dynamic rupture processes with nonlinear mechanical models of rocks and soils. Dissertation, LMU München: Fakultät für Geowissenschaften

Abstract

Earthquakes are catastrophic geohazards with severe impacts on life and property. Emerging evidence suggests that linear mechanical models are inadequate for fully explaining the origins and consequences of devastating earthquakes. There is a critical need for adequate physics-based rock models and efficient numerical algorithms to explore how nonlinearities affect our understanding of earthquake mechanisms, in-situ rock conditions along seismic paths, and the resulting ground motions. In this dissertation, I focus on enhancing the representation and simulation of nonlinear rock behaviors under dynamic loading across multiple scales, i.e., from laboratory rock samples to regional-scale earthquakes. The advancement incorporates physics-based nonlinear rock models derived from laboratory experiments, scalable software for physics-based earthquakes simulation from source to site on supercomputers, and innovative inversion algorithms aimed at accurately determining nonlinear parameters of both laboratory and in-situ rocks. The first part of the dissertation introduces two continuum damage models that aligns with observations of nonlinear behaviors in rock samples from two commonly utilized laboratory experiment setups. In the second part, I propose an algorithm based on the discontinuous Galerkin method to model wave propagation through nonlinear rock rheologies in 3D. This algorithm is designed for regional-scale simulations that involve complex geometries. I verify the algorithm against three sets of analytical solutions and confirm that the algorithm scales effectively on supercomputers. I demonstrate the applicability of the framework to simulating co-seismic wave speed changes and ground motions during the 2015 Mw 7.8 Ghorka earthquake in Kathmandu valley. In the third part, the numerical solver is extended to incorporate 3D dynamic rupture modeling. This extension allows us to illustrate the off-fault co-seismic damage patterns and the high-frequency seismic radiation. In addition, I demonstrate that in the tensile step-over configuration, localized damage zones extending from one fault can induce heterogeneous stress perturbations on a neighboring fault, thereby triggering nucleation. This capability enables detailed simulations of dynamic rupture processes, including off-fault co-seismic moduli reduction at regional scales. Based on the earthquake simulation software developed in this dissertation, I propose two Markov chain Monte Carlo (MCMC) sampling methods to perform Bayesian inversion of nonlinear material parameters. At the laboratory scale, the forward simulations are computationally inexpensive. I demonstrate, in the first part of the dissertation, that the Adaptive Metropolis MCMC algorithm can be utilized to illustrate the precision with which material parameters can be constrained from the existing experiment setups. At the regional scale, each forward simulation takes thousands of core hours on modern compute clusters. To accelerate the inversion, I propose, in the fourth part of this work, a more effective sampling algorithm, i.e., the Multi-level Delayed Acceptance (MLDA) MCMC algorithm. In the fifth part of this work, I then show how the MLDA algorithm can be optimized for nonlinear inversion of parameters in dynamic rupture models, specifically for the 2019 Mw 7.1 Ridgecrest earthquake, leveraging geological, seismic, and geodetic observations.

• Dokumententyp: Dissertationen (Dissertation, LMU München)
• Themengebiete: 500 Naturwissenschaften und Mathematik ; 500 Naturwissenschaften und Mathematik > 550 Geowissenschaften
• Fakultäten: Fakultät für Geowissenschaften
• Sprache der Hochschulschrift: Englisch
• Datum der mündlichen Prüfung: 5. Juni 2025
• 1. Bericht­erstatter:in: Igel, Heiner