Seismo Lab Brown Bag Seminar

Fault systems are inherently complex in their geometrical properties, including heterogeneous strike directions, discontinuities, and depth variations. Previous studies have shown that the complexity of fault geometry influences the initiation, arrest, and recurrence of earthquakes. To better estimate earthquake magnitudes and understand rupture processes, it is crucial to develop advanced earthquake cycle models that capture fault geometry details, although this often requires significant computational resources. Some models oversimplify by using two dimensions, thereby missing critical 3D variations in fault structure. Traditional accelerated 3D modelling methods by Fast Fourier transform have also been limited to a single planar fault.

We developed a 3D quasi-dynamic earthquake cycle model using the boundary element method, accelerated by Hierarchical matrices, to account for elastic interactions among multiple fault segments and heterogeneous stress fields over multiple earthquake cycles. This approach reduces computational complexity from O(N²) to O(N log N), where N represents the number of discretized fault elements. We have cross-validated our code with analytical solutions for static cracks (such as the penny-shaped crack and cracks with a cohesive zone) and compared it with numerical solutions for planar fault dynamics from the Southern California Earthquake Center SEAS benchmark community.

We applied this method on a step-over configuration under spatially uniform rate-weakening friction conditions and investigated how geometrical and frictional parameters influence slip behavior in a 3D model. By considering fault interaction among two faults, our results identify three slip regimes: periodic earthquakes, coexisting slow slip events (SSEs) and earthquakes, and complex earthquakes. Notably, a single planar fault under the same conditions only produces periodic earthquakes. We quantified the geometry complexity and its relationship with slip complexity. Furthermore, we reproduced moment-duration scaling for both fast and slow earthquakes, showing that the scaling for slow slip events is highly sensitive to the slip rate threshold used for event detection.

Finally, we applied our model to the 2023 Kahramanmaraş, Turkey earthquake sequence, reproducing the rupture patterns and interactions between faults.