[S-CG45] Science of slow-to-fast earthquakes

[SCG45-02] Sparse fault representation based on moment tensor interpolation

*Julien Thurin¹ (1.University of Alaska Fairbanks)

Keywords:Finite-fault, Source charaterization, Inversion, Numerical modeling

Accurate representations of large earthquake sources are crucial for understanding rupture processes and improving seismic hazard assessments. Although finite-fault models can represent complex slip distributions, they are often overparameterized and require strong regularization. Conversely, multiple point-source approaches offer considerable simplicity but struggle to reproduce near-field static displacements and high-frequency signals. We present a sparse fault parametrization that bridges the gap between classical finite-fault and point-source models by leveraging the geometric structure of the seismic moment tensor. Using a few “key” tensors as centroids of a continuous tensor field, we show that interpolating eigenvalues and eigenvectors separately preserve source type and provide smooth transitions between orientations. This interpolation framework enables us to (1) approximate detailed finite-fault models—which we demonstrate on the USGS NEIC slip for the 2024 Noto earthquake—with a few key moment tensors and (2) upscale these sparse models for forward simulations in 3D spectral-element solvers, preserving both the dynamic and static components of the wavefield. Our results confirm that this sparse representation can provide comparable displacement fields to finite-fault models despite requiring fewer parameters. These advances highlight the potential of sparse, moment-tensor-based models for rapid-response earthquake source characterization and high-fidelity forward modeling in complex 3D Earth structures.