How Can Fault Slip Inversions Be Reliable? Insights from Bayesian Analysis of the 2019 Ridgecrest Earthquakes and Afterslip

Xiong Zhao, & Junle Jiang

Submitted September 7, 2025, SCEC Contribution #14513, 2025 SCEC Annual Meeting Poster #TBD

Regularization of model parameters is frequently employed to mitigate the ill-posedness of optimization-based fault slip inversion. The Bayesian approach can reveal the full posterior model space without ad hoc regularization, albeit at a higher computational cost. Despite the widespread use of fault slip models, the impacts of model regularization, such as spatial smoothing or amplitude damping, on the physical inference of fault slip processes remain ambiguous. Here we systematically quantify the influence of regularization on inversion outcomes in full or hybrid Bayesian frameworks using a numerical sampling algorithm. The Bayesian unregularized (BUR) method directly samples slip parameters over a spatially discretized finite fault, while the Bayesian regularized (BR) method samples the regularization hyperparameters followed by constrained least-squares inversion of slip distribution. Through synthetic slip scenarios on a vertical strike-slip fault with dense geodetic data, we quantify how data (co)variance, the ratio of data spacing to fault patch size, slip complexity, and prior choices affect the inversion performance of both methods. In the shallower zone (above 8–12 km), where data has better resolving power, both BUR and BR recover overall slip features; however, BUR achieves higher accuracy and spatial resolution, while BR tends to produce spurious slip, particularly in uniform slip or multi-asperity scenarios. Applying the two methods to InSAR/GNSS inversions of the 2019 Mw 7.1 Ridgecrest earthquake, as well as early afterslip, yield similar slip patterns where observations are accurate and dense, with notable discrepancies at greater seismogenic depths. When we compare seven published coseismic slip models projected onto a common fault geometry, moment release over six primary slip zones, identified from the BUR model, collectively accounts for 65%–84% of the total moment in each respective model, with BUR on the higher end. This indicates the broad-scale consistency and fine-scale resolving capability of the BUR model. Our results facilitate quantitative interpretation of the spatial complexity and overlap of fault slip processes, as well as their implications for fault dynamics.

Key Words
Bayesian Inversion, Regularization, Ridgecrest Earthquakes, Fault Slip Model

Citation
Zhao, X., & Jiang, J. (2025, 09). How Can Fault Slip Inversions Be Reliable? Insights from Bayesian Analysis of the 2019 Ridgecrest Earthquakes and Afterslip. Poster Presentation at 2025 SCEC Annual Meeting.

Related Projects & Working Groups
Tectonic Geodesy