Constraining on- and off-fault nonlinear dynamic rupture parameters via hierarchical Bayesian inversion for the 2019 Mw 7.1 ridgecrest earthquake

We present a multilevel Bayesian inversion to quantify uncertainties and correlations among on- and off-fault dynamic rupture parameters for the 2019 M w 7.1 Ridgecrest earthquake. Our realistic rupture models incorporate strong velocity-weakening rate-and-state friction and off-fault plasticity. We invert for four parameters capturing along-strike variability in on-fault friction and off-fault plasticity. The inversion is constrained by multidisciplinary surface deformation data, including fault-parallel offsets from satellite imagery, high-rate GNSS time series, and static GNSS displacements, and is enabled by more than four million CPU hours of 3D dynamic rupture simulations. We find a strong correlation between on-fault frictional weakening and off-fault plasticity: increased inelastic deformation can compensate for more prominent velocity-weakening friction in the northwestern part of the fault. The preferred dynamic rupture models reduce velocity weakening from northwest to southeast along the main fault. This improves the fit with observed fault-parallel surface offsets, consistent with along-strike variations in fault maturity. The inversion favors low off-fault plastic cohesion, which improves the fit to observed surface offsets and is consistent with shallow damage zones imaged geophysically. We infer a shallow slip deficit (SSD) of 13.1% with a standard deviation of 5.1%. Notably, 55% of the models fall within the 8–32% SSD range estimated from previous kinematic inversions, implying that co-seismic off-fault inelasticity is crucial to explain the observed SSD. Our results demonstrate the feasibility of integrating 3D dynamic rupture simulations with multilevel Bayesian inversion to constrain nonlinear dynamic rupture processes and off-fault properties, toward probabilistic, physics-based earthquake source characterization and uncertainty quantification grounded in observational constraints.