Bayesian Inference of Fault Slip Heterogeneity and Overlap: Application to the 2019 Ridgecrest Earthquakes and Afterslip

Regularization favoring spatial smoothness is often used to stabilize the non-unique finite-fault slip inversions. Yet, its impact on the physical interpretation of complex fault slip remains unclear. Here, we evaluate this effect in full and hybrid Bayesian frameworks by comparing an unregularized (BUR) approach that samples slip parameters on a discretized fault, with a regularized (BR) approach that samples smoothing hyperparameters before constrained slip inversions. Synthetic tests demonstrate how both methods perform under varying data accuracy, spatial coverage, slip heterogeneity, and prior choices. When data are sufficiently accurate and dense, BUR models preserve slip roughness and perform better than BR models, particularly at shallower depths and for high-gradient or irregular slip patterns. We apply both methods to InSAR/GNSS-based inversions of the 2019 M w 7.1 Ridgecrest earthquake and three-month early afterslip. Our analysis shows that a layered elastic structure, near-fault data, and their uncertainties are essential for accurately resolving shallow slip heterogeneity consistent with field measurements. Through a comparison with seven published models, we find similar along-strike moment distributions, while the BUR model attributes a larger portion (48.6±10.9%) of the total moment to depths of 6.7–10 km, implying locally high static stress drops. Analysis of co- and post-seismic slip shows that spatial overlap in respective moment is ~30% for 10%-peak-slip thresholds in the BUR model, well below the ~60% overlap in the BR model. These results indicate that the unregularized approach minimizes artifacts and better captures the characteristics and relationship of different fault-slip modes, aiding the study of fault rheology and heterogeneity.