A unified framework for adaptive fault modeling: Methods and applications

Accurate fault modeling is essential for understanding earthquake rupture processes and improving seismic hazard assessment. We present a unified framework that integrates geodetic data with multidisciplinary constraints, including relocated aftershocks, geological observations, and geophysical information, to adaptively model fault geometry and slip in diverse scenarios such as multi-segment and multi-event ruptures. The framework combines adaptive fault construction with a scenario-driven Bayesian joint inversion approach. Fault geometries are first built from prior constraints, such as surface ruptures and aftershocks, and then refined through probabilistic inference when such data are inadequate. To enhance computational efficiency, we introduce a Sequential Monte Carlo Fukuda-Johnson (SMC-FJ) strategy. This separates nonlinear parameters—including geometry, data weights, and smoothing factors—from linear slip parameters, which are conditionally solved via constrained least squares. Geometry updates follow a hierarchical adjustment scheme based on point, line, and structural units, enabling flexibility across rupture complexities. Synthetic tests and four case studies, including the 2022 Menyuan, 2023 Türkiye, 2022 Luding, and 2019 Ridgecrest earthquakes, demonstrate robustness under various constraints. For the Menyuan earthquake, full Bayesian inversion confirms that the fault geometry constrained by relocated aftershocks is highly accurate and needs only minor adjustment to match the observed surface deformation. The results further show that gradual changes in fault strike and dip modulated rupture arrest and postseismic stress accumulation, highlighting the critical role of inherited geometric structure in controlling rupture termination and delayed seismic activation.