Inferring Fault Frictional and Reservoir Hydraulic Properties From Injection‐Induced Seismicity

Characterizing the rheological properties of faults and the evolution of fault friction during seismic slip are fundamental problems in geology and seismology. Recent increases in the frequency of induced earthquakes have intensified the need for robust methods to estimate fault properties. Here we present a novel approach for estimation of aquifer and fault properties, which combines coupled multiphysics simulation of injection‐induced seismicity with adaptive surrogate‐based Bayesian inversion. In a synthetic 2‐D model, we use aquifer pressure, ground displacements, and fault slip measurements during fluid injection to estimate the dynamic fault friction, the critical slip distance, and the aquifer permeability. Our forward model allows us to observe nonmonotonic evolutions of shear traction and slip on the fault resulting from the interplay of several physical mechanisms, including injection‐induced aquifer expansion, stress transfer along the fault, and slip‐induced stress relaxation. This interplay provides the basis for a successful joint inversion of induced seismicity, yielding well‐informed Bayesian posterior distributions of dynamic friction and critical slip. We uncover an inverse relationship between dynamic friction and critical slip distance, which is in agreement with the small dynamic friction and large critical slip reported during seismicity on mature faults.