Conversations between earthquakes: Dynamics and delays of the 2019 Ridgecrest rupture sequence

Abstract The overwhelming observational difficulties and the complexity of earthquake physics have rendered seismic hazard assessment largely empirical. Despite increasingly high-quality geodetic, seismic, and field observations, data-driven earthquake imaging yields stark differences and physics-based models explaining all observed dynamic complexities are elusive. Here we present data-assimilated 3D dynamic rupture models which untwine California's biggest earthquakes in more than 20 years: the moment magnitude (Mw) 6.4 Searles Valley and Mw7.1 Ridgecrest, California, sequence breaking multiple segments of the same fault system. Our models use supercomputing to find the link between the two large earthquakes. We unify the uniquely high-quality strong-motion and teleseismic, field mapping, high-rate GNSS, and space geodetic foreshock and mainshock datasets with earthquake physics. We find that the regional structure, the ambient long- and short-term stress, as well as the dynamic and static fault system interactions, are conjointly crucial to understand the dynamics and delays of the sequence. Dynamic rupture of a statically strong yet dynamically weak fault system is driven by overpressurized fluids and low dynamic friction in our models. The observed earthquake complexity results from static and dynamic stress changes acting across a non-vertical quasi-orthogonal conjugate fault structure. We demonstrate that joint physics-based and data-driven illumination of the mechanics of complex fault systems and earthquake sequences is possible when reconciling dense earthquake recordings, 3D regional structure and stress models. We foresee that physics-based interpretation of big observational data-sets characterizing complex nonlinear systems will have a transformative impact on future geohazard mitigation.