Accurate prediction of the exact time and location of earthquakes has long been a goal, but its realization is unlikely at least into the near future. Fault trenching studies and Coulomb stress modeling are important for understanding individual faults and also for long range planning and risk assessment, but they provide no data on actual ground motions that is crucial to planning in seismic regions. A drawback to many present mechanical models, as well as to trenching and Coulomb studies is that they cannot take into account blind thrusts, or faults not yet formed.

The ultimate goal of this study is to develop forward, 3D, distinct-element particle models of the San Francisco Bay Area to simulate patterns and locations of seismicity. This work describes equivalent 2D distinct-element particle models.

The distinct-element method treats a rockmass as a bonded - in tension and shear - assembly of frictional, elastic spheres. Loading forces bonds between particles to fail progressively, simulating rock fractures, and then thoroughgoing faults. For these 2D models fault traces at different depths are defined by fault surface models, made from historical hypocenter locations. Fault traces derived from these depth slices are introduced into the particle assembly as irregular lines in 2D that separate intact blocks between faults. The mechanical properties of the faults can be modified. The model is fully dynamic: Elastic strain energy release due to bond breakage, or fault slip results in earthquakes. Fault surface asperities within the models area sites of stress concentration and resist slip.

If fault traces defined from Bay Area seismicity are incorporated into these 2D and future 3D particle models, and if boundary displacements constrained by GPS, the model will likely produce seismicity at the locations of robust fault asperities. They may predict sites of future faulting, something that existing regional numerical models do not. One goal of this work is to produce an understanding of possible ground motions resulting from earthquakes in our model fault system. Accurate prediction of the exact sequence of future seismicity is very unlikely. These models may, however, give us a glimpse into possible future earthquake scenarios, and represent a state-of-the-art tool for seismic hazard identification and risk assessment.