There is no scientific consensus on how finite virtual faults should be modeled in the probabilistic seismic hazard analysis of diffuse seismicity. Often pragmatic implementation decisions are made as a matter of convenience or computational efficiency. To better understand how differences in modeling virtual faults impact seismic hazard, we evaluated several methods of modeling such faults. For our preferred model, we test the sensitivity of the hazard to differences in fault discretization density, number of random azimuths, and cutoff distance. We propose a new computationally efficient method of modeling the random orientation of virtual faults by holding the fault plane fixed in space and rotating a virtual site around the source grid point. We show that there can be relatively large differences in seismic hazard, depending on how virtual faults are modeled, and provide guidance on how virtual-fault parameters should be selected in order to achieve a specified degree of accuracy.

You do not currently have access to this article.