On the Comparison of Seismic Ground Motion Simulated by Physics-Based Dynamic Rupture and Predicted by Empirical Attenuation Equations

ABSTRACT The improvement of ground-motion prediction accuracy is crucial for seismic hazard and risk assessment and engineering practices. Empirically regressed ground-motion prediction equations (GMPEs) are widely used for such purposes in decades. However, the inherent drawbacks of GMPEs, such as the ergodic assumption, lack of near-source observation, and insufficiency to deal with the spatial correlation issue, have motivated geophysicists to find better alternatives. Recent studies on well-recorded earthquakes have illustrated that physics-based simulation (PBS) methods can provide predictions that are comparable to or ever superior to GMPE predictions. The increasing interests in applying PBSs also pose the need to statistically compare these simulations against GMPE predictions or actual observations. We notice the limitations in previous studies focusing on the predictive capability check of PBS. This article is to illustrate how more reasonable check of PBS should be conducted. We consider GMPE works in generally judging the reasonability of PBS, but PBS has the advantage in characterizing the heterogeneity of ground motion of a moderate-to-large earthquake, especially when considering the complexities in fault geometry, regional stress fields, rock properties, surface of the Earth, and site effects. We would rather recommend that, in the future, different GMPEs are only used to preliminarily judge the reasonability of PBS scenarios; then the ground motions simulated by those reasonable PBS scenarios (not limited to one) are further used for the following seismic hazard and risk assessment.