The ground‐motion median and standard deviation of empirical ground‐motion prediction equations (GMPEs) are usually poorly constrained in the near‐source region due to the general lack of strong‐motion records. Here we explore the use of a deterministic–stochastic simulation technique, specifically tailored to reproduce directivity effects, to evaluate the expected ground motion and its variability at a near‐source site, and seek a strategy to overcome the known GMPEs limitations.
To this end, we simulated a large number of equally likely scenario events for three earthquake magnitudes ( 7.0, 6.0, and 5.0) and various source‐to‐site distances. The variability of the explored synthetic ground motion is heteroscedastic, with smaller values for larger earthquakes. The standard deviation is comparable with empirical estimates for smaller events and reduces by 30%–40% for stronger earthquakes.
We then illustrate how to incorporate directivity effects into probabilistic seismic‐hazard analysis (PSHA). This goal is pursued by calibrating a set of synthetic GMPEs and reducing their aleatory variability () by including a predictive directivity term that depends on the apparent stress parameter obtained through the simulation method. Our results show that, for specific source‐to‐site configurations, the nonergodic PSHA is very sensitive to the additional epistemic uncertainty that may augment the exceedance probabilities when directivity effects are maximized.
The proposed approach may represent a suitable way to compute more accurate hazard estimates.