Uncertainty in intraevent spatial correlation of elastic pseudo-acceleration spectral ordinates

Abstract

The probabilistic nature of seismic ground motion intensity measures such as peak ground acceleration and spectral acceleration ordinates has been extensively studied during the last decades. However, their spatial correlation is mostly considered without any event-to-event variability, using a mean estimate from a number of seismic events. The present study quantitatively evaluates the event-to-event uncertainty of intraevent spatial correlations, using 39 well-recorded earthquakes. Results indicate a high event-to-event variability in the correlation model parameters, which if taken explicitly into account, would improve regional hazard and risk analyses. Event magnitude was found to be a statistically significant predictor variable of the model parameter, however it explains less than 20% of the total event-to-event variability. Moreover, clustering of site conditions, tectonic region, and fault mechanism are not statistically significant as predictor variables of the spatial correlation model parameter. Finally, this paper proposes a simple Monte Carlo approach for considering the high event-to-event variability of spatial correlation models, taking advantage of the Markov dependence of residuals for reducing the number of correlated variables to be simulated. This approach can be used with different intraevent spatial correlation models, as long as proper estimates of the dispersion of their parameters are considered.