Exploring uncertainties in probabilistic seismic hazard analysis through a stochastic source-to-site seismic model

Abstract eng:
Environmental loading is often a significant design consideration for civil engineering undertakings and it is often the case that those loading conditions dominate in terms of the uncertainties that need to be taken into account. This is particularly so when the inertial forces due to earthquakes need to be considered. This paper presents a way of quantifying uncertainties due to earthquakes by means of a stochastic source-to-site model (SSTSM) and compares the findings with those of existing empirically based methodologies. The approach described incorporates features from Random Vibration Theory (RVT) that provide for varying probabilities of exceedence and hence incorporates uncertainties into the SSTSM. The concept of model matching between the SSTSM and a mean Empirical Strong Ground Motion Prediction Equation (ESGMPE) enables the generalised uncertainty in all the significant variables to be calibrated. The significance of the distinction between epistemic and aleatory uncertainty is considered in the context of SSTSM. An example is given where this has been applied to a rock site in Western North America. The matched SSTSM provides a vehicle with which to investigate several issues that are currently of interest in the application of ESGMPEs to seismic hazard studies. One is whether it is appropriate to truncate the assumed lognormal distribution representing the uncertainty in the ESGMPE which has implications for selecting the fractile against which the design is to be delivered. The uncertainties that arise from the geotechnical effect of the local site response are explored using an approach based on transfer matrices and by examining probabilistically the properties of the spectral response. The study suggests that, within the limitations of the examples, large excursions of spectral accelerations are unlikely to occur and if they do, it is probable that it would be a single occurrence. The study also reveals some anomalies which are identified for future study. The analysis has been performed entirely in the Mathcad software environment.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket