Site-specific seismic hazard analyses

Current seismic hazard analyses are generally performed using probabilistic methods. When dealing with a specific site, the typical methodology involves using a ground motion prediction equation (GMPE) to estimate the rock outcrop ground motion and associated variability, then the ground motion is propagated to the ground surface by site response analysis.;The site response process is inherently variable. Including this uncertainty in site response analyses without modifying the input ground motion uncertainty produces double counting of the uncertainty associated with site response. In this dissertation the total uncertainty is partitioned into its several contributing components, quantifying these components, and proposing methods to perform site-specific seismic hazard analyses without double counting uncertainties.;Four random field models were developed, and an existing one was fitted to a different database. These models can be used to generate shear-wave velocity profiles for site response analyses. Two types of models are presented, using Gaussian random fields, and using Markov Chains. The first ones showed better performance, and among those a stationary Gaussian model (stationary on rho) showed the best performance, and it is the simplest among the five models.;Three GMPE's were developed, one only from surface records, one from "at-depth" records, and a third one combining surface and "at-depth" records. The results show the iv same magnitude and distance scaling for the three equations. For stations that recorded a large number of records, total uncertainty was measured by the standard deviation of the observed minus predicted, and similarly for intra-event residuals. These statistics serve as lower bounds for site-specific seismic hazard analyses, note that these standard deviations are non-ergodic. The use of a GMPE capable of predicting bedrock and surface median ground motions, allows the partition of the components of the total uncertainty at the surface into those related to the bedrock ground motion and those to site response. These components and their correlations are presented.;Measurements at a site can potentially reduce the uncertainty in the ground motion prediction from that calculated using the ergodic assumption, to that observed at single sites. This implies a reduction on the order of 25%.