Ⓔ A Site-Specific Seismological Model for Probabilistic Seismic-Hazard Assessment by Alin Radu and Mircea Grigoriu Abstract Seismic site characterization requires large numbers of ground-motion records, and such datasets are rarely available. Seismological models are generally used to enrich the existent sets of earthquake records with artificial ground acceler- ation time histories. The specific barrier model (SBM) is an earthquake source model that, as an integrating part of a seismological model, produces spectral densities of the ground-motion process as a function of the moment magnitude m and source-to-site distance r. It is calibrated to regional records and, therefore, may only provide partial information on the site seismic characteristics. We propose a statistical update of the seismological model, constructed with the SBM such that it accounts for the site ob- servations. A parametric probabilistic model is sought for the SBM spectral densities, and a Bayesian framework is used to statistically update it to local records. This new, enhanced version of the seismological model is used to simulate any number of ground-motion samples and perform site-specific probabilistic seismic-hazard analy- ses (PSHAs). A numerical example of the PSHA calculations using the statistically updated seismological model is shown for a site in southern California. Online Material: MATLAB scripts to calculate basis functions, and parameter update. Introduction Earthquakes can have a catastrophic impact on human lives, economy, and environment. The main goal of earth- quake engineering is the development of structures that expe- rience limited damage under moderate seismic events and do not collapse under large events. To achieve this goal, the input characterized by the seismic ground accelerations needs to be characterized accurately, particularly when dealing with large events. Such a characterization cannot be obtained solely from data because of the limited number of large seismic records at individual sites. Three options are available to enrich the set of records at a site: (1) use data from sites similar to that of in- terest and scale them to desired intensities (data-based seismic hazard), (2) calibrate probabilistic models for the ground ac- celeration process to actual records available at a site and view the samples of these models as likely ground accelerations (probability-based seismic hazard), and (3) use the informa- tion from site seismic records to calibrate seismological models based on geophysical considerations (physics-based seismic hazard). Douglas and Aochi (2008) present a compre- hensive classification of the seismological models in literature. Data-Based Seismic Hazard Douglas (2006) used empirical arguments to select ground acceleration records. Baker (2011) selected ground-motion records to match a response target accelera- tion spectrum conditional on the spectral acceleration ampli- tude at a period of interest. Computational methods for the conditional spectrum were described by Lin et al. (2013), and an efficient selection algorithm was proposed by Jayaram et al. (2011). Ozer and Akkar (2012) introduced a similar approach, in which a methodology was proposed to select records from a database by matching a target spectral displacement value. The selected seismic records were then scaled to obtain ground motions of various intensities. This procedure is of questionable value because it only changes the amplitude of the motion, but not its frequency content (Kafali and Grigoriu, 2010; Grigoriu, 2011). Probability-Based Seismic Hazard Postulated mathematical models are calibrated to actual records and used to produce artificial records. For example, Zentner and Poirion (2012) developed a model to produce ground-acceleration samples using the Karhunen–Loève (K-L) expansion. They estimated the marginal distribution and the second-order properties of the random variables in the K-L expansion using seismic records. Even though this model can be used to produce any number of records, the K-L expansion captures only the second moment properties 3054 Bulletin of the Seismological Society of America, Vol. 104, No. 6, pp. 3054–3071, December 2014, doi: 10.1785/0120140013