ALBERT

Description: INTRODUCTION The seismic design of structures sometimes necessitates the use of synthetic strong ground motion time histories (European Committee for Standardization 2003; International Code Council 2000). Toward this end, the stochastic method with a point-source representation of the seismic source is a fast and efficient way to generate synthetic time histories (e.g., Boore 2003). While other modeling procedures may be more refined and physically realistic (e.g., full finite-fault simulations), they generally require a larger number of input parameters for which calibration relations do not exist (see Douglas and Aochi 2008, and references therein). Even for well-recorded earthquakes, these parameters are difficult to develop calibration relations for because of (1) the observed variability in the results from earthquake source process inversion and its dependence on available data (e.g., Custódio et al. 2005) and (2) the overall lack of confidence in even the best estimates of the parameters controlling fault rupture (Monelli and Mai 2008; e.g., Monelli et al. 2009). The simple seismological models also have advantages over empirical ground motion predictive equations (GMPE) (Abrahamson and Shedlock 1997; Power et al. 2008), although the forms of the source, path, and site functions in the simple seismological models may be somewhat less flexible than empirical GMPEs when trying to fit existing data. Seismological models can provide information on the physical nature of the parameters controlling the strong motion while the GMPEs cannot (e.g., Boore 2003; Olafsson et al. 2001; R. Sigbjörnsson and Ambraseys 2003). Moreover, despite the increase in strong motion recordings over the last couple of decades, and presumably the corresponding increase in knowledge, there are still considerable differences among the median ground motions estimated using the various GMPEs. These differences are a measure of epistemic uncertainty and have not been...