Ground-motion predictions for California — Comparisons of three prediction equations

Abstract

We systematically evaluate datasets, functional forms, independent parameters of estimation, and resulting ground-motion predictions (as median and aleatory variability) of the Graizer and Kalkan (2015, 2016) (GK15) ground-motion prediction equation (GMPE) with the next generation of attenuation project (NGA-West2) models of Abrahamson and others (2014) (ASK14) and Boore and others (2014) (BSSA14) for application to earthquakes in California. This evaluation is performed in three stages: (1) by comparing attenuation, magnitude scaling, style-of-faulting effects, site response, response-spectral shape and amplitude, and standard deviations; (2) by comparing median predictions, standard deviations, and analyses of residuals with respect to near-field (within 20 kilometers [km] of the fault) and intermediate-field (50 to 70 km from the fault) records from major earthquakes in California, and (3) by comparing total, intra-event, and inter-event residual distributions among the GMPEs with respect to a near-source (within 80 km of the fault) subset of the NGA-West2 database covering 975 ground motions from 73 events in California ranging from moment magnitude 5 to 7.36. The results reveal that the scaling features of the GK15 GMPE and the ASK14 and BSSA14 GMPEs are, in general, similar in terms of distance attenuation but differ in terms of scaling with magnitude, style of faulting, and site effects. The original standard deviations of GMPEs are also different. For the near-source California subset, the three GMPEs result in standard deviations that are similar to each other. The mixed-effect residuals analysis shows that the GK15 GMPE has no perceptible trend with respect to the independent predictors.