Regionalized Ground-Motion Models for Subduction Earthquakes Based on the NGA-SUB Database

A set of global and region-specific ground-motion models (GMMs) for subduction zone earthquakes is developed based on the database compiled by the Pacific Earthquake Engineering Research Center (PEER) Next Generation Attenuation - Subduction (NGA-SUB) project. The subset of the NGA-SUB database used to develop the GMMs includes 3914 recordings from 113 subduction interface earthquakes with magnitudes varying between 5 and 9.2 and 4850 recordings from 89 intraslab events with magnitudes varying between 5 and 7.8. Recordings in the back-arc region are excluded, except for the Cascadia region. The functional form of the model accommodates the differences in the magnitude, distance, and depth scaling for interface and intraslab earthquakes. The magnitude scaling and geometrical spreading terms of the global model are used for all regions, with the exception of the Taiwan region which has a region-specific geometrical spreading scaling. Region-specific terms are included for the large distance (linear R) scaling, VS30 scaling, Z2.5 scaling, and the constant term. The nonlinear site amplification factors used in Abrahamson et al. (2016) subduction GMM are adopted. The between-event standard deviation piece of the aleatory variability model is region and distance independent; whereas, the within-event standard deviations are both region and distance dependent. Region-specific GMMs are developed for seven regions: Alaska, Cascadia, Central America, Japan, New Zealand, South America, and Taiwan. These region-specific GMMs are judged to be applicable to sites in the fore-arc region at distances up to 500 km, magnitudes of 5.0 to 9.5, and periods from 0 to 10 sec. For the Cascadia region, the region-specific model is applicable to distances of 800 km including the back-arc region. For the sites that are not in one of the seven regions, the global GMM combined with the epistemic uncertainty computed from the range of the regional GMMs should be used