Development and Verification of Non-Ergodic Ground-Motion Methodologies and Modeling Tools

Non-ergodic ground-motion models (NGMMs) have the potential of reducing the ground-motion aleatory variability significantly, which has a large impact on the seismic hazard, especially at large return periods important for critical infrastructure. This reduction in aleatory variability is accompanied by epistemic uncertainty in regions with sparse recordings or a systematic shift in the median ground motion in regions with dense recordings. Gaussian process regression (GPR)-with spatially varying coefficients for modeling the source and site systematic effects and cell-specific anelastic attenuation for modeling the systematic path effects-is a flexible and robust modeling technique used in this study for developing NGMMs. As part of this work, open-source computer tools and instructions have been developed to show the steps toward developing NGMMs in the GPR framework. Statistical software packages STAN and INLA are used and compared. The developed software packages were tested against synthetic data sets with known non-ergodic effects, and different implementations of the developed software were evaluated for scalability, universality, precision, and model complexity.