Ground motion intensity measures are used to represent various components of earthquake shaking intensity and frequency content in the form of simple parameters; examples include peak ground acceleration, Arias intensity, and pseudo-spectral acceleration (PSA). Ground motion models (GMMs) are developed to predict these intensity measures as a function of earthquake source, wave propagation path, and local geotechnical site conditions. GMMs are formulated to capture the underlying physics of source processes, wave propagation, and site response, with individual model parameters set based on various combinations of empirical ground motion data analysis and physics-based ground motion simulations. The majority of GMMs are conditioned for hard rock reference sites, with shear wave velocity (VS) = 3000 m/s, or with a time-averaged shear wave velocity in the upper 30 meters of the crust (VS30) = 760 m/s. Additional site amplification models are necessary in order to estimate GMIMs for other site conditions, including weathered rock and soil sites. As shear waves propagate vertically in the near-surface, the conservation of energy dictates that the wave amplitude must increase as the seismic velocity of the medium decreases. This amplification, or the so-called linear site effect, is usually parameterized using VS30, and sometimes site fundamental frequency or depth to bedrock, if available.

This thesis has two parts, according to subject matter. The first part of this thesis, consisting of Chapters 2, 3, and 4, focuses on seismic site characterization and site amplification in central and eastern North America (CENA) in the context of the Next Generation Attenuation-East (NGA-East) project. Chapter 2 presents a hybrid geology‐slope approach for VS30 estimation that utilized a new and expanded shear‐wave velocity (VS) measurement database for CENA. The proxy is conditioned on geologic category from newly considered large‐scale geologic maps, the extent of Wisconsin glaciation, sedimentary basin structure, and 30 arc-sec topographic gradient. Nonglaciated sites were found to have a modest natural log dispersion of VS30 (σln V= 0.36) relative to glaciated sites (σlnV = 0.66), indicating better predictability of VS30 for the former. These findings were used estimate the mean and standard deviation of VS30 for NGA‐East recording stations when measurements were not available. Chapter 3 presents empirical linear site amplification models conditioned on time-averaged shear wave velocity in the upper 30 m (VS30) for CENA, developed using a combination of least-squares, mixed effects, and Bayesian techniques. Site amplification is found to scale with VS30 for intermediate to stiff site conditions (VS30 > 300 m/s) in a weaker manner than for active tectonic regions. For stiff sites (> 800 m/s), I find differences in amplification for previously glaciated and non-glaciated regions, with non-glaciated sites having lower amplification. The models account for predictor uncertainty, which does not affect the median model, but decreases model dispersion. Lastly, Chapter 4 presents recommendations for modeling of ergodic site amplification in CENA, based primarily on results from the literature (including the model in Chapter 3), for application in the U.S. Geological Survey national seismic hazard maps. Previously, the maps have used site factors developed using data and simulations for active tectonic regions; however, results from NGA-East demonstrate different levels of site amplification in CENA. The recommended model has three terms, two of which describe linear site amplification: an empirically constrained VS30-scaling term relative to a 760 m/s reference, and a simulation-based term to adjust site amplification from the 760 m/s to the CENA reference of VS = 3000 m/s.

The second part of this thesis, consisting of Chapters 5 and 6, focuses on the development of a global GMM and site amplification model with regional adjustment factors for subduction zone regions as a part of the Next Generation Attenuation-Subduction (NGA-Sub) project. Chapter 5 presents global subduction zone GMMs for interface and intraslab events, with regionalized terms for Alaska, Cascadia, Central America. Mexico, Japan, South America, and Taiwan. The near-source saturation model, magnitude-dependent geometrical spreading, and magnitude-scaling break point are constrained using simulations and fault geometry, and the anelastic attenuation, magnitude scaling, and depth scaling terms are constrained empirically. The model is regionalized in the constant, anelastic attenuation, and depth-scaling terms, and the magnitude break-point. When applying the model to a region not considered in the study, we recommend using an appropriate range of epistemic uncertainty that captures regional variation. Chapter 6 presents a subduction-specific site amplification model, meant to be paired with the reference-rock GMM of Chapter 5. This site amplification model for subduction regions accounts for regional differences in VS30-scaling, and re-calibrates a widely used nonlinear site term for active tectonic regions.