5th Geotechnical Earthquake Engineering and Soil Dynamics Conference: Seismic Hazard Analysis, Earthquake Ground Motions, and Regional-Scale Assessment, GEESDV 2018, Texas, United States Of America, 10 - 13 June 2018, vol.2018-June, pp.523-537
© 2018 American Society of Civil Engineers.This paper presents large scale simulation-based modelling of site amplification in two regions, Central and Eastern North America (CENA) and the Western Unites States (WUS). For CENA, a parametric study is developed with 1.2 million 1-D site response analyses (0.58 million of each linear elastic and nonlinear analyses) to represent the variability and uncertainty of site geologies, material properties, and ground motions in CENA and their impacts on site amplification. These simulations are used to develop a suite of linear (L) and nonlinear (NL) site amplification functions for the response spectrum (RS) and smoothed Fourier amplitude spectrum (FAS). The functions for each spectrum are composed of independent, additive linear, and nonlinear terms. While time averaged shear wave velocity in the top 30 m of a site (VS30) is a key input parameter, it is shown that including site period (Tnat) effects into the linear site amplification functions significantly improves these functions. Nonlinear amplification can be modelled as a function of VS30 and either rock outcrop peak ground acceleration (PGA) or spectral acceleration (SA) at rock outcrop as the driver of nonlinearity. The depth to hard rock plays a similar role to Tnat. A subset of models is developed from the CENA dataset for sites with strong, shallow VS impedance ratios resulting from hard rock present less than 30 m from the ground surface. A smaller but similar study of 30 thousand site amplification simulations representative of WUS site conditions is used to develop linear and nonlinear site amplification models which show similar amplification behaviors as the models developed for CENA. These large-scale simulations are enabled through the development of highly optimized, efficient nonlinear site response software, multi-core batch analyses, and efficient data structures to analyze the nearly two terabytes of simulation data.