Influence of parametric nonlinear soil models on site response predictions

In the absence of site-specific laboratory test data, a variety of dynamic soil models can be employed in ground response analysis (GRA) to represent both the shear modulus degradation and the soil damping observed during cyclic shear stress loadings. However, significant discrepancies exists among those material curves and their implementation in nonlinear codes. In this paper, we investigate the sensitivity of GRA to a range of dynamic soil models implemented using a nonlinear Masing-based model and a modified equivalent- linear method using frequency-dependent soil parameters. Case studies were carried out considering soft soil deposits and broadband earthquakes, including downhole records at the Lotung experiment site (LSST) in Taiwan. Nonlinear predictions were found to be highly sensitive to the parametrization of dynamic soil parameters along with hysteretic functions, herein using the Masing rules approach, with a tendency to attenuate the ground surface accelerations even where moderate amplitudes of ground deformation occur.