Improvement of seismic site response analysis for small- and large-strain motions

Small-strain damping profiles developed from geotechnical laboratory testing have been observed to be smaller than the damping inferred from the observed site amplification from downhole array recordings. On the other hand, several studies have shown that current site response procedures underestimate site amplification at strains larger than about 0.1%. The high-frequency spectral decay parameter (κ₀) of earthquake motions from soil sites is investigated and the use of κ₀ to constrain the small-strain damping profile for one-dimensional site response analysis is evaluated. A relationship was developed between κ₀ at the surface and both the 30-m time averaged shear wave velocity (V [lower case s] ₃₀) and the depth to the 2.5 km/s shear wave velocity horizon (Ζ₂.₅) based on the analysis of motions from 51 sites in Kik-net array in Japan and six sites from California. This relationship demonstrates that κ₀ increases with decreasing V [lower case s] ₃₀ and increasing Ζ₂.₅. An approach is developed that uses this relationship to establish a target κ₀ from which to constrain the small-strain damping profile used in one-dimensional site response analysis. Following the study of small-strain κ₀ and its constraint on small-strain damping profile. The effects of soil nonlinearity on the high-frequency spectral decay parameter, κ, is investigated using recorded ground motions at soil sites over a range of shaking intensities. Using more than 2,500 motions from 32 sites, the ground motion data shows that κ does not vary systematically with the induced shear strain but rather remains at its small-strain value. This observation indicates that the high-frequency components of motion are consistent with small-strain damping, rather than the strain-compatible damping used in site response analysis. An attempt of using κ or the frequency-dependent, equivalent-linear method (EQL-FD) is proposed to fix the issue of underprediction of site response for large-strain motions. The proposed large-strain site response correction procedures--- κ correction method, EQL-FD method, as well as the strength correction method are systematically evaluated by using six 1D borehole arrays from Japan. The recorded and predicted surface response spectra are compared, and the results show that the strength correction moderately improves the underprediction of site response from EQL analysis at frequencies greater than 3 Hz and strains larger than 0.2%. The κ correction and EQL-FD methods both further improve the site response predictions, with the κ correction providing the most unbiased predictions of site response