The response of a site (its tendency to amplify or de-amplify ground motion) is estimated either with simulations (site response analyses) or empirical site factors. Simulations can be performed using linear, equivalent-linear (EL), or nonlinear (NL) methods. NL simulations are more complex and require elaborate constitutive models, but have proven to be an improvement over the EL method at shear strain levels greater than 0.1% (Kaklamanos et al. 2015). This strain level is often exceeded in soft soils in Eastern Canada, even for moderate ground motion. Empirical models for site factors are derived based on recordings of past earthquakes combined with 1D EL simulations. The limitations of EL simulations, especially for soft soils, and the lack of soft soil profiles in the recorded ground motion, make these site factors inaccurate and sometimes unsafe.

The objective is to provide better regression models for site factors by combining recorded data and NL simulations, which will reduce the uncertainty of the site factors, and correct the bias toward soft soil profiles. A database of NL 1D ground response analyses will be created. For several dozen soft soil profiles, similar to profiles encountered in Eastern Canada, ground response analysis will performed using input ground motions consistent with the seismic hazard of Eastern Canada. Based on this database and the data collected, a regression model will be derived for soft sites, and the bias of the existing methods will be assessed. A new metric able to characterize soil impedance will be researched, and the uncertainty of the site factors, based on soil type and degree of nonlinearity, will be characterized.

References:

Kaklamanos, J., Baise, L. G., Thompson, E. M., & Dorfmann, L. (2015). “Comparison of 1D linear, equivalent-linear, and nonlinear site response models at six KiK-net validation sites.” Soil Dynamics and Earthquake Engineering, 69, 207-219.