The Effect of Earthquake Record Scaling Technique on Embankment Dam Response

The state of the practice in dynamic analysis of structures includes the selection of earthquake records based on the mean (or modal) magnitude and distance of the design earthquake, and linearly scaling of the selected records to the target spectral acceleration at the period of significance, or matching the selected records to the uniform hazard response spectrum (UHRS). A method is presented by Baker and Cornell (2005, 2006a, and 2006b) to develop the conditional mean response spectrum of a ground motion given a target value of the spectral acceleration at the period (or period range) of significance. The shape of the spectrum is dependent on the epsilon value, where epsilon is the number of standard deviations needed for a ground motion prediction equation to return the target value of spectral acceleration. The result is referred to as the conditional mean spectrum considering epsilon (CMS-ε). The developed response spectrum falls below the UHRS at periods other than the period of significance, and is a more appropriate target for earthquake record selection and scaling. This method also provides a means for multi-component ground motion scaling (Baker and Cornell, 2006a and Abrahamson, 2006). This method was implemented to develop period-specific scenario target spectra for use in the safety assessment of one of BC Hydro’s embankment dams on Vancouver Island, British Columbia, Canada. Earthquake records were selected and matched to the developed target spectrum using CMS-ε technique. Dynamic analyses were performed using the records spectrally matched to the CMS-ε target spectrum to evaluate the performance of the dam and to assess the applied load on the wall of the powerhouse downstream of the dam shell. The analyses were re-run using records linearly scaled to the UHRS at the period of interest and also using the records spectrally matched to the entire UHRS. All the analyses were run with and without applying the vertical ground motion excitation. This paper compares the results of the dynamic analyses using the three methods of ground motion record scaling and the effect of vertical ground motion on the results