SCEC Award Number 13016 View PDF
Proposal Category Individual Proposal (Integration and Theory)
Proposal Title Validating nonlinear site response prediction methodologies for SCEC Broadband Ground Motion Simulations
Name Organization
Domniki Asimaki Georgia Institute of Technology
Other Participants
SCEC Priorities 6 SCEC Groups GMP, EEII, GMSV
Report Due Date 03/15/2014 Date Report Submitted N/A
Project Abstract
We have conducted a validation study of three site-specific response models --linear, equivalent linear and nonlinear-- by comparing ground surface predictions to observations at 17 stations of the Japanese strong motion network KIK-Net, which have recorded several strong events including the 2011 Tohoku mainshock. We used velocity profiles as the only input to all models, while additional parameters such as quality factor, density and nonlinear dynamic soil properties were estimated from empirical geotechnical engineering correlations. Thus, differences in site-specific response predictions are attributed to the model capabilities rather than to differences in input parameters. We used validation metrics that are of interest in both seismology and engineering, including bias ratios of peak ground response and visual comparisons of elastic spectra. Although we found remarkable agreement of our nonlinear site response model predictions with observations, for both time and frequency domain metrics, the number of ground motions that induced strong nonlinear effects was relatively small. For the linear response, we clearly observed degradation of the goodness-of-fit (GOF) with increasing peak ground acceleration (PGA), for maximum induced strain larger than 10^{-2}. For the equivalent linear analyses, we found that although their GOF is comparable to nonlinear predictions in the frequency range 0.5-25Hz, they perform very poorly in the high frequency range. The capability of our nonlinear model to realistically represent high frequencies of large events is important for physics-based simulations that seek to achieve increasingly high frequency deterministic predictions.
Intellectual Merit The intellectual merit lies in the validation of a nonlinear site response model that has been developed in previous years with SCEC funding, and the development of an interactive computational module that enables the SCEC community to conduct nonlinear site response predictions. We have shown that our nonlinear model can realistically predict high frequency components for strong earthquakes, and can thus reliably introduce nonlinear site response in physics-based simulations that seek to achieve increasingly high frequency deterministic predictions.
Broader Impacts A key objective of this work was to integrate research, education, and outreach. As such, it had broader impacts on graduate education through the training of one graduate student in the research; and through the use of the computational module in the Geotechnical Earthquake Engineering classroom; and it also had broader impacts on the engineering and seismological communities by providing a tool that can replace the state-of-practice site response methodology (that is, equivalent linear analysis) with a model that can perform -at no additional cost- nonlinear site response analyses more reliably, particularly for large strains and high frequency components.
Exemplary Figure Figure 6 (also shown separately at the end of the report).
Caption: (top) Goodness-of-fit (GOF) of linear, equivalent linear and nonlinear predictions at station IWTH08 in the frequency ranges 0.5-25Hz and 10-25Hz; (bottom) PGA GOF for the Tohoku mainshock (indicated above by the red arrow) in the 0.5-25Hz and 10-25Hz frequency range. By contrast to the nonlinear model, the linear and equivalent linear predictions perform very poorly in the high frequency range.