Wavelet-based simulation of scenario-specific nonstationary accelerograms and their GMPE compatibility

In seismic hazard analysis ground motion prediction equation (GMPE) plays a pivotal role because it provides the statistical distribution of hazard parameter for a chosen seismic scenario. However, GMPEs in general do not provide nonlinear response statistics, and the latter should be ideally obtained by time-history analyses of a scenario-specific suite of motions. In the present study a new wavelet-based method is proposed to simulate scenario-specific ensemble of accelerograms with realistic variability of time-frequency characteristics. Firstly, a methodology is proposed to stochastically characterize the nonstationarity of a recording process from the energy arrival curve of the wavelet coefficients of the recorded ground motion. Then a new empirical scaling model is developed to estimate the instantaneous energy arrival, with model uncertainty. Further, a reconstruction method is formulated to simulate the scenario-specific ensemble of accelerograms from the estimated scenario-specific energy arrival curves. It is found that the simulated ensemble exhibits realistic variation of time-frequency characteristics and hence, it naturally becomes comparable with GMPEs (in terms of median estimates for response spectrum and strong motion duration) developed using the same database. Finally, an algorithm is proposed to tune the estimated energy arrival such that the ensemble of simulated motions can be made compatible with the target GMPEs, both in terms of median estimates and standard deviations. It is found that the GMPE-compatible ensemble, obtained for 5% damping PSV spectra, shows good agreement with respect to PSV scaling models developed for a wide range of damping ratio.