To obtain design time series, it is common practice to select empirical recordings of ground motion and modify them by scaling or by making them spectrum-compatible. The computed nonlinear response of a system can vary greatly depending on the records selected even after making the time series spectrum-compatible. A method for time series selection based on properties of the structure and ground motion, not simply magnitude and distance, is presented. The procedure is for use in nonlinear analyses that are intended to result in a median global nonlinear response given a design event.

In developing the time series selection procedure, the structure-specific properties are considered using a proxy of the nonlinear system. Using a suite of recorded and scaled ground motions as inputs, a regression analysis is performed to develop a model for the proxy response based on the properties of the record and the proxy. Candidate scaled time series are evaluated to find those that have key record properties near their expected values and that yield a response of the proxy that is near the expected response for a seismic event. Those scaled time series with responses near the expected value are selected as the optimum time series for defining average response even if the scale factors are larger than commonly accepted.

Results for applications to structural response and slope stability are presented. The resulting time series selection methods allow for wider magnitude and distance bins for identifying candidate time series, and reduce variability in the response of the system.