Next Generation SDSU BBP Module Validation

Kim B. Olsen, & Rumi Takedatsu

Submitted August 15, 2016, SCEC Contribution #6998, 2016 SCEC Annual Meeting Poster #001

The SDSU module (BBtoolbox V1.5), merging low-frequency deterministic signals with high-frequency scattering functions (Olsen and Takedatsu, 2015), participated in and passed the SCEC Broadband Platform validation exercise 1.0, where the focus was on validating simulated median pseudo-spectral accelerations (PSA) for large earthquakes in western and eastern US and Japan, as well as NGA-west1 GMPEs. Here, we have started to prepare the SDSU module for the next generation seismic hazard analysis, by extending the validation to additional metrics important for structural engineering, namely signal duration, frequency-dependent ground motion variability (intra-event standard deviation), and correlation of spectral accelerations. Moreover, we ‘re-validate’ the SDSU module against median NGA-west2 horizontal ground motion levels.

We find that the fit to durations depends strongly on the specifics of how the values are calculated. For example, the SDSU module provides generally good fit to durations for 9 western events if evaluated by the current BBP alignment methodology for goodness-of-fit calculation, starting at 5% energy level and truncating the synthetic time series to fit the length of the data. However, such truncation is not feasible without seismic records, such as for seismic hazard analysis. Without the truncation, the durations from the SDSU module are generally too long, and we adjust the coda energy envelope of the synthetics to improve the GOF for the durations. In general, intra-event standard deviations for the Part A PSAs generated by the SDSU module using 50 realizations are in agreement with those from data. Preliminary tests show that the PSA correlation structure for the SDSU module deviates from data, generally too correlated for shorter periods and with too small correlation for longer periods. We propose a path toward better correlation structure for the SDSU module for future work. Finally, we modify the module to fit median PSAs according to NGA-west2 (SCEC Validation Exercise, Part B). Among other differences compared to NGA-West1, the NGA-West2 PSAs are generally lower for periods between 0.2 s and 1 s. Through adjusted magnitude- and region-dependent parameters for the shape of the source-time function convolved with the high-frequency scattering functions we obtain satisfactory fit for 50-realization median PSAs for Part B, mostly with similar or smaller bias for the Part A scenarios.

Citation
Olsen, K. B., & Takedatsu, R. (2016, 08). Next Generation SDSU BBP Module Validation. Poster Presentation at 2016 SCEC Annual Meeting.


Related Projects & Working Groups
Ground Motion Prediction (GMP)