Evaluation of Ground Motion Models for USGS Seismic Hazard Forecasts: NGA-Subduction

Date: 4/24/2019

Time: 09:00 AM

Room: Pike

The selection and weighting of GMMs introduces a significant source of uncertainty in probabilistic seismic hazard analysis (PSHA). The November 30, 2018 M7 Alaska earthquake sequence provides an excellent independent dataset of instrumental ground motion observations that we use to evaluate ground motion model (GMM) performance. In this study, we evaluate recently released NGA-Subduction GMMs, to inform logic-tree weights for future updates of the United States Geological Survey (USGS) National seismic hazard model project (NSHMP) seismic forecasts.

GMMs are selected and weighted based on analysis of total residuals and the log likelihood probabilistic scoring method (LLH; Scherbaum et al., 2004; 2009). The LLH method compares the mean and total standard deviation where the total standard deviation is determined from the individual standard deviations in the distribution of the intra- (within) and inter- (between) event residuals. We apply the LLH scoring method to a database of instrumental ground motions from recent earthquakes Mw 4-7 associated with the 2018 Alaska aftershock sequence available from the USGS real-time ShakeMaps. The recent ShakeMap data provides independent ground motion observations for PGA and PSA at 3.0, 1.0, and 0.3s allowing us to evaluate the predictive power GMMs implemented in the new USGS NSHMP-haz software system. The LLH and residual results provide information for selecting and weighting GMMs for the 2019 update of the Alaska seismic hazard model.

Presenting Author: Daniel E. McNamara

Authors