Probabilistic seismic hazard analysis, ground-motion model development and accelerogram selection

Dr Peter Stafford

Although the framework of Probabilistic Seismic Hazard Analysis (PSHA) is now very mature, there remain a number of components that can be enhanced. This is particularly true for the case of vector-valued seismic hazard analysis (VPSHA), and for the translation of hazard results into optimal inputs for structural analyses. Over recent years I have undertaken work towards developing the inputs and software required for conducting VPSHA.

In particular, the necessary correlation coefficients among various intensity measures have been computed from newly developed ground-motion models. Models have been published for the following ground-motion measures:

• Fourier amplitude spectrum
• Arias intensity
• Pseudo-spectral acceleration
• Duration & numbers of cycles of strong motion
• Response spectral ordinates for damping ratios other than 5% of critical

With the expansion of global databases of high-quality accelerograms various updates to the proposed models have been made, or are underway. In most cases, empirical approaches are combined with theoretical models for earthquake ground-motion in order to develop the published models.

Recently, work has been undertaken on coupling the prediction of Fourier amplitudes and measures of ground-motion duration within a Random Vibration Theory framework to enable the development of response spectral prediction equations that are highly transparent in terms of their underlying physical origins. This new approach to ground-motion model development has significant implications for the portability of the developed models.

Following scenario identification from either a scalar or vector hazard analysis, it is necessary to select suites of accelerograms for downstream structural analyses. To that end, contributions have been made in the areas of:

• Identification of optimal numbers of accelerograms for median response prediction
• Identification of optimal numbers of accelerograms for estimating the distribution of drift response
• Definition of a hazard consistent, energy-based envelope function for temporal modulation of stochastic accelerograms and non-stationary filter characteristics for spectral modulation
• Development of a hazard consistent vector-based approach to stochastic simulation of accelerograms The most recent contributions in this area of seismic hazard analyses have been related to the development of new source-scaling relationships and advanced regression approaches for removing the ergodic assumption in empirical ground-motion modeling.