Probabilistic Seismic Hazard Deaggregation of Ground Motion Prediction Models

Probabilistic seismic hazard analysis (PSHA) combines the probabilities of all earthquake scenarios with different magnitudes and distances with predictions of resulting ground motion intensity, in order to compute seismic hazard at a site. PSHA also incorporates uncertainties in ground motion predictions, by considering multiple Ground Motion Prediction ("attenuation") Models (GMPMs). Current ground motion selection utilizes probabilistic seismic hazard deaggregation to identify the distribution of earthquake scenarios that contribute to exceedance of a given spectral acceleration (Sa) level. That calculation quantifies effects of the aleatory uncertainties in earthquake events, but does not describe the epistemic uncertainties from multiple GMPMs. This paper describes ways to calculate contributions of multiple GMPMs to Sa exceedance using deaggregation calculations. Deaggregation of GMPMs plays an important role in target response spectrum computation for ground motion selection, in a similar way as assigned logic tree weights of GMPMs do in PSHA computation. Just as the deaggregation of magnitude and distance identifies the relative contribution of each earthquake scenario to Sa exceedance, the deaggregation of GMPMs tells us the probability that the exceedance of that Sa level is predicted by a specific GMPM. We can further extend deaggregation to other ground motion parameters, such as earthquake fault mechanism, to more fully quantify the parameters that contribute to Sa values of interest. The proposed methodology for deaggregation of prediction models can be immediately applicable to other procedures which require multiple prediction models in an earlier stage of total prediction and a later stage of new target computation.