On the Forecasting of Ground-Motion Parameters for Probabilistic Seismic Hazard Analysis

It is well understood that the range of application for an empirical ground-motion prediction model is constrained by the range of predictor variables covered in the data used in the analysis. However, in probabilistic seismic hazard analysis (PSHA), the limits in the application of ground-motion prediction models (GMPMs) are often ignored, and the empirical relationships are extrapolated. In this paper, we show that this extrapolation leads to a quantifiable increment in the uncertainty of a GMPM when it is used to forecast a future value of a given intensity parameter. This increment, which is clearly of epistemic nature, depends on the adopted functional form, on the covariance matrix of the regression coefficients, on the used regression technique, and on the quality of the data set. In addition, through some examples using the database of the Next Generation of Ground-Motion Attenuation Models project and some currently favored functional forms we study the increment in the seismic hazard produced by the extrapolation of GMPMs. [DOI: 10.1193/1.3525379]