An Exploration of the Nature of the Scatter in Ground-Motion Prediction Equations and the Implications for Seismic Hazard Assessment

This paper reviews the modelling of the scatter in ground-motion prediction equations, and also explores the nature of scatter in strong ground-motion, using an extended database of European accelerograms and newly derived predictive equations. Several statistical tests are used to confirm that the distribution of the residuals is genuinely lognormal over a certain range but appears to deviate from this model significantly in the tails of the distribution. An alternative model, the upper limit lognormal (ULLN) distribution is found to provide a good fit to the data; the parameters of this model include an upper-bound on the distribution, which is effectively inferred from the statis-tics of the data itself. However, it is clear that upper bounds established in this way do not provide a reliable ceiling on ground-motion amplitudes and physical constraints are required. Investigative tools that could be used to explore the physical limits of ground-motion amplitudes are discussed. In line with the underlying focus of the work, which is to explore the sometimes almost dominant influence that the scatter can exert on PSHA, and also to provide new data to prevent the overestimation of strong-motion parameters for design, the covariance in the scatter of different strong-motion parameters is also explored.