Over-Prediction of Extreme Motions by Ground-Motion Models

Abstract eng:
There is often questioning of the credibility of extreme earthquake ground-motion predictions that sometimes arise in probabilistic seismic hazard analyses. Several attempts to explain the most extreme ground-motion predictions have sought departures of the upper-tail residuals from the log-normal probability distributions that govern most of the data range. Such departures are usually found to be only marginal. The approach also has the disadvantage that extreme upper-tail residuals need not be associated with the strongest motions found in the dataset or predicted by the model. This paper reports a more direct approach. For the dataset from which the ground-motion model was derived, the actual numbers of exceedances of various accelerations (e.g. 0.5g, 0.75g, 1g…) were compared to their expected numbers found by summing the probabilities of their exceedance according to the model across the data points. The method is illustrated for models from New Zealand and Japan, for both peak ground acceleration and 5% damped response spectrum accelerations. Both models show similar trends, increasingly over-predicting the numbers of exceedances as the acceleration increases. For the Japanese model with a very large dataset of 4695 records from 269 earthquakes, the ratio of actual to predicted exceedances of peak ground acceleration values falls from close to one for accelerations up to about 0.3g to about 0.15 at 1.0g. If the over-prediction of rates of the stronger motions is shown to be a general feature of ground-motion models, the results are of profound importance for seismic hazard estimates for critical facilities.

• Export as: BibTeX | MARC | MARCXML | DC | EndNote | NLM | RefWorks
• Add to your basket