Discussion on the influence of truncation of ground motion residual distribution on probabilistic seismic hazard assessment

Recent studies on assessment of a very low annual probability of exceeding (APE) ground motions, 10 or less, have highlighted the importance of the upper bound of ground motions when very low probability results are acquired. The truncation level adopted in probabilistic seismic hazard analysis (PSHA) should be determined by an aleatory uncertainty model (i.e., distribution model) of ground motions and the possible maximum and minimum ground motion values of a specific earthquake. However, at the present time, it is impossible to establish the upper bound model for ground motions based on the source characteristics and/or ground motion propagation. McGuire suggested a truncation level be fixed at a number of epsilon = 6, or the distribution of residuals be truncated in such a manner that site intensity cannot be greater than the epicenter intensity. This study aims to find a reasonable and feasible truncation level to be used in PSHA when the physical mechanism is not available to find the extreme ground motion. A mathematical analysis of the influence of the truncation level on PSHA, case studies of sites in different seismotectonic settings, and a distribution analysis of ground motion residuals are conducted in this study. It is concluded that epsilon = 4 is the minimum acceptable value for engineering applications for APEs within 0.002 to 10(-4), and for low APEs, such as 10(-5) and 10(-6), the value of e should be no less than 5 in most regions of China.