Probability-based PGA estimations using the double-lognormal distribution: Including site-specific seismic hazard analysis for four sites in Taiwan

This study develops a probabilistic approach for examining site-specific seismic hazards using data from 438 Mw≥5.5 earthquakes and attenuation relationships local to the region around Taiwan. These are combined to generate semi-observed peak ground accelerations, in a form which can be modeled by a double-lognormal distribution. This model, which satisfies statistical goodness-of-fit, provides the relationship between the exceedance probability and a given ground motion level. The study includes site-specific seismic hazard analysis for four nuclear power plant sites in Taiwan. The results show that the seismic hazards at the four sites are not the same. While no seismic hazard analysis is without challenge, a troublesome trend appears that many applications of decision making are being influenced by the complexity of the calculation, instead of how well the fundamentals of the analysis are understood and can be verified. The study provides an analysis which is not overly complicated, is in good agreement with an empirical control, and offers transparency, traceability, and verifiability.

► Establish the statistics of PGA back-calculated using an intense earthquake catalog around Taiwan.
► Conduct statistical goodness-of-fit searching for a suitable probability model simulating the PGA.
► Results show that a double-lognormal distribution is suitable based on eight cases evaluated.
► Develop a probability approach using the double-lognormal distribution in seismic hazard assessments.
► The procedure was demonstrated during evaluations for four nuclear power plants in Taiwan.