Zoning versus faulting models in PSHA for moderate seismicity regions reliminary results for the tricastin nuclear site, France
Résumé
We are developing a site-specific logic tree approach for a probabilistic seismic hazard assessment at Tricastin, a nuclear site located in the south-east of France, where potentially active faults have been identified. The aim of the logic tree developed in this paper is to compare hazard levels computed according to three main hypotheses (branches) that consider alternative source models (zoning/faulting) as well as alternative seismicity models for the faults (Gutenberg Richter/characteristic earthquakes). A preliminary exploration of a reduced number of uncertainties, clearly shows that diffuse source zone models lead to hazard results for this site that are higher when compared to fault source models. This holds for short and long return periods and for all spectral values of the ground motion. We attribute this mainly to the source-to-site distance of the fault sources. Indeed, with the exception of a hypothetical blind "fault" modeled near the site, fault sources are all located more than 25 km from the site. Disaggregation results for the Tricastin site of a source zone scenario at 10-4 annual probability show that intermediate range magnitudes located at close distances to the site are indeed the controlling events at PGA but at higher spectral periods (0.5 s) greater magnitudes located at larger distances do contribute. Uniform hazard spectra of fault sources show also a significant pseudo spectral acceleration level at central spectral periods for return periods comparable to those of maximum magnitudes. Thus future work requires a more thorough investigation of fault source models. © 2004 OGS.