An evaluation of epistemic and random uncertainties included in attenuation relationship parameters
Attenuation relationships are generally built in a deterministic way: based on physical considerations, the overall process (including stability tests) focuses on obtaining median values (median from epistemic and random variability). According to this point of view the magnitude, depth and localization of data base earthquakes are supposed to be perfectly estimated.The objective of the paper is to present a stochastic method that allows to take into account these uncertainties of data base at the regression step of the attenuation relationship.This method is applied on a European Strong Motion Database with three steps: (i) an evaluation of data base uncertainties,(ii) a regression taking into account the fuzzy data base,(iii) an estimation of model uncertainties.Finally this paper (i) presents a method which allows to build appropriate attenuation relationships for PSHA and clearly shows that (ii) the usual methods used to build attenuation relationships overestimate random variability and underestimate epistemic variability, introducing bias in PSHA.
Attenuation Relationship random variability epistemic uncertainties
N. Humbert E. Viallet
EDF SEPTEN, Dept. of Dynamics & Earthquake Engineering, Lyon France Team Manager Dept. Of Dynamics & Earthquake Engineering, EDF SEPTEN, Lyon France
国际会议
14th World Conference on Earthquake Engineering(第十四届国际地震工程会议)
北京
英文
2008-10-12（万方平台首次上网日期，不代表论文的发表时间）