Impact of attenuation models on probabilistic seismic hazard analysis: A case study for Bursa city, Turkey


Yunatci A., ÇETİN K. Ö.

8th US National Conference on Earthquake Engineering 2006, San Francisco, CA, United States Of America, 18 - 22 April 2006, vol.2, pp.993-1002 identifier

  • Publication Type: Conference Paper / Full Text
  • Volume: 2
  • City: San Francisco, CA
  • Country: United States Of America
  • Page Numbers: pp.993-1002
  • Middle East Technical University Affiliated: Yes

Abstract

The significance of epistemic uncertainty and aleatory variability associated with empirical ground motion estimation relationships on annual seismic hazard levels is demonstrated for Bursa, a highly urbanized city in Turkey, inhabiting approximately 1.2 million people and surrounded by strike slip faulting systems located within 15 km of the city center. Two Western U.S. and two local shallow crustal attenuation relationships are adopted to assess the effects of epistemic uncertainty on the estimated seismic hazard levels. Impact of randomness in attenuation relationship predictions was studied first by ignoring the standard error term of the individual relationship then by incorporating it into the probabilistic seismic hazard assessment scheme. Following conclusions are reached as part of these studies : i) estimated seismic hazard levels by global and local relationships are non-systematically different by a factor of more than 1.5, thus the choice of attenuation model or the weights assigned to them in the logic tree assessment becomes a sensitive part of seismic hazard assessment for near field sites, ii) inclusion of aleatory variability on both peak ground acceleration and spectral acceleration values increases the hazard levels by more than a factor of two, impact being relatively greater for local attenuation relationships, iii) such large differences, reaching in some extreme cases to a factor of more than 3, exceeds the factor of safety margins we conventionally adopt in design applications, suggesting the development of a single consensus attenuation model, at least reducing the epistemic uncertainty.