Ground Motion Uncertainty and Variability (single-station sigma): Insights from Euroseistest, Greece

Thursday, 18 December 2014: 10:50 AM
Olga Joan Ktenidou, ISTerre Institute of Earth Sciences, Saint Martin d'Hères, France, Zafeiria Roumelioti, Aristotle University of Thessaloniki, Thessaloniki, Greece, Norman A Abrahamson, Pacific Gas and Electric Company, San Francisco, CA, United States, Fabrice Cotton, Helmholtz Centre Potsdam GFZ German Research Centre for Geosciences, Potsdam, Germany and Kyriazis Pitilakis, Aristotle University of Thessaloniki, Civil Engineering, Thessaloniki, Greece
Despite recent improvements in networks and data, the global aleatory uncertainty (sigma) in GMPEs is still large. One reason is the ergodic approach, where we combine data in space to make up for lack of data in time. By estimating the systematic site response, we can make site-specific GMPEs and use a lower, site-specific uncertainty: single-station sigma.

In this study we use the EUROSEISTEST database (http://euroseisdb.civil.auth.gr), which has two distinct advantages: good existing knowledge of site conditions at all stations, and careful relocation of the recorded events. Constraining the site and source parameters as best we can, we minimise the within- and between-events components of the global, ergodic sigma.

Following that, knowledge of the site response from empirical and theoretical approaches permits us to move on to single-station sigma. The variability per site is not clearly correlated to the site class. We show that in some cases knowledge of Vs30 is not sufficient, and that site-specific data are needed to capture the response, possibly due to 2D/3D effects from complex geometry.

Our values of single-station sigma are low compared to the literature. This may be due to the good ray coverage we have in all directions for small, nearby records. Indeed, our single-station sigma values are similar to published single-path values, which means that they may correspond to a fully -rather than partially- non-ergodic approach. We find larger ground motion variability for short distances and small magnitudes. This may be related to the uncertainty in the depth affecting nearby records more, or to stress drop and causing trade-offs between the source and site terms for small magnitudes.