Toward New Single-Station Sigma Models
Résumé
Single-station sigma values for more than 600 stations have been computed using the same method for all records and datasets from different tectonic regions (California, Taiwan, Japan, Turkey, Mexico, Switzerland and France). This work has been driven by the need to develop better estimates of standard deviation for non-ergodic probabilistic seismic hazard assessment in Switzerland (PEGASOS Refinement Project), and for new ground motion models being developed for eastern North America (NGA-east) and western North America (NGA-west2). The results show that the average single-station standard deviations are lower than the ergodic values of standard deviation. Moreover, the single-station withinevent standard deviations appear to have little variability across different tectonic regions. However, there is significant variability between stations and, in some instances, values of single-station standard deviation are higher than their ergodic counterpart. This variability seems to be (on the first order) controlled by path effects, and it can be modelled by probability distribution functions. Our results also show that the Vs30 dependency is weak, while a reduction of the withinevent standard deviation is apparent with increasing distance (and magnitude). We finally amalgamate all within-event residuals from the different tectonic regions in order to propose and test several within-event single-station sigma models (constant, distance-dependent, distance and magnitude-dependent).