TOWARD NON-ERGODIC AND SITE-SPECIFIC PROBABILISTIC SEISMIC HAZARD ASSESSMENT: REQUIREMENTS FOR THE NEXT GENERATION OF STRONG MOTION NETWORK AND DATABASES.

Ground-motion models used in engineering seismology are usually calibrated on global databases that are usually created by mixing data from different regions. These models also assume that the ground-motion variability observed in a global dataset is the same as the variability in ground motion at a single site-source combination. This assumption is referred to as the ergodic assumption.

New data give a unique opportunity to remove the ergodic assumption and take into account regional source, path and site specificities. Using recent data analysis performed on the EUROSEISTEST valley (Greece) and global ground-motion datasets (Kiknet, Knet, NGA2 and the European strong-motion databases) we will show the impact of source parameters, site monitoring and site-characterisation on the uncertainty of the ground motion estimates and associated hazard curves.

Our results suggest that future strong-motion networks should use higher sampling rates (to better evaluate site-specific high frequency attenuations) and record both strong and weak motions (to evaluate single-station sigma). These results also quantify the impact of a better characterisation of source parameters (depth, fault maturity, source to site distances) ans site parameters on ground-motion models.

We finally will show how new networks and high-level strong-motion databases may help to built consistent ergodic PSHA at a regional scale and non-ergodic, site specific,  PSHA.