Accounting for spatial correlation in the empirical scoring of probabilistic seismic hazard estimates

A procedure is here proposed to score probabilistic seismic hazard estimates provided by ergodic models by a comparison with accelerometric observations available during a control period. The procedure allows accounting for the potential bias induced by spatial correlation of hazard estimates due to the use of ground motion prediction equations in the modeling. Since this bias depends on the specific topological features of the studied system (relative positions of seismic sources and accelerometric control sites), numerical simulations are considered to correct it. An application of the above procedure is proposed for Italy where some probabilistic seismic hazard estimates have been provided in the last years, and where 76 accelerometric sites have been operating for more than 25 years. It is shown that, at least as the case study concerns, the amount of spatial correlation in the hazard estimates is relatively weak, it depends on the GMPE used in computations and increases with the exceedance probabilities and with the period of the acceleration response spectrum.