A scoring test on probabilistic seismic hazard estimates in Italy

Abstract

Probabilistic estimates of seismic hazard represent a basic element for seismic risk reduction strategies and they are a key element of seismic regulation. Thus, it is important to select the most effective estimates among the available ones. An empirical scoring strategy is described here and is applied to a number of time-independent hazard estimates available in Italy both at national and regional scale. The scoring test is based on the comparison of outcomes provided by available computational models at a number of accelerometric sites where observations are available for 25 years. This comparison also allows identifying computational models that, providing outcomes that are in contrast with observations, should thus be discarded. The analysis shows that most of the hazard estimates proposed for Italy are not in contrast with observations and some computational models perform significantly better than others do. Furthermore, one can see that, at least locally, older estimates can perform better than the most recent ones. Finally, since the same computational model can perform differently depending on the region considered and on average return time of concern, no single model can be considered as the best-performing one. This implies that, moving along the hazard curve, the most suitable model should be selected by considering the specific problem of concern.