Toward a New Probabilistic Framework to Score and Merge Ground-Motion Prediction Equations: The Case of the Italian Region

The ground-motion prediction equation (GMPE) is a basic component for probabilistic seismic-hazard analysis. There is a wide variety of GMPEs that are usually obtained by means of inversion techniques of datasets containing ground motions recorded at different stations. However, to date there is not yet a commonly accepted procedure to select the best GMPE for a specific case; usually, a set of GMPEs is implemented (more or less arbitrarily) in a logic-tree structure, in which each GMPE is weighted by experts, mostly according to gut feeling. Here, we discuss a general probabilistic framework to numerically score and weight GMPEs, highlighting features, limitations, and approximations; finally, we put forward a numerical procedure to score GMPEs, taking into account their forecasting performances, and to merge them through an ensemble modeling. Then, we apply the procedure to the Italian territory; in addition to illustrating how the procedure works, we investigate other relevant aspects (such as the importance of the focal mechanism) of the GMPEs to different site conditions. Online Material: Figures showing regression analysis for peak ground acceleration (PGA) values and location map, and earthquake catalog and summary table of parameters corresponding to each ground-motion prediction equation (GMPE) implemented.