hide
Free keywords:
-
Abstract:
Ground motion models (GMMs) are one of the fundamental ingredients of probabilistic seismic hazard analysis (PSHA). Usually, to capture epistemic uncertainties, alternative GMMs are adopted in PSHA applications and merged using the traditional logic tree/ensemble approach. In such approach, each GMM needs a weight that reflects its relative importance. Data-driven methods suggest using weights based on the performance of GMMs against actual observations (i.e. ground motion recordings). The classical score used for this performance evaluation is the average sample log-likelihood (LLH). LLH is defined as the average log-likelihood score of the GMM against observations. In this work, we introduce a method to move from the classical average to a weighted average, i.e., an average where the weights are not equal to 1/N (where N is the total number of observations). We based our weighted average on fragility functions, to assign the performance of the GMM according to the target infrastructure type that will be the subject of the PSHA (e.g., masonry buildings, earth dams, bridges, levees). We compared the proposed approach with the classical LLH using instrumental data from Central Italy. Finally, we discussed the advantage and drawbacks of our approach in PSHA applications and outcomes. Applying this novel approach provides different values of the relative importance of each analysed GMM and changes ranking positions in a meaningful and repeatable manner.