Issues related to typological fragility curves derivation starting from observed seismic damage

In this study typological fragility curves are proposed with a macro-seismic approach. To this scope, a stock of 56.338 residential masonry buildings struck by L'Aquila 2009 seismic sequence is analyzed, whose AeDES forms are archived within Da.D.O. platform, that is a web-gis database collecting the observed seismic damage data related to buildings surveyed after several Italian earthquakes. Moreover, issues significantly influencing the fragility curves derivation are in depth discussed. In particular, a criterion for the buildings stock completion adding undamaged and not surveyed buildings is proposed, based on the distributions known of the residential building typologies. Comparisons highlight that the database completion affects the resulting fragility curves, and in particular for low damage levels. Furthermore, it is shown how the fundamental parameters estimation, by using the Maximum Likelihood Estimation (MLE) method, is conspicuously influenced by the PGA intervals number (nint), that is an issue often ignored when fragility curves are derived. The numerical investigations show that, although a non-monotonic trend is observed, the fundamental parameters tend to converge to the asymptotic values as the nint of PGA increases, and that they are markedly dispersed when nint is low.