Localizing and quantifying potential damage in large and complex structures is one of the most challenging problems in structural health monitoring. Vibration- based finite element (FE) model updating has proven to be a powerful methodology to identify (i.e., detect, localize and quantify) structural damage. However, due to the presence of uncertainties on the identified modal parameters used in model updating, probabilistic damage identification methods are preferable. In this paper, Bayesian FE model updating is applied for the damage identification of a full-scale seven-story reinforced concrete building slice tested on the UCSD-NEES shake table. The modal parameters identified based on ambient vibration data are used to identify damage as a decrease of the effective stiffness in a number of predefined substructures of a FE model of the structure.