The integrity and durability of civil infrastructure—especially bridges—are fundamental to public safety and network resilience. Weathering, environmental degradation, and growing traffic can progressively impair structural performance, so safety evaluations must rely on numerical models whose reliability depends on how faithfully they represent reality. Finite-element (FE) models are indispensable for this task, yet they inevitably suffer from uncertainties in material properties and boundary conditions, which can compromise their predictive value. Ambient-vibration testing (AVT) offers a non-intrusive means to overcome these limitations by supplying in-service dynamic data that, when integrated into a finite-element model-updating (FEMU) procedure, systematically calibrates the most uncertain parameters and reduces modelling uncertainty.

In this paper, an AVT-driven FEMU procedure is applied to a reinforced-concrete Gerber half-joint bridge built in the 1970s. Several calibration strategies are investigated, focusing on the representation of external and internal constraints and on material properties. Two baseline FE configurations are considered: (i) abutment supports idealised by translational springs and (ii) abutments modelled explicitly. After quantifying the uncertainty associated with each boundary-condition scheme, material parameters are updated through alternative multi-parameter strategies. Calibrating the spring-supported model reduces the discrepancy between numerical and experimental modal responses more than the explicit-abutment alternative, although the latter still provides acceptable accuracy and is simpler to manage. The study demonstrates that integrating AVT data with computationally efficient, physically consistent FEMU strategies provides reliable numerical models for bridge safety assessments.