This paper presents a progressive numerical model validation of a bowstring-arch railway bridge based on the analysis of experimental data from different structural response measurements, namely, static deformations under environmental actions, modal vibrations, and transient dynamic responses under traffic loads. This work also addresses an integrated approach that uses structural health monitoring (SHM) measurements in combination with FE modelling to understand the structural behaviour of a long-span complex bridge. An in-situ, progressively-phased SHM system has provided a diverse set of data streams ranging from static and dynamic responses to the measurement of environmental and operational traffic loads. The first phase consists of defining a detailed baseline finite element (FE) model of the bridge, envisaging the initial condition of the structure immediately after construction, and its validation using modal parameters (natural frequencies and mode shapes) derived from an ambient vibration test. Since in-service deflections generally do not exercise the non-linear regime of the response of the bridge, the second phase focuses on the analysis of static response data and temperature measurements to validate the non-linear behaviour of the structural system, particularly at the bearing devices, under slow actions. The third and final phase addresses the dynamic analysis under traffic actions, which provides greater sensitivity in the detection of non-linear behaviour due to the effects of high amplitude actions induced by regular train loading profiles. In order to guarantee the accuracy of the baseline numerical model, particularly under temperature and traffic actions, it was necessary to use contact restrictions in some specific bearing devices. This improvement is in line with the structural changes detected in some bearing devices through visual inspections. As a result, an updated numerical model capable of reproducing the modal, static, and dynamic structural responses was achieved, showing a very good agreement between experimental and numerical data. In future applications, this updated numerical model will be useful for assessing the condition of the bridge under traffic loads, namely to identify damages and support the adoption of life cycle maintenance strategies based on the integration of SHM systems with (stochastic) load and failure models.