Structural health monitoring has gained more and more interest over the recent decades. As the technology has matured and monitoring systems are employed commercially, the development of more powerful and precise methods is the logical next step in this field. Especially vibration sensor networks with few measurement points combined with utilisation of ambient vibration sources are attractive for practical applications, as this approach promises to be cost-effective while requiring minimal modification to the monitored structures. Since efficient methods for damage detection have already been developed for such sensor networks, the research focus shifts towards extracting more information from the measurement data, in particular to the localisation and quantification of damage.Two main concepts have produced promising results for damage localisation. The first approach involves a mechanical model of the structure, which is used in a model updating scheme to find the damaged areas of the structure. Second, there is a purely data-driven approach, which relies on residuals of vibration estimations to find regions where damage is probable. While much research has been conducted following these two concepts, different approaches are rarely directly compared using the same data sets. Therefore, this thesis presents advanced methods for vibration-based damage localisation using model updating as well as a data-driven method and provides a direct comparison using the same vibration measurement data.The model updating approach presented in this thesis relies on multiobjective optimisation. Hence, the applied numerical optimisation algorithms are presented first. On this basis, the model updating parameterisation and objective function formulation is developed. The data-driven approach employs residuals from vibration estimations obtained using multiple-input finite impulse response filters. Both approaches are then verified using a simulated cantilever beam considering multiple damage scenarios. Finally, experimentally obtained data from an outdoor girder mast structure is used to validate the approaches. In summary, this thesis provides an assessment of model updating and residual-based damage localisation by means of verification and validation cases. It is found that the residual-based method exhibits numerical performance sufficient for real-time applications while providing a high sensitivity towards damage. However, the localisation accuracy is found to be superior using the model updating method.