Using full scale vibration testing as a means for structural identification (St-Id) of long span bridges represents an efficient technique for quantitatively characterizing the in-service mechanical attributes and behaviors of these complex systems. However, there are various uncertainties, known and unknown, involved in the experimental and identification processes that impact the reliability of St-Id and serve as a barrier to its more widespread acceptance and application in engineering practice. A full-scale vibration test on a long-span suspension bridge is taken as an example to illustrate a number of possible approaches for coping with the challenges in practical identification of the dynamic characteristics of large-scale constructed systems. This paper examines uncertainty mechanisms in the following areas: (1) experiment design and implementation, (2) data pre-processing, (3) data post processing and modal parameter estimation. The identification results obtained for the bridge spans and towers demonstrate that the developed field testing and data processing methods may provide a reliable solution for bridge characterization.