It is a normal practice to consider the reduction of modal frequencies as an indicator for structural damage but some long-term monitoring studies revealed that the structural modal frequencies exhibited substantial seasonal variation without evidence of structural damage. Therefore, in order to perform reliable and consistent assessment on the health condition of civil engineering infrastructures, it is an indispensable task to quantify the modal frequency fluctuation induced by change of ambient conditions, such as temperature and humidity. In other words, it is necessary to distinguish the inevitable ambient interference from the abnormal changes due to structural damage and deterioration. In this paper, one-year daily measurement of a 22-storey reinforced concrete building is utilized to trace the variation of its modal frequencies, which are identified using the Bayesian spectral density approach with the ambient vibration data. Then, the Timoshenko beam model is considered to construct the mathematical model for the modal frequency-ambient condition relationship. Since the Timoshenko beam model incorporates both the shear and flexural deformation, it is appropriate for the behavior of buildings with different aspect ratios and configurations. Finally, Bayesian analysis is conducted to quantify the uncertain parameters in the modal frequency-ambient condition model. Results show that explicit consideration of the ambient temperature and relative humidity is essential for long-term structural health monitoring.