Structural modal frequency is a well-accepted indicator for the purpose of structural health assessment. However, it is known that this indicator is sensitive to changing environmental conditions (such as temperature and humidity). Thus, development of modal frequency-environmental condition relation is important for making reliable inference. This paper introduces a Bayesian learning approach for modal frequency-environmental conditionrelation development based on the one-year monitoring dataset of a reinforced concrete building. By introducing a sophisticated hyperparameterizaion strategy, Bayesian learning appraoch automatically search the optimal model during hyperparameter optimazation.The appraoch is then utilized for health assessment of a reinforced concrete building based on one-year daily measurements.