This study is intended to discuss feasibility of scour detection for railway bridges by means of remote microtremor monitoring. The impact test that applies impact force to the bridge pier by dropping a hammer so as to measured vibrations of the bridge pier is a laborious and time-consuming method, and has difficulties in real time monitoring during flood. This study investigates feasibility of scour detection from microtremor monitoring as an alternative method for the conventional impact test. An in-service railway bridge has been being monitored utilizing a sensing unit including functions of vibration measurement, data processing and remote control. Bayesian modal analysis was adopted as a system identification method. Identified frequencies from the microtremor data before and after a flood event showed that target frequency was slightly decreased under increase of water surface during flood. The decreased frequency recovered once water surface level decreased. Scour assessment showed an extremely low probability of scour of the target bridge.