It is proposed in this paper a novel damage identification method of concrete-filled steel tubular arch bridge using data fusion based on information allocation theory. This method is actually a data fusion process, in which the statistic characteristics of the measured data, such as mean, standard deviation are calculated firstly, and then the bad points are excluded by the Grubbs Guidelines from the data set. After that, the measured data satisfied to certain precision are sent to a data fusion centre where the fusion computation is performed by employing a weighted average algorithm, and the damage location of a CFST arch bridge is determined by the peak value method finally. The efficiency and feasibility of the method proposed is validated by detecting both single- and multi-damage patterns of a two-span CFST arch bridge, and the effect of environmental on identification results is studied simultaneously. The results show that the proposed method is capable of detecting damage and has a good robustness.