Abstract: The development trend of disasters can be perceived through mining and analyzing the topics evolution of social media data in disasters. A method of studying the evolution of the topic communities based on the common word network is proposed, so the development trend of the disaster situation can be sensed. Firstly, according to the word frequency-inverse document frequency analysis, the key words related to the topics are selected and a social media co-word network with keywords as nodes is constructed. Thus, the topic community detection is performed on the social media co-word network based on the module optimization. Secondly, on the basis of verifying the topic detection, and the time window division, the evolution types of the topic communities in adjacent time windows are distinguished, and then the result of the topic community evolution is obtained. Finally, taking the 7.21 Beijing Heavy Rainstorm disaster event in 2012 as an example, the proposed method is used to analyze the collected microblog data. The experiment shows that the method can reflect the evolution process of the topics well. It can further reveal the development trend of the disaster, and help emergency managers understand the development process of disasters, so as to assist managers to take appropriate emergency response measures in appropriate time.