With the emergence of social media, user-generated content from people affected by disasters has gained significant importance. Thus far, research has focused on identifying categories and taxonomies of the types of information being shared among users during times of disasters. However, there is a lack of theorizing with the dynamics of and relationships between the identified concepts. In our current research, we applied probabilistic topic modelling approach to identify topics from Chennai disaster Twitter data. We manually interpreted and further clustered the topics into generic categories and themes, and traced their development over the days of the disaster. Finally, we build a process model to explore an emerging phenomenon on social media during a disaster. We argue that the conditions/activities such as collective awareness, collective concern, collective empathy and support are necessary conditions for people to feel, respond, and act as forms of collective behaviour.