Due to the increasing number of disasters, studies on responding to disasters using AI and big data technologies have received much attention. However, the diverse data collected during a disaster makes it difficult to utilize such data to understand disaster situations. Furthermore, state-of-the-art technologies are only limited to post-analysis disasters. To that end, in this paper, we first collect disaster data and generate a time-series temporal knowledge graph to establish relationships between different data types. Next, we discuss approaches to identifying critical keywords and analyzing disaster situations through graph exploration in real-time. As case studies, we select blackouts, typhoons, fires, and earthquakes to apply our approach, and the experimental results show that we can acquire disaster-specific information. Finally, we discuss how our approach can be applied to the government’s disaster management system or policies, thereby increasing the overall understanding of disasters.