Natural disasters occur frequently and pose a huge threat to China. Disaster prevention, mitigation, and disaster relief are eternal topics of human survival and development. However, in the field of disaster relief and emergency response, the relevant data increase sharply while the critical knowledge of emergency is obviously lacking. The "data-information-knowledge" transformation capacity is insufficient to meet the urgent needs of disaster prevention and reduction. Firstly taking natural disasters as the core, and around four elements of natural disaster events, disaster emergency tasks, disaster data, and methods, this paper proposes a knowledge graph construction method by combining a top-down approach and a bottom-up approach. Then, concept layer of knowledge graph is built from top to down, and the conceptual framework is formed through ontology modeling. Data layer of knowledge graph is built from bottom to top, and the relationship between entities is established through data acquisition, knowledge extraction, fusion, and storage. Finally, a flood disaster emergency knowledge graph is built to verify the validity of the proposed method. The concept layer in flood disaster emergency knowledge graph defines the conceptual levels, the attributes and the semantic relationships of flood disaster events, disaster emergency tasks, disaster data, and methods. The data layer in flood disaster emergency knowledge graph realizes the extraction of entities and relationships from multi-source data. After the knowledge fusion process, 3 054 nodes and 12 689 relationship edges are obtained and stored in the Neo4j graph database. The flood disaster emergency knowledge graph realizes the transformation from multi-source data to interrelated knowledge.