A typhoon disaster is a common meteorological disaster that seriously impacts natural ecology, social economy, and even human sustainable development. It is crucial to access the typhoon disaster information, and the corresponding disaster prevention and reduction strategies. However, traditional question and answering (Q&A) methods exhibit shortcomings like low information retrieval efficiency and poor interactivity. This makes it difficult to satisfy users’ demands for obtaining accurate information. Consequently, this work proposes a typhoon disaster knowledge Q&A approach based on LLM (T5). This method integrates two technical paradigms of domain fine-tuning and retrieval-augmented generation (RAG) to optimize user interaction experience and improve the precision of disaster information retrieval. The process specifically includes the following steps. First, this study selects information about typhoon disasters from open-source databases, such as Baidu Encyclopedia and Wikipedia. Utilizing techniques such as slicing and masked language modeling, we generate a training set and 2204 Q&A pairs specifically focused on typhoon disaster knowledge. Second, we continuously pretrain the T5 model using the training set. This process involves encoding typhoon knowledge as parameters in the neural network’s weights and fine-tuning the pretrained model with Q&A pairs to adapt the T5 model for downstream Q&A tasks. Third, when responding to user queries, we retrieve passages from external knowledge bases semantically similar to the queries to enhance the prompts. This action further improves the response quality of the fine-tuned model. Finally, we evaluate the constructed typhoon agent (Typhoon-T5) using different similarity-matching approaches. Furthermore, the method proposed in this work lays the foundation for the cross-integration of large language models with disaster information. It is expected to promote the further development of GeoAI.

Rainstorm disasters pose a significant threat to the sustainable development of urban areas, and effectively organizing diverse information sources about them is crucial for emergency management. In light of recent advances in knowledge graph theory and application technology, their notable knowledge integration and representation capabilities may offer support for dynamic monitoring and decision-making processes concerning urban rainstorm disaster events. However, conventional models do not adequately capture the spatiotemporal characteristics of these events. To fill this gap, we analyze the essence of urban rainstorm disaster events and divide their evolution into four stages, namely, pregnant, development, continuous, and decline stages. On this basis, a multilevel knowledge representation model is proposed from four layers, which are event, object–state, feature, and relationship layers, by analyzing the components and dynamic characteristics from the mechanism of urban rainstorm disaster events. The proposed model can not only express the comprehensive structure and relationships within urban rainstorm events, but also emphasize the evolution of disaster events through a series of ordered states. Moreover, we test the utility of the constructed knowledge graph through a case study of the Zhengzhou 720 rainstorm. The case study first validates that the selected machine learning models can extract the urban rainstorm disaster event information accurately by comparing them with some mainstream models. Then, it demonstrates that the knowledge graph is practical in the field of disaster knowledge representation, and disaster condition retrieval. Additionally, since the knowledge graph can show the evolution of a disaster event throughout its full life cycle, it can promote the understanding of the mechanisms of urban rainstorm disasters and pave the way for future applications of disaster prevention and reduction.