Urban flood vulnerability Knowledge-Graph based on remote sensing and textual bimodal data fusion

Flood disasters inflict extensive, serious damage on cities and society, significantly constraining urban construction and development. There is an urgent demand to reduce urban flood vulnerability, to explore the evolution mechanism of urban flood vulnerability, and to guide the construction and improvement of urban flood control resilience. An objective and accurate knowledge graph of urban flood vulnerability intuitively, quantitatively, and conveniently expresses the logical relationship between vulnerability indexes. This provides a theoretical and data-based foundation for enhancing urban flood resistance and improving the safety of urban flood control and drainage systems. To address this issue, first, the fusion extraction of text and remote sensing dual-mode data is achieved through technical means such as neural networks. Second, by using multiclass natural language processing (NLP) models, we create an objective index system for urban flood vulnerability that avoids subjective human influence. Finally, we construct an objective weight model group, calculate weights, and then, we establish a vulnerability knowledge graph. The results indicate that (1) by utilizing multidimensional remote sensing images and by adopting the robotic satellite (RoboSat) semantic segmentation model, we achieve high-precision extraction of the remote sensing parameters, such as those for urban roads, terrain, and buildings. Thus, we successfully transform remote sensing data into text data (accuracy is approximately 1 m). (2) We have confirmed the effectiveness of the subjective–objective combined weight method. (3) We introduce a novel approach to create an urban flood vulnerability index system based on bimodal objective data fusion. (4) Utilizing the flood vulnerability knowledge graph, we assess vulnerability levels within the primary urban areas of Zhengzhou City, and we propose governance strategies tailored to the current vulnerability status of each district.