Designing a Knowledge Graph System for Digital Twin to Assess Urban Flood Risk

The digital twin is widely used to create virtual counterparts of physical entities, and it is also as a promising approach for assessing urban flood risk. A key challenge of supporting such DTs is to formalize objects related to urban flood, and integrate multi-modal data and distilled expert knowledge. However, no research has applied or implemented the DT with KG to assess unban flood risk. To ad-dress the problem, we propose the UrbanFloodKG system, a knowledge graph system that models and integrates data in a digital twin formalism for urban flood risk assessment. The system provides a complete solution for constructing a knowledge graph of urban flooding events, with a data layer, graph layer, algorithm layer, and digital twin layer. It implements functions for knowledge extraction, as well as integrating knowledge representation learning models and graph neural network models to support link prediction and node classification tasks. We conduct model comparison experiments on data related to flood events in Guangzhou, which demonstrates the effectiveness of our proposed solution.