A water-potential-semantic coupled state-field inversion method for continuous perception and semantic expression of urban flooding

Emergency monitoring of urban waterlogging requires more than rapid extraction of inundation extent; it also requires continuous perception and semantic expression of water distribution, low lying connectivity, field disaster meaning and short term development potential. Existing remote sensing flood mapping provides overall inundation patterns, while DEM and hydrodynamic models characterize terrain constraints and water depth variations, and crowdsourced information offers field level disaster observations. However, these information sources are rarely modeled around a unified continuous state variable, and the coupling between field disaster reports and terrain convergence structures remains insufficient. This makes it difficult to explain inundation extent, waterlogging intensity and emergency semantic states from a unified state generation mechanism. To address these problems, this paper proposes , a water potential semantic coupled state field inversion method for continuous perception and semantic expression of urban flooding. The method represents inundation and waterlogging intensity with a continuous water potential field, expresses emergency semantics such as road blockage, deep water danger and potential expansion with a semantic state field, and jointly inverts the coupled physical and semantic states through SAR observation anchoring, DEM terrain constraints and field disaster reports. A case study of the Zhuozhou city waterlogging process in Hebei Province in China compares proposed method with ASTGCN, CS-DA and S1S2-DL. The results show that the proposed method jointly expresses flood extent, relative water depth, priority rescue areas and next period potential hazard areas, forming a coupled state field representation of water distribution, semantic risk and state evolution.