Semantic-driven dynamic integration of multi-LOD terrain and hydraulic objects for digital twin irrigation districts

Digital twin irrigation districts rely on stable multiscale 3D representation of terrain and hydraulic objects, but multi-LOD terrain transitions often cause floating, misalignment, and discontinuous dynamic behavior. Existing methods focus on static geo-metric alignment and lack semantic-aware dynamic constraint mechanisms across scales. This work proposes a semantically driven dynamic integration framework for multi-LOD terrain and hydraulic objects. We build a unified three-dimensional semantic model covering operational state, geometric form, and terrain dependency, then ex-tend LOD to a spatial constraint regulation mechanism. Differentiated strategies are designed for dynamic point, line, and surface objects via spatiotemporal mapping, pose constraints, and water-level-driven updating. Experiments in the Zhuluoba Irrigation District show that static channel boundary and centerline errors reach 0.343 m and 0.395 m, while dynamic point planimetric and elevation errors are 0.492 m and 0.245 m. The experimental results demonstrate that the proposed framework can maintain relatively stable spatial–temporal consistency across different LOD levels in the investigated irrigation district.