On Thursday, July 9, as part of a session of two consecutive seminars starting at 10:00 CET, a seminar titled “Surrogate modelling for hydromorphodynamic across scales” will be held in Grandori Room (Building 4).
The seminar will be given by Yifan Yang, Professor of River Engineering at Wuhan University.

Abstract
This seminar presents two complementary surrogate-modelling strategies that learn compact mappings between observable or simulated flow states and otherwise inaccessible or future hydromorphodynamic responses.
At the local scale, an encoder–decoder network learns the relationship between instantaneous surface-flow fields and mean near-bed hydrodynamics, thereby providing a pathway to infer scour-relevant processes from image-based or remotely sensed observations. Gradient-SHAP analysis identifies localised regions that contain the dominant hydrodynamic signatures, while a multiresolution strategy combines high-resolution in situ monitoring of critical zones with coarser, wider-area drone observations. The compressed regions of interest retain sufficient information to reconstruct near-bed fields while substantially reducing data and computational requirements, supporting the identification of scour progression, bed mobility and instability.
At the river-reach scale, an Adaptive-Horizon Graph Neural Operator learns bed updating directly on native unstructured meshes while retaining hydrodynamic feedback from a 2D hydrodynamic solver. It maps recent hydrodynamic and topographic states to future bed-elevation increments, with the effective history length and update interval adapting to the evolving flow–bed system. Trained using validated morphodynamic solutions and evaluated on several benchmarks, the model reproduces dominant bedform migration, erosion and deposition patterns while achieving approximately 100x accelerations.
Together, these studies demonstrate a cross-scale pathway from measurement-informed inference to accelerated morphodynamic forecasting, offering a foundation for rapid scenario analysis, smart scour monitoring, early warning, and hydromorphodynamic digital twins.

Bio-sketch
Yifan Yang is Professor of River Engineering at Wuhan University and a recipient of the National Science Fund for Excellent Young Scientists (Overseas). He received his PhD in Civil Engineering from the University of Auckland and previously worked as a lecturer at the University of Waikato, a postdoctoral researcher at the University of Auckland, and a civil engineer at Wellington Water. His research focuses on scour and erosion hazards around river and coastal infrastructure, data-driven hydro-morphodynamics and sediment transport, and AI-enabled flood risk management and digital twins. He has developed experimental, numerical and machine-learning approaches for bridge scour prediction, turbulent flow-field reconstruction, streambed morphology analysis, and infrastructure resilience assessment. Prof. Yang is active in international professional service through leadership roles in IAHR and ASCE EWRI committees, has received the International Leader Fellowship of the Royal Society of New Zealand, and has led or contributed to collaborative research and engineering projects in China, New Zealand, Australia, and the United States. His work bridges fundamental river mechanics, intelligent modelling and practical risk reduction.