Author(s): Clemens Cremer; Jesper Sandvig Mariegaard; Henrik Andersson; Jannik Elsasser; Faro Schafer
Linked Author(s): Jesper Sandvig Mariegaard, Clemens Cremer
Keywords: Calibration Hydrodynamic Modeling Multi-Objective-Optimization Parameter Estimation
Abstract: Calibration of hydrodynamic models is fundamental for accurate simulations, yet traditional manual methods are time-consuming and may not fully explore complex parameter spaces in two-dimensional (2D) and three-dimensional (3D) models. While machine learning (ML) has advanced optimization algorithms capable of efficiently navigating high-dimensional parameter spaces, their application in hydrodynamic modeling remains limited. This study leverages Optuna for automated calibration, testing various sampling strategies on two distinct case studies: the German Elbe River estuary and Southern North Sea models. The multi-objective optimization considers water levels at various measurement locations, aiming to enhance calibration efficiency while providing valuable insights through Pareto fronts and parameter sensitivity analysis. Results demonstrate improved calibration quality and generalizability while reducing manual effort.
Year: 2025