Assessing the Challenges of Urban Flood Modelling: A TELEMAC-2D Rain-on-Grid Framework to Quantify the Relative Impacts of Surface Calibration, Infrastructure Modelling and Operational Rules in the Emscher Catchment

Urban flood modelling in infrastructure-dense and heavily modified catchments requires enhanced process realism, operational applicability, robust diagnostic and scenario-based evaluation to reliably capture complex system interactions and support decision-making under extreme and failure conditions. This study employs a TELEMAC-2D rain-on-grid approach to simulate pluvial flood dynamics in two urban sub-catchments of the Emscher River (North Rhine-Westphalia, Germany). A stepwise model development and calibration workflow is implemented, combining and adjustments of land-use-based roughness, re-finement of SCS Curve Numbers, and the progressive integration of key hydraulic and op-erational components, including culverts, bridges, retention basins, and pumping stations. Model performance is evaluated based on hydrograph shape and volume, peak discharge and its timing, and inundation extent, with a specific focus on the relative contributions of (i) surface parameter calibration (friction coefficient-Manning’s n and run-off- Curve Numbers), (ii) explicit representation of hydraulic structures, and (iii) operational control rules under varying rainfall scenarios and antecedent moisture conditions (AMC). The analysis tests the hypothesis that structural and operational realism can contribute as much as traditional surface calibration to improve model performance and that their effec-tiveness is strongly influenced by prior wetness. Results shows that including retention basins and pumping stations along with operational rules significantly improves agree-ment with observed discharge. It shows systematic sensitivity and improvements across AMC scenarios with NSE values from -0.129 to +0.77, RMSE from 3.380 to 1.52 m³ s⁻¹, peak discharge errors from −6.20 to −0.49 m³ s⁻¹, and volume bias from −0.67 to +0.04. This shows that even with careful calibration of surface parameters (e.g., roughness and runoff coefficients), models that exclude infrastructure (e.g., pumps and retention basins) fail to accurately reproduce peak flows and recession behaviour. A targeted routing-focused cali-bration (R4) further reduced the remaining peak timing mismatch under saturated condi-tions, but introduced increased volume bias, indicating that residual discrepancies are primarily linked to simplified representation of fast urban conveyance pathways rather than surface parameterisation alone. Flood response is not determined by rainfall alone. Initial wetness and how infrastructure is operated can strongly and unpredictably change flood behaviour. Overall, the findings emphasise that for a complex and engineered urban environment, reliable urban flood simulations requires the combined consideration of hydrodynamic processes, hydrological initial conditions, and operational behaviour. The study provides practical guidance on the limits of calibration-only approaches and identifies when explicit representation of infra-structure and operational processes is essential for robust modelling.