Validation of a Semi-automatically Calibrated 1-D Open-Channel Model Against Experimental Data with Changes in Channel Geometry

Experimental and numerical study of unsteady flows in a compound open channel

“…The root mean square error (RMSE) is computed from the squared differences in water Maximum and mean absolute bias, Max (x) and Mean (x) , and RMSE calculated between water surface level measurements and simulated levels using the codes: 1D+ MAGE-ISM (with either turbulent exchange coefficient t = 0.03 or t = 0.08); 1D MAGE-DEBORD; and 1D FudaaCrue. levels [7]. Since the results of the 1D simulations on the water levels are mostly driven by the total discharge and the bed-friction coefficients (e.g., Strickler or Manning roughness coefficients [43]), these first results suggested that the turbulent exchange coefficient introduced in the ISM played a significant role in the modelling of the global head losses and could efficiently correct the well-known weaknesses of the classical 1D modelling [41].…”

“…In this area, although Navier-Stokes equations can be numerically solved in either one, two or three dimensions, within reasonable computational time, either two-dimensional (2D) or threedimensional (3D) models are still too time-consuming for some specific operational applications such as flood management (requiring real time computing) or optimisations in the design of hydraulic works (requesting hundreds of runs in order to optimize the hydraulic design or to process automatic model calibration). Therefore, there is still a deep interest for improvements in one-dimensional (1D) modelling techniques since 1D models are generally fast enough for those purposes whereas they suffer from a poor modelling of the different terms of head losses − mainly treated as bed-friction head losses as pointed out by Visse et al [7]. The influence of this poor representation of the physical processes is somehow compensated by some calibration techniques that could benefit from numerous routine data [8].…”