IAHR Document Library

« Back to Library Homepage « Proceedings of the 39th IAHR World Congress (Granada, 2022)

Automatic Calibration of a Continental Shelf Model of the North Sea Using Data Assimilation Algorithm

Author(s): Kai Chu; Alexander Breugem; Li Wang; Boudewijn Decrop

Linked Author(s): Li Wang, Boudewijn Decrop

Keywords: ICSM; TELEMAC; Automatic calibration; Data assimilation

Abstract: The iCSM is a 2D barotropic, tidal surge model developed in-house in TELEMAC-2D , focusing on the continental shelf of the North Sea. The model performance is in general decent by including the most dominant physical processes in the North Sea such as inverse barometer correction, self-attraction and loading and internal tide dissipation. Preliminary calibration was carried out by manual adjustment of spatial varying bottom roughness, which shows great potentials to further improve model predictions on water level, yet not optimal. Calibration is an essential modelling step but often requires a certain amount of time and efforts, without guarantee of optimal solutions. Therefore in this study, an automatic calibration routine is applied to further improve the predictions of water level in the North Sea. The Python module TelApy internally built-in the TELEMAC kernel is online coupled with an external Python library ADAO from the SALOME platform. The 3D-VAR data assimilation algorithm is adopted to actualize the automatic model calibration. This algorithm compares time-series of water level predicted by iCSM and measurement data at 29 coastal stations in the North Sea, leading to a cost function which is subject to minimization. The automatic calibration routine steers iterative runs, after each of which, bottom roughness is updated based on the minimization of cost function. After around 90 iterations, an optimal bottom roughness field is found, leading to the lowest cost function. After automatic calibration, the model shows reliable ability to precisely reproduce the hydrodynamics in the North Sea. For instance, the averaged root-mean-square-error (RMSE) of water level is around 10 cm at the Belgian coast. The corresponding discrepancy of M2 tidal amplitude and phase is around 2 cm and 1°. The present model accuracy makes the model a reliable tool to provide metocean conditions for offshore projects. The model is fast and can provide time series of surge and current speed up to 30 years, allowing for extreme value analysis in the process of determining design conditions.


Year: 2022

Copyright © 2024 International Association for Hydro-Environment Engineering and Research. All rights reserved. | Terms and Conditions