Main content area

Simultaneous calibration of multiple hydrodynamic model parameters using satellite altimetry observations of water surface elevation in the Songhua River

Jiang, Liguang, Madsen, Henrik, Bauer-Gottwein, Peter
Remote sensing of environment 2019 v.225 pp. 229-247
altimeters, geodesy, hydrologic models, parameter uncertainty, remote sensing, risk assessment, rivers, satellite altimetry, surface water, topography, China
Hydrodynamic modeling is an essential tool to simulate water level for flood forecasting and risk assessment. However, parameterization of hydrodynamic models is challenging due to poor knowledge of bathymetry and lack of gauge data. In this study, we present an approach for calibrating spatially distributed Strickler coefficient and river datum simultaneously by regularized inversion. Calibration was carried out using altimetry derived observations of water surface elevation in the Songhua River (China). Synthetic experiments show that spatial variability of model parameters can be well constrained by geodetic altimetry, e.g., CryoSat-2 and SARAL, and to a lesser extent by Envisat and Jason-1. However, Jason-2 can only recover some of the parameters. We also find that a higher accuracy of observations does not proportionally decrease parameter uncertainty for all cross sections. Instead, high spatial sampling density helps to identify the spatial variability of parameters. Real-world calibrations indicate that CryoSat-2 by far outperforms other altimeters in terms of parameter identification. We conclude that high spatial resolution is more important than temporal resolution and observation accuracy for calibrating parameters of large-scale river models. The findings demonstrate the added value of geodetic satellite altimetry for inversion of spatially resolved parameter fields. This study is timely because the upcoming Surface Water and Ocean Topography mission will significantly improve spatial and temporal resolution of spaceborne observations of water surface elevation, and hydraulic models need to be prepared for the uptake of these data.