Generation and analysis of stage-fall-discharge laws from coupled hydrological-hydraulic river network model integrating sparse multi-satellite data
Résumé
Current and forthcoming altimetry satellites provide increasing hydraulic visibility of river surface variabilities. However, due to unobservable river bathymetry and friction, as well as the sampling and quality of remote sensing data, discharge estimation remains a challenging problem. This paper proposes a new rigorous methodology to estimate stage-fall-discharge laws at any point of a hydrographic network including backwater affected reaches. This methodology is adapted to operational estimation of discharge from multi-mission satellite altimetry data. The cornerstone of the method is to take advantage of meaningful hydraulic signatures, including water surface slopes, simulated by a river network model optimally combined with multi-satellite noisy observations. This optimal combination of heterogeneous multi-source data with a river network model is achieved through variational data assimilation performed using the DassFlow-HiVDI hydraulic model inflowed by MGB regional hydrological model. It is newly applied to the large and complex Negro-Branco River network in the Amazon basin. An effective model is obtained by variational calibration using Jason-3 and Sentinel-3A and -3B altimetry observations between February 2016 and May 2019. A fairly good fit bellow 1 m for 80% of the altimetric observations is obtained. Using this hydrodynamic model as a fine spatio-temporal extrapolator of hydraulic quantities, and in particular of multi-scale water surface slope signatures, the parameters of a general SFD law, as well as the uncertainties estimations, can be constrained at any location within the river network domain using a Bayesian optimization method. This methodology is tested over 4 virtual stations representing various hydraulic complexities. Accurate discharge estimates and meaningful SFD parameters are obtained compared to those of the reference low Froude model at these locations. Analyses are also performed on a reference case with fine-scale bathymetry data of the Garonne River, France. The strength of the method lies in its applicability to noisy multi-satellite data while ensuring physical consistency. It should be applicable in real time to any river basin, using nadir altimetry only or combined to the forthcoming SWOT satellite data. Moreover, the method allows to take full advantage, via the hydrodynamic model, of variable water masks observations and physical segmentations of river networks. The method is implemented in the open source hydrodynamic data assimilation platform DassFlow.
Origine : Fichiers produits par l'(les) auteur(s)