Abstract's details
Variational Assimilation of SWOT Altimetry into a 1D-2D Porosity-based Hydraulic River Model with Upstream Hydrology: Toward Integrated Hydrological-Hydraulic Discharge Estimation from SWOT
Event: 2025 SWOT Science Team Meeting
Session: Hydrology: River Science Working Group
Presentation type: Poster
The high spatial density of SWOT data products enables unprecedented access to small scale variations of water surface elevation (WSE) of worldwide rivers. The informative content carried by such observations has great potential for the calibration of high resolution basin-scale hydraulic-hydrological (H&H) models of river networks and could allow the estimation of reach and cross-section scale geometry parameters and distributed inflows. However, estimating river discharge solely from WSE altimetry data remains a notoriously ill-posed inverse problem if bathymetry-friction are unknown (see Larnier et al., 2020). In this context, H&H modeling frameworks that integrate variational data assimilation (VDA) have shown promise in providing hydrologic closure to this under-constrained problem and also enable the estimation of high-dimensional spatio-temporal H&H parameters. To address these challenges, we employ a robust H&H variational data assimilation (VDA) framework proposed in Pujol et al. (2022), consisting in 1D-2D effective hydraulic model (DassFlow) integrating the GR4H state-space hydrological model (Santos et al. 2018), to assimilate SWOT altimetry for (i) optimizing sequentially or simultaneously effective hydraulic parameters and hydrological parameters and (ii) improving model realism and finally discharge estimation.
To model sub-cell cross-sectional variability in a 1Dlike model (Pujol et al. (2022)), an effective depth-independent porosity parameter was implemented based on Guinot and Soares-Frazão (2006). This parameter is key to the effective modelling of hydraulic controls and signal propagation through a river network hydraulic model based on the 1Dlike modelling approach. It can be estimated from local bathymetry surveys, channel geometry databases, but remains model-dependent and must be calibrated.
Our methodological framework enables the simultaneous inference of distributed hydraulic parameters (bathymetry, friction coefficients, longitudinal riverine porosity) and upstream hydrological parameters or inflow time series. Notably, the H\&H model is integrated into the VDA framework such that information feedback from hydraulic observables to the hydrological model can be achieved. This is key for integrated H&H discharge estimation based on SWOT observations.
The approach was tested on the Garonne river between Tonneins and La Réole (around 50km of river length). Knowledge on the real geometric variabilities of the minor bed is provided by a high-resolution expertised DEM. A 1Dlike hydraulic model of the river was built based on this accurate bathymetry data.
10 SWOT passes over the 2024-02-16 and 2024-04-21 period (65 days) provide snapshots of the waterline for a range of non-flooding upstream discharges. Assimilated SWOT data is in the format of single WSE observation points every 200m at the centerline (RiverObs algorithm output) and covers the whole reach.
A series of inference experiments were designed to evaluate the capability of the VDA method to extract informative content from SWOT WSE. Sought parameters are distributed friction and porosity, as well as hydrological parameters of the lumped upstream hydrological model. Results show improvement of the fit of the modeled waterlines to altimetry observations thanks to the calibration of finely distributed parameters and establishes a robust H&H approach for SWOT discharge estimation.
Low computation costs achieved through the 1Dlike modeling approach make it clear that this method could be scaled to larger river networks, opening the way toward leveraging the full informative content of basin-scale SWOT altimetry through our H&H VDA framework.
References:
Guinot, V., Soares-Frazão, S. (2006). Flux and source term discretization in two‐dimensional shallow water models with porosity on unstructured grids. International Journal for Numerical Methods in Fluids, 50(3), 309-345.
Larnier, K., Monnier, J., Garambois, P.-A., Verley, J. (2020) River discharge and bathymetry estimation from SWOT altimetry measurements, Inverse Problems in Science and Engineering (IPSE).
Pujol, L., Garambois, P.-A., Monnier, J. (2022) Multi-dimensional hydrological-hydraulic model with variational data assimilation for river networks and floodplains, Geoscientific Model Development.
Santos, L., Thirel, G., Perrin, C. (2018). Continuous state-space representation of a bucket-type rainfall-runoff model: a case study with the GR4 model using state-space GR4 (version 1.0). Geoscientific Model Development, 11(4), 1591-1605.
Back to the list of abstractTo model sub-cell cross-sectional variability in a 1Dlike model (Pujol et al. (2022)), an effective depth-independent porosity parameter was implemented based on Guinot and Soares-Frazão (2006). This parameter is key to the effective modelling of hydraulic controls and signal propagation through a river network hydraulic model based on the 1Dlike modelling approach. It can be estimated from local bathymetry surveys, channel geometry databases, but remains model-dependent and must be calibrated.
Our methodological framework enables the simultaneous inference of distributed hydraulic parameters (bathymetry, friction coefficients, longitudinal riverine porosity) and upstream hydrological parameters or inflow time series. Notably, the H\&H model is integrated into the VDA framework such that information feedback from hydraulic observables to the hydrological model can be achieved. This is key for integrated H&H discharge estimation based on SWOT observations.
The approach was tested on the Garonne river between Tonneins and La Réole (around 50km of river length). Knowledge on the real geometric variabilities of the minor bed is provided by a high-resolution expertised DEM. A 1Dlike hydraulic model of the river was built based on this accurate bathymetry data.
10 SWOT passes over the 2024-02-16 and 2024-04-21 period (65 days) provide snapshots of the waterline for a range of non-flooding upstream discharges. Assimilated SWOT data is in the format of single WSE observation points every 200m at the centerline (RiverObs algorithm output) and covers the whole reach.
A series of inference experiments were designed to evaluate the capability of the VDA method to extract informative content from SWOT WSE. Sought parameters are distributed friction and porosity, as well as hydrological parameters of the lumped upstream hydrological model. Results show improvement of the fit of the modeled waterlines to altimetry observations thanks to the calibration of finely distributed parameters and establishes a robust H&H approach for SWOT discharge estimation.
Low computation costs achieved through the 1Dlike modeling approach make it clear that this method could be scaled to larger river networks, opening the way toward leveraging the full informative content of basin-scale SWOT altimetry through our H&H VDA framework.
References:
Guinot, V., Soares-Frazão, S. (2006). Flux and source term discretization in two‐dimensional shallow water models with porosity on unstructured grids. International Journal for Numerical Methods in Fluids, 50(3), 309-345.
Larnier, K., Monnier, J., Garambois, P.-A., Verley, J. (2020) River discharge and bathymetry estimation from SWOT altimetry measurements, Inverse Problems in Science and Engineering (IPSE).
Pujol, L., Garambois, P.-A., Monnier, J. (2022) Multi-dimensional hydrological-hydraulic model with variational data assimilation for river networks and floodplains, Geoscientific Model Development.
Santos, L., Thirel, G., Perrin, C. (2018). Continuous state-space representation of a bucket-type rainfall-runoff model: a case study with the GR4 model using state-space GR4 (version 1.0). Geoscientific Model Development, 11(4), 1591-1605.