News

Once the altimetry and watermasks datasets selected, the computation of the river geometry is computed using the dedicated workflow presented in Figure 1 in the case of ICESat-2 data for the Z and ExtractEO watermasks to estimate W. The main concept behind this methodology is to merge wse and widths timeseries, by first interpolating in time on common timestamps and then fit a powerlaw W=aZ^b to filter noise in both the widths and wse observations. Finally priors for the two remaining unknowns for the hydraulic model, namely the bathymetry (or equivalently the unobserved flow area A₀ in part 1 – Figure 1) and the friction coefficient are estimated using usual methods (inversion of the Manning equation, geomorphological laws, etc.). 

Figure 1 : Example workflow of the method. 

Figure 2 : Example of a cross-section geometry in the (t,z) plane estimated using our method. (Left) timeseries of Z and W with Z interpolated on the W timestamps, (right) (Z,W) observations in red dots and fitted powerlaw a(Z-Z₀)^b in dashed blue line. 

The final mesh (geometry) of the model obtained using SWORD v17b and a fusion of ICESat-2 + Sentinel 3 altimetric data with widths computed from ExtractEO watermasks is illustrated on Figure 3. This mesh consists in 11 segments with a total of 5229 cross-sections. 

Figure 3 : mesh of the 1D hydraulic model for the Garonne (La Réole) basin, which contains 11 river segments and a total of 5229 Cross-sections. 

To go further, another application example has been published in Water Resources Research) (see Larnier et al, 2025) with full analysis of the assimilation of a fusion of in-situ and EO data to improve the estimation of discharges along the Maroni River network using a coupled hydrological-hydraulic model. 

References 

Larnier, K., Garambois, P.-A., Emery, C., Pujol, L., Monnier, J., Gal, L., et al. (2025). Estimating channel parameters and discharge at river network scale using hydrological-hydraulic models, SWOT and multi-satellite data. Water Resources Research, 61, e2024WR038455. https://doi.org/10.1029/2024WR038455  

How to use satellite observations to build a 1D hydraulic model, part 2 : data and preprocessing 

As presented in part 1, the definition of the geometry of a 1D hydraulic model is twofold: a) the river network planform (on a map) and b) the geometry of the cross-sections. 

1. River network 

In our methodology, the definition of the river network (planform) and the location of the cross-sections are predefined using the SWOT a priori River Database (SWORD, Altenau et al, 2020), which contains the river network segmented in connected reaches (river portions of a few kilometers) and each reach is composed of river nodes approximately distant of 250m. The cross-sections of our hydraulic model are located at each SWORD node and the river network is computed from the connectivity of the SWORD reaches.  

2. Cross-section geometry 

For each cross-section, the geometry in the (t,z) plane (see Figure 1) are represented as tuples of (Z,W) where Z is a water surface elevation (wse) and W the top width. These tuples of (Z,W)_i=0,N are estimated by merging two different sourcces of data: (ii) altimetric data for H and watermasks to estimate top width W. 

For the altimetric data, our methodology can leverage any of the altimetry mission (for instance EnviSat, the Jason missions, Sentinel-3 SRAL, IceSat-2, SWOT).  

Watermasks are usually computed either from optical images (Sentinel 2, Landsat missions, Pleiades.) or radar images (Sentinel 1, RadarSat, etc.). To avoid the cloud‑related issues that can occur with optical imagery, Sentinel‑1 SAR data were here favoured. Over the Garonne River Basin, 4 paths were used, and around 240 products were processed for 2024 (about 30 dates per path). Both ascending and descending orbits were used to build these water masks, as illustrated in Figure 2. This processing was carried out using ICUBE-SERTIT’s ExtractEO (Maxant et al. 2022) chain, applied to the full dataset. 

Figure 2 : The Garonne River Basin (red contour) and main river reaches (yellow lines), and four S1 paths obtained on a given data (6-day window) both from ascending and descending orbits 

A first filtering step was then performed to exclude smaller rivers based on their width. Given the 20 m resolution of the images, some rivers could not be accurately detected and were therefore removed. 

It was also observed that a significant portion of the water mask was not detected due to vegetation shadowing (see Figure 3). Combining ascending and descending paths, separated by one-day intervals, we finally obtained 60 complete masks for 2024, which are being used for river network hydraulic model set-up. 

Figure 3 : Impact on vegetation shadowing on single path depending on descending or ascending course (central figures) and improvement achieved by using both jointly ascending and descending tracks (right panel). 

Note that the methodology is generic and can be further re-employed for other river network models and other basins, and that the shadowing will depend on vegetation height and river geometry.  

Next part (Part 3) will conclude this series of news with the presentation of the core components of the method and the application to the Garonne River basin. 

References 

Altenau, E. H., Pavelsky, T. M., Durand, M. T., Yang, X., Frasson, R. P. d. M., & Bendezu, L. (2021). The surface water and ocean topography (SWOT) mission river database (SWORD): A global river network for satellite data products. Water Resources Research, 57, e2021WR030054. https://doi.org/10.1029/2021WR030054  

Maxant, J.; Braun, R.; Caspard, M.; Clandillon, S. ExtractEO, a Pipeline for Disaster Extent Mapping in the Context of Emergency Management. Remote Sens. 2022, 14, 5253. https://doi.org/10.3390/rs14205253 

1D hydraulic models are powerful tools to conduct fine-scale analysis of the main river flows. The setup of such models relies on the definition of the river geometry that describes both the river planform (in curvilinear 1D network) and the topography of the riverbed along the river network. Thus defining the model geometry is twofold: (i) the definition of the river network and the location of the cross-sections (computation nodes) along this network and (ii) the geometry of each cross-section (see Figure 1). Usually both these two aspects of the geometry are computed from field survey data for the riverbed topography and a DEM (Digital Elevation Model) for the floodplain. In this series of articles we present a new methodology to leverage multi-sources Earth Observation (EO) data to estimate the geometry. This methodology is crucial for areas where collecting field data is difficult due to several reasons (topography or vegetation, political or transboundary issues, etc.). 

Figure 1: River geometry: (left) river network in solid lines and location of the cross-sections (red dots), (right) geometry in (t,z) plane for the considered cross-section in the left pane. 

In the next part of this series, the application of this methodology will be presented on the Garonne River basin in France. Part 2 will introduce the EO data of interest and how to preprocess them and part 3 will present the core components of the method with its application to build a 1D hydraulic model the Garonne River basin. 

Figure 2 : Considered network of the Garonne River basin, south-west of France.