PhD: Spatial data fusion and scaling for small water reservoir monitoring

Centre National D'Etudes Spatiales (CNES)

Montpellier, France 🇫🇷


The water reservoirs smaller than 1.5 ha represent up to 60 % of surface water bodies  [France : INPE-MTES] being just a small fraction of the total surface water bodies volume. However these small reservoirs disseminated all over surfaces are of great importance for a smart local use of water (cf. Varennes de l’eau agricole), to preserve ecosystems and to mitigate hydroclimatic risks. Water reservoir monitoring from space is effective or planned (e.g. SWOT) for large reservoirs. However, these missions are not adapted for small water reservoirs since their detection, the monitoring of their water volumes and storage capacities require both very high spatial resolution and high revisiting frequency. Monitoring the water stocks of small reservoirs along seasons is hence still dramatically missing at the global scale. The detection of such small reservoirs is not an issue since their contours can exist in geodatabases (e.g. BD-Topage-France) or it can be processed using a small number of spatially highly resoluted images (Pleiades, Pleiades Neo). Nevertheless, the characterization of their individual geometry and the monitoring of the waterstock is still a challenge. 

The main question addressed by the PHD is thus: are we able to monitor (geometry, water volume) small surface water bodies by fusioning spatial data? up to which accuracy, and in which contexts? 

To answer this question, the PhD work aims at developing data fusion and spatial downscaling methods allowing the extension to small water bodies of the capacity of contemporain satellite missions to observe waters’ surface (e.g SWOT).  Assuming that the maximum contour of any small water body is known, the two main steps of the PhD could be the following. In the first step, spatial downscaling methods will be developed for water surface dynamic estimation at the reservoir scale. This action aims to transform coarse reservoir locations into water surfaces time series with two targeted outputs: water area estimation along time, and water contour along time. This action will be based on “pan-sharpening like” approaches using highly resoluted multispectral images, pixel analytical spectral unmixing optimization, pixel mixing spatial decomposition optimization or convolutional neural networks (super-resolution) techniques. To get water surface spatial contours from coarse pixel resolution (e.g. SurfWater output from Sentinel image series), additive image processing algorithms from vision domain (contours, edges, snakes) are candidates. 

A second step aims at transforming water surface dynamics to water storage dynamics at water body scale. Two ways are possible. 

A first direct way will  estimate an explicit bathymetry by fusionning different informations: water surface and contours dynamics from previous step coupled with altimetric/bathymetric data, spatial laser altimetry (ICESAT-2, GEDI) on water surfaces, very high spatial resolution terrestrial DEM (e.g. CO3D) on water body surroundings or at dry season, and using bathymetric algorithms from image spectra in shallow waters [4SM, Morel 2017]. A spatial estimation method based on spatially constrained radial functions from [Delenne, Bailly et al. 2021] could be extended for that purpose. 

A second more indirect way will try to transform water surface dynamics to water volume dynamics with implicit bathymetry through a parametric hypsometric curve with parameters estimated from the same data source as previously exposed.   

All methods will be trained and evaluated on a set of contrasted regions (Occitania-France, PACA-France, CapBon-Tunisia, Karnataka-India) benefiting from historical ground truth data (small reservoir water height monitoring at daily time step) and high resolution waterbodies bathymetries (Bathymetric drone available at LISAH-lab). Reference airborne or spatial data with higher resolutions for benchmarks are also available in these test sites: Bathymetric LiDAR (PACA-Fr), Venµs data – SMARA project (CapBon-Tn ; Karnataka-In).


For more Information about the topics and the co-financial partner (found by the lab !), contact Directeur de thèse :

Then, prepare a resumé, a recent transcript and a reference letter from your M2 supervisor/ engineering school director and you will be ready to apply online !


Région Occitanie (at the cross-section of the 2 défis clés régionaux : “WoC” (Water) and “OT” (Observation de la Terre et Territoires en transition)

Infos pratiques


Mot du recruteur

More details on CNES website :




You ad could be here!