Monitoring water volume dynamics in West African lakes and reservoirs using altimeters and optical satellite sensors

Lakes, reservoirs, and small water bodies play a pivotal role in West African drylands. They are widely distributed across the landscape, making them a primary source of water for both people and livestock. However, due to their generally small size and strong temporal variability, their hydrological dynamics remain poorly understood at the regional scale. Moreover, these water bodies are highly sensitive to both climatic and anthropogenic forcing, exhibiting complex and sometimes counter-intuitive dynamics, such as the increase in surface runoff observed in the Sahel despite a decrease in precipitation during and the after the major droughts of the 1970s and 1980s. Understanding the past and present dynamics of these water bodies is therefore crucial to anticipate their future evolution in a context of environmental change and rapid population growth.

Recent satellite missions provide an unprecedented view of small water bodies at large scales by combining high spatial resolution, high temporal frequency, and novel observations of water level and volume. This study relies on recent altimetric sensors coupled with surface water extent derived from optical imagery to investigate the dynamics of water levels and volumes across thousands of lakes within the study area.

Water level dynamics are first estimated using Sentinel-3 SRAL data for lakes intersected by satellite tracks, and then extended spatially by more than one order of magnitude using SWOT observations. We show that SWOT-derived water levels are in excellent agreement with in-situ measurements collected in Niger, Burkina Faso, and Senegal, as well as with water level estimates from other satellite sensors (Girard et al., 2025).

The analysis of dry-season water level dynamics allows to identify distinct hydrological behaviours at the regional scale, and to highlight the influence of anthropogenic water withdrawals in agricultural reservoirs, as well as connections between lakes, the river network, and/or groundwater (de Fleury et al., 2023).

Water volume variations are subsequently obtained by combining water level data with water surface areas. The latter are estimated appliying a U-Net convolutional neural network to optical imagery from Sentinel-2 and the Landsat archive. This approach, specifically developed for the study region, provides accurate estimates of water area for the different types of lakes encountered. These include water bodies covered by vegetation and extremely bright lakes characterized by high suspended sediment loads and very fine particles (de Fleury et al., 2025).

The resulting elevation-area relationships are then used to reconstruct past changes in water volume from Landsat-derived water surface areas (1984 to present) for more than 2,000 lakes and reservoirs (Girard et al., 2026). This analysis reveals long-term changes and trends in hydrological dynamics in relation to environmental drivers (i.e. precipitation, temperature, and land use/land cover) and anthropogenic activities (e.g. reservoir construction and management).

 References

• Girard, L. Kergoat, J. S. Paiva, R. Yonaba and M. Grippa (2026) 40-Year Volume Changes of West African Lakes Derived from SWOT and Optical Imagery. Submitted to WRR
• Girard et al. (2025b) https://doi.org/10.1109/JSTARS.2025.3570859
• de Fleury et al (2025) https://doi.org/10.1016/j.rsase.2024.101412
• de Fleury et al. (2023) https://doi.org/10.5194/hess-2022-367