Absolute Water Volume Estimation from Multi-Sensor approach using SWOT and Sentinel-2

Water resources are under unprecedented pressure driven by climate change, societal demands, and geopolitical tensions. To address these issues in France, the FR2030 project supported by the Ministry for Ecological Transition, aims to improve water management. In this context, authorities have identified 18,500 water bodies requiring regular monitoring through satellite data. These water bodies of varying nature range from 3ha to several hundred hectares.

Current satellite missions provide data well-suited for regular monitoring thanks to their revisit frequency and spatial resolution, enabling observation of a large number of water bodies. The launch of SWOT in 2023 expanded significantly the number of observable water bodies through its near-global coverage, opening up new possibilities for monitoring water resources.

One of FR2030’s objectives is to provide volume measurements that decision-makers, such as prefectures, regional environmental agencies (DREAL) and other authorities, can rely on to act quickly in crisis situations. Most methods focus on estimating volume variations as this approach is more straightforward. However, end users also need absolute quantitative measurements.

The first developed approach is based on the hypsometric law commonly used for volume estimation (Crétaux et al., 2016). While SWOT provides height and surface data, its surface measurements lack the precision required for quantitative monitoring making a multi-sensor approach preferable. The hypsometric curve is derived by combining Sentinel-2 surface data (Peña-Luque et al, 2021) with water surface elevation data from SWOT_L2_HR_LakeSP_Prior products. Lake bottom information obtained from a DEM and dam base data (e.g. DEM4Water) is needed to compute absolute volume to correct the bias. This 2D approach already provides valuable insights for user but requires prior data.

A second method was developed to overcome this limitation. Water body contours are extracted from multiple clear Sentinel-2 surface images each linked to a water surface elevation from SWOT. Using 3D reconstruction, we derive bathymetry (Khazaei et al., 2022) discretized along a height scale. Water columns at the target elevation are then used to compute lake volume. This innovative 3D approach relying only on surface and height remote sensing data already shows strong potential. Its preliminary results are consistent with established datasets and methods. The method delivers in-situ validated results with an initial error of just 25% on absolute volumes. With several limitations already identified, this approach is on track for significant improvements.

These two approaches illustrate the potential for developing a global framework for dynamic monitoring of reservoir water storage under time constraints. By combining multi-sensor satellite data and advanced reconstruction techniques, they enable direct estimation of absolute water volumes, an innovative breakthrough compared to traditional methods focused on relative variations. While further validation and optimization are required, these methods open promising perspectives for decision-makers with actionable insights at scales relevant for resource management.

References:

• Crétaux et al., 2016, Lake volume monitoring from space: https://doi.org/10.1007/s10712-016-9362-6
• Peña-Luque et al, 2021, Sentinel-1&2 Multitemporal Water Surface Detection Accuracies, Evaluated at Regional and Reservoirs Level: https://doi.org/10.3390/rs13163279
• DEM4Water: https://github.com/CNES/dem4water
• Khazaei et al., 2022, GLOBathy, the global lakes bathymetry dataset: https://doi.org/10.1038/s41597-022-01132-9