Estimation of water elevation and volume changes over mountain lakes using Sentinel-1/2 and SWOT data

Monitoring changes in lake water elevation and volume in mountainous regions is crucial for hydro-electricity management, natural hazard mitigation and quantifying water resources. The SWOT mission is revolutionising lake monitoring from space since its swath altimetry sensor provides data with a global spatial coverage and a better spatial resolution than previous nadir altimeters. This opens the door to the study of smaller lakes. However some challenges remain, especially in mountainous regions. It includes the impact of layover in steep terrain and the poor quality of the SWOT water area products. These limitations hinder the derivation of hypsometric curves (i.e. the area/height relationship) that are required to calculate changes in lake volume. One way to overcome this is to measure lake areas using the segmentation of SAR and optical images. However, mountain lake areas from SAR images are affected by layover and shadow effects while those derived from optical images are distorted by cloud cover.

In this context we first evaluate the quality of the SWOT Pixel Cloud (L2_HR_PIXC) products over several mountain lakes. We developed a processing chain to filter PIXC data based on quality flags, backscatter signals and spatio-temporal statistics. The filtered water surface elevations are validated using in-situ data from the Lacs Sentinelles network in the French Alps and from the FOEN network in Switzerland.

Next, we present a methodology for combining Sentinel-1 and Sentinel-2 areas with SWOT surface elevation using the hypsometric function, and a spatio-temporal interpolation of the lake water levels. Lake surface areas are derived from the Sentinel-1 OASIS index, a highly sensitive measure for water body detection, while Sentinel-2 based lake surface areas are extracted from the CNES Surfwater products, available on the hydroweb.next open access platform. We illustrate our strategy on three different mountain lakes of various sizes and in different environments: Joux Lake in the Jura Mountains, Switzerland (8-9 km²) ; the Lauvitel Lake in the Ecrins Massifs, French Alps (1-3 km²) and the Rosolin Lake, a supra-glacial lake in the Vanoise Massif, French Alps (0.01-0.03 km²). This work is a step towards a better quantification of changes in lake elevation and volume in complex terrains.