A novel approach for water level changes with SAR amplitude data: first results using Sentinel-1 imagery on Trasimeno Lake, Italy

Traditional in-situ monitoring is often limited to major reservoirs in developed regions. However, rising water scarcity necessitates monitoring smaller, isolated water bodies critical for local agricultural systems. Remote sensing has emerged as an efficient alternative to complement or replace gauge stations. Satellite altimetry missions offer high accuracy, but they can be constrained by coarse spatial resolution and revisit time. Consequently, SAR imagery has been widely exploited. Interferometric techniques use phase to detect level changes but are limited to vegetated wetlands or sub-wavelength changes [1]. On the other hand, amplitude-based methods, which rely on shoreline backscatter differences, are often dependent on accurate DEMs [2]. This research aims to introduce a novel approach for estimating water level changes using SAR amplitude data, without relying on prior morphological information. The approach assumes that the horizontal shift Δ of a shoreline and its water level change Δh are geometrically dependent through the local coastal slope i, under the hypothesis that locally the coastal morphology can be approximated with a plane. From the satellite perspective, the level change on this plane is captured as a variation in the sensor-to-target distanced. By combining d and Δ with other parameters which describe the geometric configuration of the satellite-coast interaction (satellite azimuth, SAR local incidence angle, coastal aspect), a final observation equation is formulated to link the unknown water level change to the measured distance. This scheme can be applied to different coastal zones around the lake, assuming variable slopes, but the same water level change between two epochs, providing redundancy for the implementation. The model is developed first by applying an image-matching technique on coregistered SAR images to detect shoreline displacements in the range direction (d). Then, the displacements are used as input for a least squares approach, which incorporates initial assumptions regarding geometrically known parameters and preliminary estimates of the unknown values, yielding estimates of both the water level changes (Δh) between epochs and the slope of each coastal zone portion (i). A preliminary analysis was focused on Trasimeno Lake in Umbria, Italy, using a stack of 30 Sentinel-1 (S1) SLC images (IW mode, VV polarisation) acquired in 2022 on the same orbit, coregistered using the pyGMSTAR library [3]. When compared to the in-situ data, the differences with the estimated values achieved an accuracy of 4 cm and a NMAD of 9 cm, demonstrating the method's potential using S1 mid-resolution imagery. Other tests are under development to improve the overall performance and support the future integration of the method for enhancing water level monitoring in different basins.

 

[1] Aminjafari, S., Brown, I., Mayamey, F. V., & Jaramillo, F. (2024). Tracking centimeter-scale water level changes in Swedish lakes using D-InSAR. Water Resources Research, 60, e2022WR034290

[2] Lee, S., Kim, D.-j., Li, C., Yoon, D., Song, J., Kim, J., & Kang, K. (2024). A new model for high-accuracy monitoring of water level changes via enhanced water boundary detection and reliability-based weighting averaging. Remote Sensing of Environment, 313, 114360

[3] Pechnikov, A. (2024). PyGMTSAR (Python InSAR) (Version 2024.2.8)