Preliminary analysis of SWOT over Low Elevation Coastal Zones: challenges and solutions

The SWOT KaRIn instrument provides revolutionary observations for quantifying surface water storage but faces several challenges in Low Elevation Coastal Zones (LECZ). Water bodies in these regions, such as estuaries, often feature intricate channel networks and complex terrain, with the highly dynamic nature of water surfaces, driven by factors such as tides and wind, impairing SWOT KaRIn’s ability to accurately measure water levels and extent. Vegetation cover further complicates observations, as the Ka-band wavelength used by KaRIn exhibits limited penetration through emergent vegetation, such as mangroves and salt marshes, which are prevalent in LECZs. Complex tropospheric conditions also reduce measurement accuracy, as the low-frequency radiometer on SWOT struggles to provide reliable tropospheric path delay corrections near or over land. Furthermore, the typically low topographic roughness in LECZs exacerbates layover errors, particularly at high incidence angles, where radar signals from multiple locations interfere.

To evaluate these challenges, we analyzed SWOT data over the LECZ of Germany and identified its limitations in such environments. Many of these challenges are evident in the data, leading to inconsistencies in measurements and errors in the original classification provided by the SWOT PIXC data. To address these issues, we tested various methods, including deep learning-based approaches. Specifically, we evaluated two modeling schemes: one without using the estimated height from SWOT and another incorporating height. In the first scheme, we integrated SWOT InSAR measurements and auxiliary data, trained the model using open water and land pixels, and applied it to the data to improve the classification of water and land in pixel cloud points. We assessed the performance of our refined classification by generating river profiles from the newly classified data, comparing them with profiles based on the original classification, and calculating the root mean square (RMS) of variations. A more consistent (less variable) river profile indicates better results. Our findings show that the refined classification significantly enhances vector products for both rivers and lakes, enabling more precise estimation of surface water storage in LECZs.