Santiago Pena-Luque (CNES, France); Renato Frasson (JPL, US); Mohammad J. Tourian (University of stuttgart, Germany); Tati de Almeida (UNB, Brazil); Daniel Moreira (SGB, Brazil); Fabrice Papa (IRD, France)

The Surface Water and Ocean Topography (SWOT) mission, a collaboration between NASA and CNES, offers a novel opportunity to advance the monitoring of inland water bodies through high-resolution elevation and extent data, even under persistent cloud cover. This is particularly relevant to Brazil, a country with extensive freshwater resources but significant challenges in monitoring water bodies at scale due to accessibility, cost, and the limitations of traditional methods. SWOT’s Ka-band Radar Interferometer (KaRIn) overcomes many of these constraints by enabling high-resolution, cloud-independent measurements of water surface extent and elevation.
While promising, the utility of many SWOT science products will depend directly on the quality of the water classification available in the PixelCloud. Initial studies indicate that the classification accuracy can be challenged under certain conditions, such as in the presence of specular ringing, dark water, and highly reflective surfaces like sandbanks.
In this context, this project proposes a systematic evaluation of water masks derived from the SWOT PixelCloud product, comparing them with established optical-based products such as OPERA (NASA/JPL) and SURFWATER (CNES). The analysis will encompass a range of global regions representing different environmental conditions, with a primary focus on the Amazon Basin. The methodology involves the spatio-temporal co-location of SWOT and Sentinel-2 data, rasterization of SWOT point clouds, and pixel-based comparisons. Classification agreement will be assessed using the kappa coefficient, and accuracy will be validated against reference samples derived from high-resolution optical imagery. To further validate classification accuracy, reference samples will be generated through the visual interpretation of high-resolution imagery from Pléiades, where available.
Additional analyses aim to explore both intrinsic (e.g., viewing geometry, signal-to-noise ratio) and extrinsic (e.g., land cover, water body size) factors influencing classification performance. Although the study is ongoing, preliminary results will be available by the time of the meeting. The findings will support SWOT users in adopting its products for reservoir monitoring and flood analysis and contribute to broader strategies for integrating radar and optical approaches in water resources management.