Victor Pellet (Laboratoire de Météorologie Dynamique, École Polytechnique, France); Filipe Aires (LIRA, Paris Observatory, France)

One expectation for the SWOT (Surface Water and Ocean Topography) satellite is that it can provide information on surface waters, including beneath clouds and possibly vegetation, at high spatial resolution, which optical sensors cannot achieve. However, SWOT observation errors do exist, e.g., due to specular reflection. It is necessary to filter these errors. This data filtering can amount up to 44% of the considered pixels in this study. They drastically limit the use of the SWOT data. Consequently, filtered pixels need to be filled in some way, to obtain clean and spatially continuous water extent maps from SWOT. We developed an approach to interpolate SWOT data so that all the SWOT observation time steps can be exploited. We focus on a part of the Negro River in the Amazon basin. First, pixels are filtered using echo nadir and specular ringing of the water area fraction variable, from the L2 KaRIn high-rate raster product under conditions of low coherence, degraded classification information, and incident angle. Second, we interpolate the filtered pixels using 1) a hydro/topography-based “Floodability Index” (FI), a proxy for the probability of a pixel being inundated compared to its adjacent pixels (Nguyen and Aires, 2023) and 2) a water occurrence derived by SWOT observation. To do so, we determined spatio-temporally varying thresholds (on the FI and water occurrence) to be the water/non-water pixels, based either on a ROC-curve analysis or on a water area-based optimization. The quality of this spatial interpolation is assessed using a confusion matrix that compares the actual SWOT estimates with the interpolated values. Our interpolation method improves the true positive water detection rate from 62% to 85-86% when compared to the simple adjunction of permanent water, meaning that it is able to capture dynamic information from the SWOT available pixels. The interpolation reduced the difference in water area between the descent and ascent paths. This could enable a clearer detection of temporal dynamics in surface water. The new interpolated SWOT water maps can better capture the seasonality of flooded/saturated or forested riverine wetlands and peatlands (based on the “The Global Lakes and Wetlands Database” (Lehner et al., 2024)). The new, interpolated and completed SWOT water maps can more easily be used by the hydrology community. We expect to improve the interpolation strategy in the future and apply it at the global scale.

Nguyen, T. H., & Aires, F. (2023). A global topography-and hydrography-based floodability index for the downscaling, analysis, and data-fusion of surface water. Journal of Hydrology, 620, 129406.
Lehner, B., Anand, M., Fluet-Chouinard, E., Tan, F., Aires, F., Allen, G. H., ... & Thieme, M. (2024). Mapping the world’s inland surface waters: An update to the Global Lakes and Wetlands Database (GLWD v2). Earth System Science Data Discussions, 2024, 1-49.