Towards daily river monitoring from space: Reach-Reg, a first approach for spatiotemporal water level reconstruction using the SWOT geometry

The Surface Water and Ocean Topography (SWOT) mission recently overcame the spatial limitation of satellite altimetry by using the wide-swath interferometry to simultaneously observe Water Surface Elevations (WSE) of inland waters within a 120 km wide swath. Thanks to this unique measurement principle, many river reaches are observed multiple times (typically two to four times) during SWOT’s 21-day repeat cycle. Nevertheless, even with these multiple observations, the resulting temporal resolution remains insufficient to capture the rapid sub-weekly dynamics of flood events. In this study, we present a novel approach to bridge this gap by exploiting the unique SWOT data geometry to densify multi-mission observations at Virtual Stations (VS) into a daily product.

Our method, named Reach-Reg, introduces a first approach to generate daily WSE time series from multi-mission satellite altimetry. Reach-Reg employs WSE data from Sentinel-3A/B, Sentinel-6A, and SWOT (provided by DAHITI, https://dahiti.dgfi.tum.de/). Our method introduces a chained regression approach that utilizes WSE from concurrent SWOT overflights to establish linear relationships between neighbouring river reaches. These regressions enable the precise transfer of the WSE measurements from various VS to a central Reference Station (RS). To ensure physical consistency, Reach-Reg employs a Manning-based time-lag correction using at-a-station hydraulic geometry simplification and calibration based on cross-validation. The daily product is achieved through the automated outlier rejection, smoothing, aggregation, and interpolation. Crucially, Reach-Reg also provides a rigorous uncertainty estimate.

We evaluated Reach-Reg across 95 RS on eight rivers (Elbe, Ganges, Mississippi, Missouri, Oder, Rhine, Po, Solimões) spanning four continents. These rivers represent diverse hydrological regimes and bed morphologies. Despite this variability, the method achieved consistent high performance, with mean RMSE of 0.30 m, normalised RMSE of 4.2%, and Nash-Sutcliffe Efficiency of 0.94. Notably, Reach-Reg significantly outperforms existing multi-mission densification techniques by improving both temporal resolution (daily vs. 2–5 days) and vertical accuracy (0.30 m vs. ≈1 m mean RMSE). Furthermore, this approach is computationally efficient (processing one station in ≈1 minute), open-source, and based solely on altimetry, ensuring global transferability. The introduction of daily WSE from multi-mission satellite altimetry will enable better understanding and modelling of river dynamics. It will also make it possible to issue WSE forecasts even in ungauged basins.

The research has been carried out in frame of the project no. BPN/BEK/2024/1/00047 within the Bekker Programme of the Polish National Agency for Academic Exchange. The Python implementation of the Reach-Reg method is available on GitHub (https://github.com/MichalHalicki4/Reach-Reg) and the daily WSE time series can be obtained from the Zenodo repository (https://doi.org/10.5281/zenodo.17928117).