Construction and validation of data-scarce river channel topography using integrated multi-source remote sensing

Abstract

[Objective] Topographic data of river channels is fundamental for hydrological monitoring. However, in data-scarce regions, existing method struggle to obtain high-accuracy river channel information, which has become a key bottleneck for hydrological observation. Using a 7-km river segment upstream and downstream of the Benzilan hydrological station on the Jinsha River as the study area, this study applies an integrated multi-source remote sensing approach to construct river channel topography, aiming to provide new technical pathways and solutions for terrain modelling in complex river segments and data-scarce settings. [Methods] Based on an unmanned-aerial-vehicle-derived digital elevation model(UAV-DEM) and measured cross-sectional elevation data from the hydrological station, river channel topography within water-level fluctuation ranges from dry and flood seasons was constructed by integrating surface water elevation data from the Surface Water and Ocean Topography(SWOT) satellite and shoreline data extracted from high-resolution remote sensing imagery. Additionally, comparative analysis was conducted with the river channel topographic data(GF7-DEM) generated from GF-7 stereo imagery during the dry season. [Results] The result showed that after correction, the GF7-DEM showed improved accuracy, with RMSE reduced from 0.717 8 m to 0.601 9 m, MAE reduced from 0.561 5 m to 0.476 2 m, and STD from 0.717 5 m to 0.596 m, and R² increased from 0.909 5 to 0.930 1, indicating enhanced representation capability for large-scale river channel morphology.(2) The multi-source remote sensing-derived DEMs—RBF-DEM, EBK-DEM, and IDW-DEM all showed good spatial consistency when compared with UAV-DEM. The RBF method showed optimal performance, achieving the lowest errors(MAE=1.063 9 m, RMSE=1.330 5 m), the least dispersion(STD=1.284 9 m), and the relatively high correlation.(3) For large-scale river channel morphology representation, the corrected GF7-DEM(RMSE=0.601 9 m, MAE=0.476 2 m, STD=0.596 m, and R²=0.930 1) outperformed the RBF-DEM(RMSE=1.330 5 m, MAE=1.063 9 m, STD=1.284 9 m, and R²=0.792 9) in both accuracy and stability. However, when GF7 data were incomplete, the RBF-DEM served as a more effective alternative for terrain reconstruction in data-scarce regions. [Conclusion] GF7-DEM demonstrates strong capability in representing river channel topography, but it has high data acquisition requirements, necessitating the use of dry-season data for DEM generation and offering limited spatial coverage. In contrast, SWOT satellite data are abundant. Together with GF7-DEM, they complement each other to form an approach for acquiring topographic data in data-scarce regions. This provides fundamental support for research applications and management in different fields, including remote sensing-based runoff monitoring, river hydrodynamic process simulation, digital twin watersheds, river habitat evolution.

Keywords

data-scarce river segments / topographic construction / GF-7 satellite / SWOT satellite / accuracy validation / hydrology / unmanned aerial vehicle / multi-source remote sensing