• Proposed an optimized SAR-based method for river mapping in steep mountainous area. • 3 algorithms included to reduce influence of spatial heterogeneity, layover, shadows. • Achieved water body extraction along a 768 km stretch of the lower Jinsha River. Accurate river mapping in steep mountains is essential for disaster warning, water resources, and ecological management, yet remains challenging due to complex terrain and dynamic hydrology. Conventional methods provide limited spatial coverage, while optical remote sensing is often hindered by cloud and fog. Although Synthetic Aperture Radar (SAR) enables all-weather observation, current water extraction techniques perform poorly in mountains owing to three major challenges: backscatter heterogeneity, layover, and shadow distortions. To address these issues, this study proposes an optimized water extraction method using dual-polarization SAR data. Our method combines: (1) A window-adaptive Otsu thresholding method based on the Sentinel-1 Dual-Polarized Water Index (SDWI) to mitigate spatial backscatter heterogeneity; (2) The synergistic use of ascending and descending orbit SAR data to reduce the layover effect; and (3) DEM-based edge detection combined with blocking elevation thresholding to eliminate shadows. Based on this framework, we achieved high-precision water body extraction for a 768-km reach of the lower Jinsha River mainstream. Comparative analyses with traditional water extraction methods (threshold segmentation and Otsu’s method) and machine learning models (Random Forest and Support Vector Machine) demonstrate the superior accuracy of our framework. It achieved the highest performance, with a user’s accuracy and Kappa coefficient reaching 96.99% and 95.53%, respectively. Furthermore, we discuss how the proposed method effectively mitigates the impacts of SAR geometric distortion and spatial heterogeneity in backscattering coefficients. Ablation experiments were also conducted to validate the contribution of different modules within the framework. This study developed a technical framework for high-precision automatic extraction of river water body in steep mountainous regions using SAR, which holds significant value for flood risk assessment.
Wang et al. (Sat,) studied this question.