Satellite-Based Discharge Estimation in Morphologically Dynamic Rivers: A Segmented Modeling Approach for the Ganjiang River Basin

River discharge estimation is critical for flood forecasting and water resources management, yet traditional gauging methods are often limited in spatial coverage. Accurate estimation of river discharge from satellite observations remains challenging in large rivers where hydraulic controls and anthropogenic disturbances induce non-stationary width–discharge relationships. In this study, multi-decadal river width time series derived from multi-sensor satellite imagery (Landsat-5/7/8 and Sentinel-1/2) were employed to estimate discharge in the Ganjiang River Basin, China, using the Google Earth Engine (GEE) platform. Particular emphasis was placed on quantifying the impacts of backwater effects and channel morphological changes on inversion accuracy. Results indicate that: (1) satellite-based width–discharge scaling performs robustly in morphologically stable reaches, yielding high accuracy at the Ji’an, Xiajiang, and Zhangshu stations (R² > 0.92$; NSE > 0.90); (2) in contrast, performance at the Waizhou station is strongly degraded by complex hydromorphological dynamics, where intensified backwater effects from Poyang Lake during the wet season weaken the functional coupling between river width and discharge (R² decreases to 0.59), and pronounced channel incision associated with historical sand mining (mean bed lowering of 2.97 m) introduces additional non-stationarity into the rating relationship; and (3) to account for these time-varying controls, a segmented modeling framework was implemented to explicitly reflect periods of morphological adjustment, substantially improving discharge estimates at Waizhou and increasing both R² and NSE to 0.90 from 2012 to 2019. These findings highlight the importance of considering morphodynamic evolution and hydraulic boundary conditions explicitly for reliable satellite-based discharge estimation in dynamically evolving river–lake systems.