Abstract: In mountainous rivers, the hydrological monitoring network is sparse, and the identification and analysis of large-scale river morphology form the foundation for runoff remote sensing monitoring in data-scarce regions.In this study, water-surface vectors of a mountain river were extracted from remote sensing data.An indicator system for cross-sectional morphological characteristics was established, and K-means clustering method was employed to identify the classification characteristics of cross sections and river reaches, as well as their combined classification features.The results show that: ① according to the cluster analysis results, cross-sectional morphology in the upper Jinsha River can be classified into five types, among which the symmetric canyon-controlled type (Type 0) accounts for 76.98%;the morphological characteristics are significantly associated with the hydrodynamic environment.② Reach morphological complexity increases with reach length (for reaches of 300~1 500 m, the number of clustered classes increases to 19).Short reaches are dominated by a single hydrodynamic process, whereas long reaches are strongly affected by terrain-flow interactions.③ The mean ratio of water-surface width between the dry season and the flood season is 0.52, and a rightward bias of flood-season flow is observed (mean reach morphology index is 1.18), which is consistent with the Coriolis effect.These results provide a scientific basis for runoff estimation in data-scarce regions, hydrodynamic modeling of mountain rivers, ecological restoration, and flood-management practices.