Abstract
Classifying river morphology is crucial for fluvial geomorphology and hydrology. River morphology reflects hydrodynamic and sedimentary processes, providing critical insights into the diversity of global river systems. This study establishes a global framework for river morphology classification based on remote sensing and topographic data. Using the Global Inland Water Dynamics Characterization dataset and the global digital elevation model ASTER GDEM V3, a river spatial image decomposition process was developed, dividing global river data into tens of thousands of image blocks containing dynamic imagery and elevation information. A ResNet-50 deep neural network was employed to construct an image-elevation fusion classification model, classifying global rivers into five major types: meandering rivers, braided rivers, straight rivers, anastomosing rivers, and anabranching rivers. These types were further divided into 17 subtypes to capture finer morphological variations. The spatial distribution patterns and morphological features of these river types were analyzed, providing a comprehensive understanding of the global distribution of river planforms. This framework advances the knowledge of river systems at a global scale and lays the foundation for future studies in fluvial geomorphology and hydrology.