<p>The micro-morphological structure of riverbeds represents the basic unit of bed surface morphology. Bedload clusters, the most common microforms in mountain rivers, constitute the main component of the surface armouring layer and significantly affect bed stability, sediment transport, and flow conditions. In this study, a fully resolved computational fluid dynamics–discrete element method (CFD-DEM) coupling model, based on the immersed boundary method (IBM) and large-eddy simulation (LES), was developed to simulate the evolution of uniform bedload clusters over rough beds under varying flow intensities. The mechanisms of cluster formation, disintegration, and their influence on flow structures were examined. Results show that cluster formation occurs in two modes: collision-dominated and wake-capture, with the latter tending to evolve into streamwise-aligned patterns. Disintegration arises from self-breakup or collision-induced breakup, with transverse and streamwise velocity fluctuations as key triggers. Larger clusters generate more extensive wake low-pressure zones with increased peak negative pressure, revealing a self-organizing trend in spatial evolution. Clusters with complex morphologies are prone to vortex breakdown, which enhances energy dissipation. Moreover, clusters significantly modify local bed shear stress distributions: a high-stress zone develops at the upstream edge and a low-stress zone downstream due to wake shielding. Importantly, the peak stress magnitude is independent of cluster planar shape. These findings clarify the dynamic interactions between bedload clusters and turbulent flow, providing insights into sediment transport processes and riverbed stability in natural rivers.</p>