{A numerical scheme is developed to simulate the transport of natural gravel. Starting with computerized tomographic (CT) scans of natural grains, our method approximates the shapes of these grains by “gluing” spheres of different sizes together with overlaps. The conglomerated spheres move using a Discrete Element Method (DEM) which is coupled with a Lattice Boltzmann Method (LBM) fluid solver, forming the first complete workflow from particle shape measurement to high resolution simulations with hundreds of distinct shapes. The simulations are quantitatively benchmarked by flume experiments. The numerical tool is used to further validate a recently proposed modified sediment transport relation, which takes particle shape effects into account, including the competition between hydrodynamic drag and material friction. Unlike a physical experiment, our simulations allow us to vary the hydrodynamic drag coefficient of the natural gravel independently of the material friction. Our studies support the modified sediment transport relation. The simulations also provide insights on the particle-level kinematics, such as particle orientations, in the bedload transport process. Particles below the bed surface prefer to orient with their shortest axes perpendicular to the bed surface, but the tendency goes down as the packing fraction decreases far from the bed surface. The particles rotate freely in the dilute particle flow regime. }