Observing and modeling bedload sediment transport at the grain-scale

Empirical sediment transport models have common characteristics suggestive of the underlying physics, but mechanistic explanations for these characteristics are lacking due to an incomplete understanding of the fundamental physical mechanisms involved. Hydrodynamic interactions at the grain-scale are thought to be key, however, it is a major challenge to either observe or model these processes. In order to improve our understanding of grain-scale dynamics in sediment entrainment and transport we are studying the detailed mechanics of fluid-grain interactions using a combination of laboratory flume experiments, advanced numerical simulations, and granular mechanics theory. 

The flume experiments are conducted with an emphasis on exploring differences and similarities in the behaviour of glass spheres, a common theoretical tool, to naturally sourced river gravel. Using high-speed cameras coupled with computer-vision based particle tracking, we tracked the majority of grains in the grain bed and water column, with 130,000 glass sphere track paths longer than 10 particle diameters. In particular, we introduce a newly developed a machine learning based particle tracking of the natural grains, with 30,000 gravel track paths longer than 10 mean particle diameters. Fluid flow fields are also observed using particle image velocimetry (PIV). We present the comparison of our detailed observations of granular dynamics between spheres and natural gravel, with a focus on how grain shape impacts fluid-grain and grain-grain interactions.

Using a discrete-element plus Lattice-Boltzmann fluid method (LBM-DEM) we simulate a small portion of the laboratory flume with high temporal and spatial resolution. This method tracks discrete particles interacting with each other through contact laws while mechanically coupled to a dynamic interstitial fluid. We discuss the ability of our simulations to emulate our experiments, the benefits of which are twofold. First, where the simulations work well, we use them to observe grain-scale dynamics that would be difficult or impossible to measure in a laboratory setting or in the field. Second, we learn from situations in which the experiments and simulations diverge, leading to improvements in both the simulations and our understanding of how fluid-grain interactions influence sediment transport.