Large Eddy Simulation of Sediment Transport and Hydrodynamics at River Bifurcations: Using a highly scalable spectral element based CFD code

Marcelo Garcia, University of Illinois at Urbana-Champaign

Usage Details

A fundamental morphological element present in most river systems is a bifurcation, where a river divides into two different channels. Bifurcations are mostly formed naturally, like the dendritic networks in deltas or the in-stream bifurcations in braided rivers, though over the years bifurcations (also called diversions) have also been built for various engineering purposes like river connectivity, flood protection, and river diversion. For example, in the last few years there has been concerted effort to study the viability of diverting flow (and sediment) from the lower Mississippi River to mitigate potential loss of coastal land in Louisiana, and various designs and locations of diversions are being studied to find an optimal one. Understanding the fundamental mechanics in play at channel diversions will go a long way towards designing better engineering solutions.

At a bifurcation (or diversion) water and sediment from the original channel gets distributed between the two branching channels. Laboratory experiments and field observations have shown that the sediment discharge distribution tends to favor the lateral channel, even in cases where the opposite trend is exhibited by the water discharge distribution. This phenomenon is commonly known as the Bulle effect, after Henri Bulle who conducted the first experiments in 1926 to study the aforementioned non-linear phenomenon. Subsequently there have been a few experimental and numerical studies, but none of them categorically points out the fundamental dynamics behind the anomalous distribution of solids. The current study attempts to fill the gap in fundamental understanding of the Bulle Effect and related phenomena such as secondary flows and vorticity-driven sediment transport. Contrary to previous studies, the simulations in the proposed work will resolve all the relevant turbulence eddies of the flow (as we will be doing LES). This will not only provide an accurate description of the dynamics of the flow, but it will also help to model sediment transport more accurately by using a Lagrangian model for the sediment.

The simulations would be conducted using Nek5000, a highly scalable eddy-resolving computational fluid dynamics (CFD) code based on the spectral element method. Nek5000 has been known to scale very well, not just over number of processes but also across different computing machines. Nek5000 has shown strong scaling up to a million ranks. The simulation results will be compared with published experiments and additional experiments that will be conducted at the Ven Te Chow Hydrosystems Laboratory.

Apart from helping the fundamental understanding of the fluid dynamics of river bifurcations, it will also provide insights that will help improve numerical models of bifurcations used for field-scale simulations. The Reynolds number of the planned simulations are of the order 105~106, which is comparable to laboratory experiments (though not comparable to those encountered in the field). The current study pushes the limit of the scale at which eddy-resolving numerical simulations have been used to study problems in river mechanics, thus the need to use a sustained petascale computing resource like Blue Waters. Finally, bifurcations are not only found in rivers but also in other places, like the carotid bifurcation in the human body. Thus, the current study will also contribute to our general understanding of dynamics and transport at bifurcations.