Multi-Channel Networks (Bifurcation and Confluence)

Cabernet2D can compute not just a single simple channel, but a complex river network that includes bifurcations (branches) and confluences, all within a single computational domain. This makes it possible to compute flow division at a bifurcation point (the distribution of discharge between the main channel and a branch), which was difficult with Nays2DH.

Example computation of a multi-channel network with a bifurcation

Fig. 21 An example computation of a multi-channel network with a bifurcation point (main channel and branch channel). The discharge in the main channel and the branch is distributed while varying over time. (Data courtesy of Minami & Yoshitake, 2024)

Switching the calculation mode

In the GUI calculation condition “Type (j_discharge),” choose the mode that matches the shape of the channel network you are computing.

Setting

Description

  1. Simple channel

A simple channel with a single inflow and a single outflow.

  1. Complex channel

A complex channel network (including bifurcations and confluences) with multiple inflows and/or outflows.

When (1) is selected, you can set multiple inflow and outflow boundaries individually on the grid. The location and condition of each boundary are set in the GUI’s boundary condition screen by selecting the corresponding edge(s) of the grid.

Types of outflow boundary condition

For each outflow boundary (in the i-direction and the j-direction), choose one of the following water-level conditions.

Setting

Description

Wall condition (zero flux)

Free outflow (constant water-surface slope)

Constant water surface elevation (m)

Time series of water surface elevation (m)

Free outflow (zero depth gradient)

Critical depth

Uniform flow (Manning normal depth). The target is the sum of the discharge from all inflow boundaries that feed this outlet, and the uniform-flow water level is computed from Manning’s formula at every time step. When choosing this condition, also set the “Slope in i/j-direction at outflow boundary” (the slope used in the uniform-flow calculation).

Note

Condition (6), uniform flow, uses the sum of the discharge from all inflow boundaries as its target discharge. This means it is fundamentally intended for a simple channel with a single upstream and downstream end (or a single outlet for the whole network). In a bifurcating network where you want to give each outlet its own, independent uniform-flow water level (for example, the outlet of a branch channel downstream of a bifurcation), this condition — which uses the total inflow — can over- or under-estimate the water level. Depending on the shape of the network you are computing, consider combining it with another condition (such as a time-series water level) and adjusting it as needed.

You can also enable “Use Riemann invariant condition for outflow boundary” (hRiemann_i/j), which determines the boundary velocity consistently from the internal Riemann invariant under the specified water-level condition (available for the constant, time-series, and uniform-flow water-level conditions).

Practical tips

  • In a network with bifurcations and confluences, the stability of the computation depends strongly on how each outlet’s water-level condition is set. In particular, at small-scale settings such as an experimental flume, where the water depth is small, it is important to set the “Reference depth” — which serves as the basis for the initial value and search range of the water level — to match the representative water depth scale of the channel (for example, on the order of centimeters).

  • In the multi-channel mode, the “Upstream suspended sediment supply ratio” in the calculation conditions is not used; instead, the supply ratio set individually for each inflow boundary is used. If there are multiple inflow boundaries, set this value individually for each one.

  • Because the area around a bifurcation or confluence is where the flow changes most abruptly, consider also using the quasi-3D flow calculation method (Quasi-3D Flow Calculation Method) if you want to accurately reproduce the local scour and deposition pattern there. Depending on the shape of the bifurcation, however, additional work on numerical stabilization may be required.