## Lateral Structure on River Bends

Written by Sebastian Ferraro | April 7, 2021

Figure 3.  Lateral structure split into four smaller lateral structures.

Splitting the lateral structure allows each lateral structure to deliver flow to a set of cross sections by assigning the tailwater flow to go to “a point between two XS’s”. Here we set up the multiple lateral structures so that each one only delivered flow between 2 cross sections.  Each lateral structure will deliver flow between its specified cross sections, transferring flow across the downstream cross sections. When you compare the results across the lateral structure and both the upstream and downstream end of the bend, you see that without the lateral structure there is a significant backwater behind the bend (upstream) and a lower WSE on the downstream end (Figure 4). With the lateral structure in place, the energy gradient is reduced significantly and results in a more reasonable water surface profile in this area.  Since the lateral structure uses the weir equation, it is important to select a weir coefficient that minimizes the head drop over the spine of the bend.  For this example, we used a weir coefficient of 1.2 for each of the four lateral structures.  While traditionally we think of weir coefficients in the 2.5 to 4 range, using the weir equation to simulate overland flow requires a smaller coefficient.  This post by Chris Goodell includes a good explanation of weir coefficients for different applications.  But be careful not to use a weir coefficient that is too low, as the head drop over the weir could be too great.  Of course the amount of head drop is subjective so consider a sensitivity analysis and calibration if possible.

Figure 4.  Differences with and without lateral structure on the bend.

Still, a strong argument could be made to make this portion of the model a 2D area and let HEC-RAS figure out the flow transfer around the bend and over the high ground at the spine of the bend-especially considering the uncertainty of the weir coefficient to use.  But if you do have to keep the model 1D, this would be a suitable method to simulate this effect.