Braided Floodplain Channel Simulation

Dear All,

I have an interesting SWMM application for you'all: let's say I want to
model a braided channel design in a riparian wetland with EXTRAN. The
braided channel will have 3 channels designed to have roughly the same
HGL. However, these channels will routinely flood and I want to make
sure I account for both the flow split between the channels at low flow
and the storage when the channels go overbank. Ultimately, I want to
create a water depth-duration relationship for the channels and the
overbank area.

Right now I am thinking I will just model the same area twice: the first
time I confirm that I'm getting the flow split I want at low flows. The
second time I just model the entire area as a storage accounting for the
storage of the low flow channels as well as the overbank area.

Any ideas on this? If you think this an application for another model
please let me know. I am still considering the use of RMA2 for this
problem.

Thanks for any and all suggestions.
Scott Dierks, PE
SDierks@ALNM.COM

Here's my 2 cents- Maybe use HEC-RAS? Or use cross-sections in
EXTRAN?...

Good luck
Lillian Lindsay
LLindsay@HNTB.COM

It might be helpful to know a little more about specifically what you're trying to find out. Since the concern is low flows, are you trying to model sediment transport? days of inundation for the wetland? other?

Also, it would be useful to compare the needed accuracy of the hydraulics of the model with the uncertainty/accuracy of the other inputs. For example, its quite likely that the unknowns on the hydrology are at an order of magnitude more severe than any unknowns you might have on specific stages due to flow splits in the braiding, inaccuracy in channel lengths, etc, etc.

The unsteady module of HEC-RAS (UNET-based) will simulate floodplain storage as will SWMM.

With HEC-RAS you have several options to simulate the braided channels

1) as distinct river reaches - This approach requires a set of cross-sections for each channel. Advantages of this approach include: (1) the flow split will be calculated automatically and (2) the length of channel vs length of overbank will be accurate. Disadvantages include: (1) the HGL will only match at the upstream and downstream ends of the wetland - where you have a single-thread stream and (2) it may be difficult to accurately simulate storage in the wetland - you will have 3 sets of cross-sections through the wetland and may have gaps or overlaps in the floodplain.

2) as multiple topographic lows within a single river reach - This approach requires a single set of cross-sections through the wetland. Advantages include: (1) the HGL will match at all cross-sections through the wetlands. Disadvantages include: (1) If the channels are very sinuous, this choice will not accurately simulate channel length vs overbank length, (2) HEC-RAS calculates Q-channel, Q-left overbank, Q-right overbank. You can "trick" the model into giving you the flow splits by using ineffective flow areas (on floodplain surfaces) and choosing bank stations to place the left and right channels onto the left and right floodplain.

There may be other approaches for simulating braided channels with HEC-RAS or other model. I've tried the above and neither is are perfect. Depends on what you really want to get from the model for design.

When designing the channels it will be important to consider what happens as the hydrograph rises and falls - Specifically, how will flow exit the channels, move across the floodplain, and reenter the channels? Braided channels are susceptible to headcuts and avulsions. Is this acceptable for your design? If not, consider how will you minimize potential for occurrence in your design?

Just some food for thought. Good Luck!
Janet Corsale, P.E.
Inter-Fluve, Inc.

Scott,

Using EXTRAN set up the channels and floodplain as a 2-dimensional grid
type network, no different than doing a storm sewer network. The
channels are modeled as irregular channels that extend from top of bank
to top of bank. The floodplain is broken up into grids, not necessarily
rectangular. Each floodplain grid is assigned a node in the center with
an invert elevation equal to the average ground elevation of that grid
and a ground elevation set a little higher than the expected highest
flood elevation to avoid surcharging the pipe. Each node has an
overland flow conduit to the four nodes (grids) bordering it. The
length of each connecting conduit is the actual distance between nodes.
The width is only half the actual width so the area of all the grids
comes out correct. The floodplain conduit between the channel flow
reach and the adjacent floodplain grid is modeled as length from the
center of the floodplain node to the top of channel bank x width to give
correct area for the wedge shaped area between the node and edge of
channel reach. Set the invert of the connecting conduit at the average
ground elevation of the floodplain node at each end. It had straight
canals instead of braided channels. If you choose, some of the
floodplain can be included with the channel, but if most of the flow
occurs in the floodplain, choose the channel upstream and downstream
invert elevations so they represent overland flow slope. Attached is a
sample file of a segment of the St. Johns River in Florida done hastily
for demonstration purposes that was never calibrated. The floodplain
"n" values are for sawgrass. I have a Visual SWMM version if you would
like it.

One other comment on meandering streams. Depending upon direction of
meander, Mannings "n" decreases up to the top of bank, then increases
above that. USGS stream gage measurement on the Ocklawaha River,
Florida show velocities increase with rising stages up to the top of
bank, then they decrease as the flow spills over the top of banks. I
can't recall the direction of the meander where they made this
measurement, but I expect it was made where the river flowed at more of
a right angle to the floodplain.

Robert Freeman
RFreeman@SJRWMD.COM