Flow loss in irregular, open channel
I am troubleshooting an issue where the peak flow in an irregular, open channel is decreasing from upstream to downstream. For example, three conduits in series moving from upstream to downstream have peak inflows of 4,300-cfs, 3,197-cfs, and 2,178-cfs. This continues along the network with a peak inflow at the upstream-most end of 5,691-cfs to a peak flow at the outfall of 2,466-cfs. Losing over half of the flow along the network is greatly impacting my model and providing what seems to be inaccurate or impossible results.
The network was created by importing a HEC-RAS model. It is relatively straightforward and includes a stream with a few culverts. The transects assigned to each conduit vary with respect to transect width, shape, cross-sectional area, manning's n, etc. based on the conditions on site as a natural channel would. I am using the Modified Green-Ampt infiltration model with dynamic wave routing. The storm that I am routing through the model is an SCS Type II, 100-year flood (based on Atlas-14). The model runs for the entire duration of the storm to capture the peak all the way through the network. To help eliminate some possible questions:
Any ideas for additional troubleshooting or possible solutions are welcome! The input file content has been copied and pasted below for reference.
First, runoff from the upstream subcatchment seems very peaky to me. The 100 yr storm runoff lags the rainfall by only 5 mins. I would expect more attenuation for that sized watershed.
Second, plot a profile of the channel for the peak values. Culverts CJ12089 and CJ8307 are causing storage in the channel which then limits your peak flows.
Third, review the transects for bank stations. A few such as 10320 have very shallow (1 ft or so) channels so your effective roughness of channel plus overbank may be too high.
I agree with David about your runoff. You only have two Subcatchments with a peak of 6250 cfs for the upper catchment. The peak flows in your upper transect seem consistent with that peak flow.
Which are your three links in series with the drop in peak flows? I am looking at your flows from upstream to downstream and they seem consistent.
A suggestion for the routing parameters you used. You are using a fixed time step of 0.5 seconds with the Ignore term turned on for the inertial terms. I would suggest using KEEP to fully use the storage in your transects. I would also suggest using a time step of 10 seconds to fully use SWMM5 ability to calculate a good CFL time step for you.
I also see based on what David mentions, a few large offset depths in your model. Most of your offsets are 0 but you have a few over 50 feet. You do not have a lot of long links and the continuity error is small so you can adjust your time step.
Ann - In your original post you mention "Losing over half of the flow along the network". There is no actual loss of volume in your model, only an attenuation of peak flow - volume is conserved (within SWMM's usual numeric tolerances, that is). If I am misunderstanding your concern, and it is the degree of attenuation that you believe is "inaccurate or impossible", then these observations from David and Robert will be helpful.
I would also run the model in unsteady HEC-RAS and see what you get.
Thank you for the feedback. I made several adjustments based on the recommendations below and determined that the culverts/road crossings act like small dams and create a storage/backwater effect that impacts peak flow. This made me think I was somehow losing flow or that there the irregular transects were not interacting well with each other, but, like Peter said, the overall volume is conserved. As a check, I removed the culverts from the model and peak flows were consistent from upstream to downstream and all routing instabilities were eliminated. I will also look into the other recommendations regarding routing time steps and calibrating runoff from the subcatchments to improve the model.