Dam Breach Modeling Q & A

Written by Chris Goodell, P.E., D. WRE | WEST Consultants
Copyright © RASModel.com. 2010. All rights reserved.
Some questions and answers related to dam breach modeling in HEC-RAS…

Question. The Sunny Day model has a consistent water surface elevation from the very start of the model – it only decreases once the breach occurs. How is HEC-RAS setting this starting WSEL?

Answer. You define the starting water surface elevation either by equalizing the flow from your outlet works with the reservoir inflow, or by setting an initial conditions water surface elevation in your flow editor and a pilot flow through the dam equal to the reservoir inflow at the beginning of the simulation.

Question. My breach models show a dramatic decrease in max Q from the cross-section immediately downstream of the dam to the end of the model. I know that HEC-RAS has an inherent storage routine that attenuates the flow throughout the model but is it reasonable to have a result that shows a beginning max Q of 12,370 cfs and an ending Q of 275 cfs (the reach is approx. 3.8 miles long with a slope of 0.02 ft/ft upstream and 0.001 ft/ft downstream)? This is an arroyo about 800-900 ft. wide, Manning’s at .055.

Answer. I would be skeptical of those results. Perhaps there is an error somewhere in the simulation, or you have a lot of flow leaving the system. Sometimes, if your model is not properly constructed, you can develop a large “wall of water” in profile view. A lot of times this is due to poorly defined HTAB parameters. This will create an artificial pool of water behind the wall, which will drastically attenuate your flood wave. Look in the profile plot and animate through your simulation. If you see an unexplainable wall of water backing up flow, that would be the cause.

Question. My models are stable but still have inherent errors (max iterations) and critical depth defaults to varying degrees. Does this have a significant effect on the model results? Changing parameters at this point to reduce inherent errors most likely will cause instability.

Answer. Max iterations are not necessarily a problem as long as the associated errors are small and it is not causing visible instabilities or obvious errors in your results. I try to get rid of all max iterations where possible. If not possible, I try to get the errors below 0.1 ft as much as I can (my own rule of thumb). RAS does not typically default to critical depth in unsteady flow (like it does in steady flow). But it sounds like you have areas that have flow close to critical depth. This can cause instability problems. If you believe flow should be close to critical depth in these locations, try turning on the Mixed Flow option and adjusting your LPI factor. If you do not believe flow should be near critical in these locations (most of the time in natural streams you should not see critical or supercritical flow), then you may be underestimating your Manning’s n values. Manning’s n values for the front end of dam breach flood waves and steep reaches are frequently underestimated. Check Jarrett’s equation if in a steep reach. Your reach slope of 2% is quite high. An n value of 0.055 is possibly too low during the low flow period preceding the dam breach flood.

Question. Does the number of vertices defining a cross section matter, in another words, does the model run better with cross sections that have fewer vertices but still accurately define the section, vs. similar sections that have many redundant vertices?


Answer. Better definition is usually advantageous. RAS does not like to have long horizontal portions of cross sections which is common for coarsely-defined cross sections. It can cause numerical problems. These days, having the maximum number of points in a cross section (500) typically does not noticeably slow down computation speed. I recommend getting as much detail as you can in your cross sections.