Unsteady Flow HEC‑RAS Model Troubleshooting

The following sections describe various issues that should be considered when troubleshooting an unsteady flow HEC‑RAS model.

Time Step Issues

  1. Model stability can be very sensitive to the computational time step. Lowering the computation time step may reduce computational instabilities and make the model more stable.
  2. Unsteady flow HEC‑RAS computational time steps are generally equal to or greater than the representative cross section spacing distance / maximum flow velocity.
  3. Too large of a computational time step can cause the hydrograph peak flow to missed at some of the cross sections, causing the model to become unstable.
  4. Too small of a computational time step can cause the model to become unstable. For example, defining too short of a time step can cause a stable model to suddenly become unstable. In addition, the computational run times can get overly long.
  5. A practical rule of thumb is to define the computational time step equal to the inflow hydrograph time to rise / 24. For example, if the time to rise for the inflow hydrograph (from base flow to peak flow) is 4 hours, then set the computational time step to 4 / 24 = 0.1666 hrs or 10 minutes.
  6. However, for dam failure models a much shorter time step is required to account for the steep rise in the flood wave being routed down the river. Typical time steps for dam failure models range from 1 minute on down to 1 second due to the fast flood wave velocity and change in discharge.
  7. A trial and error method can be used to test various computational time steps to see what the largest time step works while providing accurate results and minimal convergence errors. For example, try 10 minutes, then 5 minutes, then 2 minutes, etc.

Cross Section Issues

  1. Place additional cross-sections at locations where the model cannot converge on several time steps. However, if the cross sections are placed too close together, then the numerical solution will cause wave steepening and the model will go unstable on the rising limb of the routed flood wave.
  2. Generally, additional cross sections are required (i.e., cross sections closer together) for the following situations:
    • Transition zones where flow is flowing out of the channel into the overbank area, or vice versa.
    • Vertical slope changes, where the flow is going from flat mild slope to steep slope, or vice versa.
    • Steep river reaches where supercritical flow is possible.
  3. It is better to use real terrain geometry for constructing additional cross sections rather than just interpolating cross sections.
  4. Check the ineffective flow areas top elevations at roadway crossing structures. If a roadway structure is overtopped, the ineffective flow area on the downstream side of the structure should also overtop and not be blocked.
  5. Ineffective flow areas should be marked as “permanent” so that during the routing, the ineffective flow areas do not “suddenly disappear” during an iteration or time step. If the water surface temporarily overtops the ineffective flow area, this flow area will suddenly cause the large increase in conveyance area and cause the model solution to oscillate by having an ineffective flow area for one iteration or time step and then having no ineffective flow area in the next iteration or time step.
  6. Inserting a pilot channel at wide flat cross sections can stabilize the model during low flow conditions.
  7. Adjust the cross section hydraulic parameters starting elevation (used for computing hydraulic flow property tables for each cross section) so that the starting elevation matches the cross-section invert.
  8. Revise the hydraulic parameters for every cross-section and structure to provide additional refinement to provide smooth conveyance curves. Abrupt changes in the computed conveyance curves can be reduced by adding additional horizontal Manning’s roughness locations.

Boundary Condition Issues

  1. A HEC‑RAS unsteady flow model cannot have a zero base flow. There must be water present in the model.
  2. A HEC‑RAS unsteady flow model cannot go dry during a simulation. Adding additional flow to the initial base flow, keeping it as small as possible, may improve stability during low flow conditions. In addition, make certain that the defined initial flow matches the defined inflow hydrograph value at start up.
  3. A minimum flow threshold value can be used at inflow boundaries to prevent the base flow from falling below the defined threshold.
  4. If the inflow hydrograph time to rise is too short, this can cause the model to experience a numerical shock. For example, change the flow from 100 cfs to 10,000 cfs in just a few time steps can cause the model to fail. Take the rising limb hydrograph data and manually stretch it out for a longer time period.
  5. Inconsistent initial conditions can cause the model to go unstable immediately upon model start up.
  6. Initial condition flows must be consistent with the boundary condition flows at time zero. However, internal reaches, such as in a dendritic model, do not require initial flows to be specified. The HEC‑RAS software will compute these initial flows automatically based upon the flows coming into each connected junction.
  7. Downstream boundaries cannot have a flat or adverse (negative) slope. If necessary, extend the downstream end of the model to provide positive (downward) slope for the channel bed. Otherwise, the software will not be able to compute an accurate loop rating curve (aka, hysteresis loop) at the downstream boundary of the model.
  8. Storage areas initial water surface elevations need to be consistent with the initial flows and gate settings defined.

Computational Options Issues

  1. Increase the maximum number of iterations from the default value of 20 to the maximum allowable value of 40. Increasing the number of iterations will generally improve the convergence accuracy of the model—especially when the model has lateral weirs and storage areas.
  2. Adjusting the Theta implicit weighting factor (0.6 to 1.0) can improve model stability or increase the accuracy of the output. Increasing this value towards 1.0 will increase model stability, decreasing this value towards 0.6 will increase model accuracy. A default Theta weighting factor value of 1.0 is used by HEC‑RAS.
  3. Do not use the “Convert Bridges to Lids” option. This option often causes model instability and was removed in HEC‑RAS version 5.0.
  4. Decrease the computational tolerances (increase the values) used in the software to determine when convergence has occurred. While the default values are good for most unsteady flow models, slightly increasing these values might suddenly help in getting the model to stabilize. The default values used for unsteady flow routing are:
    • Water surface elevation = 0.02 ft
    • Storage area elevation = 0.05 ft
    • Discharge = BLANK (not used)

Other Issues

  1. If a drop structure is present in the HEC‑RAS model, the best way to represent this is as an inline structure.
  2. Lateral and inline structure stability factors can improve model stability. To improve model stability, try to get the stability factors close to a value of 1.0.
  3. Check the weir coefficient used for roadway decks.
  4. If the model crashes at the beginning of a simulation, check the initial conditions such as discharge values, storage area elevations, and downstream boundary conditions. If the model crashes during the simulation, check the computed water surface profiles to pinpoint the model stability locations.
  5. For models that start up unstable, define the model so that the river network is flooded (i.e., high downstream boundary condition). Then, gradually lower the downstream water surface elevation to match initial conditions for model.

Other Tips

  1. Try running the unsteady flow model with steady state flow conditions (i.e., a hydrograph with a constant flow value over time). This will help you identify what other issues could be causing the model to fail in its computations.
  2. Eliminate structures (i.e., roadway crossings, inline structures, lateral structures) in the model. Perhaps one of the structures is causing the model to fail. If the model suddenly runs after removing the structures, then place the structures back in, one at a time, to see which structure(s) are causing the instability.
  3. Replace any steep reaches with an artificial inline structure. This will help in transitioning the flow from an upper reach to a lower reach, bypassing the steep reach where supercritical flow could be occurring which often causes model instability.
  4. Use the Modified Puls Routing option to represent steep river sections or low flow conditions.
  5. Use the Variable Time Step option where the HEC-RAS computational engine dynamically recomputes the required time step during the simulation based upon the Courant number specified.

About the Author Chris Maeder

Chris Maeder

Chris is an experienced civil engineering and software technology leader, with over 30 years industry experience. With proven expertise in global software development, he has built engineering teams that adapt quickly, focus on what’s important and, most importantly, deliver. He is a licensed professional civil engineer with extensive experience in water resource engineering. He has performed and supervised engineering projects in urban stormwater drainage, transportation and roadway drainage, storm sewer design, detention pond design, stormwater quality, green infrastructure, watershed management planning, wastewater sewers, water distribution networks, pump stations, FEMA flood studies, bridge and culvert design, bridge scour and armoring, dam failure analysis, seepage and groundwater modeling, and environmental permits.

Related Articles