A few new 1D only computational options have been added for version 6.0, and one existing option has been removed.  The General tab, selected from the Computational Options and Tolerances window, contains the following new 1D options:

  1. 1D Finite Volume Solver. HEC has developed a new 1D Finite Volume solution algorithm for solving the 1D Shallow Water equations.  The current 1D Finite Difference scheme has trouble in the following situations:
  • Can’t handle starting or going dry in a cross section (XS).
  • Low flow model stability issues with irregular XS data.
  • Extremely rapidly rising hydrographs.
  • Mixed flow regime (i.e., flow transitions).
  • Stream junctions do not transfer momentum.

The new 1D Finite Volume algorithm has the following positive attributes:

  • Can start with channels completely dry or they can go dry during a simulation (wetting/drying).
  • Very stable for low flow modeling.
  • Can handle extremely rapidly rising hydrographs without going unstable.
  • Handles subcritical to supercritical flow and hydraulic jumps better – No special option to turn on.
  • Junction analysis is performed as a single 2D cell when connecting 1D reaches (continuity and momentum is conserved through the junction).

Additionally, the new 1D Finite Volume solution scheme breaks up the cross section into separate cells for the left overbank, main channel, and the right overbank.  The current finite difference scheme combines the properties of the left and tight overbank into a single flow area called the floodplain.  Additionally, the hydraulic properties of the left and right floodplain are combined and the reach lengths are averaged together.

Stream junctions are also handled very differently in the new 1D Finite Volume scheme.  The 1D finite difference scheme had two options.  The default option was to assume all cross sections bounding the junction had the same water surface elevation.  The second option allowed the user to compute the water surface across the junction using a steady flow energy solution each time step.  For the 1D Finite Volume solution, the junction is treated as a single 2D cell and the 2D shallow water equations are solved through the junction.  This solution allows for more complex junctions within a 1D framework.  Also, this approach allows for the transfer of momentum through the junctions as well as a more detailed calculation of the water surface elevations around the junction.  The 1D model must be georeferenced in order for this to work correctly, as that is how the software forms the 2D cell correctly.  See the example in Figure 5-9 below.

Figure 5-9. 2D Cell used for Modeling Junctions in new 1D Finite Volume Scheme.
 

To use the new 1D finite volume solution scheme, select Computational Options and Tolerances from the Unsteady Flow Analysis window.  Then from the HEC-RAS Unsteady Computation Options and Tolerances editor, select the General tab (Figure 5-10).


Figure 5-10. General Tab from the Unsteady Flow Analysis Computational Options and Tolerances.

As shown in Figure 5-10, to turn on the new 1D Finite Volume solution scheme, simply select the Finite Volume (new approach) radio button from the 1D Numerical Solution area in the lower right-hand side of the window.  The user can also control the number of cores used to solve the problem.  Most of the time, the solution will be solved the fastest with a single core.

An example of a model that was started completely dry, and containing bridges, is shown in Figure 5-11.


Figure 5-11. Example Model using 1D Finite Volume and Starting Dry.
 

In general, the 1D Finite Volume solution algorithm is more robust (greater model stability) that the existing 1D finite Difference solution scheme.  However, there are some draw backs to the new 1D Finite Volume solution scheme.  These deficiencies are:

  1. Users cannot use lidded cross sections with the new 1D Finite Volume solution scheme. This also means that it cannot handle pressurized flow, even with the Priessman slot option turned on.  If the model contains lidded cross sections, as soon as the water hits the high point of the lids’ low chord it will go unstable.
  2. The 1D Finite Volume solution scheme is sensitive to the volume of water between any two cross sections. The equations are written from a “volume” perspective.  If two cross sections are very close together, then there is very little volume between those two cross sections.  The 1D Finite Volume scheme will require smaller time steps in order to handle the change in volume over the time step for this type of situation.  The 1D Finite Difference scheme handles this better, because the equations are written in terms of change in water surface and velocity, which may be a small change for cross sections that are closer together.  Therefore, for the 1D Finite Volume scheme to work well with larger time steps, users may need to remove cross sections that are very close together.
  3. The 1D Finite Volume scheme is computationally slower than the 1D Finite Difference solution scheme. This is because the 1D Finite difference scheme combined the properties of the left and right overbank together (area, wetted perimeter, average length between cross sections, etc.) and solved the equations for a main channel and a single floodplain.  The new 1D Finite Volume solution scheme keeps the left and right overbank properties completely separate.  Thus the equations are written for a separate left overbank, main channel, and right over bank, and then solved.  This is more computational work for every time step, but it is computationally more accurate.
  4. The 1D Finite Volume Solver cannot be used with the Modified Puls Routing Option. The Modified Puls Routing option cannot be used with the 1D Finite Volume solution scheme.  This was done on purpose, as it was deemed users would not need the Modified Puls Routing option since the Finite Volume routing method can handle dry channels, extremely low flows, rapidly rising flows, and very steep channels.  Modified Puls routing was added to HEC-RAS because the original 1D Finite difference solver has trouble in the areas previously described.


2.   Maximum number of Iterations without improvement. This option is off by default.  If the user turns it on, the maximum numerical error computed during the 1D Iterations will be monitored.  If the error does not improve within the specified number of iterations, then the 1D solver stops iterating and goes on to the next time step.  For example, let’s say the default maximum number of iterations set to 20.  If the “Maximum number of iterations without improvement” is set to 5, then during any time step, if the iteration scheme does not continue to improve the numerical solution for 5 iterations in a row it will stop and go to the next time step, using whichever previous iteration was the best solution.  In general, 5 is a good number to start with for this option, but the user may want to try lowering it.  This option will improve computational speed for data sets that iterate a lot.  However, if the user turns this option on and sets the value too low, then the result may be an increase the model instability. 

3. 1D Equation Matrix Solver. HEC-RAS uses a matrix solution solver called “Skyline” which uses Gaussian elimination for reducing the size of the matrix.  The 1D Equation Matrix Solver has been optimized for dendritic river systems and is very fast.  However, sometimes HEC-RAS models can be very large and have many interconnections (loops in the stream network, or many interconnected storage areas).  The HEC-RAS team has added an option to solve the 1D matrix with the “PARDISO” solver, which is used by the 2D solver.  This solver has the benefit of being able to use multiple cores.  Results from experiments conducted at HEC have found that the Skyline Matrix solver is still faster for dendritic systems.  However, large models with lots of lateral structures, storage areas, and loops in the reaches may be solved faster using the PARDISO solver.  Users can try the Skyline Matrix Solver out to see which one works better on a specific data set.  However, the HEC-RAS team do not have a lot of experience in using this solver on the 1D side.  So if using the 1D Equation Matrix Solver be aware of the risk.  In other words, don’t just compare the computational times, also compare the results to make sure they are the same.

Note:  The HEC-RAS team removed the option to “Convert 1D Energy Bridges to Cross Sections with Lids.”  This option was not used often, and in some cases caused model stability issues.  So now all bridges are pre-processed into a family of curves.  If this option was turned on in a model built and computed in an older version of HEC-RAS, then this change may produce different computed results in the vicinity of that bridge.