[Scott Miller]

The same weir station is listed twice for cross sections that are 18 feet apart. Start by double checking that. Were they the same in the previous models, or did they differ by even a small amount?

[Scott Miller]

It’s not much information to go on. How did the roadway change? What are the culvert and flows? Any other changes?

[Scott Miller]

From about station 0 to 300 and then 1800 to the end, the weir is below the terrain. That is what the error is about. Either shorten the weir, or raise the ends of the weir above the terrain. To raise the ends of the weir above the terrain, click on the terrain profile button and copy the profile table. Paste it to a spreadsheet and add a few millimeters to the elevation. This helps stay above the headwater and tailwater elevations for each cell, which look like steps in the graphic.

Remove any rows where profile elevations are lower than the weir elevation, and set stations at the ends of the weir to the weir elevation (at about 300 and 1800). Open the weir/embankment data and paste the station-elevation numbers there. Double check to make sure the new elevations do not cut the corners of any of those steps. That should take care of the error.

The breach should not go lower than the headwater terrain elevation, or there will be a similar error.

[Scott Miller]

Yes.

[Scott Miller]

It does not look like there are very many vertices on the TIN. The number of points my ArcGIS defaults to is 1,5000,000. With not very many vertices, the resulting cross section will be very angular.

Take a look at the vertical and horizontal units on the cross section. GeoRAS may be having difficulty with the vertical exaggeration.

[Scott Miller]

The friction locations can be changed in Cross Section Data. Add or change the value in the last column. If the cross section distance is not there, add a row. The locations and values can also be changed in the Graphical Cross Section Editor.

I don’t think there is a direct way to control the interpolation. It is linear between the end cross sections.

[Scott Miller]

Try increasing the “Maximum Number of Points” when converting from raster to TIN in 3D Analyst. That ought to smooth the channel out. I usually bump it up by a factor of 10. It also helps to clip (data management) the raster to the area where the cross sections will be cut. Check the “Use Input Features…” box. Each of these approaches puts a greater density of vertices where they are needed.

[Scott Miller]

Does the model go unstable every run, regardless what the initial flow is? Set the output intervals to 1 minute. Then run the unsteady model and look at the animation of the profile. This can reveal where instabilities start.

(Click on the animation/”play” button, then right click the profile plot for “full extent”… the view ought to zoom in to the channel profile. Click on animation again to get the animation tool bar back.)

The initial flow should almost always match the initial flow in the hydrograph. It may be useful to set a minimum flow in the boundary conditions. Make it a small fraction of typical flows, but not so small the channel goes dry anywhere.

Low flows can be problematic where they arrive at a steep reach. A cross-section may dewater in a time step. It may help to put more cross sections in where the bed slope is steep or the channel changes rapidly. Make sure the distance between cross sections and the time step make sense for velocities in the channel, and that the velocities are reasonable.

[Scott Miller]

The model is 1D only, right? What are the low end flows? How long does it run in model time? How well does the initial flow match the first flows to enter the reach?

The steep downstream end may be contributing to instability. If the model runs for long enough to output a few minutes of data, take a look at what the profile does during that time.

[Scott Miller]

Culverts in HEC are modeled using FHWA standard equations for culvert hydraulics. They can be partially filled with material, which give the culvert a flat bottom.

In my state, we have a stream crossing guideline that encourages/requires stream simulation culverts. There has to be a channel in the culvert matching the upstream and downstream channel. For this reason I use a bridge with abutments to model a box culvert. I believe using a culvert routine would misrepresent the channel cross section within the culvert.

A storage area with pumping could be used to model diversion and dewatering. A partially obstructed, upstream culvert may serve as a coffer dam. Be careful not to fully dewater the channel if the model is unsteady.

[Scott Miller]

It would be good to know whether the two-cross section model would work. My initial thoughts, once I understood the model configuration, were about how few cross sections that is… and that there are issues with the river slope. Does a two cross section model with stage boundary conditions work?

An alternative to full 2D would be 1D main channel / 2D overbank. It’s probably the best way to use HEC RAS to track overbank flows. You have the road profile to use as lateral structure elevations. Calibrate up to the road becoming overtopped, and use your best flow or hydrologic data for higher flows. That lee side gauge ought to help.

[Scott Miller]

Got it. I was referring to the channel in the HEC RAS model. I don’t expect this approach will work well. There’s head to drive flow across the road, but there is also a downstream component to the flow, parallel to the road. 1D cross sections should always be cut perpendicular to the flow direction.

Here are two approaches to compare:

Broad crested weir. Solve analytically for flow across the road. Break the road into discreet weirs each with it’s own elevation. Adjust the road elevations or the gauge stage to account for the slope of the main water course. Figure cross flow is going to affect the weir coefficient. That may be enough to develop a stage-discharge curve and get to the answers.

Full 2D model. Include the main water course along the length of road. Use the stage-discharge for the water course side gauge to determine the flow in the main watercourse for the upstream boundary condition. Augment that, if needed, with the weir flow stage-discharge curve. Use the road profile for a 2D flow area break line. Elevation data as coarse as a USGS DEM ought to be usable at this scale.

I’m just not sure whether there’s sufficient drop on the land side of the road to model as a weir. I hope this helped.

[Scott Miller]

I’m reading that it is an unsteady 1D model with two cross sections 24 feet apart, the width of the road. The channel, crossing the road, is as much as three miles wide. Flow is lost from the main water course, without return flow along that length of the road, and that’s what the model is intended to get at. Is that correct?

[Scott Miller]

You can make a model with a stage time series boundary at both the upstream and downstream end.

If this is a 1D model, you have to specify an initial flow just to get things up and running but after that RAS will compute the flow.

Wesley,

This is fitting gradually varied flow to a prismatic channel, assuming uniform friction. Do I understand correctly? Or does the cross section vary?

[Scott Miller]

Open the cross-section data and click on the button for the graphical cross-section editor. You can use the LB and RB buttons to quickly move the bank stations.

[Author]

Posts